Focusing on the time series of regional hydrologic variables, a class of statistical, state-space models (SSM) is evaluated as an efficient alternative to the use of process-based hydrologic models driven by global climate model (GCM) results. The SSM-based approach is also developed to systematically integrate key physical and empirical parameters affecting hydrologic changes, e.g., the sensitivity of global mean surface temperature (GMST) with respect to anthropogenic forcings, the regional temperature and precipitation responses based on the increase of GMST, and the effect of regional temperature and precipitation on hydrologic variables. These SSM parameters are informed by GCM projections and hydrologic projections from the Variable Infiltration Capacity (VIC) model results to 2100. Multidecadal probabilistic projections are provided by fitting the SSM to historical observations (starting from 1900s). This SSM-based method is applied to project annual streamflow of the watersheds in the Colorado River Basin and seasonal streamflow in the California Central Valley watersheds. The method is assessed through cross-validation with observed data and comparative analyses of projected trends and uncertainty with the results from other approaches such as the VIC model driven by downscaled GCM climate projections. Using the SSM-based method to update future projections with additional observed data is also studied to facilitate flexible planning and decision-making. By developing and testing SSMs as a simplified, surrogate approach to assess future hydrologic changes, this work suggests an efficient alternative to facilitate decision-making in regional water resource management.
