Scalable flood-risk analysis for New York State culvert infrastructure reveals patterns of dependence

Floodwater flowing over roads can disrupt traffic, threaten public safety, and compromise structural stability. Culverts–structures that convey water beneath roads–are critical for flood mitigation, yet statewide rehabilitation would cost billions of dollars. We evaluated reliability by quantifying the probability of hydraulic capacity exceedance of large culverts serving 2300 catchments across New York State. Using terabytes of geospatial data and a scalable, uncertainty-aware modeling framework, we examined spatial and temporal risk patterns at a systems-level. Our results indicate that about 16% (with an uncertainty range of 1–47%) of culverts have hydraulic capacities below the 50-year design discharge, and 2–4% are expected to exceed their capacity annually, assuming independence of exceedance events. The findings show that culverts that are a) located along main roads, b) exposed to inland rather than coastal rainfall temporal patterns, c) situated within small watersheds, and d) constructed more recently tend to be more hydraulically reliable, whereas greater antecedent soil moisture is associated with reduced system reliability. Finally, we find that the probability of capacity exceedance increases over time across all combined land-use and precipitation projection scenarios. The open-source framework developed here enables scaling this analysis to other regions globally. About 16% (with an uncertainty range of 1–47%) of culverts in New York State have capacities below the 50-year design discharge, which could lead to flooding, as revealed by large-scale hydrologic-hydraulic modeling across 2,300 catchments.