RESULTS


      In the sections below we present the HRV results pertaining to each disturbance process, cover type, habitat of special interest, and wildlife indicator species. Results pertaining to HRV departure (i.e., Fire Regime Condition Class) are included in theses sections where appropriate. In addition, we present the results pertaining to the effects of scale (i.e., landscape extent) and landscape context (i.e., location) on landscape structure, wildlife habitat, and HRV departure. Lastly, we present a complete list of HRV tables for each cover type for the entire UPL and for each of the four major geographic sub-landscapes (i.e., quadrants).


      Among the methods (see Methods for details) used to derive the results reported below, it is worth briefly noting our method of computing disturbance return intervals and rotation periods. Unless otherwise noted, all return intervals represent the population mean return interval to an individual 25-m cell; that is, the average interval between disturbances to an individual cell. The population mean return interval can be calculated for each cell and displayed graphically as a surface or it can be averaged across all eligible cells and reported as a single statistic. Note, the latter statistic is equivalent to the rotation period (i.e., number of years required to disturb an area equivalent to the total eligible area), which is simply a nonspatial representation of the cell-specific mean return interval because it does not depend on the explicit spatial distribution of disturbances; rather, it depends only on the total area disturbed each time step in relation to the total eligible area.

 

    Disturbance Processes & Dynamics

    Vegetation Patterns & Dynamics

    Habitats of Special Interest

    Wildlife Indicator Species

    Effects of Scale and Context

    Historic Range of Variability Tables


Disturbance Processes & Dynamics


      This report focuses on the effects of two major natural disturbances: wildfire and insects/diseases; the impacts of other natural disturbances during the reference period were likely localized in time or space and therefore probably had far less impact on vegetation patterns over broad spatial and temporal scales than did fire and insects/diseases. In the sections below, we briefly describe the simulated disturbance regime (i.e., spatial extent and distribution, frequency and temporal variability) associated with each of these disturbance processes. In these sections, we refrain from describing variations among vegetation types - this will be accomplished in the section on Vegetation Patterns and Dynamics. In addition, although each disturbance process is discussed separately, reflecting the fact that each disturbance process was implemented as a separate process in RMLANDS, the model did allow for synergisms and feedbacks among disturbances (see RMLANDS - Model Parameterization). Finally, it is important to realize that while the information below is presented as “results”, it could have easily been presented in the methods section as “model calibration”. Key spatial and temporal aspects of the disturbance regime were evaluated during preliminary “calibration” runs, and subsequent adjustments were made to model parameters to effect desired changes. Thus, while the information presented below does in fact represent results (output) of the simulation, it also represents a set of targets used to calibrate the model (i.e., adjust model parameters to achieve desired results). While this may seem a bit circular, it was a necessary process for a complex model such as RMLANDS. Plus, our real emphasis was on quantifying the vegetation patterns and dynamics resulting from these disturbance processes.


Wildfire


      Roughly 99% (655,553 ha) of the landscape was eligible for wildfire disturbance, and this included all vegetated cover types and successional stages despite their pronounced variation in susceptibility to wildfires. As expected, the frequency and extent of simulated wildfires varied markedly among decades (Figure-frequency, Figure-extent). On average, once every two decades, >10% of the eligible area was burned, and roughly once every 120 years, >20% was burned (Figure-recurrence), inclusive of both high- and low-mortality affected areas. Under an extreme case from a single simulation run, the total area disturbed in a single decade was 249,110 ha (38% of eligible area)(Figure-map). This was the result of many individual fires, including a couple very large fires (>50,000 ha) and many intermediate and small fires. In addition, the majority of the area disturbed was from high-mortality fire burning principally through pinyon-juniper woodlands, mountain shrublands, and semi-desert communities.

 

Wildfire Movie - Click here to view a movie depicting wildfire disturbances on the San Juan National Forest (patterns are similar for the UPL) over an 800-year (10-year time steps) simulation representing the reference period disturbance regime. NOTE, this is a 38 Mb Microsoft Media file (.avi) that requires appropriate movie viewing software (e.g., Quick Time Player).


      As expected (given the model structure), the variation in frequency and extent of wildfires was strongly related to climate (Figure-initiations, Figure-extent). It is worth noting that while the number of wildfire initiations was strongly and linearly related to the climate modifier (representing, in this case, the average drought index), the total area disturbed was much less so. This reflects that fact that while wildfires were much more likely to start during drought periods, they were not always guaranteed to expand into large fires. Thus, even under drought periods, there were many decades in which fires burned relatively little area. However, all decades in which wildfires burned >25% of the eligible landscape were during droughts (i.e., climate modifier >1). Interestingly, the most extreme decade of burning (~ 38% of eligible) was during a moderate (not extreme) drought (climate modifier = 1.2).


      Under this wildfire regime, the mean return interval between fires (of any mortality level) to a single 25-m cell varied widely across the forest from 22 years to >800 years, although very little (<2%) area escaped disturbance altogether over the course of an 800-year simulation (Figure-return). As expected, there was a distinct spatial pattern of variation in mean return intervals (Figure-map). In general, there was a somewhat bimodal distribution in mean return intervals related to elevation. Return intervals were relatively short in low-elevation semi-desert grasslands and mid-elevation mountain shrublands and warm, dry forests due to perennially dry conditions and abundant and contiguous fine fuels in these cover types. Return intervals were relatively long in low-to-mid elevation pinyon-juniper woodlands and high-elevation forests, but for different reasons. Pinyon-juniper woodlands lack the consistent fine fuels needed to facilitate fire spread and the moist, cool conditions at higher elevations reduce the flammability of fuels, making ignition and spread difficult under most climatic conditions. Overall, the wildfire rotation period for the eligible area of the landscape averaged 100 years across simulation runs (413 and 132 years for low- and high-mortality wildfires, respectively) and rotation periods were relatively consistent among simulation runs (Table-rotation). Note, the relatively long rotation period for low-mortality wildfires compared to high-mortality wildfires reflects the paucity (<8% of landscape) of cover types (i.e., ponderosa pine forest and warm, dry mixed conifer forest) that support predominantly low-mortality regimes.


      It is also worth noting that contemporary observations indicate that fire is more frequent and extensive in the SW quadrant than elsewhere. Although the cause of this pattern is poorly understood, some plausible reasons include: (1) prevailing winds from the SW tend to blow fires upslope on the west side of the UPL, enhancing their spread; whereas, SW winds blow fires downslope on the east side, impeding their spread; (2) summer lightning storms may be "funneled" into this portion of the UPL because of relatively lower topography just to the SW, between the Wilson Mountains on the south and the La Sal Mountains on the north (Eric Brantingham, GMUG National Forest, personal communication 5/05); and (3) plant productivity (and therefore rates of fuel accumulation) also may be greater in this quadrant because winter storms also are funneled between the Wilsons and La Sals. In accordance with these observations, we simulated a 20% greater frequency of fires in the SW quadrant and 10% greater frequency of fires in the NW quadrant compared to the NE and SE quadrants.


Pinyon Decline


      Roughly 41% (269,495 ha) of the landscape was eligible for pinyon decline disturbance, and this included all cover types containing the host species (Pinus edulis). The host cover types are quite extensive on both sides of the plateau and cover vast contiguous areas within the boundaries of the project area, extending to the lower and upper elevation limits of this species’ natural distribution, and this clearly influenced the simulated disturbance regime. As expected, the frequency and extent of simulated pinyon decline varied markedly among decades (Figure-frequency, Figure-extent). Major epidemics were relatively infrequent and distinctly episodic in occurrence. In most decades <1% of the host area was affected by an outbreak; however, roughly once every 100 years a major epidemic affecting >20% of the host area would occur (Figure-recurrence) The frequency of these major outbreaks was highly variable among simulation runs. Perhaps once in every 800-year simulation an extreme epidemic affecting as much as 75-80% of the host would occur; although epidemics of this magnitude did not occur in all simulations. Under an extreme case from a single simulation run, the total area disturbed in a single decade was 207,511 ha (77% of eligible)(Figure-map). Typical of other major outbreaks, in this case disturbance patches coalesced into extensive areas covering most of the host distribution, although there were noticeable gaps of significant extent that were not disturbed, and the majority of the disturbance was low mortality due to the immunity of the co-dominant juniper component of the host forest types (pinyon-juniper woodlands).

 

Pinyon Decline Movie - Click here to view a movie depicting pinyon decline epidemics on the Uncompahgre Plateau Landscape over an 800-year (10-year time steps) simulation representing the reference period disturbance regime. NOTE, this is a 30 Mb Microsoft Media file (.avi) that requires appropriate movie viewing software (e.g., Quick Time Player).


      As expected (given the model structure), the variation in frequency and extent of epidemics was strongly related to climate, although the relationship was much stronger for the number of initiations (Figure-initiations) than the area disturbed (Figure-extent). The area disturbed exhibited somewhat of a threshold relationship with climate (representing, in this case, the cumulative consecutive years of drought index), whereby epidemics affecting a substantial proportion (>30%) of the host area occurred only during decades with a drought index > 1, but the magnitude (area disturbed) of epidemics varied dramatically in relation to the climate index above this threshold. This reflects that fact that while major epidemics were limited to periods of extended drought, local outbreaks during such periods were not always guaranteed to expand into major epidemics. Thus, even under extended drought conditions, there were many decades in which pinyon decline affected relatively little area. The most extreme decade of pinyon decline (77% of eligible) was during a moderately severe drought (climate modifier = 1.5). Interestingly, the three most extreme epidemics (~57-77% of eligible) did not occur during the most extreme droughts, reflecting in part the stochastic nature of the disturbance process.


      Under this pinyon decline disturbance regime, the mean return interval between outbreaks (of any mortality level) to a single 25-m cell varied widely across the forest from 57 years to >800 years, and on average ~3% of the host area escaped disturbance altogether over the course of an 800-year simulation (Figure-return). As expected, the mean return interval varied spatially across the forest in relation to the distribution of the host cover types (Figure-map). In general, return intervals were shortest where the host concentration was greatest and increased with the degree of patch isolation; smaller, disjunct and more isolated patches of the host cover type exhibited much longer return intervals than average. This can be attributed to the decreased effectiveness of spread into these patches from outbreaks initiated elsewhere in the landscape. Overall, the pinyon decline rotation period for the eligible area of the landscape averaged 197 years across simulation runs (231 and 1,321 years for low- and high-mortality disturbances, respectively); however, the observed rotation periods were quite variable among simulation runs (Table-rotation). This variability was most likely due to the relatively infrequent and episodic nature of the major outbreaks, especially the extreme outbreaks affecting >70% of the host area which did not occur in every simulation.


      Not surprisingly, the simulated landscape susceptibility to pinyon decline fluctuated over time in response to changing climatic conditions and vegetation conditions. The coefficient of variation in landscape susceptibility was 46%, and this was almost the highest variability among the simulated insects/pathogens (Table-hrv-susceptibility). Based on the available vegetation data, the current landscape is well within the simulated HRV in susceptibility to pinyon decline. Overall, the current landscape condition falls at the 59th percentile of the HRV distribution (i.e., 0% departure index). However, the magnitude of departure varies dramatically across the forest and relatively large areas exhibit very high susceptibility (Figure-susceptibility map). However, given the lack of reliable field data on current stand age and condition in the pinyon-juniper woodland cover types, these findings must be viewed with caution. Yet, observations of the recent pinyon decline in the project area and surrounding region seem to substantiate our findings that susceptibility is quite high in some locations.


Pine Beetle


      Less than 8% (51,367 ha) of the landscape was eligible for pine beetle disturbance, and this included all cover types containing the host species (Pinus ponderosa). The host cover types are patchily distributed, often consisting of disjunct patches interspersed among other cover types. There are relatively few large, contiguous areas of the host cover types, with the exception of a few areas in the Southwest Quadrant, and this clearly influenced the simulated disturbance regime. As expected, the frequency and extent of simulated pine beetle epidemics varied markedly among decades (Figure-frequency, Figure-extent). In most decades, pine beetles were at endemic levels (i.e., no epidemic). However, 1-3 times per 100 years an epidemic occurred. On average, once every 100 years an epidemic affecting >10% of the host area would occur, and roughly once every 250-300 years a major epidemic affecting >20% of the host area would occur (Figure-recurrence), inclusive of both high- and low-mortality affected areas. Under an extreme case from a single simulation run, the total area disturbed in a single decade was 13,869 ha (27% of eligible)(Figure-map). Typical of other major outbreaks, in this case disturbance patches of widely varying sizes up to roughly 200 ha were widely dispersed across the forest, but were more concentrated in some areas, and the majority of the disturbance was low mortality.

 

Pine Beetle Epidemics Movie - Click here to view a movie depicting pine beetle epidemics on the San Juan National Forest (patterns are similar for the UPL) over an 800-year (10-year time steps) simulation representing the reference period disturbance regime. NOTE, this is a 41 Mb Microsoft Media file (.avi) that requires appropriate movie viewing software (e.g., Quick Time Player).


      As expected (given the model structure), the variation in frequency and extent of epidemics was strongly related to climate (Figure-initiations, Figure-extent). The area disturbed exhibited a positive and relatively linear relationship with climate (representing, in this case, the cumulative consecutive years of drought index). Interestingly, extreme drought decades (climate modifier = 1.98) always resulted in fairly major epidemics affecting 10-25% of the host area, but there was considerable variation in the magnitude of epidemic for decades with moderate or mild droughts.


      Under this pine beetle disturbance regime, the mean return interval between outbreaks (of any mortality level) to a single 25-m cell varied widely across the forest from 44 years to >800 years, and on average ~3% of the host area escaped disturbance altogether over the course of an 800-year simulation (Figure-return). There did not appear to be any predictable spatial pattern of variation in mean return intervals (Figure-map). Instead, the pattern seemed somewhat random, suggesting that it was largely chance that determined which areas got disturbed more frequently than others. This was somewhat borne out by the lack of strong correlation among separate simulation runs in the spatial pattern of variation in mean return intervals. Overall, the pine beetle rotation period for the eligible area of the landscape averaged 222 years across simulation runs (277 and 1,119 years for low- and high-mortality disturbances, respectively), and rotation periods were relatively consistent among simulation runs (Table-rotation).


      The simulated landscape susceptibility to pine beetle outbreaks fluctuated over time in response to changing climatic conditions and vegetation conditions, although the variability was relatively low; the coefficient of variation in landscape susceptibility was only 22% (Table-hrv-susceptibility). The low variability was likely due to the relatively poor representation and low variability over time of the most susceptible stage of stand development, the stem exclusion stage, in the host cover types. Based on the available vegetation data, the current landscape is in a state of moderately high susceptibility compared to the simulated HRV (i.e., 79th percentile of the HRV distribution; 17% departure index) and the magnitude of departure varies dramatically across the forest with relatively large areas exhibiting high susceptibility (Figure-susceptibility map) The current state of departure appears to be due to the preponderance of stands in the stem exclusion stage - the most susceptible to pine beetle outbreaks - and the paucity of stands in the fire-maintained open canopy stage, which have a relatively low susceptibility to pine beetle outbreaks.


Douglas-fir Beetle


      Only 2% (16,008 ha) of the landscape was eligible for Douglas-fir beetle disturbance, and this included all cover types containing the host species (Pseudotsuga menziesii). The host cover types are quite patchily distributed and generally consist of disjunct patches interspersed among other cover types. There are essentially no large, contiguous areas of the host cover types. The scarce and patchy distribution of the host cover types clearly influenced the simulated disturbance regime. As expected, the frequency and extent of simulated Douglas-fir beetle epidemics varied markedly among decades (Figure-frequency, Figure-extent). In most decades <1% of the host area was affected by an outbreak; however, at least once every 100 years an epidemic affecting >2% of the host area would occur, and roughly once every 400 years an epidemic affecting >3% would occur (Figure-recurrence), inclusive of both high- and low-mortality affected areas. Under an extreme case from a single simulation run, the total area disturbed in a single decade was 592 ha (~4% of eligible)(Figure-map). Typical of other major outbreaks, in this case disturbance patches of widely varying sizes up to roughly 100 ha were widely dispersed across the forest, but were more concentrated in some areas, and the majority of the disturbance was low mortality.


      By design, the variation in frequency and extent of Douglas-fir beetle epidemics was not related to climate. It was determined that climate is not a major factor controlling the spatial and temporal occurrence of outbreaks. Instead, it was determined that variation is driven more by changes in population demographics (e.g., changes in reproduction and survival due to predators and parasitoides) and interactions with other disturbance processes (e.g., spruce budworm outbreaks, wildfires, and windthrow). Consequently, aside from the explicit interactions with spruce budworm outbreaks and wildfires, we treated these other factors as a source of random variation.


      Under this Douglas-fir beetle disturbance regime, the mean return interval between outbreaks (of any mortality level) to a single 25-m cell varied widely across the forest from 89 years to >800 years, although on average the majority of the host area escaped disturbance altogether over the course of an 800-year simulation (Figure-return). As might be expected, given the relatively minor extent of disturbance, there did not appear to be any predictable spatial pattern of variation in mean return intervals (Figure-map). Instead, the pattern seemed more or less random, suggesting that it was largely chance that determined which areas got disturbed more frequently - or even at all - than others. This was somewhat borne out by the lack of strong correlation among separate simulation runs in the spatial pattern of variation in return intervals. Overall, the Douglas-fir beetle rotation period for the eligible area of the landscape averaged 1,247 years across simulation runs (1,354 and 15,779 years for low- and high-mortality disturbances, respectively); however, as might be expected given the relatively low level of disturbance, the observed rotation periods were highly variable among simulation runs (Table-rotation). The rotation period and return intervals for Douglas-fir beetle are surprisingly long, but this is consistent with the patchy distribution of the host tree and expert opinion regarding the frequency and extent of epidemics.


      The simulated landscape susceptibility to Douglas-fir beetle outbreaks fluctuated over time in response to changing vegetation conditions and interactions with other disturbance agents (specifically, spruce budworm epidemics and wildfires), although the variability was remarkably low; the coefficient of variation in landscape susceptibility was only 13%, the lowest of any of the simulated insects/pathogens (Table-hrv-susceptibility). The low variability was likely due to the preponderance of late-successional stands (all highly susceptible to outbreaks) and the regular occurrence of wildfires and spruce budworm outbreaks - which often trigger Douglas-fir beetle outbreaks - in the host cover types maintained over time during the reference period. Based on the available vegetation data, the current landscape is in a state of extremely low susceptibility compared to the simulated HRV. Overall, the current landscape condition falls at the 0th percentile of the HRV distribution (i.e., 100% departure index), although the magnitude of departure varies spatially across the forest (Figure-susceptibility map). The low susceptibility we calculated for the current landscape is likely due to the absence of recorded recent spruce budworm outbreaks or wildfires in the host cover types - agents that were partially responsible for maintaining relatively high susceptibility levels under the HRV simulations. However, it should be noted, that recent spruce budworm activity observed in the project area (not taken into account in our data) may significantly increase the susceptibility to Douglas-fir beetle outbreaks over the next several years.


Spruce Beetle


      Roughly 7% (46,780 ha) of the landscape was eligible for Spruce beetle disturbance, and this included all cover types containing the host species (Picea engelmannii); although susceptibility varied dramatically among cover types in relation to the abundance of the host species. The host cover types are extensive and contiguous in the higher elevations but are increasingly fragmented and interspersed with other cover types as elevation decreases. As expected, the frequency and extent of simulated spruce beetle epidemics varied markedly among decades (Figure-frequency, Figure-extent). Major epidemics were relatively infrequent and distinctly episodic in occurrence. In most decades <1% of the host area was affected by an outbreak; however, once every 200 years a major epidemic affecting >20% of the host area would occur, and roughly once every 300-400 years a major epidemic affecting >50% would occur (Figure-recurrence), inclusive of both high- and low-mortality affected areas. It should be noted that the relatively infrequent occurrence of major outbreaks and the relatively high percentage of the host area disturbed when a major outbreak occurs is a consequence of the relatively small area of eligible cover types (<47,000 ha). The relatively small extent and sparse distribution of the host cover types reduces the likelihood that conditions will exist that allow for both initiation and extensive spread. However, if the “right” conditions are achieved and a major outbreak occurs, then it is likely that a large percentage of the host will get disturbed given that major outbreaks in the region often encompass up to 100,000 ha before declining. Under an extreme case from a single simulation run, the total area disturbed in a single decade was 31,810 ha (68% of eligible)(Figure-map). Typical of other major outbreaks, this outbreak consisted of a single, large, contiguous disturbance patch, plus a number of smaller, disjunct patches nearby, with the majority of the disturbance being high mortality.

 

Spruce Beetle Epidemics Movie - Click here to view a movie depicting spruce beetle epidemics on the San Juan National Forest (patterns are similar for the UPL) over an 800-year (10-year time steps) simulation representing the reference period disturbance regime. NOTE, this is a 46 Mb Microsoft Media file (.avi) that requires appropriate movie viewing software (e.g., Quick Time Player).


      By design, the variation in frequency and extent of spruce beetle epidemics was not related to climate. It was determined that climate is not a major factor controlling the spatial and temporal occurrence of outbreaks. Instead, it was determined that variation is driven more by changes in population demographics (e.g., changes in reproduction and survival due to predators and parasitoides) and interactions with other disturbance processes (e.g., wildfires and windthrow). In particular, spruce beetle outbreaks often initiate following major windthrow events - which we did not model in RMLANDS. Consequently, aside from the explicit interaction with wildifire, we treated these other factors as a source of random variation.


      Under this spruce beetle disturbance regime, the mean return interval between outbreaks (of any mortality level) to a single 25-m cell varied widely across the forest from 62 years to >800 years, although on average ~20% of the host area escaped disturbance altogether over the course of an 800-year simulation (Figure-return). There was a distinct spatial pattern of variation in mean return intervals over the course of a single 800-year simulation (Figure-map). Specifically, return intervals generally increased with decreasing elevation, reflecting the increasing scarcity and fragmented distribution of the preferred host species (Picea engelmannii) at lower elevations. In general, return intervals were shortest where the preferred host species was abundant and contiguous in distribution. In addition, the return interval variation was highly contagious or clumped in distribution, reflecting the chance occurrence of multiple outbreaks in the same area over the course of the simulation. However, it is important to note that this spatial pattern of variation, while consistently contagious among separate simulation runs, also varied somewhat in geographic distribution among simulation runs, reflecting the stochastic nature of where outbreaks occurred. Overall, the spruce beetle rotation period for the eligible area of the landscape averaged 323 years across simulation runs (3,214 and 359 years for low- and high-mortality disturbances, respectively); however, as might be expected given the relatively episodic nature of this disturbance, the observed rotation periods were highly variable among simulation runs (Table-rotation). Recall that the preponderance of high-mortality disturbance of the host tree (Picea engelmannii) does not equate to stand replacement, in this case, because most susceptible stands contain a mixture of host and non-host species. Thus, most high-mortality spruce beetle epidemics do not result in stand replacement, unless they co-occur with high-mortality spruce budworm and/or Douglas-fir beetle outbreaks. In addition, the rotation periods varied markedly among eligible cover types, reflecting the relative abundance of the preferred host species among cover types.


