Changes in snow depth and soil moisture from regional climate model simulations of future scenario climates

Ten-year simulations by the RegCM2 regional climate model driven by NCEP/NCAR reanalysis (1979-88), Hadley Centre global model contemporary climate results, and Hadley Centre global model future scenario climate results have been analyzed for possible future changes in the hydrological cycle in the US under anthropogenically driven climate change. Simulations of snow depth patterns (snow water equivalent - SWE) for the current climate agree with observations in the high spatial variability and southern snow boundary in the mountainous western US, the sharp gradient at the front range of the Rocky Mountains, the north-south gradient of SWE across the Midwest, the lake-effect snowfall in the lee of the Great Lakes, and the southward protrusion of the snow pattern over the Appalacians. Peak values of observed spatial patterns are not reproduced, but spot checks of values in regions of low SWE gradient (e.g., Minneapolis, Sioux Falls) give values quite close to observed. Simulated changes to SWE due to increase in greenhouse gases expected by about middle of the 21st century are uniformly negative (less snow under future climates) and most notable in mountainous areas where decreases of up to 50% are common in both winter and spring. The scenario climate retains high spatial variability in the mountainous west in winter and spring. The 20 mm seasonal SWE line in winter moves from the Iowa-Missouri border to the Iowa Minnesota border in the scenario climate. Lake-effect snows are substantially reduced in the scenario climate. Spring season in the scenario climate has very little SWE east of the front range of the Rocky Mountains.

SWE errors in the mountainous west tend to be the result of temperature bias (rather than precipitation bias). One consequence of this is that the annual cycle of SWE tends to be distorted relative to observed, so accurate simulation of SWE climatology and its change in future climates is very sensitive to the temperature simulation.

The Illinois Water Survey 8-year (1981-88) dataset of observed soil moisture offers an opportunity to evaluate the model values of soil moisture in the top 10 cm and the root zone. Comparisons of time series shows that the model captures the seasonal cycle quite well for both layers. In the top 10 cm, the magnitudes produced by the model are quite good except that the fall recharge period is slow to develop. For the root layer the model has a 20-45% dry bias. The eight-year climatology for the observations compared with the 10-year climatology of the model confirms the high-quality of the simulation for January-October and deficient recharge in November and December. In the root layer the climatological values confirm generally low values throughout the year, with deficiencies of 10-20% in January-June and 25-30% for July-December. Calculated values of runoff tend to agree reasonably well with observations. Plots of climate change show mostly increases of 5-10% in the root zone with more variability in the top layer, but mostly higher levels of soil moisture under climate change.