CEE’s Konstantinos Andreadis and Colleagues Publish Paper on New Flood-prediction Model in Journal of Hydrology

Content

Floods in the U.S. trigger $4.5 billion in damages and 16 deaths annually. To address this critical national challenge, Assistant Professor Konstantinos Andreadis of the Civil and Environmental Engineering (CEE) Department has collaborated with two colleagues on a pioneering paper in the Journal of Hydrology. The research team’s state-of-the-art, data-driven methodology, which is called the “Quantile Regression Forest” model, precisely predicts annual-peak-daily streamflow in natural basins, properly quantifies their flood-risk assessments, and enables appropriate strategies for adapting to flood conditions. See https://www.sciencedirect.com/science/article/abs/pii/S0022169425005712.

Image

According to the research team’s Journal of Hydrology paper, “The model effectively captures hydrological relationships and achieves realistic calibration to observed conditions. This approach provides actionable insights for water-resources management and flood-risk assessment.”

Andreadis published the Journal of Hydrology paper with his two research colleagues, CEE Ph.D. student Kwan-Hyuck Kim, who was the lead author, and Assistant Professor Fiachra O’Loughlin of the School of Civil Engineering at the University College Dublin in Ireland.

As the Andreadis team summarizes the need for its research, “Floods are one of the most destructive natural disasters in the United States, causing more than $201.3 billion in damage since 1980, with 44 major events and 738 deaths….”

According to the team’s Journal of Hydrology paper, “Flood risk is characterized by flood-inundation areas influenced by hydroclimatic extremes such as peak-streamflow events. Predicting peak-streamflow discharge in ungauged basins upstream of dams or reservoirs is critical for forecasting inflows, aiding operational management, and mitigating downstream flood risk.” 

Image

The paper goes on to explain that “A key approach to quantifying flood risk involves the estimation of flood-inundation areas, typically using hydrodynamic models, as they form the basis of risk assessments by illustrating the spatial and temporal extent of flooding under different scenarios.”

The paper says that inputs to these models include data such as topography, river geometry, and boundary conditions. However, peak streamflow represents the most crucial parameter because “it directly influences flood magnitude, inundation extent, and recurrence intervals.” 

According to the Journal of Hydrology paper, “Despite significant advances in hydrological modeling, predicting peak streamflow remains an important but challenging task for the prediction of flood-inundation area, particularly across diverse hydrological, climatic, and physiographic contexts.”

In that context, Andreadis and his colleagues have tackled the major weaknesses in current hydrodynamic models by developing their Quantile Regression Forest model to predict annual peak-daily streamflow accurately in ungauged basins while also incorporating uncertainty quantification and variable-influence analysis. 

“Results reveal peak precipitation as the dominant driver of flood magnitude (>50% importance) in streamflow prediction, alongside significant contributions from other explanatory variables [such as wetlands, antecedent rain, and shrubs],” as the Journal of Hydrology paper explains. 

Andreadis and his colleagues conclude in their paper that their study addresses the major limitations of most hydrodynamic models by employing “a data-driven, event-based Quantile Regression Forest model to predict peak streamflow across the contiguous United States…The Quantile Regression Forest model leverages the nonlinear capabilities of machine learning to improve predictive accuracy and reliability.” (June 2025)