      The simulated landscape susceptibility to spruce beetle outbreaks fluctuated over time in response to changing vegetation conditions and interactions with other disturbance agents (e.g., large windthrow events, which we treated as a purely stochastic occurrence in our simulations); indeed, variability over time was the highest of any of the simulated insects/pathogens - the coefficient of variation in landscape susceptibility was 49% (Table-hrv-susceptibility). The high variability was likely due to the relative scarcity of the host cover types, which resulted in major fluctuations in the preponderance of late-successional stands (all highly susceptible to outbreaks). Based on the available vegetation data, the current landscape is well within the simulated HRV in susceptibility to spruce beetle epidemics. Overall, the current landscape condition falls at the 56th percentile of the HRV distribution (i.e., 0% departure index), although the magnitude of departure varies spatially across the forest and there are large areas of very high susceptibility (Figure-susceptibility map). These areas of high susceptibility correspond to areas with a preponderance of contiguous spruce-fir forest in the late-seral stages of development - the most susceptible to spruce beetle outbreaks.


Spruce Budworm


      Roughly 7% (46,780 ha) of the landscape was eligible for Spruce budworm disturbance, and this included all cover types containing the host species (principally Pseudotsuga menziesii, Abies lasiocarpa, and Picea engelmannii); although susceptibility varied slightly among cover types in relation to the abundance of the preferred host species - the true firs and Douglas-fir. As expected, the frequency and extent of simulated spruce budworm epidemics varied markedly among decades, although the relationship was much stronger for the area disturbed (Figure-frequency, Figure-extent). In most decades, spruce budworm populations were at endemic levels (i.e., no epidemic). However, 2-5 times per 100 years an epidemic occurred. On average, once every 30 years an epidemic affecting >10% of the host area would occur, and roughly once every 100 years a major epidemic affecting >30% of the host area would occur (Figure-recurrence), inclusive of both high- and low-mortality affected areas. In many cases, major epidemics extended over two decades; thus, the maximum area disturbed during a single epidemic was much greater than the maximum area disturbed in any single decade. Under an extreme case from a single simulation run, the total area disturbed in a single epidemic (extending over two decades) was 35,533 ha (76% of eligible)(Figure-map). Typical of other major outbreaks, in this case disturbance patches coalesced into extensive areas covering much of the host distribution, although there were noticeable gaps of varying extent that were not disturbed, and the majority of the disturbance was low mortality.

 

Spruce Budworm Epidemics Movie - Click here to view a movie depicting spruce budworm epidemics on the San Juan National Forest (patterns are similar for the UPL) over an 800-year (10-year time steps) simulation representing the reference period disturbance regime. NOTE, this is a 30 Mb Microsoft Media file (.avi) that requires appropriate movie viewing software (e.g., Quick Time Player).


      As expected (given the model structure), the variation in extent of epidemics was strongly related to climate, even though, surprisingly, there was no relationship between climate and frequency of epidemics (Figure-initiations, Figure-extent). The area disturbed exhibited a strong positive and linear relationship with climate (representing, in this case, the cumulative consecutive years of wet index). Interestingly, extremely wet decades (climate modifier = 1.98) always resulted in fairly major epidemics affecting 20-60% of the host area.


      Under this spruce budworm disturbance regime, the mean return interval between outbreaks (of any mortality level) to a single 25-m cell varied widely across the forest from 36 years to >800 years, although almost no eligible area escaped disturbance altogether over the course of an 800-year simulation (Figure-return). There did not appear to be any predictable spatial pattern of variation in mean return intervals (Figure-map). Instead, the pattern seemed somewhat random, suggesting that it was largely chance that determined which areas got disturbed more frequently than others. This was somewhat borne out by the lack of strong correlation among separate simulation runs in the spatial pattern of variation in return intervals. Overall, the spruce budworm rotation period for the eligible area of the landscape averaged 104 years across simulation runs (109 and 2,094 years for low- and high-mortality disturbances, respectively), and rotation periods were relatively consistent among simulation runs (Table-rotation).


      The simulated landscape susceptibility to spruce budworm outbreaks fluctuated over time in response to changing climate and vegetation conditions; the coefficient of variation in landscape susceptibility was 34% (Table-hrv-susceptibility). Based on the available vegetation data, the current landscape is well within the simulated HRV in susceptibility to spruce budworm epidemics. Overall, the current landscape condition falls at the 56th percentile of the HRV distribution (i.e., 0% departure index), although the magnitude of departure varies spatially across the forest and there are large areas of very high susceptibility (Figure-susceptibility map) These areas of high susceptibility correspond to areas with a preponderance of contiguous spruce-fir forest in the late-seral stages of development - the most susceptible to spruce budworm outbreaks. It should be noted that recent spruce budworm activity observed in the project area (not taken into account in our data) supports our contention that areas of high susceptibility exist in the current landscape despite the relatively average overall susceptibility of the landscape.


Vegetation Patterns & Dynamics


      We recognized 23 distinct vegetation types ("cover types") on the UPL and surrounding area for the purposes of RMLANDS simulations (see RMLANDS - Vegetation Classification). However, several of these types were treated as “static” in our simulations; i.e., they did not have separate stand conditions or undergo successional changes, at least over the spatial and temporal scales (e.g., 25-m cell size, 10-year time step) considered in our simulations, even though they may have been subject to one or more disturbance process. For example, mountain grasslands were subject to wildfire but were assumed to remain constant in composition and structure due to the relatively rapid (i.e. 1-2 years) recovery of vegetation following fire. In the sections below, we limit our discussion to only those cover types that we treated as dynamic in RMLANDS. For each of these cover types, we begin the section with a statement of our confidence in the results based on the sources of uncertainty in the model as it affects that cover type. We then briefly describe the simulated disturbance regime (i.e., spatial extent and distribution, frequency and temporal variability) associated with each relevant disturbance process, the vegetation dynamics resulting from the interplay between these disturbance processes and succession, and we conclude each section with an examination of the cover type’s current departure from the simulated HRV. Lastly, we describe the range of variation and departure of the current landscape in overall landscape structure, in which all cover types and stand conditions are considered jointly as a single patch mosaic.


Semi-Desert Grassland [cover type description]


      Semi-desert grassland is relatively common on the UPL, encompassing 35,847 ha and comprising 5.4% of the landscape (Table-areal coverage).


      Sources of Uncertainty.--Unfortunately, the paucity of empirical data on disturbance regimes and succession processes in this cover type forced us to rely almost exclusively on expert opinion to parameterize the model. Consequently, we did not have objective targets against which to verify model outcomes. In particular, we estimated the mean fire return interval (rotation period) for the reference period to be quite short (50-60 years). We reasoned that the low shrubs, grasses and forbs would provide abundant and contiguous fine fuels, especially during wet decades, and that the annually dry conditions and abundant ignition sources (i.e., lightning) would allow for frequent and extensive fires to occur. We reasoned that fuel bed conditions and fire return intervals would be similar to low-elevation ponderosa pine forest (i.e., driven by the availability of fine fuels), but that the drier conditions of the semi-desert environment would make productivity lower and therefore return intervals likely 50% longer. The main impact of the short rotation fire regime was to shift the seral stage and age distributions towards the younger classes. A secondary impact was to exert influence on the disturbance regimes in surrounding cover types; in particular, to shorten the fire return interval in adjoining pinyon-juniper woodlands. Given our uncertainty in the fire rotation period for this cover type, the results reported below should be viewed with extreme caution.


      Wildfire.--The frequency and extent of simulated wildfires in semi-desert grassland varied markedly among decades (Figure-initiations, Figure-extent). In most decades, >5% of the semi-desert grassland burned, inclusive of both high- and low-mortality affected areas, and roughly once per 100 years, >35% of the area burned (Figure-recurrence). Under this wildfire regime, the mean return interval between fires (of any mortality level) to a single 25-m cell varied widely from 23 years to >800 years, with a mean and median of roughly 60 years, and almost no area escaped disturbance altogether over the course of an 800-year simulation (Figure-return). Note, the mean return interval was 68 years and 600 years for high- and low-mortality disturbances, respectively, due to the preponderance of high-mortality wildfires. As expected, mean return intervals varied spatially across the landscape (Figure-map). The shortest return intervals were found in the patches of extensive, contiguous grassland, and return intervals were longer in the more disjunct grassland patches and on the periphery of the grassland distribution where the grasslands were juxtaposed to less flammable pinyon-juniper woodland types.


      Age Structure.--The age structure and dynamics of semi-desert grassland reflected the interplay between disturbance and succession processes. The survivorship distribution represents the percent of stands that survived to any age, where age represents time since stand origin, not necessarily the age of the oldest individual plants. On average (over time), roughly 50% of the semi-desert grassland was >40 years since stand origin, although at any point in time this varied from 18% to 85% (Figure-survivorship). On average, 25% of the semi-desert grassland survived to >90 years, 5% survived to >230 years, and <1% survived a stand-replacing disturbance for >400 years.


      Seral-stage Distribution.--The distribution of area among stand conditions within semi-desert grassland fluctuated markedly over time, as expected (Figure-conditions). For example, the percentage of semi-desert grassland in the early grass-forb stage varied from 2% to 71%, reflecting the dynamic nature of this cover type when considered over century-long periods (Figure-hrv). The seral-stage distribution appeared to be in dynamic equilibrium (i.e., the percentage in each stand condition varied about a stable mean), and appeared to reach equilibrium relatively quickly (i.e., <100 years). The spatial configuration of stand conditions fluctuated markedly over time as well, although there was considerable variation in the magnitude of variability among configuration metrics (Table-hrv). The early grass-forb condition was particularly dynamic, followed by the mid grass-shrub condition and then the shrub-grass condition.


      HRV Departure.--Unfortunately, due to the lack of data on current stand age and condition in this cover type, we were unable to evaluate departure of the current landscape from the simulated HRV.


Semi-Desert Savannah [cover type description]


      Semi-desert savannah is relatively uncommon on the UPL, encompassing 5,847 ha and comprising <1% of the landscape (Table-areal coverage).


      Sources of Uncertainty.--As with semi-desert grassland, the paucity of empirical data on disturbance regimes and succession processes in this cover type forced us to rely almost exclusively on expert opinion to parameterize the model. Consequently, we did not have objective targets against which to verify model outcomes. In particular, we estimated the mean fire return interval (rotation period) for the reference period to be quite short (50-60 years). We reasoned that the low shrubs, grasses and forbs would provide abundant and contiguous fine fuels, especially during wet decades, and that the annually dry conditions and abundant ignition sources (i.e., lightning) would allow for frequent and extensive fires to occur. We reasoned that fuel bed conditions and fire return intervals would be similar to low-elevation ponderosa pine forest (i.e., driven by the availability of fine fuels), but that the drier conditions of the semi-desert environment would make productivity lower and therefore return intervals likely 50% longer. The main impact of the short rotation fire regime was to shift the seral stage and age distributions towards the younger classes. A secondary impact was to exert influence on the disturbance regimes in surrounding cover types; in particular, to shorten the fire return interval in adjoining pinyon-juniper woodlands. Given our uncertainty in the fire rotation period for this cover type, the results reported below should be viewed with extreme caution.


      Wildfire.--The frequency and extent of simulated wildfires in semi-desert savannah varied markedly among decades (Figure-initiations, Figure-extent). In most decades, >5% of the semi-desert savannah burned, inclusive of both high- and low-mortality affected areas, and roughly once per 100 years, >25% of the area burned (Figure-recurrence). Under this wildfire regime, the mean return interval between fires (of any mortality level) to a single 25-m cell varied widely from 24 years to >800 years, with a mean and median of 80 and 89 years, respectively, and almost no area escaped disturbance altogether over the course of an 800-year simulation (Figure-return). Note, the mean return interval was 89 years and 843 years for high- and low-mortality disturbances, respectively, due to the preponderance of high-mortality wildfires. As expected, mean return intervals varied spatially across the forest (Figure-map), but given the scarcity of this cover type it was impossible to discern whether there were any meaningful patterns.


      Age Structure.--The age structure and dynamics of semi-desert savannah reflected the interplay between disturbance and succession processes. The survivorship distribution represents the percent of stands that survived to any age, where age represents time since stand origin, not necessarily the age of the oldest individuals in the stand. On average (over time), roughly 50% of the semi-desert savannah was >60 years since stand origin, although at any point in time this varied from 12% to 78% (Figure-survivorship). On average, 25% of the semi-desert savannah survived to >130 years, 5% survived to >330 years, and <1% survived a stand-replacing disturbance for >500 years.


      Seral-stage Distribution.--The distribution of area among stand conditions within semi-desert savannah fluctuated markedly over time, as expected (Figure-conditions). For example, the percentage of semi-desert savannah in the herb-shrub-tree stage varied from 40% to 99%, reflecting the dynamic nature of this cover type when considered over century-long periods (Figure-hrv). However, given the scarcity of this cover type in the project area, it was not surprising that the range of variation was so wide. The seral-stage distribution appeared to be in dynamic equilibrium (i.e., the percentage in each stand condition varied about a stable mean), and appeared to reach equilibrium relatively quickly (i.e., <100 years). The spatial configuration of stand conditions fluctuated markedly over time as well, although there was considerable variation in the magnitude of variability among configuration metrics (Table-hrv). The herb-annual and herb-perennial conditions were particularly dynamic, with coefficients of variation typically several-fold or even an order of magnitude greater than the herb-shrub-tree stage.


      HRV Departure.--Unfortunately, due to the lack of data on current stand age and condition in this cover type, we were unable to evaluate departure of the current landscape from the simulated HRV.


Pinyon-Juniper Woodland [cover type description]


      Pinyon-juniper woodland is the dominant cover type on the UPL, encompassing 132,689 ha and comprising ~20% of the landscape (Table-areal coverage).


      Sources of Uncertainty.--At the time of our simulation experiments, there was very little local empirical data available on disturbance regimes and succession processes in pinyon-juniper woodlands from which to parameterize the model. Recently published data from nearby Mesa Verde indicated a 400-yr mean fire return interval (rotation period) for this cover type (Floyd et al. 2004), and data collected from 138 pinyon-juniper stands (inclusive of all PJ types) on the UPL and surrounding areas (Eisenhart 2004) provided a target stand age distribution suggestive of a much shorter fire rotation period. We decided to parameterize the model to be consistent with the Eisenhart stand age data, after adjusting the data to represent stand ages in 1880. The results reported below are generally consistent with our interpretation of these adjusted data. However, recent unpublished data by Shinneman and Baker (in review) based on 27 random plots in pinyon-juniper woodlands (inclusive of all PJ types) on the UPL suggest a much older age distribution than previously reported. The differences between studies reflect important methodological differences. In particular, Shinneman and Baker estimated stand age based on the age of the oldest tree, pinyon or juniper, in a stand, whereas Eisenhart based stand age on multiple lines of evidence, which included the age of the oldest pinyon pine trees (but not juniper trees) in the stand. Shinneman and Baker noted that due to the shorter life expectancy of pinyon pine compared to juniper, there may be a significant bias towards younger stands in Eisenhart’s data; they estimated the bias to be 127 years (B. Baker, pers. commun.). Accordingly, Shinneman and Baker estimated a mean fire return interval of 400-600 years - considerably longer than we simulated. The net result of this is that in our simulation we may have significantly underestimated the fire rotation period in this cover type, resulting in a significant shift in the seral stage and age distributions towards the younger classes. Given the uncertainty in this regard, the numerical/quantitative results reported below should be viewed with extreme caution. Nevertheless, the overall picture that emerges is entirely consistent with empirical studies, viz., that historical fires in pinyon-juniper vegetation tended to recur at very long intervals (decades to centuries), that fires were predominantly stand-replacing, and that re-forestation after fire was a slow process.


      Wildfire.--The frequency and extent of simulated wildfires in pinyon-juniper woodland varied markedly among decades (Figure-initiations, Figure-extent). In most decades, <2% of the pinyon-juniper woodland burned, inclusive of both high- and low-mortality affected areas; however, roughly once per 100 years, >10% of the area burned and roughly once every 300 years, >20% of the area burned (Figure-recurrence). Under this wildfire regime, the mean return interval between fires (of any mortality level) to a single 25-m cell varied widely from 24 years to >800 years, with a mean and median of 197 and 267 years, respectively, and roughly 5% of the area escaped disturbance altogether over the course of an 800-year simulation (Figure-return). Note, the mean return interval was 216 years and 2,148 years for high- and low-mortality disturbances, respectively, due to the preponderance of high-mortality wildfires. As expected, mean return intervals varied spatially across the forest (Figure-map). In particular, return intervals were shorter at the upper and lower elevations of the pinyon-juniper distribution where the cover type was juxtaposed with more flammable cover types.


      Pinyon Decline.--The frequency and extent of simulated pinyon decline epidemics in pinyon-juniper woodland varied among decades in an episodic fashion (Figure-initiations, Figure-extent). In most decades, pinyon decline was at endemic levels and <1% of the pinyon-juniper woodland was disturbed. However, roughly once every 100 years an epidemic affecting >20% of the host area would occur, and roughly once every 300-400 years a major epidemic affecting >40% of the host area would occur (Figure-recurrence), inclusive of both high- and low-mortality affected areas. The mean return interval between epidemics (of any mortality level) at a single location (25-m cell) varied from 57 years to >800 years, with a mean and median of 184 and 200 years, respectively, and 2-3% of the eligible area escaped disturbance altogether over the course of an 800-year simulation (Figure-return). Note, the mean return interval was 1,220 years and 216 years for high- and low-mortality disturbances, respectively, due to the preponderance of low-mortality outbreaks. As expected, mean return intervals varied spatially across the forest (Figure-map). In general, return intervals were shorter where the host species was abundant and contiguous in distribution and decreased on the periphery of its distribution where it was more patchy and interspersed with non-host cover types.


      Age Structure.--The age structure and dynamics of pinyon-juniper woodland reflected the interplay between disturbance and succession processes. The survivorship distribution represents the percent of stands that survived to any age, where age represents time since stand origin, not necessarily the age of the oldest trees in the stand. On average (over time), roughly 50% of the pinyon-juniper woodland was >150 years since stand origin, although at any point in time this varied from 28% to 72% (Figure-survivorship). On average, 25% of the pinyon-juniper woodland survived to >320 years, 5% survived to >620 and ~1% survived a stand-replacing disturbance for >800 years.


      Seral-stage Distribution.--The distribution of area among stand conditions within pinyon-juniper woodland fluctuated markedly over time, as expected (Figure-conditions). For example, the percentage of pinyon-juniper woodland in the tree-dominated stage varied from 1% to 54%, reflecting the dynamic nature of this cover type when considered over century-long periods (Figure-hrv). Note, the principal affect of pinyon decline was to shift stands from the tree-dominated stage to the shrubs-trees stage through the reduction or loss of the pinyon component from stands. Hence, the shrubs-trees condition includes some stands with very old junipers. The seral-stage distribution appeared to be in dynamic equilibrium (i.e., the percentage in each stand condition varied about a stable mean), and appeared to reach equilibrium relatively quickly (i.e., <100 years). The spatial configuration of stand conditions fluctuated markedly over time as well, although there was considerable variation in the magnitude of variability among configuration metrics (Table-hrv). The herb-dominated condition was particularly dynamic, with coefficients of variation typically two to several times greater than the later seral stages.


      HRV Departure.--Unfortunately, due to insufficient spatial data on current stand age and condition in this cover type, we were unable to evaluate departure of the current landscape from the simulated HRV.


Pinyon-Juniper-Sagebrush Woodland [cover type description]


      Pinyon-juniper-sagebrush woodland is less common than pinyon-juniper woodland, but nonetheless very common, on the UPL, encompassing 89,918 ha and comprisng almost 14% of the landscape (Table-areal coverage). In general, the vegetation dynamics in pinyon-juniper-sagebrush woodland were very similar to those of pinyon-juniper woodland. Nevertheless, we provide a complete description of the results for this cover type below and highlight the notable differences.


      Sources of Uncertainty.--At the time of our simulation experiments, there was very little local empirical data available on disturbance regimes and succession processes in pinyon-juniper-sagebrush woodlands from which to parameterize the model. Recently published data from nearby Mesa Verde indicated a 400-yr mean fire return interval (rotation period) for pinyon-juniper woodlands (Floyd et al. 2004), but the study did not include pinyon-juniper-sagebrush woodlands. Data collected from 138 pinyon-juniper stands on the UPL and surrounding areas (Eisenhart 2004) provided a target stand age distribution suggestive of a much shorter fire rotation period, but the study did not differentiate among pinyon-juniper woodland types. We reasoned that pinyon-juniper-sagebrush woodlands have a much better developed and more continuous and flammable fuel bed in the sagebrush understory than pinyon-juniper woodlands, and therefore estimated the mean fire return interval to be 200 years, intermediate between the longer (400 yrs) interval observed for pinyon-juniper woodlands (Floyd et al. 2004) and the comparatively short (75-100 yrs) interval observed for mountain shrublands (Floyd et al. 2000). In addition, we sought to be consistent with the Eisenhart stand age data, after adjusting the data to represent stand ages in 1880. Given the uncertainty in these parameters, however, the numerical/quantitative results reported below should be viewed with extreme caution. Nevertheless, the overall picture that emerges is entirely consistent with empirical studies, viz., that historical fires in pinyon-juniper vegetation tended to recur at very long intervals (decades to centuries), that fires were predominantly stand-replacing, and that re-forestation after fire was a slow process.


      Recent unpublished data by Shinneman and Baker (in review) from 27 pinyon-juniper stands on the UPL suggest a much older age distribution than reported by Eisenhart (2004), although the study also did not differentiate among pinyon-juniper woodland types. The differences between studies reflect important methodological differences. In particular, Shinneman and Baker estimated stand age based on the age of the oldest tree, pinyon or juniper, in a stand, whereas Eisenhart based stand age on multiple lines of evidence, which included the age of the oldest pinyon pine trees (but not juniper trees) in the stand. Shinneman and Baker noted that due to the shorter life expectancy of pinyon pine compared to juniper, there may be a significant bias towards younger stands in Eisenhart’s data; they estimated the bias to be 127 years (B. Baker, pers. commun.). Accordingly, Shinneman and Baker estimated a mean fire return interval of 400-600 years - considerably longer than we simulated - although they did not differentiate among pinyon-juniper woodland types as noted above. Nevertheless, the net result is that in our simulation we may have significantly underestimated the fire rotation period in this cover type, resulting in a significant shift in the seral stage and age distributions towards the younger classes. Given the uncertainty in this regard, the results reported below should be viewed with extreme caution.


      Wildfire.--The frequency and extent of simulated wildfires in pinyon-juniper-sagebrush woodland varied markedly among decades (Figure-initiations, Figure-extent). In most decades, <5% of the pinyon-juniper-sagebrush woodland burned, inclusive of both high- and low-mortality affected areas; however, roughly once per 100 years, >10% of the area burned and roughly once every 300 years, >25% of the area burned (Figure-recurrence). Under this wildfire regime, the mean return interval between fires (of any mortality level) to a single 25-m cell varied widely from 27 years to >800 years, with a mean and median of 176 and 200 years, respectively, and roughly 1% of the area escaped disturbance altogether over the course of an 800-year simulation (Figure-return). Note, the mean return interval was 198 years and 1,841 years for high- and low-mortality disturbances, respectively, due to the preponderance of high-mortality wildfires. As expected, mean return intervals varied spatially across the forest (Figure-map). In particular, return intervals were shorter at the lower elevations of the distribution where the cover type was juxtaposed with more flammable cover types such as semi-desert grassland.


