Use of observed hydroclimatic trends to constrain projections of snowmelt season runoff in the Rio Grande headwaters

Abstract

AbstractA method is developed for choosing 21st Century streamflow projections among widely varying results from a large ensemble of climate model‐driven simulations. We quantify observed trends in climate–streamflow relationships in the Rio Grande headwaters, which has experienced warming temperature and declining snowpack since the mid‐20th Century. Prominent trends in the snowmelt runoff season are used to assess corresponding statistics in downscaled global climate model projections. We define “Observationally Consistent (OC)” simulations as those that reproduce historical changes to linear statistics of diminished snowpack–streamflow coupling in the headwaters and an associated increase in the contribution of spring season (post‐peak snowpack) precipitation to streamflow. Only a modest fraction of the ensemble of simulations meets these consistency metrics. The subset of OC simulations projects significant decreases in headwaters flow, whereas the simulations that poorly replicate historical trends exhibit a much wider range of projected changes. These results bolster confidence in model‐based projections of declining runoff in the Rio Grande headwaters in the snowmelt runoff season and offer an example of a methodology for evaluating model‐based projections in basins with similar hydroclimates that have experienced pronounced climate changes in the recent historical record.