Drivers and Buffers of Stream Flashiness in the Mid‐Atlantic United States

Abstract

Flashiness is a hydrologic signature used to describe the responsiveness of streamflow to precipitation. In practise, flashiness has most often been correlated with rainfall intensity and the presence of urban land cover or imperviousness. However, a number of watershed characteristics can influence flashiness, necessitating further investigation of how this particular watershed behaviour and its associated drivers vary at regional scales. We investigated how ten measurable watershed characteristics are empirically related to flashiness of 195 gauged streams in the Mid‐Atlantic region using linear regression analysis. The selected gauges had drainage areas ranging in size from 15 to 250 km2. To explore multiple quantifications of flashiness, we determined both the Richards‐Baker flashiness index, calculated from the continuous hydrograph, and number of peaks‐over‐threshold, for all watersheds for a 10‐year period. As found in other studies, the percentage of watershed development was strongly correlated with stream flashiness, though significant inverse relationships also existed for forest cover, water and wetlands and carbonate geology. Differences in flashiness values between rural and suburban watersheds were not significant, with a significant increase in flashiness (p < 0.01) occurring at 80% development or ~ 25% imperviousness. In addition, we found significant differences in relationships between watershed characteristics and flashiness when differentiated by physiographic regions. For the study area and each of its five physiographic sub‐regions, we identified the four most important predictors through multiple regression. The overall mid‐Atlantic model suggested that development, wetlands and compactness ratio as the most important predictors (adj R2 = 0.57). Significant predictors of stream flashiness varied among regions; however, wetlands served as a significant buffer of flashiness in four of five sub‐regions demonstrating the importance of wetlands in modulating flash‐flooding. In all physiographic regions, empirical models of flashiness values outperformed the Mid‐Atlantic regional model, underscoring the value of local landscape characteristics in augmenting or modulating flashy watershed responses. Our findings suggest the potential for improvements to operational flash‐flood forecasting using hydrologic characteristics of the landscape.