      Pinyon Decline.--The frequency and extent of simulated pinyon decline epidemics in pinyon-juniper-sagebrush woodland varied among decades in an episodic fashion (Figure-initiations, Figure-extent). In most decades, pinyon decline was at endemic levels and <1% of the pinyon-juniper-sagebrush woodland was disturbed. However, roughly once every 100 years an epidemic affecting >20% of the host area would occur, and roughly once every 300-400 years a major epidemic affecting >40% of the host area would occur (Figure-recurrence), inclusive of both high- and low-mortality affected areas. The mean return interval between epidemics (of any mortality level) at a single location (25-m cell) varied from 57 years to >800 years, with a mean and median of 200 years, and 2-3% of the eligible area escaped disturbance altogether over the course of an 800-year simulation (Figure-return). Note, the mean return interval was 1,324 years and 234 years for high- and low-mortality disturbances, respectively, due to the preponderance of low-mortality outbreaks. As expected, mean return intervals varied spatially across the forest (Figure-map). In general, return intervals were shorter where the host species was abundant and contiguous in distribution and decreased on the periphery of its distribution where it was more patchy and interspersed with non-host cover types.


      Age Structure.--The age structure and dynamics of pinyon-juniper-sagebrush woodland reflected the interplay between disturbance and succession processes. The survivorship distribution represents the percent of stands that survived to any age, where age represents time since stand origin, not necessarily the age of the oldest trees in the stand. On average (over time), roughly 50% of the pinyon-juniper-sagebrush woodland was >140 years since stand origin, although at any point in time this varied from 24% to 69% (Figure-survivorship). On average, 25% of the pinyon-juniper-sagebrush woodland survived to >290 years, 5% survived to >550 years, and <1% survived a stand-replacing disturbance for >800 years. Overall, the age structure of pinyon-juniper-sagebrush woodland was slightly “younger” than pinyon-juniper woodland, reflecting the shorter wildfire return intervals.


      Seral-stage Distribution.--The distribution of area among stand conditions within pinyon-juniper-sagebrush woodland fluctuated markedly over time, as expected (Figure-conditions). For example, the percentage of pinyon-juniper-sagebrush woodland in the tree-dominated stage varied from 1% to 52%, reflecting the dynamic nature of this cover type when considered over century-long periods (Figure-hrv). Note, the principal affect of pinyon decline was to shift stands from the tree-dominated stage to the shrubs-trees stage through the reduction or loss of the pinyon component from stands. Hence, the shrubs-trees condition includes some stands with very old junipers. The seral-stage distribution appeared to be in dynamic equilibrium (i.e., the percentage in each stand condition varied about a stable mean), and appeared to reach equilibrium relatively quickly (i.e., <100 years). The spatial configuration of stand conditions fluctuated markedly over time as well, although there was considerable variation in the magnitude of variability among configuration metrics (Table-hrv). The herb-dominated condition was particularly dynamic, with coefficients of variation typically two to several times greater than the later seral stages.


      HRV Departure.--Unfortunately, due to insufficient spatial data on current stand age and condition in this cover type, we were unable to evaluate departure of the current landscape from the simulated HRV.


Pinyon-Juniper-Oak-Serviceberry Woodland [cover type description]


      Pinyon-juniper-oak-serviceberry woodland is much less common than the other pinyon-juniper woodland types, but nonetheless relatively common, on the UPL, encompassing 36,910 ha and comprising roughly 6% of the landscape (Table-areal coverage). In general, the vegetation dynamics in pinyon-juniper-oak-serviceberry woodland were very similar to those of pinyon-juniper woodland. Nevertheless, we provide a complete description of the results for this cover type below and highlight the notable differences.


      Sources of Uncertainty.--At the time of our simulation experiments, there was very little local empirical data available on disturbance regimes and succession processes in pinyon-juniper-oak-serviceberry woodlands from which to parameterize the model. Recently published data from nearby Mesa Verde indicated a 400-yr mean fire return interval (rotation period) for pinyon-juniper woodlands (Floyd et al. 2004), but the study did not distinguish pinyon-juniper-oak-serviceberry woodlands. Data collected from 138 pinyon-juniper stands on the UPL and surrounding areas (Eisenhart 2004) provided a target stand age distribution suggestive of a much shorter fire rotation period, but the study did not differentiate among pinyon-juniper woodland types. We reasoned that pinyon-juniper-oak-serviceberry woodlands have a much better developed and more continuous and flammable fuel bed in the oak-serviceberry understory than pinyon-juniper woodlands, and therefore estimated the mean fire return interval to be 200 years, intermediate between the longer (400 yrs) interval observed for pinyon-juniper woodlands (Floyd et al. 2004) and the comparatively short (75-100 yrs) interval observed for mountain shrublands (Floyd et al. 2000). In addition, we sought to be consistent with the Eisenhart stand age data, after adjusting the data to represent stand ages in 1880. Given the uncertainty in these parameters, however, the numerical/quantitative results reported below should be viewed with extreme caution. Nevertheless, the overall picture that emerges is entirely consistent with empirical studies, viz., that historical fires in pinyon-juniper vegetation tended to recur at very long intervals (decades to centuries), that fires were predominantly stand-replacing, and that re-forestation after fire was a slow process.


      Recent unpublished data by Shinneman and Baker (in review) from 27 pinyon-juniper stands on the UPL suggest a much older age distribution than reported by Eisenhart (2004), although the study also did not differentiate among pinyon-juniper woodland types. The differences between studies reflect important methodological differences. In particular, Shinneman and Baker estimated stand age based on the age of the oldest tree, pinyon or juniper, in a stand, whereas Eisenhart based stand age on multiple lines of evidence, which included the age of the oldest pinyon pine trees (but not juniper trees) in the stand. Shinneman and Baker noted that due to the shorter life expectancy of pinyon pine compared to juniper, there may be a significant bias towards younger stands in Eisenhart’s data; they estimated the bias to be 127 years (B. Baker, pers. commun.). Accordingly, Shinneman and Baker estimated a mean fire return interval of 400-600 years - considerably longer than we simulated - although they did not differentiate among pinyon-juniper woodland types as noted above. Nevertheless, the net result is that in our simulation we may have significantly underestimated the fire rotation period in this cover type, resulting in a significant shift in the seral stage and age distributions towards the younger classes. Given the uncertainty in this regard, the results reported below should be viewed with extreme caution.


      Wildfire.--The frequency and extent of simulated wildfires in pinyon-juniper-oak-serviceberry woodland varied markedly among decades (Figure-initiations, Figure-extent). In most decades, <5% of the pinyon-juniper-oak-serviceberry woodland burned, inclusive of both high- and low-mortality affected areas; however, roughly once per 100 years, >15% of the area burned and roughly once every 300 years, >25% of the area burned (Figure-recurrence). Under this wildfire regime, the mean return interval between fires (of any mortality level) to a single 25-m cell varied widely from 24 years to >800 years, with a mean and median of 135 and 160 years, respectively - somewhat less than pinyon-juniper woodland - and roughly 2% of the area escaped disturbance altogether over the course of an 800-year simulation (Figure-return). Note, the mean return interval was 153 years and 1,382 years for high- and low-mortality disturbances, respectively, due to the preponderance of high-mortality wildfires. As expected, mean return intervals varied spatially across the forest (Figure-map). In particular, return intervals were shorter at the higher elevations of the distribution where the cover type was juxtaposed with more flammable cover types such as mountain shrubland and ponderosa pine-oak forest.


      Pinyon Decline.--The frequency and extent of simulated pinyon decline epidemics in pinyon-juniper-oak-serviceberry woodland varied among decades in an episodic fashion (Figure-initiations, Figure-extent). In most decades, pinyon decline was at endemic levels and <1% of the pinyon-juniper-oak-serviceberry woodland was disturbed. However, roughly once every 100 years an epidemic affecting >20% of the host area would occur, and roughly once every 300-400 years a major epidemic affecting >30% of the host area would occur (Figure-recurrence), inclusive of both high- and low-mortality affected areas. The mean return interval between epidemics (of any mortality level) at a single location (25-m cell) varied from 62 years to >800 years, with a mean and median of 228 and 200 years, respectively, although ~7% of the eligible area escaped disturbance altogether over the course of an 800-year simulation (Figure-return). Note, the mean return interval was 1,536 years and 269 years for high- and low-mortality disturbances, respectively, due to the preponderance of low-mortality outbreaks. As expected, mean return intervals varied spatially across the forest (Figure-map). In general, return intervals were shorter where the host species was abundant and contiguous in distribution and decreased on the periphery of its distribution where it was more patchy and interspersed with non-host cover types.


      Age Structure.--The age structure and dynamics of pinyon-juniper-oak-serviceberry woodland reflected the interplay between disturbance and succession processes. The survivorship distribution represents the percent of stands that survived to any age, where age represents time since stand origin, not necessarily the age of the oldest trees in the stand. On average (over time), roughly 50% of the pinyon-juniper-oak-serviceberry woodland was >100 years since stand origin, although at any point in time this varied from 24% to 74% (Figure-survivorship). On average, 25% of the pinyon-juniper-oak-serviceberry woodland survived to >210 years, 5% survived to >490 years, and <1% survived a stand-replacing disturbance for >720 years. Overall, the age structure of pinyon-juniper-oak-serviceberry woodland was “younger” than other pinyon-juniper woodland types, reflecting the shorter wildfire return intervals.


      Seral-stage Distribution.--The distribution of area among stand conditions within pinyon-juniper-oak-serviceberry woodland fluctuated markedly over time, as expected (Figure-conditions). For example, the percentage of pinyon-juniper-oak-serviceberry woodland in the tree-dominated stage varied from 1% to 39%, reflecting the dynamic nature of this cover type when considered over century-long periods (Figure-hrv). However, given the scarcity of this cover type in the project area, it was not surprising that the range of variation was so wide. Note also that the principal affect of pinyon decline was to shift stands from the tree-dominated stage to the shrubs-trees stage through the reduction or loss of the pinyon component from stands. Hence, the shrubs-trees condition includes some stands with very old junipers. The seral-stage distribution appeared to be in dynamic equilibrium (i.e., the percentage in each stand condition varied about a stable mean), and appeared to reach equilibrium relatively quickly (i.e., <100 years). The spatial configuration of stand conditions fluctuated markedly over time as well, although there was considerable variation in the magnitude of variability among configuration metrics (Table-hrv). The herb-dominated condition was particularly dynamic, with coefficients of variation typically two to several times greater than the later seral stages.


      HRV Departure.--Unfortunately, due to insufficient spatial data on current stand age and condition in this cover type, we were unable to evaluate departure of the current landscape from the simulated HRV.


Mountian Shrubland [cover type description]


      Mountain shrubland is the second most common cover type on the UPL, encompassing 113,083 ha and comprising roughly 17% of the landscape (Table-areal coverage).


      Sources of Uncertainty.---At the time of our simulation experiments, there was very little local empirical data available on disturbance regimes and succession processes in mountain shrublands from which to parameterize the model. However, data from nearby Mesa Verde indicated a ~100-yr mean fire return interval (rotation period) during the reference period (Floyd et al. 2000). We reasoned that the fire rotation period for the UPL would be shorter than that observed on Mesa Verde due to fewer topographic barriers to fire spread from surrounding areas. Accordingly, we estimated the mean fire return interval to be ~75 years. The results reported below are generally consistent with this interpretation. The only major source of uncertainty regarding our treatment of this cover type that we are aware of is regarding the vegetation transition model; more specifically, with the duration of the herb-shrubs stage during succession. We allowed stands to persist in this early-seral stage for 20-30 years following a high-mortality fire. Recent discussions have led us to believe that stands may not persist in this stage (as we have defined it) for more than 10 years. The net result of this is that in our simulation we may have overestimated the proportion of area in this seral stage. Nevertheless, overall we have moderately high confidence in the reliability of the results for this cover type.


      Wildfire.--The frequency and extent of simulated wildfires in mountain shrubland varied markedly among decades (Figure-initiations, Figure-extent). Wildfire was fairly common in this cover type. In most decades, >5% of the mountain shrubland burned, inclusive of both high- and low-mortality affected areas, and roughly 2 times per 100 years, >20% of the area burned (Figure-recurrence). Under this wildfire regime, the mean return interval between fires (of any mortality level) to a single 25-m cell varied widely from 23 years to >800 years, with a mean and median of 67 years, and essentially no eligible area escaped disturbance altogether over the course of an 800-year simulation (Figure-return). Note, the mean return interval was 76 years and 667 years for high- and low-mortality disturbances, respectively, due to the preponderance of high-mortality wildfires. As expected, mean return intervals varied spatially across the forest (Figure-map). In particular, return intervals were shorter where the cover type was interspersed with highly flammable cover types such as mesic sagebrush and ponderosa pine-oak forest, and in the Southwest Quadrant where wildfire was more common.


      Age Structure.--The age structure and dynamics of mountain shrubland reflected the interplay between disturbance and succession processes. The survivorship distribution represents the percent of stands that survived to any age, where age represents time since stand origin, not necessarily the age of the oldest individuals in the stand. On average (over time), roughly 50% of the mountain shrubland was >45 years since stand origin, although at any point in time this varied from 19% to 78% (Figure-survivorship). On average, 25% of the mountain shrubland survived to >90 years, 5% survived to >200 years, and <1% survived a stand-replacing disturbance for >330 years.


      Seral-stage Distribution.--The distribution of area among stand conditions within mountain shrubland fluctuated over time, as expected (Figure-conditions). For example, the percentage of mountain shrubland in the late shrub-dominated stage varied from 18% to 71%, reflecting the dynamic nature of this cover type when considered over century-long periods (Figure-hrv). The seral-stage distribution appeared to be in dynamic equilibrium (i.e., the percentage in each stand condition varied about a stable mean), and appeared to reach equilibrium with roughly 200 years. The spatial configuration of stand conditions fluctuated markedly over time as well, although there was considerable variation in the magnitude of variability among configuration metrics (Table-hrv). Patch area and extent (i.e., radius of gyration) and the proximity index (a measure of patch isolation) exhibited the greatest variability and the herb-shrub condition was particularly dynamic relative to the early and late shrub-dominated stages.


      HRV Departure.--Unfortunately, due to the lack of data on current stand age and condition in this cover type, we were unable to evaluate departure of the current landscape from the simulated HRV.


Ponderosa Pine-Oak Forest [cover type description]


      Ponderosa pine-oak forest is relatively common on the UPL, encompassing 37,976 ha and comprising almost 6% of the landscape (Table-areal coverage).


      Sources of Uncertainty.--At the time of our simulation experiments, there was very little local empirical data available on disturbance regimes and succession processes in ponderosa pine-oak forests from which to parameterize the model. However, data from the adjoining San Juan National Forest indicated a composite mean fire interval of 6-13 years in lower-elevation ponderosa pine forests (Romme et al. 2003, Grissino-Mayer et al. 2004). While there is little uncertainty regarding the accuracy of this estimate, there is considerable confusion over its meaning given the many ways return intervals can be calculated and the potential biases in field sampling methods (Baker and Ehle 2001). The composite mean fire interval just referred to represents the frequency with which fire recurred anywhere within a sampling area of approximately 100 ha. However, not all of these fires burned the entire sampling area; on the contrary, most probably burned only a portion of the area, but what that portion was is unknown. To help resolve this uncertainty, and to help reduce the effect of small, localized fires on the composite mean fire interval, Romme et al. (2003) also computed a filtered composite mean fire interval, based only on fires that were recorded on >25% of recorder trees. These are the fires that presumably affected a greater proportion of the stand. The filtered composite mean fire intervals were approximately twice as long as the unfiltered, but filtering did not resolve the fundamental issue of what fraction of a sampling area actually was affected by each historic fire event. Moreover, the composite mean fire interval is profoundly influenced by the size of the sampled area: a larger area will encompass more historic fires and will therefore have a shorter composite mean fire interval.


      Because of the problems outlined above, we did not use composite mean fire intervals directly to parameterize RMLANDS. Instead, we calculated individual-tree fire intervals from the original data used to develop the composite mean fire intervals reported in Romme et al. (2003) and Grissino-Mayer et al. (2004). This individual-tree mean fire interval is equivalent to rotation at the landscape level if certain assumptions are valid (Baker and Ehle 2003). This calculation resulted in a 15-25 yr mean fire return interval (rotation period) for low-mortality fires during the reference period. However, we further adjusted our estimate of mean fire return interval to account for an important potential bias in the empirical data. Specifically, the empirical data was collected from extensive, relatively homogeneous stands, which we reasoned would be more likely to support frequent fires (due to continuous fuels facilitating fire spread) than small, disjunct stands or stands interspersed with less flammable cover types. Consequently, we reasoned that the mean fire return interval would be somewhat longer for the cover type as a whole, and our simulation results are consistent with this reasoning.


      It is important to emphasize that the results we report here reflect historical fire occurrence and the effects of fires throughout the ponderosa pine cover type, not just in areas of relatively homogeneous topography and vegetation structure where most of the local empirical fire history data were obtained; and that we report fire intervals at the scale of an individual 25-m pixel, and not as composite fire intervals within larger sampling areas. Isolated patches of ponderosa pine forest, surrounded by less flammable vegetation types, have substantially longer fire intervals in RMLANDS simulations -- as would be expected -- and consequently the mean fire intervals that we report for this vegetation type are longer than most published reports based on composite mean fire intervals.


      Recently, Peter Brown and Wayne Shepperd completed an analysis of fire history in ponderosa pine forests of the UPL (unpublished report to the Rocky Mountain Research Station, U.S. Forest Service, 2004). Although it was too late to incorporate their findings into our parameterization of RMLANDS, we note that their results were generally consistent with the broad conclusions of Romme et al. (2003) and Grissino-Mayer et al. (2004) in the San Juan Mountains.


      A major source of uncertainty regarding our treatment of wildfire in this cover type is the estimate of severity (or mortality) following wildfire. There is no empirical data from which to estimate the proportion of burned area exhibiting a high-mortality (stand-replacing) effect for fires during the reference period. There is considerable widespread evidence from Arizona and New Mexico (Cooper 1960, Swetnam and Baisan 1996, Fule et al. 1997) and eastern Oregon, Washington, and western Montana (Arno 1980, Agee 1993, Everett et al. 1994, Heyerdahl et al. 2001) that fire regimes in ponderosa pine forest during this period were principally low-mortality fires that maintained open forests of many size and age classes. In contrast, high-mortality fire was an important component of the historical fire regime in ponderosa pine forests of the northern Front Range in Colorado (e.g., Brown et al. 1999, Ehle and Baker 2003, Sherriff 2004). While most experts agree that at least some high-mortality fire was likely in ponderosa pine forests, there is no agreement on how much high-mortality occurred in any particular location. There is considerable evidence of high-mortality fires in ponderosa pine forests from recent historical fires (e.g., more than half of the ponderosa pine forest that burned in the 2002 Missionary Ridge fire on the adjoining San Juan National Forest exhibited high mortality), but the effects are confounded with 20th century land use practices. We reasoned that some high mortality was likely and estimated it to vary spatially and temporally with vegetation conditions. Except for very young pine stands, which we assumed to be highly vulnerable to mortality from fire, we estimated the percentage of high-mortality to be 1-5%, depending on pre-existing vegetation conditions. Clearly, this estimate affects the seral-stage distribution; in particular, the proportion of area maintained in the early seral stages of development.


      Another important source of uncertainty regarding our treatment of this cover type is the vegetation transition model itself; more specifically, with the discrete stand conditions (or seral stages) we defined for this cover type. Although the model we used in the simulations was developed in collaboration with a local expert team, recent discussions have led us to revise this model for future applications. In particular, the revisions are designed to better account for the ecological dominance of the shrubs during the early stages of development and to better accommodate wide variations among sites in tree cover during the intermediate and later stages of development. A specific concern centers on whether stands commonly develop a true stem exclusion condition before developing a mature overstory. The net result of this is that future simulations may produce a somewhat different picture of the seral-stage distribution under the same disturbance regime.


      Overall, despite the sources of uncertainty discussed above, this is a relatively well-studied cover type for which we have excellent empirical data on the major disturbance regime (i.e., wildfire) from a nearby location (San Juan National Forest) and locally (Brown and Shepperd). Hence, we have moderately high confidence in the reliability of the results for this cover type, subject to the constraint that the vegetation transition model adequately represents meaningful stand conditions and successional pathways.


      Wildfire.--The frequency and extent of simulated wildfires in ponderosa pine-oak forest varied markedly among decades (Figure-initiations, Figure-extent). Wildfire was quite prevalent in this cover type. In most decades, >10% of the ponderosa pine-oak forest burned, inclusive of both high- and low-mortality affected areas, and roughly 3-4 times per 100 years, >30% of the area burned (Figure-recurrence). Under this wildfire regime, the mean return interval between fires (of any mortality level) to a single 25-m cell varied widely from 22 years to >800 years, with a mean and median of roughly 45 years, and almost no eligible area escaped disturbance altogether over the course of an 800-year simulation (Figure-return). Note, the mean return interval was 552 years and 50 years for high- and low-mortality disturbances, respectively, due to the preponderance of low-mortality wildfires. In addition, the mean return interval between low-mortality fires as measured by the sample-based approach [in which each recorded interval between low-mortality fires in a cell was treated as an independent observation, in order to approximate the method of most dendrochronological fire history studies and that of Romme et al. (2003)] was somewhat shorter (median = 30 years; Figure-return). As expected, mean return intervals varied spatially across the forest (Figure-map). In general, return intervals increased with elevation, reflecting the moister, cooler conditions at higher elevations. In addition, ponderosa pine-oak stands embedded in a neighborhood containing cover types with longer return intervals (e.g., aspen, cool moist mixed-conifer forest) exhibited longer return intervals, reflecting the importance of landscape context on fire regimes.


      Pine Beetle.--The frequency and extent of simulated pine beetle epidemics in ponderosa pine-oak forest varied markedly among decades (Figure-initiations, Figure-extent). In most decades, pine beetles were at endemic levels and none of the ponderosa pine-oak forest was disturbed. However, 1-3 times per 100 years an epidemic affecting >5% of the host area occurred. On average, once every 50 years an epidemic affecting >10% of the host area would occur, and roughly once every 200-300 years a major epidemic affecting >20% of the host area would occur (Figure-recurrence), inclusive of both high- and low-mortality affected areas. The mean return interval between epidemics (of any mortality level) at a single location (25-m cell) varied from 47 years to >800 years, with a mean and median of 213 and 200 years, respectively, although 2-3% of the eligible area escaped disturbance altogether over the course of an 800-year simulation (Figure-return). Note, the mean return interval was 1,062 years and 265 years for high- and low-mortality disturbances, respectively, due to the preponderance of low-mortality outbreaks. As expected, mean return intervals varied spatially across the forest, but at a fine grain and in a seemingly random pattern (Figure-map).


