Econometric panel approaches to isolating the causal impact of impervious cover on annual peak floods

Complex patterns in flood changes across the United States (US) have been observed but attributing the causes of these changes is difficult. Here, we illustrate the benefits and limitations of an econometric approach that isolates a single driver of change: impervious cover. From Houston, Texas, to Ellicott City, Maryland, increases in impervious cover have been blamed for recent extreme floods across the US. However, the magnitude of the causal relationship between increases in impervious surfaces and changes in flood flows is poorly understood. Causal diagrams and panel data approaches, both common in economics, can exploit data that vary temporally and spatially and allow one to control for confounders that affect both impervious surface and flood flows. Using data from 754 USGS stream gages across the conterminous US over 1974­-2012, we estimate the causal impact of impervious cover on annual maximum streamflows. All gages have minimal upstream regulations (less than six months of dam storage) and at least 1% impervious cover in their basins at some point during the record. The panel model includes terms which account for both time-varying and time-invariant confounders. Increasing impervious cover is only one of many factors that may be causing observed changes in floods. These methods can be more broadly applied to estimate the impact of other drivers of flood change, as well as to answer other causal questions within hydrology.