Data Driven Assessment of Spillway Performance and Downstream Flood Risk Using Long Term Hydrological and Structural Records

Abstract

Spillways play a critical role in dam safety by regulating extreme flood events and mitigating downstream flood hazards. However, traditional spillway performance assessments often rely on design-based assumptions and stationary hydrological conditions, which may not adequately represent long-term variability observed in historical records. In this study, a data-driven framework is developed to evaluate spillway hydraulic performance and associated downstream flood risk using long-term hydrological and structural datasets. The proposed approach integrates observed inflow hydrographs, reservoir water levels, spillway discharge records, and downstream stage–discharge data collected over multiple decades. First, key performance indicators including spillway discharge efficiency, reservoir surcharge frequency, and exceedance probability of critical downstream water levels are extracted from historical records. Subsequently, non-stationary flood frequency analysis is employed to quantify changes in extreme inflow characteristics and their implications for spillway operation. Reliability-based metrics are then used to assess the probability of spillway capacity exceedance under observed flood conditions. To enhance the assessment, data-driven modeling techniques are applied to identify nonlinear relationships between inflow magnitude, reservoir storage dynamics, and downstream flood response. This allows for a comparative evaluation of historical operational performance versus current design assumptions. Downstream flood risk is quantified by combining hydraulic response indicators with frequency-based exceedance metrics derived from observed data. The results demonstrate that spillway performance and downstream flood risk are strongly influenced by long-term hydrological variability and operational practices. The analysis reveals periods in which observed flood events approached or exceeded spillway design thresholds, highlighting potential underestimation of downstream flood hazards when relying solely on stationary assumptions. The proposed framework provides a practical and transferable methodology for dam safety evaluation using readily available historical datasets. By emphasizing observed system behavior rather than purely theoretical design conditions, this study contributes to improved spillway safety assessment and downstream flood risk management. The findings support the adoption of data-driven approaches as complementary tools for decision-making in dam operation, flood mitigation, and infrastructure resilience planning.