      Age Structure.--The age structure and dynamics of ponderosa pine-oak forest reflected the interplay between disturbance and succession processes. The survivorship distribution represents the percent of stands that survived to any age, where age represents time since stand origin, not necessarily the age of the oldest trees in the stand. On average (over time), roughly 50% of the ponderosa pine-oak forest was >340 years since stand origin, although at any point in time this varied from 9% to 65% (Figure-survivorship). On average, 25% of the ponderosa pine-oak forest survived to >540 years, and roughly 6% survived a stand-replacing disturbance for >800 years. The relatively “old” age structure of this cover type may seem surprising at first; however, most wildfires in this cover type were low-mortality fires that did not result in stand initiation and the rotation period for high-mortality pine beetle epidemics was >1,000 years.


      Seral-stage Distribution.--The distribution of area among stand conditions within ponderosa pine-oak forest fluctuated over time, as expected (Figure-conditions). For example, the percentage of ponderosa pine-oak forest in the late-seral stages (i.e., understory reinitiation and shifting mosaic conditions) varied from 13% to 47%, reflecting the dynamic nature of this cover type when considered over century-long periods. However, fluctuations in the seral-stage distribution were much less pronounced in this cover type than in most others. The seral-stage distribution appeared to be in dynamic equilibrium (i.e., the percentage in each stand condition varied about a stable mean) and appeared to reach equilibrium within roughly 200 years. The spatial configuration of stand conditions fluctuated markedly over time as well, although there was considerable variation in the magnitude of variability among configuration metrics (Table-hrv). Patch area and extent (radius of gyration) and the proximity index (a measure of patch isolation) exhibited the greatest variability and the understory reinitiation condition was particularly dynamic relative to the earlier and later stages.


      HRV Departure.--Our estimate of the current seral-stage distribution was never observed under the simulated HRV (Figure-hrv). The most notable departure was in the fire-maintained open canopy (FMO) condition. The current landscape contains no ponderosa pine-oak forest in the FMO condition, yet this condition was always well represented (30-71%) under the simulated HRV. Conversely, the stand initiation and stem exclusion conditions are over-represented in the current landscape. In particular, the combined stand initiation and stem exclusion conditions comprise 82% of the current extent of this cover type, but never exceeded 35% under the simulated HRV. Overall, based on the five separate stand conditions (i.e., without aggregating the late-seral stages), the seral-stage departure index was 95% (Table-hrv). When the late-seral stages were aggregated, the seral-stage departure index declined to 75%. The current seral-stage configuration deviated similarly (94%) from the simulated HRV and with the exception of the interspersion and juxtaposition index was relatively consistent among metrics. Despite difficulties in classifying the late-seral stages in the current landscape, there is consistent evidence that, in general, the current landscape contains a greater number of larger, geometrically more complex and clumped (less isolated) stands in the stand initiation, stem exclusion and understory reinitiation stages than existed under the simulated HRV.


Ponderosa Pine-Oak-Aspen Forest [cover type description]


      Ponderosa pine-oak-aspen forest is much less common than ponderosa pine-oak forest on the UPL, encompassing 3,047 ha and comprising only 0.5% of the landscape (Table-areal coverage). In general, the vegetation dynamics in ponderosa pine-oak-aspen forest were very similar to those of ponderosa pine-oak forest. Nevertheless, we provide a complete description of the results for this cover type below and highlight the notable differences.


      Sources of Uncertainty.–For purposes of model parameterization and verification we treated this cover type the same as ponderosa pine-oak forest. This was necessary because the available empirical data and, to a large extent, the scientific literature, did not distinguish between pure ponderosa pine and pine-aspen forest. We reasoned that the less flammable aspen component of these stands would function to retard fire and therefore lengthen the fire rotation period, but we have no empirical basis from which to confirm or refute this hypothesis. Nevertheless, our simulation results are consistent with this logic, whereby the seral-stage and age distributions are shifted slightly towards the older age classes. Otherwise, the sources of uncertainty regarding our treatment of wildfire described previously for ponderosa pine-oak forest apply to this cover type as well (see previous discussion).


      Another important source of uncertainty regarding our treatment of this cover type is the vegetation transition model itself; more specifically, with the discrete stand conditions (or seral stages) we defined for this cover type. Although the model we used in the simulations was developed in collaboration with a local expert team, recent discussions indicate a lack of consensus on the stand conditions and the process of succession in this cover type. These discussions have led us to revise this model for future applications. In particular, the revisions are designed to better account for the ecological dominance and patchy distribution of the shrubs and aspen during the early stages of development and to better accommodate wide variations among sites in tree cover during the intermediate and later stages of development. A specific concern centers on whether stands commonly develop a true stem exclusion condition before developing a mature overstory. The net result of this is that future simulations may produce a somewhat different picture of the seral-stage distribution under the same disturbance regime.


      Overall, despite the sources of uncertainty discussed above, this is a relatively well-studied cover type (if lumped together with pure ponderosa pine-oak forest) for which we have excellent empirical data on the major disturbance regime (i.e., wildfire) from a nearby location (San Juan National Forest). Hence, we have moderately high confidence in the reliability of the results for this cover type, subject to the constraint that the vegetation transition model adequately represents meaningful stand conditions and successional pathways.


      Wildfire.--The frequency and extent of simulated wildfires in ponderosa pine-oak-aspen forest varied markedly among decades (Figure-initiations, Figure-extent). Wildfire was quite prevalent in this cover type, but somewhat less so than in ponderosa pine-oak forest. In most decades, >10% of the ponderosa pine-oak-aspen forest burned, inclusive of both high- and low-mortality affected areas, and roughly 2-3 times per 100 years, >30% of the area burned (Figure-recurrence). Under this wildfire regime, the mean return interval between fires (of any mortality level) to a single 25-m cell varied widely from 22 years to >800 years, with a mean and median of 47 and 44 years, respectively, and almost no eligible area escaped disturbance altogether over the course of an 800-year simulation (Figure-return). Note, the mean return interval was 932 years and 50 years for high- and low-mortality disturbances, respectively, due to the preponderance of low-mortality wildfires. In addition, the mean return interval between low-mortality fires as measured by the sample-based approach (in which each recorded interval between low-mortality fires in a cell was treated as an independent observation, in order to approximate the method of most dendrochronological fire history studies) was somewhat shorter (median = 30 years; Figure-return). As expected, mean return intervals varied spatially across the forest, but due to the scarcity of this cover type it was not possible to discern any meaningful patterns (Figure-map).


      Pine Beetle.--The frequency and extent of simulated pine beetle epidemics in ponderosa pine-oak-aspen forest varied markedly among decades (Figure-initiations, Figure-extent). In most decades, pine beetles were at endemic levels and none of the ponderosa pine-oak-aspen forest was disturbed. However, 1-3 times per 100 years an epidemic affecting >5% of the host area occurred. On average, once every 50 years an epidemic affecting >10% of the host area would occur, and roughly once every 300 years a major epidemic affecting >20% of the host area would occur (Figure-recurrence), inclusive of both high- and low-mortality affected areas. The mean return interval between epidemics (of any mortality level) at a single location (25-m cell) varied from 62 years to >800 years, with a mean and median of 218 and 200 years, respectively, although 2-3% of the eligible area escaped disturbance altogether over the course of an 800-year simulation (Figure-return). Note, the mean return interval was 1,221 years and 262 years for high- and low-mortality disturbances, respectively, due to the preponderance of low-mortality outbreaks. As expected, mean return intervals varied spatially across the forest, but at a very fine grain and in a seemingly random pattern (Figure-map).


      Age Structure.--The age structure and dynamics of ponderosa pine-oak-aspen forest reflected the interplay between disturbance and succession processes. The survivorship distribution represents the percent of stands that survived to any age, where age represents time since stand origin, not necessarily the age of the oldest trees in the stand. On average (over time), roughly 50% of the ponderosa pine-oak-aspen forest was >360 years since stand origin, although at any point in time this varied from 3% to 66% (Figure-survivorship). On average, 25% of the ponderosa pine-oak-aspen forest survived to >550 years, and roughly 6% survived a stand-replacing disturbance for >800 years. The relatively “old” age structure of this cover type may seem surprising at first; however, most wildfires in this cover type were low-mortality fires that did not result in stand reinitiation and the rotation period for high-mortality pine beetle epidemics was >1,000 years.


      Seral-stage Distribution.--The distribution of area among stand conditions within ponderosa pine-oak-aspen forest fluctuated over time, as expected (Figure-conditions). For example, the percentage of ponderosa pine-oak-aspen forest in the late-seral stages (i.e., understory reinitiation and shifting mosaic conditions) varied from 10% to 59% (slightly greater range of values than in ponderosa pine-oak forest), reflecting the dynamic nature of this cover type when considered over century-long periods. However, the fluctuations in the seral-stage distribution were much less pronounced in this cover type than in most others. The seral-stage distribution appeared to be in dynamic equilibrium (i.e., the percentage in each stand condition varied about a stable mean) and appeared to reach equilibrium within roughly 200 years. The spatial configuration of stand conditions fluctuated markedly over time as well, although there was considerable variation in the magnitude of variability among configuration metrics (Table-hrv). Patch area and extent (radius of gyration) and the proximity index (a measure of patch isolation) exhibited the greatest variability and the understory reinitiation condition was particularly dynamic relative to the earlier and later stages.


      HRV Departure.--Our estimate of the current seral-stage distribution was never observed under the simulated HRV (Figure-hrv). As with ponderosa pine-oak forest, the most notable departure was in the fire-maintained open canopy (FMO) condition. The current landscape contains no ponderosa pine-oak-aspen forest in the FMO condition, yet this condition was always well represented (31-79%) under the simulated HRV. Conversely, the stand initiation and stem exclusion conditions are over-represented in the current landscape. In particular, the stem exclusion condition comprises 42% of this cover type in the current landscape, but never exceeded 12% under the simulated HRV. There was a notable difference in departure between ponderosa pine-oak forest with and without aspen. The current landscape contains much less ponderosa pine-oak-aspen in the stand initiation and stem exclusion stages (58% vs. 82%) and much more (42% vs. 18%) in the late-seral (specifically, understory reinitiation) stage than in the pure ponderosa pine cover type. This is likely due to differences between cover types in the rate of succession during the early- and mid-seral stages. In the ponderosa pine with aspen, the stand initiation and stem exclusion stages are dominated by aspen, which transitions to the understory reinitiation stage between 80-120 years after stand origin - owing to the relatively short-lived aspen and early break-up of the aspen-dominated canopy. In contrast, the conifer canopy in the pure ponderosa pine forest takes 150-250 years to break up in the absence of other major disturbances. Given the paucity of wildfires over the past century, most ponderosa pine-oak-aspen stands have already transitioned to the understory reinitiation stage, whereas a larger proportion of the pure ponderosa pine stands have not yet transitioned. Overall, based on the five separate stand conditions (i.e., without aggregating the late-seral stages), the seral-stage departure index for ponderosa pine-oak-aspen was 96% (Table-hrv). When the late-seral stages were aggregated, the seral-stage departure index declined to 75%. The current seral-stage configuration deviated slightly less (90%) from the simulated HRV and, with the exception of the clumpiness index, was relatively consistent among metrics. Despite difficulties in classifying the late-seral stages in the current landscape, there is consistent evidence that, in general, the current landscape contains a greater number of larger, geometrically more complex and clumped (less isolated) stands in the stand initiation, stem exclusion and understory reinitiation stages than existed under the simulated HRV.


Warm-Dry Mixed-Conifer Forest [cover type description]


      Warm dry mixed-conifer forest is relatively uncommon on the UPL, encompassing 9,288 ha and comprising only 1.4% of the landscape (Table-areal coverage).


      Sources of Uncertainty.--At the time of our simulation experiments, there was very little local empirical data available on disturbance regimes and succession processes in warm dry mixed-conifer forests from which to parameterize the model. However, data from the adjoining San Juan National Forest indicated a composite mean fire interval of 19-30 years for three sites in warm, dry mixed conifer forest (Grissino-Mayer et al. 2004). Wu (1999) reported 8-18 year composite mean fire intervals for three other sites in the eastern SJNF. While there is little uncertainty regarding the accuracy of these estimates, there is considerable confusion over their meaning given the many ways return intervals can be calculated and the potential biases in field sampling methods (Baker and Ehle 2001). For example, the composite mean fire interval just referred to calculated by Grissino-Mayer represents the frequency with which fire recurred anywhere within a sampling area of approximately 100 ha. However, not all of these fires burned the entire sampling area; on the contrary, most probably burned only a portion of the area, but what that portion was is unknown. To help resolve this uncertainty, and to help reduce the effect of small, localized fires on the composite mean fire interval, Grissino-Mayer et al. (2004) also computed a filtered composite mean fire interval, based only on fires that were recorded on >25% of recorder trees. These are the fires that presumably affected a greater proportion of the stand. The filtered composite mean fire intervals were considerably longer than the unfiltered, but filtering did not resolve the fundamental issue of what fraction of a sampling area actually was affected by each historic fire event. Moreover, the composite mean fire interval is profoundly influenced by the size of the sampled area: a larger area will encompass more historic fires and will therefore have a shorter composite mean fire interval.


      Because of the problems outlined above, we did not use composite mean fire intervals directly to parameterize RMLANDS. Instead, we calculated individual-tree fire intervals from the original data used to develop the composite mean fire intervals reported in Grissino-Mayer et al. (2004). This individual-tree mean fire interval is equivalent to rotation at the landscape level if certain assumptions are valid (Baker and Ehle 2003). This calculation resulted in a 25-35 yr mean fire return interval (rotation period) for low-mortality fires during the reference period. However, we further adjusted our estimate of mean fire return interval to account for an important potential bias in the empirical data. Specifically, the empirical data was collected from extensive, relatively homogeneous stands, which we reasoned would be more likely to support frequent fires (due to continuous fuels facilitating fire spread) than small, disjunct stands or stands interspersed with less flammable cover types. Consequently, we reasoned that the mean fire return interval would be somewhat longer for the cover type as a whole, and our simulation results are consistent with this reasoning.


      It is important to emphasize that the results we report here reflect historical fire occurrence and the effects of fires throughout the ponderosa pine cover type, not just in areas of relatively homogeneous topography and vegetation structure where most of the local empirical fire history data were obtained; and that we report fire intervals at the scale of an individual 30-m pixel, and not as composite fire intervals within larger sampling areas. Isolated patches of ponderosa pine forest, surrounded by less flammable vegetation types, have substantially longer fire intervals in RMLANDS simulations -- as would be expected -- and consequently the mean fire intervals that we report for this vegetation type are longer than most published reports based on composite mean fire intervals.


      Recently, Peter Brown and Wayne Shepperd completed an analysis of fire history in ponderosa pine forests of the UPL (unpublished report to the Rocky Mountain Research Station, U.S. Forest Service, 2004). Although it was too late to incorporate their findings into our parameterization of RMLANDS, we note that their results were generally consistent with the broad conclusions of Romme et al. (2003) and Grissino-Mayer et al. (2004) in the San Juan Mountains.


      A major source of uncertainty regarding our treatment of wildfire in this cover type is the estimate of severity (or mortality) following wildfire. There is no empirical data from which to estimate the proportion of burned area exhibiting a high-mortality (stand-replacing) effect for fires during the reference period, although it is generally believed that this cover type was characterized by a mixed-severity regime (Romme et al. 2003). The available empirical data suggest that fires during this period were principally low-mortality fires that maintained open forests of many size and age classes, but that high-mortality effects were probably common as well. It has been established that high-mortality fire was an important component of the historical fire regime in ponderosa pine and ponderosa pine - Douglas-fir forests of the northern Front Range in Colorado (e.g., Brown et al. 1999, Ehle and Baker 2003, Sherriff 2004). While most experts agree that at least some high-mortality fire was likely in warm, dry mixed-conifer forests, there is no agreement on how much high-mortality occurred in any particular location. There is considerable evidence of high-mortality fires in warm dry mixed-conifer forests from recent historical fires (e.g., roughly half of the warm, dry mixed-conifer forest that burned in the 2002 Missionary Ridge fire on the adjoining San Juan National Forest exhibited high mortality), but the effects are confounded with 20th century land use practices. We reasoned that some high mortality was likely and estimated it to vary spatially and temporally with vegetation conditions. Except for very young pine stands, which we assumed to be highly vulnerable to mortality from fire, we estimated the percentage of high-mortality to be 5-15%, depending on pre-existing vegetation conditions. Clearly, this estimate affects the seral-stage distribution; in particular, the proportion of area maintained in the early seral stages of development.


      Another important source of uncertainty regarding our treatment of this cover type is the vegetation transition model itself; more specifically, with the discrete stand conditions (or seral stages) we defined for this cover type. Although the model we used in the simulations was developed in collaboration with a local expert team, recent discussions have led us to revise this model for future applications. In particular, the revisions are designed to better account for the ecological dominance of the shrubs during the early stages of development and to better accommodate wide variations among sites in tree cover during the intermediate and later stages of development. A specific concern centers on whether stands commonly develop a true stem exclusion condition before developing a mature overstory. The net result of this is that future simulations may produce a somewhat different picture of the seral-stage distribution under the same disturbance regime.


      Overall, despite the sources of uncertainty discussed above, this is a relatively well-studied cover type, although not nearly so much as pure ponderosa pine forest, for which we have some empirical data on the major disturbance regime (i.e., wildfire) from a nearby location (San Juan National Forest). Hence, we have moderately high confidence in the reliability of the results for this cover type, although less so than for pure ponderosa pine forest, subject to the constraint that the vegetation transition model adequately represents meaningful stand conditions and successional pathways.


      Wildfire.--The frequency and extent of simulated wildfires in warm dry mixed-conifer forest varied markedly among decades (Figure-initiations, Figure-extent). Wildfire was quite prevalent in this cover type. In most decades, >10% of the warm dry mixed-conifer forest burned, inclusive of both high- and low-mortality affected areas, and roughly once per 100 years, >30% of the area burned (Figure-recurrence). Under this wildfire regime, the mean return interval between fires (of any mortality level) to a single 25-m cell varied widely from 23 years to >800 years, with a mean and median of 63 and 67 years, respectively, and almost no eligible area escaped disturbance altogether over the course of an 800-year simulation (Figure-return). Note, the mean return interval was 592 years and 71 years for high- and low-mortality disturbances, respectively, due to the preponderance of low-mortality wildfires. In addition, the mean return interval between low-mortality fires as measured by the sample-based approach (in which each recorded interval between low-mortality fires in a cell was treated as an independent observation, in order to approximate the method of most dendrochronological fire history studies) was somewhat shorter (median = 40 years). As expected, mean return intervals varied spatially across the forest, but due to the scarcity of this cover type it was not possible to discern any meaningful patterns (Figure-map).


      Spruce Beetle.--Spruce beetles can venture into warm dry mixed-conifer forest, but due to the scarce and patchy distribution of suitable host trees in this cover type, epidemics are relatively insignificant and have negligible impact on this cover type (Figure-initiations, Figure-extent). Only occasionally (e.g., once every 200 years) did spruce beetles disturb >2% of the warm dry mixed-conifer forest, inclusive of both high- and low-mortality affected areas (Figure-recurrence). Consequently, mean return intervals between epidemics (of any mortality level) at a single location (25-m cell) typically exceeded 800 years (Figure-return).


      Spruce Budworm.--The frequency and extent of simulated spruce budworm epidemics in warm dry mixed-conifer forest varied markedly among decades (Figure-initiations, Figure-extent). In almost two-thirds of the decades, spruce budworm populations were at endemic levels and none of the warm dry mixed-conifer forest was disturbed. However, 2-3 times per 100 years, epidemics occurred affecting >10% of the area, inclusive of both high- and low-mortality affected areas, and approximately once every 100 years, >25% of the warm dry mixed-conifer forest was disturbed (Figure-recurrence). The mean return interval between epidemics (of any mortality level) at a single location (25-m cell) varied from 38 years to >800 years, with a mean and median of 100 years, although very little eligible area (<1%) escaped disturbance altogether over the course of an 800-year simulation (Figure-return). Note, the mean return interval was 1,046 years and 106 years for high- and low-mortality disturbances, respectively, due to the preponderance of low-mortality outbreaks. As expected, mean return intervals varied spatially across the forest, but at a very fine grain and in a seemingly random pattern (Figure-map).


      Douglas-fir Beetle.--The frequency and extent of simulated Douglas-fir beetle epidemics in warm dry mixed-conifer forest varied markedly among decades (Figure-initiations, Figure-extent). In most decades <1% of the warm dry mixed-conifer forest area was affected by an outbreak; however, 2-3 times per 100 years an epidemic affecting >1% of the host area would occur, and once every 300-400 years an epidemic would affect >3% of the host area (Figure-recurrence), inclusive of both high- and low-mortality affected areas. The mean return interval between epidemics (of any mortality level) at a single location (25-m cell) varied from 89 years to >800 years, with a mean and median of >800 years, and almost half of the area escaped disturbance altogether over the course of an 800-year simulation (Figure-return). Mean return intervals varied spatially across the forest, but at a very fine grain and in a seemingly random pattern (Figure-map).


      Pine Beetle.--The frequency and extent of simulated pine beetle epidemics in warm dry mixed-conifer forest varied markedly among decades (Figure-initiations, Figure-extent). In most decades, pine beetles were at endemic levels and none of the warm dry mixed-conifer forest was disturbed. However, 1-3 times per 100 years an epidemic occurred affecting >5%. On average, once every 100 years an epidemic affecting >10% of the host area would occur, and roughly once every 400 years a major epidemic affecting >20% of the host area would occur (Figure-recurrence), inclusive of both high- and low-mortality affected areas. The mean return interval between epidemics (of any mortality level) at a single location (25-m cell) varied from 62 years to >800 years, with a mean and median of 266 years, although roughly 5% of the eligible area escaped disturbance altogether over the course of an 800-year simulation (Figure-return). Note, the mean return interval was 1,311 years and 331 years for high- and low-mortality disturbances, respectively, due to the preponderance of low-mortality outbreaks. As expected, mean return intervals varied spatially across the forest, but at a very fine grain and in a seemingly random pattern (Figure-map).


      Age Structure.--The age structure and dynamics of warm dry mixed-conifer forest reflected the interplay between disturbance and succession processes. The survivorship distribution represents the percent of stands that survived to any age, where age represents time since stand origin, not necessarily the age of the oldest trees in the stand. On average (over time), roughly 50% of the warm dry mixed-conifer forest was >360 years since stand origin, although at any point in time this varied from 8% to 62% (Figure-survivorship). On average, 25% of the warm dry mixed-conifer forest survived to >560 years, and roughly 7% survived a stand-replacing disturbance for >800 years. The relatively “old” age structure of this cover type may seem surprising at first; however, most wildfires in this cover type were low-mortality fires that did not result in stand reinitiation.


