Stationarity assumptions in streamflow sensitivity to precipitation may bias future projections

Streamflow elasticity to precipitation is a metric which is used to estimate how responsive rivers are to changes in precipitation. It is commonly used to anticipate future impacts of climate change on streamflow and is assumed to be constant in time, despite evidence that this relationship varies with climatological and landscape changes. To assess the need for a more flexible definition, we present a large-sample non-stationary regional regression approach to estimate long-term trends and variability in interannual streamflow elasticity to precipitation in the USA. We find that elasticity is highly variable in water-limited catchments year-to-year, indicating high sensitivity to climate variability in arid regions. Statistically significant long-term trends in elasticity exist in some regions, but trend magnitude is generally small. We demonstrate that a single average estimate of elasticity may be a poor indicator of streamflow sensitivity to climate change. Consideration of the variability of response is essential for elasticity to be a useful hydrologic signature.