Risk and Resilience of Distributed Stormwater Infrastructure at Scale: Assessment of 8000 Catchments Across New York State

Omid Emamjomehzadeh; Omar Wani

This study presents a comprehensive, uncertainty-aware framework, along with an open-source code repository, for assessing the hydraulic resilience and failure risk of thousands of large culverts at the state level. The framework combines 1-meter digital elevation data, extracts catchment morphological attributes, and utilizes historical and projected rainfall and land use data to model both spatial and temporal variations in culvert performance. Our large-scale simulations determine the percentage of culverts with hydraulic capacities below the 5-year, 25-year, 50-year, and 100-year design discharge levels. Further analysis shows that culverts on main roads, with larger drainage areas, constructed more recently, or exposed to rainfall temporal pattern II, exhibit greater resilience on average. The study also finds that the exceedance risk increases over time due to changes in land use and precipitation. The magnitude of this risk increase varies significantly depending on the chosen projection scenarios. These insights provide practical guidance for infrastructure planning, enabling road managers to assess and prioritize culvert upgrades under current and future climate conditions. All data, methods, and open-source code developed in this study are publicly available to promote scalability to other regions.

Innovative Curve Number Methodology for Estimating Per- and polyfluoroalkyl substances (PFAS) Mass Discharge in Stormwater

Hanadi Rifai; William Vines; Charles Newell

Per- and polyfluoroalkyl substances (PFAS) contamination in stormwater represents a significant environmental concern due to its potential for contaminating natural and engineered water systems. While mechanistic models for estimating runoff and pollutant mass flux have been developed, there is a lack of simple hydrologic and pollutant mass discharge tools for simulating PFAS loads from small watershed areas for use by non-expert modelers, site managers, and decision makers. This paper develops a novel decision support system (DSS), STORME-PFAS, that relies on updated curve number methodology to estimate PFAS mass flux from single storm events and annually. STORME-PFAS, in contrast with conventional DSS, integrates the databases within the modeling tools in a readily accessible interface built within Excel spreadsheets and simulates three hydrologic variants: single storm mass discharge, annual mass discharge, and discrete sampling mass discharge. The single storm event variant converts measured rainfall to modeled runoff via calculations using modified TR-55 curve number methodology. The annual runoff variant estimates runoff for a 1-year period using historic precipitation data and the Guswa curve number exponential distribution method. The modeled variants provide users with modified curve number tables using a non-standard initial abstraction ratio and guidance with data for modeling runoff from concrete surfaces. Validation against field data from a military site with measured PFAS in stormwater demonstrated correspondence with SWMM modeling for the same site. Sensitivity analyses and Monte Carlo modeling were used to develop ranges for runoff volumes and PFAS mass flux with their probability of occurrence for the modeled site.
 

Event Selection Matters: Impacts of Rainfall-Runoff Data Filtering on Curve Number Calibration and Prediction

John Ramirez-Avila; Sandra Ortega Achury; Jonathan Lasco; Jesus Ortiz; Diego Galindo; Carlos Gonzalez Murillo

The Curve Number (CN) method remains one of the most widely used approaches for estimating direct runoff in hydrologic engineering, flood studies, and watershed management. Despite its widespread application, uncertainty remains regarding how rainfall-runoff event selection influences CN calibration and subsequent prediction performance. While the original development of the method emphasized large storm events, contemporary calibration practices frequently rely on continuous datasets that include storms of varying magnitudes. This difference raises concerns about parameter representativeness and calibration bias. This study evaluated the influence of event selection on CN calibration across 145 watersheds in the southeastern United States using rainfall-runoff observations from the Caravan dataset. Three event pools were investigated: (1) all rainfall-runoff events (Pall), (2) events with precipitation depths ≥25.4 mm, and (3) annual maximum precipitation events (Pmax). Curve Numbers were calibrated using annual and seasonal datasets under λ = 0.05, λ = 0.20, and optimized λ conditions, and performance was evaluated using multiple error and bias metrics. Results showed that calibrations based on annual maximum precipitation events generally produced more representative runoff predictions, whereas continuous event records introduced systematic bias in calibrated CN values. Applying a rainfall-depth threshold improved agreement between continuous and event-based calibrations and enhanced predictive performance. Seasonal analyses further suggested that hydroclimatic conditions influence calibration behavior. These findings highlight event selection as a critical component of CN calibration and provide practical guidance for improving runoff estimation and hydrologic model reliability