      Seral-stage Distribution.--The distribution of area among stand conditions within warm dry mixed-conifer forest fluctuated over time, as expected (Figure-conditions). For example, the percentage of warm dry mixed-conifer forest in the late-seral stages (i.e., understory reinitiation and shifting mosaic conditions) varied from 22% to 69%, reflecting the dynamic nature of this cover type when considered over century-long periods. However, the fluctuations in the seral-stage distribution were much less pronounced in this cover type than in most others - with the exception of the ponderosa pine forest types. The seral-stage distribution appeared to be in dynamic equilibrium (i.e., the percentage in each stand condition varied about a stable mean) and appeared to reach equilibrium within roughly 200 years. The spatial configuration of stand conditions fluctuated markedly over time as well, although there was considerable variation in the magnitude of variability among configuration metrics (Table-hrv). Patch area and the proximity index (a measure of patch isolation) exhibited the greatest variability and the understory reinitiation conditions was particularly dynamic relative to the other stages.


      HRV Departure.--Our estimate of the current seral-stage distribution was never observed under the simulated HRV (Figure-hrv). The most notable departure was in the fire-maintained open canopy (FMO) condition. The current landscape contains no warm dry mixed-conifer forest in the FMO condition, yet this condition was always well represented (21-63%) under the simulated HRV. Conversely, the current landscape contains 63% of the cover type in the stem exclusion condition, yet this condition never exceeded 13% under the simulated HRV. Overall, based on the five separate stand conditions (i.e., without aggregating the late-seral stages), the seral-stage departure index was 94% (Table-hrv). When the late-seral stages were aggregated, the seral-stage departure index declined slightly to 90%. The current seral-stage configuration deviated similarly (91%) from the simulated HRV and was relatively consistent among metrics, with the lowest departure index equal to 71%. In general, the current landscape contains a greater number of larger, geometrically more complex and clumped (less isolated) stands in the stand initiation and stem exclusion stages than existed under the simulated HRV. In addition, despite difficulties in classifying the late-seral stages in the current landscape, there is evidence that stands in the shifting mosaic (old-growth) stage in the current landscape, while fewer in number and more isolated from each other, are larger and geometrically more complex than existed under the simulated HRV.


Warm-Dry Mixed-Conifer with Aspen Forest [cover type description]


      Warm dry mixed-conifer-aspen forest is the least common vegetation type on the UPL, encompassing 1,056 ha and comprising only 0.2% of the landscape (Table-areal coverage). In general, the vegetation dynamics in warm dry mixed-conifer-aspen forest were very similar to those of warm dry moist mixed-conifer forest. Nevertheless, we provide a complete description of the results for this cover type below and highlight the notable differences.


      Sources of Uncertainty.–For purposes of model parameterization and verification we treated this cover type the same as warm dry mixed-conifer forest. This was necessary because the available empirical data and, to a large extent, the scientific literature, did not distinguish between pure mixed-conifer and mixed conifer-aspen forest. We reasoned that the less flammable aspen component of these stands would function to retard fire and therefore lengthen the fire rotation period, but we have no empirical basis from which to confirm or refute this hypothesis. Nevertheless, our simulation results are consistent with this logic, whereby the seral-stage and age distributions are shifted slightly towards the older age classes. Otherwise, the sources of uncertainty regarding our treatment of wildfire described previously for warm dry mixed-conifer forest apply to this cover type as well (see previous discussion).


      Another important source of uncertainty regarding our treatment of this cover type is the vegetation transition model itself; more specifically, with the discrete stand conditions (or seral stages) we defined for this cover type. Although the model we used in the simulations was developed in collaboration with a local expert team, recent discussions indicate a lack of consensus on the stand conditions and the process of succession in this cover type. These discussions have led us to revise this model for future applications. In particular, the revisions are designed to better account for the ecological dominance and patchy distribution of the shrubs and aspen during the early stages of development and to better accommodate wide variations among sites in tree cover during the intermediate and later stages of development. A specific concern centers on whether stands commonly develop a true stem exclusion condition before developing a mature overstory. The net result of this is that future simulations may produce a somewhat different picture of the seral-stage distribution under the same disturbance regime.


      Overall, despite the sources of uncertainty discussed above, this is a relatively well-studied cover type (if lumped together with pure warm dry mixed-conifer forest), although not nearly so much as pure ponderosa pine forest, for which we have some empirical data on the major disturbance regime (i.e., wildfire) from a nearby location (San Juan National Forest). Hence, were it not for its poor representation on the UPL (1,056 ha), we would have moderately high confidence in the reliability of the results for this cover type, although less so than for ponderosa pine forests, subject to the constraint that the vegetation transition model adequately represents meaningful stand conditions and successional pathways. However, due to its poor representation on the UPL, we place low confidence in the results.


      Wildfire.--The frequency and extent of simulated wildfires in warm dry mixed-conifer-aspen forest varied markedly among decades (Figure-initiations, Figure-extent). Wildfire was quite prevalent in this cover type. In most decades, >10% of the warm dry mixed-conifer-aspen forest burned, inclusive of both high- and low-mortality affected areas, and roughly once per 100 years, >35% of the area burned (Figure-recurrence). Under this wildfire regime, the mean return interval between fires (of any mortality level) to a single 25-m cell varied widely from 35 years to 800 years, with a mean and median of roughly 66 years, and no eligible area escaped disturbance altogether over the course of an 800-year simulation (Figure-return). Note, the mean return interval was 505 years and 79 years for high- and low-mortality disturbances, respectively, due to the preponderance of low-mortality wildfires. In addition, the mean return interval between low-mortality fires as measured by the sample-based approach (in which each recorded interval between low-mortality fires in a cell was treated as an independent observation, in order to approximate the method of most dendrochronological fire history studies) was somewhat shorter (median = 40 years). As expected, mean return intervals varied spatially across the forest (Figure-map), but given the scarcity of this cover type it was impossible to discern whether there were any meaningful patterns.


      Spruce Beetle.--Spruce beetles can venture into warm dry mixed-conifer-aspen forest, but due to the scarce and patchy distribution of suitable host trees in this cover type, epidemics are relatively insignificant and have negligible impact on this cover type (Figure-initiations, Figure-extent). Only occasionally (e.g., once every 200 years) did spruce beetles disturb >2% of the warm dry mixed-conifer-aspen forest, inclusive of both high- and low-mortality affected areas (Figure-recurrence). Consequently, mean return intervals between epidemics (of any mortality level) at a single location (25-m cell) typically exceeded 800 years (Figure-return).


      Spruce Budworm.--The frequency and extent of simulated spruce budworm epidemics in warm dry mixed-conifer-aspen forest varied markedly among decades (Figure-initiations, Figure-extent), and were slightly less prevalent than in warm dry mixed-conifer forest due to the lower density of suitable host trees. In almost two-thirds of the decades, spruce budworm populations were at endemic levels and none of the warm dry mixed-conifer-aspen forest was disturbed. However, 2-3 times per 100 years, epidemics occurred affecting >10% of the area, inclusive of both high- and low-mortality affected areas, and approximately once every 100 years, >25% of the warm dry mixed-conifer-aspen forest was disturbed (Figure-recurrence). The mean return interval between epidemics (of any mortality level) at a single location (25-m cell) varied from 38 years to >800 years, with a mean and median of 133 years, although very little eligible area (<1%) escaped disturbance altogether over the course of an 800-year simulation (Figure-return). Note, the mean return interval was 1,425 years and 149 years for high- and low-mortality disturbances, respectively, due to the preponderance of low-mortality outbreaks. As expected, mean return intervals varied spatially across the forest (Figure-map), but given the scarcity of this cover type it was impossible to discern whether there were any meaningful patterns.


      Douglas-fir Beetle.--The frequency and extent of simulated Douglas-fir beetle epidemics in warm dry mixed-conifer-aspen forest varied markedly among decades (Figure-initiations, Figure-extent), and were slightly more prevalent than in warm dry mixed-conifer forest. In most decades <1% of the warm dry mixed-conifer-aspen forest area was affected by an outbreak; however, 2-3 times per 100 years an epidemic affecting >1% of the host area would occur, and once every 300-400 years an epidemic would affect >6% (Figure-recurrence), inclusive of both high- and low-mortality affected areas. The mean return interval between epidemics (of any mortality level) at a single location (25-m cell) varied from 80 years to >800 years, with a mean and median of >800 years, and a large portion of the area (~40%) escaped disturbance altogether over the course of an 800-year simulation (Figure-return). Mean return intervals varied spatially across the forest, but at a very fine grain and in a seemingly random pattern (Figure-map), but given the scarcity of this cover type it was impossible to discern whether there were any meaningful patterns.


      Pine Beetle.--The frequency and extent of simulated pine beetle epidemics in warm dry mixed-conifer-aspen forest varied markedly among decades (Figure-initiations, Figure-extent), and were slightly less prevalent than in warm dry mixed-conifer forest due to the lower density of suitable host trees. In most decades, pine beetles were at endemic levels and none of the warm dry mixed-conifer-aspen forest was disturbed. However, 1-3 times per 100 years an epidemic affecting >5% occurred. On average, once every 100 years an epidemic affecting >10% of the host area would occur, and roughly once every 400 years a major epidemic affecting >20% of the host area would occur (Figure-recurrence), inclusive of both high- and low-mortality affected areas. The mean return interval between epidemics (of any mortality level) at a single location (25-m cell) varied from 62 years to >800 years, with a mean and median of 361 and 400 years, respectively, and roughly 5% of the eligible area escaped disturbance altogether over the course of an 800-year simulation (Figure-return). Note, the mean return interval was 1,969 years and 417 years for high- and low-mortality disturbances, respectively, due to the preponderance of low-mortality outbreaks. As expected, mean return intervals varied spatially across the forest (Figure-map), but given the scarcity of this cover type it was impossible to discern whether there were any meaningful patterns.


      Age Structure.--The age structure and dynamics of warm dry mixed-conifer-aspen forest reflected the interplay between disturbance and succession processes. The survivorship distribution represents the percent of stands that survived to any age, where age represents time since stand origin, not necessarily the age of the oldest trees in the stand. On average (over time), roughly 50% of the warm dry mixed-conifer-aspen forest was >340 years since stand origin, although at any point in time this varied from 6% to 63% (Figure-survivorship). On average, 25% of the warm dry mixed-conifer-aspen forest survived to >540 years, and roughly 6% survived a stand-replacing disturbance for >800 years. The relatively “old” age structure of this cover type may seem surprising at first; however, most wildfires in this cover type were low-mortality fires that did not result in stand reinitiation.


      Seral-stage Distribution.--The distribution of area among stand conditions within warm dry mixed-conifer-aspen forest fluctuated over time, as expected (Figure-conditions). For example, the percentage of warm dry mixed-conifer-aspen forest in the late-seral stages (i.e., understory reinitiation and shifting mosaic conditions) varied from 10% to 80% (slightly greater range of values than in warm dry mixed-conifer forest), reflecting the dynamic nature of this cover type when considered over century-long periods. The seral-stage distribution appeared to be in dynamic equilibrium (i.e., the percentage in each stand condition varied about a stable mean) and appeared to reach equilibrium within roughly 100 years. The spatial configuration of stand conditions fluctuated markedly over time as well, although there was considerable variation in the magnitude of variability among configuration metrics (Table-hrv). Patch density, edge density, core area metrics and the proximity index (a measure of patch isolation) exhibited the greatest variability.


      HRV Departure.--Our estimate of the current seral-stage distribution was never observed under the simulated HRV (Figure-hrv). As with warm dry mixed-conifer forest, the most notable departure was in the fire-maintained open canopy (FMO) condition. The current landscape contains no warm dry mixed-conifer-aspen forest in the FMO condition, yet this condition was always well represented (11-72%) under the simulated HRV. Similarly, the current landscape contains roughly 1% of the cover type in the stand initiation stage, yet this low a percentage was almost never observed under the simulated HRV. Conversely, the current landscape contains 46% of the cover type in the stem exclusion condition, yet this condition never exceeded 16% under the simulated HRV. There was a notable difference in seral-stage departure between warm dry mixed-conifer forest with and without aspen. The current landscape contains much less mixed-conifer with aspen in the earlier seral stages (i.e., stand initiation and stem exclusion stages) than in the pure conifer cover type (47% versus 69%); a greater proportion of the mixed-conifer-aspen forest has succeeded to the understory reinitiation stage. This is likely due to differences between cover types in the rate of succession during early seral stages. For example, in the mixed-conifer-aspen type, the stem exclusion stage is dominated by aspen, which transitions to the understory reinitiation stage between 80-120 years after stand origin - owing to the relatively short-lived aspen and early break-up of the aspen-dominated canopy. In contrast, the conifer canopy in the pure conifer forest takes 150-250 years to break up in the absence of other major disturbance processes. Hence, given the paucity of wildfires over the past century, a greater proportion of mixed-conifer-aspen stands have already transitioned to the understory reinitiation stage compared to the pure conifer stands. Overall, based on the five separate stand conditions (i.e., without aggregating the late-seral stages), the seral-stage departure index was 97% (Table-hrv). When the late-seral stages were aggregated, the seral-stage departure index declined to 74%. The current seral-stage configuration deviated less dramatically (77%) from the simulated HRV and varied considerably among metrics, with the lowest departure index equal to 27%. However, given the scarcity of this cover type, it was not possible to reliably interpret the specific nature of the configuration departure.


Cool-Moist Mixed-Conifer Forest [cover type description]


      Cool moist mixed-conifer forest is relatively uncommon on the UPL, encompassing 1,124 ha and comprising only 0.2% of the landscape (Table-areal coverage).


      Sources of Uncertainty.–With respect to disturbance regimes and vegetation dynamics, cool moist mixed-conifer forests (with and without aspen) probably are the least well understood of all of the forest types in the South Central Highlands Section. At the time of our simulation experiments, there was essentially no local empirical data available on disturbance regimes and succession processes in cool moist mixed-conifer forests from which to parameterize the model. Thus, we were obligated to rely principally on expert opinion. In general, based on a number of observations (see Romme et al. 2003), we reasoned that the wildfire disturbance regime during the reference period was characterized by infrequent, high-severity fires, much like the higher elevation spruce-fir forests, but that the mean fire return interval would be somewhat shorter than in spruce-fir forests due to the lower elevation of the mixed-conifer forest and the propensity for lower elevation fires to spread into this cover type under the right climatic conditions. Thus, while we had no objective basis on which to verify model outcomes for this cover type, we sought to observe patterns intermediate between the more objective targets established for the higher elevation spruce-fir forests and the lower elevation warm dry mixed-conifer and ponderosa pine forest. Our simulation results are consistent with this objective.


      Overall, given the lack of empirical data on disturbance regimes and vegetation dynamics in this cover and its poor representation on the UPL, we have low confidence in the reliability of the results for this cover type.


      Wildfire.--The frequency and extent of simulated wildfires in cool moist mixed-conifer forest varied markedly among decades (Figure-initiations, Figure-extent). In about one third of the decades, less than 1% of the cool moist mixed-conifer forest burned, inclusive of both high- and low-mortality affected areas. However, on average, 2-3 times per 100 years, >10% of the area burned, and roughly once every 200 years, >30% of the cool moist mixed-conifer forest burned (Figure-recurrence). Under this wildfire regime, the mean return interval between fires (of any mortality level) to a single 25-m cell varied widely from 36 years to >800 years, with a mean and median of roughly 100 years, and only a negligible area escaped disturbance altogether over the course of an 800-year simulation (Figure-return). Note, the mean return interval was 183 years and 244 years for high- and low-mortality disturbances, respectively, due to the preponderance of high-mortality wildfires. As expected, mean return intervals varied spatially across the forest (Figure-map), but given the scarcity of this cover type it was impossible to discern whether there were any meaningful patterns.


      Spruce Beetle.--The frequency and extent of simulated spruce beetle epidemics in cool moist mixed-conifer forest varied markedly among decades (Figure-initiations, Figure-extent). In most decades, less than 1% of the cool moist mixed-conifer forest was disturbed, inclusive of both high- and low-mortality affected areas. However, on average, roughly once per 200 years, >10% of the area was disturbed, and at least once every 400 years, >50% of the area was disturbed (Figure-recurrence). The mean return interval between epidemics (of any mortality level) at a single location (25-m cell) varied widely from 89 years to >800 years, with a mean and median of 383 and 400 years, respectively, although 5-10% of the eligible area escaped disturbance altogether over the course of an 800-year simulation (Figure-return). Note, the mean return interval was 415 years and 3,724 years for high- and low-mortality disturbances, respectively, due to the preponderance of high-mortality outbreaks. Recall that high-mortality spruce beetle outbreaks do not result in stand initiation in this cover type unless they are concomitant with high-mortality outbreaks of other insect disturbance agents. Not surprisingly, mean return intervals varied spatially across the forest (Figure-map), but given the scarcity of this cover type it was impossible to discern whether there were any meaningful patterns.


      Spruce Budworm.--The frequency and extent of simulated spruce budworm epidemics in cool moist mixed-conifer forest varied markedly among decades (Figure-initiations, Figure-extent). In almost two-thirds of the decades, spruce budworm populations were at endemic levels and none of the cool moist mixed-conifer forest was disturbed. However, roughly 3 times per 100 years, epidemics occurred affecting >10% of the area, inclusive of both high- and low-mortality affected areas, and approximately once every 100 years, >35% of the area was disturbed (Figure-recurrence). The mean return interval between epidemics (of any mortality level) at a single location (25-m cell) varied from 36 years to >800 years, with a mean and median of roughly 100 years, and only a negligible area escaped disturbance altogether over the course of an 800-year simulation (Figure-return). Note, the mean return interval was 1,432 years and 101 years for high- and low-mortality disturbances, respectively, due to the preponderance of low-mortality outbreaks. As expected, mean return intervals varied spatially across the forest (Figure-map), but given the scarcity of this cover type it was impossible to discern whether there were any meaningful patterns.


      Douglas-fir Beetle.--The frequency and extent of simulated Douglas-fir beetle epidemics in cool moist mixed-conifer forest varied markedly among decades (Figure-initiations, Figure-extent). In most decades <1% of the cool moist mixed-conifer forest area was affected by an outbreak; however, at least once every 100 years an epidemic affecting >1% of the host area would occur (Figure-recurrence), inclusive of both high- and low-mortality affected areas. The mean return interval between epidemics (of any mortality level) at a single location (25-m cell) varied from 200 years to >800 years, with a mean and median of >800 years (Figure-return). Mean return intervals varied spatially across the forest (Figure-map), but given the scarcity of this cover type it was impossible to discern whether there were any meaningful patterns.


      Age Structure.--The age structure and dynamics of cool moist mixed-conifer forest reflected the interplay between disturbance and succession processes. The survivorship distribution represents the percent of stands that survived to any age, where age represents time since stand origin, not necessarily the age of the oldest trees in the stand. On average (over time), roughly 50% of the cool moist mixed-conifer forest was >110 years since stand origin, although at any point in time this varied from 20% to 76% (Figure-survivorship). On average, 25% of the cool moist mixed-conifer forest survived to >220 years, 5% survived to >420, and <1% survived a stand-replacing disturbance for >600 years.


      Seral-stage Distribution.--The distribution of area among stand conditions within cool moist mixed-conifer forest fluctuated over time, as expected (Figure-conditions). For example, the percentage of cool moist mixed-conifer forest in the late-seral stages (i.e., understory reinitiation and shifting mosaic conditions) varied from 13% to 69%, reflecting the extremely dynamic nature of this high-elevation forest cover type when considered over century-long periods. The seral-stage distribution appeared to be in dynamic equilibrium (i.e., the percentage in each stand condition varied about a stable mean) and appeared to reach equilibrium within 100 years. The spatial configuration of stand conditions fluctuated markedly over time as well, although there was considerable variation in the magnitude of variability among configuration metrics (Table-hrv). Patch area, core area and the proximity index (a measure of patch isolation) exhibited the greatest variability over time.


      HRV Departure.--Our estimate of the current seral-stage distribution was never observed under the simulated HRV (Figure-hrv). The current landscape contains only 3% of the cool moist mixed-conifer forest in the stand initiation condition, yet this condition was always better represented (4-55%) under the simulated HRV. Similarly, the current landscape contains 18% of the cover type in the late-seral stages (i.e., understory reinitiation and shifting mosaic conditions), yet these stages were usually better represented (13-69%) under the simulated HRV. Conversely, the current landscape contains a large pulse (79%) of cool moist mixed-conifer forest in the stem exclusion stage, although it never exceeded 61% in this stage under the simulated HRV. Overall, based on the four separate stand conditions (i.e., without aggregating the late-seral stages), the seral-stage departure index was 98% (Table-hrv). When the late-seral stages were aggregated, the seral-stage departure index remained the same. The current seral-stage configuration deviated somewhat less (81%) from the simulated HRV, although it varied considerably among metrics, with the lowest departure index equal to 32%. Not surprisingly given the current seral-stage distribution, the current landscape contains a greater number of larger, more extensive, geometrically more complex and less isolated stands in the stem exclusion stage than existed under the simulated HRV.


Cool-Moist Mixed-Conifer with Aspen Forest [cover type description]


      Cool moist mixed-conifer-aspen forest is roughly four times more common than cool moist mixed-conifer forest, but is still nonetheless relatively uncommon on the UPL, encompassing 4,540 ha and comprising only 0.7% of the landscape (Table-areal coverage). In general, the vegetation dynamics in cool moist mixed-conifer-aspen forest were very similar to those of cool moist mixed-conifer forest. Nevertheless, we provide a complete description of the results for this cover type below and highlight the notable differences.


      Sources of Uncertainty.–As note above, with respect to disturbance regimes and vegetation dynamics, cool moist mixed-conifer forests (with and without aspen) probably are the least well understood of all of the forest types in the South Central Highlands Section. At the time of our simulation experiments, there was essentially no local empirical data available on disturbance regimes and succession processes in cool moist mixed-conifer-aspen forests from which to parameterize the model. Thus, we were obligated to rely principally on expert opinion. In general, we treated this cover type similar to cool moist mixed-conifer forest without aspen. However, we reasoned that the less flammable aspen component of these stands would function to retard fire and therefore lengthen the fire rotation period, but we have no empirical basis from which to confirm or refute this hypothesis. Nevertheless, our simulation results are consistent with this logic, whereby the seral-stage and age distributions are shifted slightly towards the older age classes. Otherwise, the sources of uncertainty regarding our treatment of wildfire described previously for cool moist mixed-conifer forest apply to this cover type as well (see previous discussion).


      Overall, given the lack of empirical data on disturbance regimes and vegetation dynamics in this cover and its relatively poor representation on the UPL, we have low confidence in the reliability of the results for this cover type.


      Wildfire.--The frequency and extent of simulated wildfires in cool moist mixed-conifer-aspen forest varied markedly among decades (Figure-initiations, Figure-extent). In about one third of the decades, less than 1% of the cool moist mixed-conifer-aspen forest burned, inclusive of both high- and low-mortality affected areas. However, on average, 2-3 times per 100 years, >10% of the area burned, and roughly once every 200 years, >30% of the cool moist mixed-conifer-aspen forest burned (Figure-recurrence). Under this wildfire regime, the mean return interval between fires (of any mortality level) to a single 25-m cell varied widely from 40 years to >800 years, with a mean and median of roughly 115 years, and negligible area escaped disturbance altogether over the course of an 800-year simulation (Figure-return). Note, the mean return interval was 189 years and 359 years for high- and low-mortality disturbances, respectively, due to the preponderance of high-mortality wildfires. As expected, mean return intervals varied spatially across the forest (Figure-map), but given the scarcity of this cover type it was impossible to discern whether there were any meaningful patterns.


