Optimal Fingerprinting Approach to Detect Anthropogenic Signal in the Regional Hydrologic Cycle

Abstract:

Anthropogenic signals such as responses to the increasing greenhouse gas concentrations have been robustly detected in temperature data at global and sub-continental regional scales. Observed changes in several large-scale hydro-climatic quantities such as precipitation and snow cover extent are also attributable to human influence but with somewhat lower confidence. However, the hydrologic impacts of climate change are often most acutely felt at regional and local scales. Detection and attribution studies of regional hydrologic quantities are challenging for a number of reasons including sparse observational records, low signal-to-noise ratios, and additional uncertainties associated with hydrologic modeling. Here, we use the VIC hydrologic model to study stream flow changes over the period 1951-2005 in four river basins in British Columbia with a total area of approximately 420,000 km**2. For each basin, we drive VIC with multiple ensembles of statistically downscaled CMIP5 simulations, including 40 simulations with natural external forcing only (NAT), a further 40 simulations with all anthropogenic and natural forcings combined (ALL) and ~5000 years of preindustrial control simulations. Detection and attribution is separately performed to analyze regional maximum and minimum temperature, normalized snow water equivalent, seasonal river flow changes as well as flow timing. Results show that the anthropogenic signals are detected in minimum temperature, normalized snow water equivalent, and summer runoff. Changes in other hydro-climatic variables are found to be consistent with the responses to ALL forcings but inconsistent with NAT.