On the Use of Snow and Climate Information in Statistical Seasonal Streamflow Forecasting

Seasonal streamflow predictions provide a critical management tool for water managers in the American Southwest. These forecasts rely primarily on observations of snowpack and precipitation accumulation, which help to quantify the hydrologic memory that enables relatively accurate streamflow forecasts at seasonal lead times. In recent decades, persistent prediction errors for spring and summer runoff volumes have been observed in a number of watersheds in the American Southwest. While mostly driven by interannual to decadal precipitation variability, these errors also relate to the influence of increasing temperature on snow and streamflow in these basins. Here we show that incorporating seasonal temperature forecasts from operational global climate prediction models into streamflow forecasting models adds prediction skill for watersheds in the headwaters of the Colorado and Rio Grande River basins. Current dynamical seasonal temperature forecasts now show sufficient skill to reduce streamflow forecast errors in snowmelt-driven and temperature-sensitive basins. Such predictions can increase the reliability of streamflow forecasting and water management systems in the face of continuing warming as well as decadal-scale temperature variability, and thus help to mitigate the impacts of climate non-stationarity on streamflow predictability.