      Spruce Beetle.--The frequency and extent of simulated spruce beetle epidemics in cool moist mixed-conifer-aspen forest varied markedly among decades (Figure-initiations, Figure-extent). In most decades, less than 1% of the cool moist mixed-conifer-aspen forest was disturbed, inclusive of both high- and low-mortality affected areas. However, on average, roughly once per 200 years, >10% of the area was disturbed, and at least once every 400 years, >50% of the area was disturbed (Figure-recurrence). The mean return interval between epidemics (of any mortality level) at a single location (25-m cell) varied widely from 80 years to >800 years, with a mean and median of 345 and 400 years, respectively, and 5-10% of the eligible area escaped disturbance altogether over the course of an 800-year simulation (Figure-return). Note, the mean return interval was 390 years and 3,379 years for high- and low-mortality disturbances, respectively, due to the preponderance of high-mortality outbreaks. Recall that high-mortality spruce beetle outbreaks do not result in stand initiation in this cover type unless they are concomitant with high-mortality outbreaks of other insect disturbance agents. Not surprisingly, mean return intervals varied spatially across the forest (Figure-map), but given the scarcity of this cover type it was impossible to discern whether there were any meaningful patterns.


      Spruce Budworm.--The frequency and extent of simulated spruce budworm epidemics in cool moist mixed-conifer-aspen forest varied markedly among decades (Figure-initiations, Figure-extent). In almost two-thirds of the decades, spruce budworm populations were at endemic levels and none of the cool moist mixed-conifer-aspen forest was disturbed. However, roughly 3 times per 100 years, epidemics occurred affecting >10% of the area, inclusive of both high- and low-mortality affected areas, and approximately once every 100 years, >40% of the area was disturbed (Figure-recurrence). The mean return interval between epidemics (of any mortality level) at a single location (25-m cell) varied from 36 years to >800 years, with a mean and median of 88 years, and negligible area escaped disturbance altogether over the course of an 800-year simulation (Figure-return). Note, the mean return interval was 1,335 years and 91 years for high- and low-mortality disturbances, respectively, due to the preponderance of low-mortality outbreaks. As expected, mean return intervals varied spatially across the forest (Figure-map), but given the scarcity of this cover type it was impossible to discern whether there were any meaningful patterns.


      Douglas-fir Beetle.--The frequency and extent of simulated Douglas-fir beetle epidemics in cool moist mixed-conifer-aspen forest varied markedly among decades (Figure-initiations, Figure-extent). In most decades <1% of the cool moist mixed-conifer-aspen forest area was affected by an outbreak; however, at least once per 100 years an epidemic affecting >1% of the host area would occur (Figure-recurrence), inclusive of both high- and low-mortality affected areas. The mean return interval between epidemics (of any mortality level) at a single location (25-m cell) varied from 133 years to >800 years, with a mean and median of >800 years (Figure-return). Mean return intervals varied spatially across the forest (Figure-map), but given the scarcity of this cover type it was impossible to discern whether there were any meaningful patterns.


      Age Structure.--The age structure and dynamics of cool moist mixed-conifer-aspen forest reflected the interplay between disturbance and succession processes. The survivorship distribution represents the percent of stands that survived to any age, where age represents time since stand origin, not necessarily the age of the oldest trees in the stand. On average (over time), roughly 50% of the cool moist mixed-conifer-aspen forest was >120 years since stand origin, although at any point in time this varied from 18% to 79% (Figure-survivorship). On average, 25% of the cool moist mixed-conifer-aspen forest survived to >230 years, 5% survived to >440, and <1% survived a stand-replacing disturbance for >620 years.


      Seral-stage Distribution.--The distribution of area among stand conditions within cool moist mixed-conifer-aspen forest fluctuated over time, as expected (Figure-conditions). For example, the percentage of cool moist mixed-conifer-aspen forest in late-seral stages (i.e., understory reinitiation and shifting mosaic conditions) varied from 28% to 89%, reflecting the extremely dynamic nature of this high-elevation forest cover type when considered over century-long periods. The seral-stage distribution appeared to be in dynamic equilibrium (i.e., the percentage in each stand condition varied about a stable mean) and appeared to reach equilibrium within roughly 200 years, although it was unclear whether the shifting mosaic stage ever reached a true equilibrium. The spatial configuration of stand conditions fluctuated markedly over time as well, although there was considerable variation in the magnitude of variability among configuration metrics (Table-hrv). Patch area, core area and the proximity index (a measure of patch isolation) exhibited the greatest variability over time, and the stand initiation and shifting mosaic stages were particularly dynamic relative to the other stages.


      HRV Departure.--Our estimate of the current seral-stage distribution was rarely observed under the simulated HRV (Figure-hrv). Specifically, the current landscape contains a paucity of stands in the early- (i.e., stand initiation stage) and late-seral stages (i.e., understory reinitiation and shifting mosaic conditions) and an overabundance of stands in the stem exclusion stage compared to the simulated HRV. There was a notable difference in seral-stage departure between cool moist mixed-conifer forest with and without aspen. Notably, the current landscape contains much less of the mixed-aspen type in the stem exclusion stage than the pure conifer type (54% versus 79%) and, consequently, much more in the late-seral stages (45% versus 18%). This is likely due to differences between cover types in the rate of succession from the stem exclusion stage to the understory reinitiation stage. In the mixed-conifer-aspen forest, the stem exclusion stage is dominated by aspen, which transitions to the understory reinitiation stage between 80-120 years after stand origin - owing to the relatively short-lived aspen and early break-up of the aspen-dominated canopy. In contrast, the conifer canopy in the pure conifer forest takes 150-250 years to break up in the absence of other major disturbance processes. Hence, given the paucity of large wildfires over the past century, most mixed-conifer-aspen stands have already transitioned to the understory reinitiation stage, whereas a larger proportion of the pure conifer stands have not yet transitioned. Overall, based on the four separate stand conditions (i.e., without aggregating the late-seral stages), the seral-stage departure index was 70% (Table-hrv) - considerably less than observed for cool moist mixed-conifer forest. When the late-seral stages were aggregated, the seral-stage departure index increased slightly to 77%. The current seral-stage configuration deviated slightly less (63%) from the simulated HRV (and considerably less than observed for cool moist mixed-conifer forest), although there was considerable variability (25-99%) among metrics. Interestingly, despite the preponderance of stands in the stem exclusion stage, the spatial configuration of these stands did not deviate greatly from the simulated configuration. Not surprisingly, stands in the stand initiation stage were fewer in number, smaller, geometrically less complex and more isolated than existed under the simulated HRV. Unfortunately, the nature of the configuration departure in late-seral stages varied depending on whether the late-seral stages were treated separately or combined into a single class.


Pure Aspen Forest [cover type description]


      Pure aspen forest is relatively common on the UPL, encompassing 31,465 ha and comprising almost 5% of the landscape (Table-areal coverage).


      Sources of Uncertainty.---At the time of our simulation experiments, there was very little local empirical data available on disturbance regimes and succession processes in pure aspen forest from which to parameterize the model. However, data from the adjoining San Juan National Forest indicated a 150-yr mean fire return interval (rotation period) for high-mortality fires during the reference period (Romme et al. 2003). The results reported below are generally consistent with this interpretation. We are not aware of any major source of uncertainty regarding our treatment of this cover type. Consequently, and because it is well represented on the UPL, overall we have high confidence in the reliability of the results for this cover type.


      Wildfire.--The frequency and extent of simulated wildfires in aspen forest varied markedly among decades (Figure-initiations, Figure-extent). In almost half of the decades, <5% of the aspen forest burned, inclusive of both high- and low-mortality affected areas. However, on average, roughly 2-3 times per 100 years, >10% of the area burned, and roughly once every 200 years, >30% of the pure aspen forest burned (Figure-recurrence). Under this wildfire regime, the mean return interval between fires (of any mortality level) to a single 25-m cell varied widely from 28 years to >800 years, with a mean and median of roughly 100 years, and negligible area escaped disturbance altogether over the course of an 800-year simulation (Figure-return). Note, the mean return interval was 144 years and 342 years for high- and low-mortality disturbances, respectively, due to the preponderance of high-mortality wildfires. As expected, mean return intervals varied spatially across the forest (Figure-map). In general, return intervals increased with elevation, reflecting the moister, cooler conditions at higher elevations. In addition, pure aspen stands embedded in a neighborhood containing cover types with shorter return intervals (e.g., mountain shrubland) exhibited shorter return intervals, reflecting the importance of landscape context on fire regimes.


      Age Structure.--The age structure and dynamics of pure aspen forest reflected the interplay between disturbance and succession processes. The survivorship distribution represents the percent of stands that survived to any age, where age represents time since stand origin, not necessarily the age of the oldest trees in the stand. On average (over time), roughly 50% of the pure aspen forest was >90 years since stand origin, although at any point in time this varied from 20% to 79% (Figure-survivorship). On average, 25% of the pure aspen forest survived to >170 years, 5% survived to >340, and <1% survived a stand-replacing disturbance for >500 years.


      Seral-stage Distribution.--The distribution of area among stand conditions within pure aspen forest fluctuated markedly over time, as expected (Figure-conditions). For example, the percentage of pure aspen forest in the late-seral stages (i.e., understory reinitiation and shifting mosaic conditions) varied from 33% to 83%, reflecting the extremely dynamic nature of this cover type when considered over century-long periods. The seral-stage distribution appeared to be in dynamic equilibrium (i.e., the percentage in each stand condition varied about a stable mean), and appeared to reach equilibrium with roughly 100 years. The spatial configuration of stand conditions fluctuated markedly over time as well, although there was considerable variation in the magnitude of variability among configuration metrics (Table-hrv). Patch area, core area and the proximity index (a measure of patch isolation) exhibited the greatest variability over time.


      HRV Departure.--Our estimate of the current seral-stage distribution was rarely observed under the simulated HRV (Figure-hrv). Specifically, the current landscape contains ~2% of this cover type in the stand initiation condition, whereas it varied from 1 to 55% under the simulated HRV. Similarly, the current landscape contains 34% of this cover type in the late-seral stages (i.e., understory reinitiation and shifting mosaic conditions), whereas it varied from 33 to 83% under the simulated HRV. The current landscape contains a pulse (64%) of pure aspen forest in the stem exclusion stage, whereas the landscape never exceeded 52% in this stage under the simulated HRV. Overall, based on the four separate stand conditions (i.e., without aggregating the late-seral stages), the seral-stage departure index was 75% (Table-hrv). When the late-seral stages were aggregated, the seral-stage departure index increased to approximately 100%. The current seral-stage configuration deviated much less (55%) from the simulated HRV, although there was considerable variability (18-83%) among metrics. In general, the current landscape contains a greater number of larger, more extensive, geometrically more complex and less isolated stands in the stem exclusion condition and, conversely, fewer, smaller and geometrically less complex and less clumped (more isolated) stands in the stand initiation condition than existed under the simulated HRV. The interpretation of the late-seral stages was confounded by the difficulties in distinguishing between stands in the understory reinitiation and shifting mosaic conditions in the current landscape.


Spruce-Fir Forest [cover type description]


      Spruce-fir forest is relatively uncommon on the UPL, encompassing 7,787 ha and comprising only 1.2% of the landscape (Table-areal coverage).


      Sources of Uncertainty.---At the time of our simulation experiments, there was very little local empirical data available on disturbance regimes and succession processes in spruce-fir forest (with and without aspen) from which to parameterize the model. However, data from the adjoining San Juan National Forest indicated a 300-350 yr mean fire return interval (rotation period) for high-mortality fires during the reference period (Romme et al. 2003), although there was no distinction between spruce-fir forest with and without aspen. The results reported below deviate from the expected rotation period for reasons mainly having to do with differences in landscape context (discussed below). Taking these differences into account, we believe that our results are generally consistent with the empirical data. We are not aware of any other major source of uncertainty regarding our treatment of this cover type. Consequently, we have moderately high confidence in the reliability of the results for this cover type.


      Wildfire.--The frequency and extent of simulated wildfires in spruce-fir forest varied markedly among decades (Figure-initiations, Figure-extent). In almost half of the decades, less than 2-3% of the spruce-fir forest burned, inclusive of both high- and low-mortality affected areas. However, on average, almost once per 100 years, >20% of the area burned, and roughly once every 300-400 years, >40% of the area burned (Figure-recurrence). Under this wildfire regime, the mean return interval between fires (of any mortality level) to a single 25-m cell varied widely from 26 years to >800 years, with a mean and median of 136 and 160 years, respectively, and negligible area escaped disturbance altogether over the course of an 800-year simulation (Figure-return). Note, the mean return interval was 182 years and 647 years for high- and low-mortality disturbances, respectively, due to the preponderance of high-mortality wildfires. Although the mean return interval for high-mortality fires was shorter than might be expected, a large proportion of the cover type exhibited much longer return intervals in the range 200-400 years. In addition, the same parameterization resulted in a 329-year mean return interval for high-mortality fires on the adjoining SJNF, where the spruce-fir forest is much more extensive and contiguous. The mean return interval on the UPL was shorter than expected, on average, due to the relatively high degree of interspersion and juxtaposition of spruce-fir stands with more flammable cover types such as aspen and mountain shrubland. As expected, mean return intervals exhibited a high degree of spatial variation across the forest owing to differences in landscape context (Figure-map). In general, return intervals increased with elevation, reflecting the moister, cooler conditions at higher elevations, and spruce-fir stands embedded in a neighborhood containing cover types with shorter return intervals (e.g., aspen, mountain shrubland) exhibited shorter return intervals, reflecting the importance of landscape context on fire regimes.


      Spruce Beetle.--The frequency and extent of simulated spruce beetle epidemics in spruce-fir forest varied markedly among decades (Figure-initiations, Figure-extent). In most decades, <1% of the spruce-fir forest was disturbed, inclusive of both high- and low-mortality affected areas. However, on average, roughly once every 200-300 years a major epidemic affecting >40% of the area occurred (Figure-recurrence). The mean return interval between epidemics (of any mortality level) at a single location (25-m cell) varied widely from 62 years to >800 years, with a mean and median of 250 and 267 years, respectively, and roughly 2-3% of the area escaped disturbance altogether over the course of an 800-year simulation (Figure-return). Note, the mean return interval was 269 years and 2,413 years for high- and low-mortality disturbances, respectively, due to the preponderance of high-mortality outbreaks. Recall that high-mortality spruce beetle outbreaks do not result in stand initiation in this cover type unless they are concomitant with high-mortality spruce budworm outbreaks. Not surprisingly, mean return intervals varied spatially across the forest, but at a fine grain and in a seemingly random pattern (Figure-map).


      Spruce Budworm.--The frequency and extent of simulated spruce budworm epidemics in spruce-fir forest varied markedly among decades (Figure-initiations, Figure-extent). In almost two-thirds of the decades, spruce budworm populations were at endemic levels and none of the spruce-fir forest was disturbed. However, roughly 3 times per 100 years, epidemics occurred affecting >10% of the area, and approximately once every 300-400 years, >50% of the area was disturbed (Figure-recurrence). The mean return interval between epidemics (of any mortality level) at a single location (25-m cell) varied from 40 years to >800 years, with a mean and median of roughly 116 years, and negligible area escaped disturbance altogether over the course of an 800-year simulation (Figure-return). Note, the mean return interval was 3,635 years and 118 years for high- and low-mortality disturbances, respectively, due to the preponderance of low-mortality outbreaks. As expected, mean return intervals varied spatially across the forest, but at a very fine grain and in a seemingly random pattern (Figure-map).


      Age Structure.--The age structure and dynamics of spruce-fir forest reflected the interplay between disturbance and succession processes. The survivorship distribution represents the percent of stands that survived to any age, where age represents time since stand origin, not necessarily the age of the oldest trees in the stand. On average (over time), roughly 50% of the spruce-fir forest was >120 years since stand origin, although at any point in time this varied from 7% to 87% (Figure-survivorship). On average, 25% of the spruce-fir forest survived to >220 years, 5% survived to >430 years, and <1% survived a stand-replacing disturbance for >600 years.


      Seral-stage Distribution.--The distribution of area among stand conditions within spruce-fir forest fluctuated markedly over time, as expected (Figure-conditions). For example, the percentage of spruce-fir forest in late-seral stages (i.e., understory reinitiation and shifting mosaic conditions) varied from 9% to 80%, reflecting the extremely dynamic nature of this high-elevation forest cover type when considered over century-long periods. Despite the large fluctuations, the seral-stage distribution appeared to be in dynamic equilibrium (i.e., the percentage in each stand condition varied about a stable mean), and appeared to reach equilibrium with roughly 100-200 years. The spatial configuration of stand conditions fluctuated markedly over time as well, although there was considerable variation in the magnitude of variability among configuration metrics (Table-hrv). Patch area, core area and the proximity index (a measure of patch isolation) exhibited the greatest variability over time.


      HRV Departure.--Surprisingly, our estimate of the current seral-stage distribution was close to being within the 25-75th percentile range of variation under the simulated HRV (Figure-hrv). The current landscape contains 36% of this cover type in the stand initiation condition, which represents the 80th percentile of the HRV distribution; otherwise, the stem exclusion and combined late-seral stages (i.e, understory reinitiation and shifting mosaic conditions) are within their respective 25-75th percentile ranges of variation under the simulated HRV. Overall, the degree of HRV departure depends on whether the late-seral stages area aggregated. Based on the four separate stand conditions (i.e., without aggregating the late-seral stages), the seral-stage departure index was 20% (Table-hrv). This was due principally to the paucity of stands in the shifting mosaic stage in the current landscape compared to the simulated HRV. However, given the unreliability of the current database in distinguishing between stands in the late-seral stages, these results must be viewed with caution. When the late-seral stages were aggregated, the seral-stage departure index declined dramatically to 7%. The current seral-stage configuration deviated only modestly (38%) from the simulated HRV, although there was considerable variation (13-67% deviation) among metrics. In general, the current landscape contains fewer, but larger and more extensive patches of late-seral forest forest than existed under the simulated HRV. Further interpretation of configuration departure was confounded by the difficulties in distinguishing between stands in the understory reinitiation and shifting mosaic conditions in the current landscape.


Spruce-Fir with Aspen Forest [cover type description]


      Spruce-fir-aspen forest is three times more common than spruce-fir forest on the UPL, encompassing 22,984 ha and comprising 3.5% of the landscape (Table-areal coverage). In general, the vegetation dynamics in spruce-fir-aspen forest were very similar to those of spruce-fir forest. Nevertheless, we provide a complete description of the results for this cover type below and highlight the notable differences.


      Sources of Uncertainty.---At the time of our simulation experiments, there was very little local empirical data available on disturbance regimes and succession processes in spruce-fir forest (with and without aspen) from which to parameterize the model. However, data from the adjoining San Juan National Forest indicated a 300-350 yr mean fire return interval (rotation period) for high-mortality fires during the reference period (Romme et al. 2003), although there was no distinction between spruce-fir forest with and without aspen. In general, we treated this cover type similar to spruce-fir forest without aspen. We reasoned that the less flammable aspen component of these stands would function to retard fire and therefore lengthen the fire rotation period, but that this would be compensated for by the lower elevation of these stands and their juxtaposition to cover types that burn more frequently, yet we have no empirical basis from which to confirm or refute this hypothesis. As with spruce-fir forest, the results reported below deviate from the expected rotation period for reasons mainly having to do with differences in landscape context (discussed below). Taking these differences into account, we believe that our results are generally consistent with the empirical data. We are not aware of any other major source of uncertainty regarding our treatment of this cover type. Consequently, we have moderately high confidence in the reliability of the results for this cover type.


      Wildfire.--The frequency and extent of simulated wildfires in spruce-fir-aspen forest varied markedly among decades (Figure-initiations, Figure-extent). In almost half of the decades, less than 2-3% of the spruce-fir-aspen forest burned, inclusive of both high- and low-mortality affected areas. However, on average, almost once per 100 years, >20% of the area burned, and roughly once every 200 years, >30% of the area burned (Figure-recurrence). Under this wildfire regime, the mean return interval between fires (of any mortality level) to a single 25-m cell varied widely from 24 years to >800 years, with a mean and median of 137 years and 160 years, respectively, and negligible area escaped disturbance altogether over the course of an 800-year simulation (Figure-return). Note, the mean return interval was 181 years and 691 years for high- and low-mortality disturbances, respectively, due to the preponderance of high-mortality wildfires. Although the mean return interval for high-mortality fires was shorter than might be expected, a large proportion of the cover type exhibited much longer return intervals in the range 200-400 years. In addition, the same parameterization resulted in a 266-year mean return interval for high-mortality fires on the adjoining SJNF, where the spruce-fir-aspen forest is much more extensive and contiguous. The mean return intervals were shorter than expected, on average, due to the juxtaposition of spruce-fir-aspen stands with more flammable cover types such as aspen and mountain shrubland. As expected, mean return intervals varied spatially across the forest (Figure-map). In general, return intervals increased with elevation, reflecting the moister, cooler conditions at higher elevations. In addition, spruce-fir-aspen stands embedded in a neighborhood containing cover types with shorter return intervals (e.g., aspen, mountain shrubland) exhibited shorter return intervals, reflecting the importance of landscape context on fire regimes.


      Spruce Beetle.--The frequency and extent of simulated spruce beetle epidemics in spruce-fir-aspen forest varied markedly among decades (Figure-initiations, Figure-extent). In most decades, <1% of the spruce-fir-aspen forest was disturbed, inclusive of both high- and low-mortality affected areas. However, on average, roughly once every 200-300 years a major epidemic affecting >40% occurred (Figure-recurrence). The mean return interval between epidemics (of any mortality level) at a single location (25-m cell) varied widely from 62 years to >800 years, with a mean and median of 248 and 267 years, respectively, and very little eligible area (2-3%) escaped disturbance altogether over the course of an 800-year simulation (Figure-return). Note, the mean return interval was 275 years and 2,474 years for high- and low-mortality disturbances, respectively, due to the preponderance of high-mortality outbreaks. Recall that high-mortality spruce beetle outbreaks do not result in stand initiation in this cover type unless they are concomitant with high-mortality spruce budworm outbreaks. Not surprisingly, mean return intervals varied spatially across the forest and exhibited a somewhat contagious or clumped distribution (Figure-map). In general, the areas of most extensive and connected spruce-fir-aspen forest exhibited the shortest return intervals, but there was considerable variation among runs owing to the stochastic nature of spruce beetle outbreaks.


      Spruce Budworm.--The frequency and extent of simulated spruce budworm epidemics in spruce-fir-aspen forest varied markedly among decades (Figure-initiations, Figure-extent). In almost two-thirds of the decades, spruce budworm populations were at endemic levels and none of the spruce-fir-aspen forest was disturbed. However, roughly 3 times per 100 years, epidemics occurred affecting >10% of the area, and approximately once every 300-400 years, >50% of the area was disturbed (Figure-recurrence). The mean return interval between epidemics (of any mortality level) at a single location (25-m cell) varied from 40 years to >800 years, with a mean and median of roughly 113 years, and negligible area escaped disturbance altogether over the course of an 800-year simulation (Figure-return). Note, the mean return interval was 3,663 years and 111 years for high- and low-mortality disturbances, respectively, due to the preponderance of low-mortality outbreaks. The mean return interval was very similar to that observed for spruce-fir forest because the stand conditions most susceptible to spruce budworm [i.e., understory reinitiation and shifting mosaic] are very similar for these two cover types. In particular, the aspen component (non-host species) declines to mere remnant status in the late understory reinitiation and shifting mosaic stages in the spruce-fir-aspen type, so that the density of preferred host trees (Abies lasiocarpa) is very similar between these cover types when in these stages. As expected, mean return intervals varied spatially across the forest, but at a very fine grain and in a somewhat random pattern - although return intervals were generally shorter where the concentration of the host species was greatest (Figure-map).


      Age Structure.--The age structure and dynamics of spruce-fir-aspen forest reflected the interplay between disturbance and succession processes. The survivorship distribution represents the percent of stands that survived to any age, where age represents time since stand origin, not necessarily the age of the oldest trees in the stand. On average (over time), roughly 50% of the spruce-fir-aspen forest was >110 years since stand origin, although at any point in time this varied from 12% to 84% (Figure-survivorship). On average, 25% of the spruce-fir forest survived to >220 years, 5% survived to >420 years, and <1% survived a stand-replacing disturbance for >620 years. This highlights the stochastic nature of disturbances, in which some areas by chance alone escaped catastrophic disturbance for very long periods.


      Seral-stage Distribution.--The distribution of area among stand conditions within spruce-fir-aspen forest fluctuated markedly over time, as expected (Figure-conditions). For example, the percentage of spruce-fir-aspen forest in the late-seral condition (i.e., understory reinitiation and shifting mosaic stages) varied from 18% to 87%, reflecting the extremely dynamic nature of this high-elevation forest cover type when considered over century-long periods. Despite the large fluctuations, the seral-stage distribution appeared to be in dynamic equilibrium (i.e., the percentage in each stand condition varied about a stable mean), and appeared to reach equilibrium with roughly 100-200 years. The spatial configuration of stand conditions fluctuated markedly over time as well, although there was considerable variation in the magnitude of variability among configuration metrics (Table-hrv). Patch area, extent (radius of gyration), core area and the proximity index (a measure of patch isolation) exhibited the greatest variability over time.


      HRV Departure.--Surprisingly, our estimate of the current seral-stage distribution was close to being within the 25-75th percentile range of variation under the simulated HRV (Figure-hrv). The only notable departure was in the stand initiation condition. Specifically, the current landscape contains ~1% of this cover type in the stand initiation condition, which represents the 4th percentile of the HRV distribution; otherwise, the stem exclusion and combined late-seral stages (i.e, understory reinitiation and shifting mosaic conditions) are within their respective 25-75th percentile ranges of variation under the simulated HRV. Overall, the degree of HRV departure depends on whether the late-seral stages are aggregated. Based on the four separate stand conditions (i.e., without aggregating the late-seral stages), the seral-stage departure index was 60% (Table-hrv). This was due principally to the paucity of stands in the shifting mosaic stage in the current landscape compared to the simulated HRV. However, given the unreliability of the current database in distinguishing between stands in the late-seral stages, these results must be viewed with caution. When the late-seral stages were aggregated, the seral-stage departure index declined dramatically to 28%. There was a notable difference in seral-stage departure between spruce-fir forest with and without aspen. Notably, the current landscape contains much less of the mixed-aspen type in the stand initiation condition than the pure conifer type (1% versus 36%) and, consequently, much more in the late-seral stages (68% versus 34%). This is likely due to differences between cover types in the rate of succession from the early- to late-seral stages. In the spruce-fir-aspen forest, the stand initiation and stem exclusion stages are dominated by aspen, which transitions to the understory reinitiation stage between 80-120 years after stand origin - owing to the relatively short-lived aspen and early break-up of the aspen-dominated canopy. In contrast, the conifer canopy in the pure conifer forest takes 150-250 years to break up in the absence of other major disturbance processes. Hence, given the paucity of large wildfires over the past century, most spruce-fir-aspen stands have already transitioned to the understory reinitiation stage, whereas a larger proportion of the pure conifer stands have not yet transitioned.


      The current seral-stage configuration deviated moderately (50%) from the simulated HRV, although there was considerable variation (24-92% deviation) among metrics and, in general, the departure was greater than observed for spruce-fir forest (Table-hrv). In general, the current landscape contains fewer, smaller, less extensive, geometrically less complex and more isolated stands in the stand initiation condition, and a greater number of stands in the late-seral stages than existed under the simulated HRV. Further interpretation of configuration departure was confounded by the difficulties in distinguishing between stands in the understory reinitiation and shifting mosaic conditions in the current landscape.


Mesic Sagebrush [cover type description]


      Mesic sagebrush is relatively common on the UPL, encompassing 44,712 ha and comprising 6.8% of the landscape (Table-areal coverage).


      Sources of Uncertainty.--Unfortunately, the paucity of empirical data on disturbance regimes and succession processes in this cover type forced us to rely almost exclusively on expert opinion to parameterize the model. Consequently, we did not have objective targets against which to verify model outcomes. In particular, given the broad elevational distribution of this cover type and its interspersion with other cover types, we estimated the mean fire return interval (rotation period) for the reference period to be quite variable and to reflect its landscape context. Where stands occur adjacent to dry conifer forest and mountain shrublands, we reasoned that the mean fire return interval would be relatively short (e.g., 60-100 yrs), and where stands are interspersed with pinyon-juniper woodlands, we reasoned that the mean fire return interval would be considerably longer (e.g., 100-200 yrs). Our simulation results are generally consistent with this interpretation. Given the lack of empirical data and our uncertainty regarding the historical fire regime in this cover type, we have low confidence in the reliability of the results reported for this cover type.


      Wildfire.--The frequency and extent of simulated wildfires in mesic sagebrush varied markedly among decades (Figure-initiations, Figure-extent). In most decades >5% of the mesic sagebrush burned, inclusive of both high- and low-mortality affected areas, and roughly once per 100 years, >20% of the area burned (Figure-recurrence). Under this wildfire regime, the mean return interval between fires (of any mortality level) to a single 25-m cell varied from 24 to 800 years, with a mean and median of roughly 80 years, and negligible area escaped disturbance altogether over the course of an 800-year simulation (Figure-return). Note, the mean return interval was 92 years and 873 years for high- and low-mortality disturbances, respectively, due to the preponderance of high-mortality wildfires. As expected, mean return intervals varied spatially across the forest (Figure-map). In general, return intervals were shorter in areas of concentrated mesic sagebrush and in patches embedded in a neighborhood containing cover types with shorter return intervals (e.g., ponderosa pine and mountain shrubland). Mesic sagebrush patches interspersed with less flammable cover types (e.g., pinyon-juniper woodlands and cool moist mixed-conifer forest) exhibited longer return intervals. Thus, landscape context appeared to have an important influence on wildfire return intervals in mesic sagebrush.


      Age Structure.--The age structure and dynamics of mesic sagebrush reflected the interplay between disturbance and succession processes. The survivorship distribution represents the percent of stands that survived to any age, where age represents time since stand origin, not necessarily the age of the oldest individuals in the stand. On average (over time), roughly 50% of the mesic sagebrush was >55 years since stand origin, although at any point in time this varied from 12% to 77% (Figure-survivorship). On average, 25% of the mesic sagebrush survived to >120 years, 5% survived to >280 years, and <1% survived a stand-replacing disturbance for >450 years. This highlights the stochastic nature of disturbances, in which some areas by chance alone escaped catastrophic disturbance for relatively long periods.


      Seral-stage Distribution.--The distribution of area among stand conditions within mesic sagebrush fluctuated markedly over time, as expected (Figure-conditions). For example, the percentage in the shrubs-herbs condition varied from 19% to 79%, reflecting the extremely dynamic nature of this cover type when considered over century-long periods (Figure-hrv). The seral-stage distribution appeared to be in dynamic equilibrium (i.e., the percentage in each stand condition varied about a stable mean), and appeared to reach equilibrium with 100 years. The spatial configuration of stand conditions fluctuated markedly over time as well, although there was considerable variation in the magnitude of variability among configuration metrics (Table-hrv). Patch area, core area and the proximity index (a measure of patch isolation) exhibited the greatest variability over time and, in general, the herbs-shrubs stage was more dynamic than the shrubs-herbs stage.


      HRV Departure.--Unfortunately, due to the lack of data on current stand age and condition in this cover type, we were unable to evaluate departure of the current landscape from the simulated HRV.


Landscape structure


      Landscape composition, measured as the proportion of the landscape in each of 63 distinct and dynamic patch types (defined by unique combinations of cover type and stand condition), fluctuated markedly over time (Table-hrv). The average coefficient of variation (representing the 90 percentile range of variation about the median) across patch types was 186%, although it ranged broadly from 54% to 655%. There were no obvious discernable factors (e.g., cover type, seral stage, elevation) explaining the variability among patch types; however, the less common patch types (i.e., those comprising a smaller proportion of the landscape on average over time) were more likely to exhibit the greatest variability over time (Figure-cv) and the earliest seral stage of each cover type tended to be the most dynamic. Interestingly, despite the high degree of dynamism exhibited by most patch types, the overall diversity of the mosaic, as represented by the Simpson’s diversity index (which is a function of the number of patch types and the equatability in the distribution of area among patch types), was surprisingly stable with a coefficient of variation of only 2%.


      Landscape configuration, measured by 19 different metrics that characterize the spatial character and arrangement, position, or orientation of patches within the landscape, was much less dynamic than landscape composition (Table-hrv). The average coefficient of variation across metrics was only 33% (range 2-112%). There were a couple of notable patterns of variation. First, metrics associated with patch size (e.g., area, core area) and isolation (e.g., proximity index) were more dynamic than metrics associated with patch geometry (e.g., edge density, shape index, core area index) and patch context (e.g., total edge contrast index, interspersion and juxtaposition index). Thus, the principal dynamics were associated with the grain of the patch mosaic. Second, in general, the area-weighted metrics were more dynamic than their unweighted counterparts. The area-weighted metrics give more weight to the larger patches when calculating the mean and are therefore not influenced greatly by variations affecting small patches. Thus, these metrics are less affected by the fine-grained heterogeneity created by the disturbance processes (e.g., abundant, small scattered patches) and provide a better overall measure of changes in the coarse-grain structure of the landscape. Taken in aggregate, these results suggest that the principal dynamics were associated with changes in the size and continuity of the large patches in the landscape; i.e., that large contiguous patches were periodically broken up (fragmented) by disturbance events, but eventually coalesced to reform a coarse-grain mosaic, only to be broken up again.


      The current landscape structure deviates substantially (74% departure index) from the simulated HRV, and departure is greater for the composition of the landscape (80% departure index) than the spatial configuration of the landscape (68% departure index) (Table-hrv). However, it is important to note that the landscape composition departure estimate is based on only forested cover types for which there is sufficient spatial data on stand age and/or condition from which to generate a reliable HRV departure estimate. The forested cover types represent only 20% of the area on the UPL comprised of dynamic patch types. Consequently, these departure estimates must be viewed with caution when referencing departure of the entire UPL. With respect to the composition of the landscape, more than half of the forested patch types (21/40) are completely outside their HRV (i.e., 100% departure index), while one-eighth of the patch types (5/40) are within their 25-75th percentile range of variation (i.e., 0% departure index). When the late-seral stages (i.e., understory reinitiation and shifting mosaic) are combined into a single, aggregated “late-seral” stage within each forest cover type to reflect the inadequate discrimination between these conditions in the current landscape database, the proportion of patch types completely outside their HRV declines to 42% (13/31) and the landscape composition departure index declines to 74%. The principal cause of the decline is our estimate of departure for the shifting mosaic condition in most of the cover types. Specifically, our estimates of the current percentage of the landscape (or cover type) in the shifting mosaic condition are completely outside their simulated HRV (i.e., 100% departure index); whereas, the combined late-seral stages are often within their respective 25-75th percentile ranges of variation (i.e., 0% departure index).


      With respect to the configuration of the landscape, more than half of the metrics (10/19) are completely outside their HRV (i.e., 100% departure index), while only five metrics are completely within their 25-75th percentile range of variation (i.e., 0% departure index) (Table-hrv). In general, the current landscape contains fewer, larger, more extensive, geometrically more complex and less isolated patches than existed under the simulated HRV. Although the current landscape contains less edge than existed under the simulated HRV, the edge that exists has higher average contrast than observed under the HRV. This latter finding is likely due to the fact that most of the edge in the current landscape is derived from abutting patches of different cover types and often involves high-contrast edges between forest and nonforest cover types (e.g., water, barren); whereas, under the simulated HRV proportionately more of the edge is due to abutting patches of different seral stages of the same or similar cover types - which have lower contrast. Overall, the current landscape appears to be less structurally diverse than existed under the simulated HRV.


Habitats of Special Interest


      In addition to the 23 distinct vegetation types, we also defined several new classes based on aggregations of particular cover types and stand conditions (seral stages) for purposes of FRAGSTATS landscape pattern analysis (see FRAGSTATS - Analyses). These new classes represented ‘habitats of special interest’ to land managers. In the sections below, we briefly describe the spatial and temporal variability of each habitat over time under the HRV scenario. Note, these results were based entirely on the landscape pattern metrics computed using FRAGSTATS.


High Elevation Late-seral Conifer Forest


      High elevation late-seral conifer forest comprised anywhere from roughly 1% to 5% of the landscape over time under the simulated HRV (Table-hrv). The spatial configuration of this class varied markedly over time, although there was considerable variation in the magnitude of variability among metrics. The area-weighted versions of mean patch size, radius of gyration, core area, proximity index and shape index exhibited the greatest variability; all had coefficients of variability >100%. Despite the high variability of these metrics, their absolute values paint a picture of a very coarse-grained mosaic of geometrically complex patches maintained over time. The median area-weighted mean patch size, for example, was more than 3,700 ha, with a median traversability (i.e., radius of gyration) of more than 3 km and a median shape index of more than 14. Interpreted in combination with the unweighted metrics, these results indicate that this habitat was broadly distributed across the high-elevation landscape at all times and contained a mixture of large, contiguous patches resulting from prolonged disturbance-free periods intermixed with numerous small (e.g., 10's of hectares) remnant patches embedded within or surrounded by younger successional stands. It is important to note that even though the habitat mosaic was coarse-grained and dominated by a relatively smaller number of very large patches, the mosaic shifted in space over time such that the location of the large patches varied over the course of the 800-year simulation.


      High elevation late-seral conifer forest comprises a little more than 3% of the current landscape, which is the 48th percentile of the HRV distribution (Table-hrv). This represents a 0% departure index. Accordingly, the current configuration is largely within the 25-75th percentile range of variation (i.e., 0% departure), although there was considerable variation (0-100% departure) among metrics. In general, the unweighted versions of most metrics exhibited 100% departure, whereas the area-weighted versions were typically well within their 25-75th percentile ranges of variation (i.e., 0% departure). This was due to the absence of numerous small remnant patches in the current landscape that made up a large portion of the patches in the simulated landscape. Overall, it appears that the coarse configuration of this habitat in the current landscape is well within the simulated range of variation, but that the fine-scale heterogeneity is either lacking or has not been adequately mapped.


Aspen-dominated Forest


      The seral-stage distribution in aspen-dominated forest fluctuated markedly over time, as expected (Table-hrv). This habitat was a dominant feature of the landscape. On average, 6.2% of the landscape was comprised of aspen-dominated forest, which is roughly 30% more than can be accounted for by pure aspen forest alone. Thus, the early seral stages of the mixed conifer-aspen forest types accounted for about one-third (on average, and considerably more at times) of the aspen present in the landscape at any point in time. The spatial configuration of seral stages fluctuated markedly over time as well, although there was considerable variation in the magnitude of variability among metrics. Patch area, core area and the proximity index (a measure of patch isolation), and especially the area-weighted versions of these metrics, exhibited the greatest variability over time, with coefficients of variability typically >>100%. In general, the early-seral stage was most dynamic, followed by the mid-seral stage and late-seral stage. This was not too surprising since the early- and mid-seral stages were influenced by the promotion of aspen following disturbances in the mixed conifer-aspen stands, yet the late-seral stage was restricted to pure aspen forest only. Accordingly, while the area-weighted mean patch size of late-seral aspen forest varied from roughly 160 to 700 ha over time, the early- and mid-seral stages varied from roughly 10 to 10,000 ha over time. Interpreted collectively, these metrics paint a picture of a highly dynamic distribution of aspen-dominated forest over time, in which late-seral patches were relatively small and isolated (being restricted in space by the distribution of pure aspen forest), while early- and mid-seral stages fluctuated dramatically over time, expanding to encompass large, contiguous and geometrically complex patches for an 80- to 120-year period following large stand-replacing disturbances, and contracting after the conifers regained dominance of the stands.


      Our estimate of the current seral-stage distribution of aspen-dominated forest was almost never observed under the simulated HRV, producing a seral-stage departure index of 99% (Table-hrv). Specifically, the current landscape contains more in the mid-seral stage and less in the early- and late-seral stages compared to the HRV. The current seral-stage configuration deviated much less (66% departure) from the simulated HRV, although there was considerable variability (7-99% departure) among metrics. In general, despite being more abundant than observed under the simulated HRV, the spatial configuration of stands in the mid-seral stage appears to be largely within the simulated HRV. Conversely, owing to their scarcity in the current landscape, both early- and late-seral stands appear to be smaller, geometrically less complex and more isolated from each other than observed under the simulated HRV. Overall, aspen-dominated forest of all seral stages appears to be less interspersed with other patch types in the current landscape than existed under the simulated HRV.


Low Elevation Fire-maintained Open Canopy Forest


      Low elevation fire-maintained open canopy forest comprised anywhere from roughly 2% to 5% of the landscape over time under the simulated HRV and thus was never a dominant feature of the landscape at any time (Table-hrv). The spatial configuration of this class varied markedly over time, although there was considerable variation in the magnitude of variability among metrics. The area-weighted versions of mean patch size, core area and proximity index exhibited the greatest variability; all had coefficients of variability >>100%. Despite the high variability of these metrics, their absolute values paint a picture of a fairly coarse-grained mosaic of geometrically complex patches maintained over time. The median area-weighted mean patch size, for example, was more than 840 ha, with a median traversability (i.e., radius of gyration) of roughly 1.2 km and a median shape index of almost 12. Interpreted in combination with the unweighted metrics, these results indicate that this habitat was broadly distributed across the low elevation ponderosa pine zone at all times and contained a mixture of large, contiguous patches maintained by frequent low-mortality fires intermixed with numerous smaller (e.g., 10's of hectares) patches interspersed with a mixture of young successional stands and late-seral stands in the shifting mosaic condition. The large patches were located wherever there were large contiguous areas of ponderosa pine, e.g., on mesas in relatively unbroken terrain. The smaller patches were more likely to be found where the ponderosa pine intermixed with pinyon-juniper woodlands, mountain shrublands, pure aspen forest, and cool, moist mixed-conifer forest where fires were less frequent (thus allowing the ponderosa pine to succeed to the shifting mosaic condition more often) or more severe (thus causing stand replacement).


      Low elevation fire-maintained open canopy forest is absent from the current landscape. This represents a 100% departure index and is in fact well below the minimum percentage of the landscape (2.3%) observed under the simulated HRV (Table-hrv). The class configuration departure is undefined since the class is absent from the current landscape.


Oak-serviceberry-dominated Shrublands


      The seral-stage distribution in oak-serviceberry-dominated shrublands fluctuated moderately over time, as expected (Table-hrv). This habitat was a dominant feature of the landscape at all times owing to the preponderance of mountain shrublands, which comprise roughly 17% of the landscape. Given the relative abundance of pinyon-juniper-oak-serviceberry woodlands (5.6% of landscape) and the paucity of stand-replacing disturbances in ponderosa pine-oak forest, the majority of oak-serviceberry-dominated shrublands were maintained as persistent mountain shrublands, although periodically following large fires in the pinyon-juniper-oak-serviceberry woodlands the extent would increase by as much as a third. However, the principal dynamic was the shifting distribution between early- and late-seral stages. The spatial configuration of seral stages fluctuated moderately over time as well, although there was considerable variation in the magnitude of variability among metrics. The area-weighted versions of patch area, core area and the proximity index (a measure of patch isolation) exhibited the greatest variability over time, with coefficients of variability >100%, although many of the metrics had remarkably low coefficients of variability. In general, the early-seral stage was slightly more dynamic than the late-seral stage. This was not too surprising since the early-seral stage was influenced by the promotion of oak-serviceberry following disturbances in pinyon-juniper-oak-serviceberry woodland and ponderosa pine-oak forest, but the late-seral stage was restricted to mountain shrubland only. Accordingly, while the area-weighted mean patch size of late-seral shrublands exhibited a 10-fold range of variation over time (580 to 5,770 ha), the early-seral stage exhibited a 25-fold range of variation over time (1,037 to 25,560 ha). Interpreted collectively, these metrics paint a picture of a relatively balanced shifting mosaic (over time) of early- and late-seral shrublands, in which occasionally, following large-scale disturbances, early-seral patches would predominate and form relatively large, contiguous patches.


      Unfortunately, due to insufficient spatial data on current stand ages and conditions in the corresponding cover types, we were unable to evaluate departure of the current landscape from the simulated HRV.


Sagebrush-dominated Shrublands


      Sagebrush-dominated shrubland comprised anywhere from roughly 7.4% to 13.8% of the landscape over time under the simulated HRV and thus was always a prevalent feature of the landscape (Table-hrv). The minimum coverage was governed by the areal extent of mesic sagebrush cover, which comprised 6.8% of the landscape. Variability in extent above this lower threshold was governed by the frequency and extent of high-mortality wildfires in pinyon-juniper-sagebrush woodlands, which comprised an additional 13.6% of landscape. Thus, the maximum possible extent of this habitat was 20.4% of the landscape, but this could only occur if 100% of the pinyon-juniper-sagebrush was burned within a 50-year period, an unlikely event given the broad spatial distribution of this cover type. The spatial configuration of this class varied markedly over time, although there was considerable variation in the magnitude of variability among metrics. The area-weighted versions of patch area, core area and the proximity index (a measure of patch isolation) and correlation length (radius of gyration) exhibited the greatest variability over time, with coefficients of variability >>100%. The absolute values of these metrics paint a picture of a fairly coarse-grained mosaic of geometrically complex patches maintained over time. The area-weighted mean patch size, for example, ranged from roughly 476 to 7,565 ha, with a median traversability (i.e., radius of gyration) of more than 1 km and a median shape index of more than 9. Overall, the configuration metrics indicate that this habitat was broadly distributed across the landscape at all times - anchored by the distribution of persistent mesic sagebrush cover, and consisted of mostly moderately sized but scattered (and therefore somewhat isolated) patches of persistent mesic sagebrush combined with periodic pulses of more extensive sagebrush following stand-replacing disturbances in pinyon-juniper-sagebrush woodlands.


      Unfortunately, due to insufficient spatial data on current stand ages and conditions in the corresponding cover types, we weree unable to evaluate departure of the current landscape from the simulated HRV.


Wildlife Indicator Species


      We selected four wildlife indicator species with uniquely different habitat associations and life history attributes to predict how a wide range of wildlife species might respond to changes in habitat conditions resulting from disturbance and succession processes over time (see HABIT@ - Species Models). In the sections below, for each indicator species we briefly describe the spatial and temporal variability in habitat capability under the HRV scenario and the degree of departure of the current landscape. Note, these results are not especially revealing as such because they represent a single disturbance scenario. The power of the habitat capability analysis resides in its use in comparing among alternative disturbance scenarios; e.g., comparing habitat capability between the HRV scenario and one or more land management scenarios. Indeed, this is our intended use in the next phase of this project.


Pine marten


      Pine marten are year-round residents that prefer the interior portions of late-successional, high-elevation conifer forests (principally spruce-fir and cool moist mixed-conifer forest) for foraging and denning. We included the pine marten as an example of a habitat specialist associated with late-successional high-elevation conifer forests, and expected habitat capability for this species to be quite sensitive to stand-replacing disturbances, especially those resulting in the fragmentation of contiguous late-successional forest and concomitant loss of interior habitat.


      Not surprisingly, the landscape capability index varied considerably over time at the scale of the Southwest Quadrant (Figure-LC index), fluctuating between 2-24 and exhibiting a 115% coefficient of variation (Table-LC index). Apparently, the highly contagious spatial pattern in vegetation successional stages imposed by the dominant stand-replacing disturbance process in high-elevation conifer forests (i.e., wildfire) coupled with the relatively small extent of high-elevation conifer forests in this sub-landscape (7,387 ha) resulted in substantial fluctuations in the large patches of interior, late-seral forest habitat needed by this species. Thus, at the scale of the Southwest Quadrant, it appears that pine marten habitat goes through “boom” and “bust” cycles in response to major disturbance events.

 

Pine Marten Movie - Click here to download a movie depicting the shifting mosaic in habitat capability for this species on the Columbine District, San Juan National Forest (patterns are similar for the UPL) over an 800-year simulation under the HRV scenario. This is a spatial representation of the habitat capability index that is summarized in the landscape capability index reported above. NOTE, this is a 30 Mb Microsoft Media file (.avi) that requires appropriate movie viewing software (e.g., Quick Time Player).


The pine marten landscape capability response curve is an alternative way of summarizing the fluctuations in the LC index (Figure-LC response) that is particularly useful for comparing among scenarios; it is provided here as a reference for future comparisons with alternative land management scenarios.


      The current landscape contains an amount of pine marten habitat that is low but still within the simulated HRV (Table-LC index). This is not surprising given the species’ preference for interior high-elevation late-seral conifer forests and our findings on HRV departure for this habitat. As already noted, the current landscape contains a coarse-grained mosaic of high-elevation late-seral conifer forest patches that is within the simulated HRV.


Three-toed Woodpecker


      Three-toed woodpeckers are year-round residents that inhabit late-successional, high-elevation conifer forests, similar to pine marten, but reach peak densities for a period of five to seven years following wildfires and severe bark beetle outbreaks in conifer (and to a lesser extent aspen) stands. We included the three-toed woodpecker as an example of a species associated with transient post-disturbance habitats, and expected habitat capability for this species to vary dramatically, spatially and temporally, in response to wildfire and bark beetle disturbances.


      As predicted, the landscape capability index varied considerably over time at the scale of the UPL (Figure-LC index), fluctuating between 114-728 and exhibiting a 106% coefficient of variation (Table-LC index). At the scale of the Southwest Quadrant, the landscape capability index fluctuated between 9-149 and exhibited a 226% coefficient of variation - roughly twice that of the pine marten. The dramatic fluctuations in habitat capability reflected the periodic pulses of high quality habitat following large-scale disturbance events in the mid- and high-elevation conifer forests. In addition, as expected, the peaks in habitat capability were usually short-lived (i.e., one time step) and followed by a rapid return to the pre-disturbance habitat capability level. Given the rapid response in habitat capability to disturbances, this species would be a particular good indicator of alterations to the disturbance regime.

 

Three-toed Woodpecker Movie - Click here to download a movie depicting the shifting mosaic in habitat capability for this species on the Columbine District, San Juan National Forest (patterns are similar for the UPL) over an 800-year simulation under the HRV scenario. This is a spatial representation of the habitat capability index that is summarized in the landscape capability index reported above. NOTE, this is a 32 Mb Microsoft Media file (.avi) that requires appropriate movie viewing software (e.g., Quick Time Player).


The three-toed woodpecker landscape capability response curve is an alternative way of summarizing the fluctuations in the LC index (Figure-LC response) that is particularly useful for comparing among scenarios; it is provided here as a reference for future comparisons with alternative land management scenarios.


      The current landscape contains a relatively low level of three-toed woodpecker habitat compared to the simulated HRV, although the degree of departure varies markedly among geographic extents (Table-LC index). The 42% departure index for the UPL is not too surprising given that this species utilizes late-seral high-elevation conifer forests, similar to the pine marten, which is in moderate supply in the current landscape, but reaches peak densities following wildfires and beetle outbreaks, which have not occurred in this landscape - at least not in significant extent - for some time.


Olive-sided Flycatcher


      Olive-sided flycatchers are neotropical migrants that select high-contrast edges between early- and late-successional, mid- and high-elevation conifer forests. We included the olive-sided flycatcher as an example of an edge specialist, and expected habitat capability for this species to fluctuate widely in response to the changing amount and distribution of high-contrast edges created by stand-replacing disturbances.


      As predicted, the landscape capability index varied somewhat over time at the scale of the Southwest Quadrant (Figure-LC index), fluctuating between 696-1142 and exhibiting a 37% coefficient of variation (Table-LC index). The fluctuations in habitat capability reflected the periodic pulses of high-quality edge habitat following large-scale disturbance events in the mid- and high-elevation conifer forests. In particular, the heterogeneous pattern of tree mortality following both wildfire and insect outbreaks resulted in an abundance of ideal edge habitat for this species. Interestingly, the coefficient of variation in the landscape capability index was considerably lower for this species compared to all others. This was likely due to the relatively consistent supply of high-contrast edges bordering permanent openings such as meadows, barren areas, and lakes and ponds that acted like a buffer against major fluctuations in available habitat. In addition, in contrast to the very transient nature of three-toed woodpecker habitat following disturbances, high-quality olive-sided flycatcher habitat tended to slowly degrade over several decades in response to the gradual succession process operating in the disturbance opening. Given the recent emphasis on fragmentation-sensitive species, which often involves avoiding creating high-contrast edges, this species would be a particularly good indicator of habitat conditions likely to be jeopardized by a narrow focus on fragmentation-sensitive species.

 

Olive-sided Flycatcher Movie - Click here to download a movie depicting the shifting mosaic in habitat capability for this species on the Columbine District over an 800-year simulation under the HRV scenario (patterns are similar for the UPL). This is a spatial representation of the habitat capability index that is summarized in the landscape capability index reported above. NOTE, this is a 32 Mb Microsoft Media file (.avi) that requires appropriate movie viewing software (e.g., Quick Time Player).


The olive-sided flycatcher landscape capability response curve is an alternative way of summarizing the fluctuations in the LC index (Figure-LC response) that is particularly useful for comparing among scenarios; it is provided here as a reference for future comparisons with alternative land management scenarios.


      The current landscape contains less olive-sided flycatcher habitat than was ever observed under the simulated HRV (Table-LC index). This is not surprising given the species’ preference for high-contrast edges and the paucity of recent disturbances. Much of the current high-quality habitat is associated with relatively permanent edges bordering water bodies, meadows and barren openings.


Elk


      Elk are year-round residents that undergo short-distance elevational migrations between summer and winter ranges. They are found in association with high-elevation conifer and aspen forests and subalpine meadows during the summer; preferring the conifer and aspen forests as hiding cover and using the forest openings and meadows for foraging. Stand-replacing disturbances have transient and opposite effects on foraging and hiding cover; initially increasing forage availability but decreasing hiding cover. Moreover, local resource requirements (for forage and cover) can be met in a wide variety of cover types. We included the elk as a generalist species that prefers a well-interspersed mosaic of early- and late-successional communities, and expected habitat capability for this species to be somewhat robust to vegetation changes caused by disturbance and succession processes.


      The landscape capability index varied much more than we expected at the scale of the Southwest Quadrant (Figure-LC index), fluctuating between 6-40 and exhibiting a 119% coefficient of variation (Table-LC index). Note, the relatively low values of the landscape capability index for this species are potentially misleading. Recall that the landscape capability (LC) index places non-overlapping homeranges on the landscape until the landscape is saturated with homeranges and then randomly keeps or rejects each homerange based on the probability given by the homerange capability (HC) or population capability (PC) index value of the focal (center) cell. This results in the number of realized homeranges supported by the landscape (i.e. the LC index). However, this homerange tiling process does not provide an interpretable index value for non-territorial species like the elk. For social species such as the elk, this index grossly underestimates the potential number of individuals supported by the landscape. Instead, we used it here as a relative index of habitat capability. In this case, the index indicates that the shifting mosaic of vegetation successional stages over time provided a relatively stable but highly variable amount of elk habitat distributed across the landscape over time.

 

Elk Movie - Click here to download a movie depicting the shifting mosaic in habitat capability for this species on the Columbine District over an 800-year simulation under the HRV scenario (patterns are similar for the UPL). This is a spatial representation of the habitat capability index that is summarized in the landscape capability index reported above. NOTE, this is a 32 Mb Microsoft Media file (.avi) that requires appropriate movie viewing software (e.g., Quick Time Player).


The elk landscape capability response curve is an alternative way of summarizing the fluctuations in the LC index (Figure-LC response) that is particularly useful for comparing among scenarios; it is provided here as a reference for future comparisons with alternative land management scenarios.


      The current landscape contains a relatively low habitat capability for elk compared to the simulated HRV (Table-LC index). This is not surprising given the species’ preference for an interspersion of cover and forage. Optimal habitat conditions are provided by a heterogeneous mosaic of vegetation succession stages that provide both cover and forage in close proximity. Given the paucity of recent disturbances, the resulting coarse-grained vegetation mosaic offers comparatively few opportunities for these interspersion and juxtaposition requirements to be met.



Effects of Scale and Context


      We quantified the dynamics in landscape structure and wildlife habitat capability in relation to landscape extent (scale) and context (i.e., geographic location) by examining the relative variability and the degree of similarity in the observed range of values in landscape metrics among several sub-landscapes (Figure-map).


      As expected, the range of variability in landscape structure and wildlife habitat capability for selected indicator species increased with decreasing landscape extent (Figure-scale-land, Figure-scale-wildlife); smaller landscapes exhibited greater variability, or dynamism, over time. Not surprisingly, the variability in landscape composition among individual landscapes was greatest for the smallest extent (watershed). In other words, while the overall range of variability in landscape composition was consistently greater at the smallest extent, there was substantial variation among watersheds in the magnitude of their variability, indicating that landscapes become increasingly unique or idiosyncratic at smaller extents (see below). Interestingly, this was only true for the landscape composition metrics where, for example, the Mesa Creek watershed exhibited three-fold greater variability than the Horsefly Creek watershed (Table-scale-land). For unknown reasons this pattern of increasing idiosyncratic behavior at smaller extents was not evident for landscape configuration, where the differences among landscapes were greatest at the intermediate (quadrant) extent.


      Despite the limitation of having only three landscape extents (watershed, quadrant, and Forest), there was some indication of a nonlinear relationship between extent and variability (or dynamism), at least with respect to the dynamism in landscape composition (Figure-scale-land). Specifically, the magnitude of variability in landscape composition increased only modestly as the landscape extent decreased from the forest scale (659,246 ha) to the quadrant scale (average = 164,812 ha), but increased dramatically as the landscape extent decreased to the watershed scale (average = 37,929 ha). This nonlinear relationship was even more apparent when the landscape composition was based on the reclassified vegetation map in which cover types and stand conditions were aggregated into a smaller set of patch types (Figure-scale-land-reclass). A similar nonlinear relationship was evident for three-toed woodpecker habitat capability, the only species for which we were able to complete the habitat capability analysis at all three scales, in which there was a slight increase in variability between the forest and quadrant scales and then an abrupt increase in the average variability between the quadrant and watershed scales; however, the nonlinear relationship was due largely to the extreme dynamism in the Dominguez Creek watershed (Figure-scale-wildlife). Note, despite the apparent nonlinear relationship between extent and variability in landscape composition (and wildlife habitat capability), there was nonetheless still substantial variability at the forest scale. Indeed, the overall coefficient of variation in landscape composition at the forest scale was 186%. Moreover, based on the observed scaling relationship, we can infer (through extrapolation) that even larger extents would still exhibit substantial variability in landscape structure over time.


      Variability in landscape composition was consistently and considerably greater than the variability in landscape configuration at all scales (Table-scale-land). The average coefficient of variability in landscape composition metrics varied from 186% (Forest) to 1,552% (Mesa Creek watershed), whereas the average coefficient of variability in landscape configuration metrics varied from 31% (Southeast Quadrant) to 74% (Southwest Quadrant). In general, landscape composition was several times more variable than landscape configuration, indicating that while the composition of patch types (i.e., unique combinations of cover type and seral stage) across the landscape was highly dynamic, the spatial configuration of the patch mosaic was relatively stable at all scales.


      As expected, each landscape (quadrant or watershed) differed in its absolute range of variability in landscape structure due to its unique landscape context. The average similarity (across metrics) among quadrants was 41% for composition and 30% for configuration (Table-context-district); among watersheds it was 26% for composition and 36% for configuration (Table-context-watershed). The patterns were similar for the reclassified vegetation map in which cover types and stand conditions were aggregated into a smaller set of patch types (Table-context-district-reclass, Table-context-watershed-reclass). In addition, although the magnitude of variability (CV’s) in each metric was generally similar across quadrants and watersheds of similar extent (Table-scale-land), the percent similarity in the absolute range of variability was highly variable among metrics (0-87% average similarity across quadrant comparisons and 0-97% average similarity across watershed comparisons). In other words, while different landscapes of similar extent were equally dynamic in a relative manner across a wide range of metrics, the absolute range of variation varied markedly between landscapes for some metrics but not for others. Thus, some landscape metrics were particularly sensitive to landscape context; although it was not clear why particular metrics were more sensitive than others.


      Lastly, while the relative and absolute range of variability in landscape structure was strongly related to landscape extent and context, the degree of departure of the current landscape from the simulated HRV was not. The overall landscape structure departure index was remarkably consistent across landscapes at all extents, ranging from 56-74%, although the component landscape composition and configuration departure indices exhibited slightly different relationships (Table-scale-hrv; Figure-scale-hrv). Specifically, landscape composition departure exhibited a slight positive relationship with landscape extent, with the greatest departure at the largest (forest) extent; whereas landscape configuration departure exhibited no relationship with landscape extent.



Historic Range of Variability Tables


      This section includes the complete set of HRV tables for the whole forest and each of the quadrants, including, for each extent, the landscape-level HRV table and a class-level HRV table for each cover type or habitat of special interest. Note, many of the tables associated with the whole forest were referenced in the previous sections. However, the tables for each of districts have not been referenced elsewhere. Given the number and variety of HRV tables listed below and the difficulty in understanding the relationship among these tables, it is essential to understand the basic organizational framework within which these tables were constructed.


      Landscape-level tables.–At the landscape level, the entire landscape (either the whole forest or one of the quadrants) is considered as a single patch mosaic comprised of many patch types (unique combinations of cover type and stand condition). The landscape composition metrics refer to the percentage of the landscape comprised of each patch type, while the landscape configuration metrics refer to the spatial character and arrangement, position, or orientation of patches within the landscape. The landscape table reports the range of variation and departure of the current landscape from the HRV for each metric, and provides a summary of the overall composition and configuration departure for the landscape.


      Class-level tables.–At the class level, each cover type is considered separately. The class composition metrics refer to the percentage of the landscape comprised of each stand condition class (or seral stage) associated with the corresponding cover type, while the class configuration metrics refer to the spatial character and arrangement, position, or orientation of these patches within the landscape. Thus, the class composition metrics are simply a subset of the landscape-level composition metrics - those of the corresponding cover type. The class configuration metrics, on the other hand, are uniquely different from the corresponding landscape metrics. The class configuration metrics represent the spatial character and arrangement, position, or orientation of a single patch type - a single condition class or seral stage of the corresponding cover type. In contrast, the landscape configuration metrics consider all patch types simultaneously. The class table reports the range of variation and departure of the current landscape from the HRV for each class metric, and provides a summary of the overall seral-stage and configuration departure for the corresponding cover type. Note, for some cover types the HRV departure statistics are not reported due to the lack of data on current stand age and condition.


      Reclass tables.–At each organizational level (landscape and class) and for each landscape extent (whole forest and each quadrant), we also reclassified cover types and stand conditions by aggregating patch types into a smaller set of classes in order to better represent certain habitats of special interest. Here, the landscape mosaic was based on a redefined and smaller set of patch types of special interest. The composition and configuration metrics at the landscape and class levels were computed from this redefined landscape for each landscape extent. These tables are flagged below as “reclass tables”.


Landscape Tables:

 

          Uncompahgre Plateau Landscape

          Southeast Quadrant

          Southwest Quadrant

          Northeast Quadrant

          Northwest Quadrant

          Uncompahgre Plateau Landscape (reclass table)

          Southeast Quadrant (reclass table)

          Southwest Quadrant (reclass table)

          Northeast Quadrant (reclass table)

          Northwest Quadrant (reclass table)                

Class Tables:


      Uncompahgre Plateau Landscape

 

          Semi-desert Grassland

          Semi-desert Savannah

          Pinyon-Juniper Woodland

          Pinyon-Juniper-Sagebrush

          Pinyon-Juniper-Oak-Serviceberry

          Mountian Shrubland

          Ponderosa Pine-Oak Forest

          Ponderosa Pine-Oak-Aspen Forest

          Warm-Dry Mixed-Conifer Forest

          Warm-Dry Mixed-Conifer with Aspen Forest

          Cool-Moist Mixed-Conifer Forest

          Cool-Moist Mixed-Conifer with Aspen Forest

          Pure Aspen Forest

          Spruce-Fir Forest

          Spruce-Fir with Aspen Forest

          Mesic Sagebrush

          High-elevation Conifer Forest (reclass table)

          Low-elevation Conifer Forest (reclass table)

          Pinyon-Juniper Woodlands Combined (reclass table)

          Aspen-dominated Stands (reclass table)

          Oak-dominated Stands (reclass table)

          Sagebrush-dominated Stands (reclass table)


      Southeast Quadrant

 

          Semi-desert Grassland

          Semi-desert Savannah

          Pinyon-Juniper Woodland

          Pinyon-Juniper-Sagebrush

          Pinyon-Juniper-Oak-Serviceberry

          Mountian Shrubland

          Ponderosa Pine-Oak Forest

          Ponderosa Pine-Oak-Aspen Forest

          Warm-Dry Mixed-Conifer Forest

          Warm-Dry Mixed-Conifer with Aspen Forest

          Cool-Moist Mixed-Conifer Forest

          Cool-Moist Mixed-Conifer with Aspen Forest

          Pure Aspen Forest

          Spruce-Fir Forest

          Spruce-Fir with Aspen Forest

          Mesic Sagebrush

          High-elevation Conifer Forest (reclass table)

          Low-elevation Conifer Forest (reclass table)

          Pinyon-Juniper Woodlands Combined (reclass table)

          Aspen-dominated Stands (reclass table)

          Oak-dominated Stands (reclass table)

          Sagebrush-dominated Stands (reclass table)


      Southwest Quadrant

 

          Semi-desert Grassland

          Semi-desert Savannah (not present)

          Pinyon-Juniper Woodland

          Pinyon-Juniper-Sagebrush

          Pinyon-Juniper-Oak-Serviceberry

          Mountian Shrubland

          Ponderosa Pine-Oak Forest

          Ponderosa Pine-Oak-Aspen Forest

          Warm-Dry Mixed-Conifer Forest

          Warm-Dry Mixed-Conifer with Aspen Forest (not present)

          Cool-Moist Mixed-Conifer Forest (not present)

          Cool-Moist Mixed-Conifer with Aspen Forest (not present)

          Pure Aspen Forest

          Spruce-Fir Forest

          Spruce-Fir with Aspen Forest

          Mesic Sagebrush

          High-elevation Conifer Forest (reclass table)

          Low-elevation Conifer Forest (reclass table)

          Pinyon-Juniper Woodlands Combined (reclass table)

          Aspen-dominated Stands (reclass table)

          Oak-dominated Stands (reclass table)

          Sagebrush-dominated Stands (reclass table)


      Northeast Quadrant

 

          Semi-desert Grassland

          Semi-desert Savannah

          Pinyon-Juniper Woodland

          Pinyon-Juniper-Sagebrush

          Pinyon-Juniper-Oak-Serviceberry

          Mountian Shrubland

          Ponderosa Pine-Oak Forest

          Ponderosa Pine-Oak-Aspen Forest (not present)

          Warm-Dry Mixed-Conifer Forest

          Warm-Dry Mixed-Conifer with Aspen Forest (not present)

          Cool-Moist Mixed-Conifer Forest

          Cool-Moist Mixed-Conifer with Aspen Forest (not present)

          Pure Aspen Forest

          Spruce-Fir Forest

          Spruce-Fir with Aspen Forest

          Mesic Sagebrush

          High-elevation Conifer Forest (reclass table)

          Low-elevation Conifer Forest (reclass table)

          Pinyon-Juniper Woodlands Combined (reclass table)

          Aspen-dominated Stands (reclass table)

          Oak-dominated Stands (reclass table)

          Sagebrush-dominated Stands (reclass table)


      Northwest Quadrant

 

          Semi-desert Grassland

          Semi-desert Savannah

          Pinyon-Juniper Woodland

          Pinyon-Juniper-Sagebrush

          Pinyon-Juniper-Oak-Serviceberry

          Mountian Shrubland

          Ponderosa Pine-Oak Forest

          Ponderosa Pine-Oak-Aspen Forest (not present)

          Warm-Dry Mixed-Conifer Forest

          Warm-Dry Mixed-Conifer with Aspen Forest (not present)

          Cool-Moist Mixed-Conifer Forest (not present)

          Cool-Moist Mixed-Conifer with Aspen Forest

          Pure Aspen Forest

          Spruce-Fir Forest

          Spruce-Fir with Aspen Forest

          Mesic Sagebrush

          High-elevation Conifer Forest (reclass table)

          Low-elevation Conifer Forest (reclass table)

          Pinyon-Juniper Woodlands Combined (reclass table)

          Aspen-dominated Stands (reclass table)

          Oak-dominated Stands (reclass table)

          Sagebrush-dominated Stands (reclass table)