An Analysis of Simulated California Climate Using Multiple Dynamical and Statistical Techniques

Four dynamic regional climate models (University of California, Santa Cruz’s RegCM3; the
University of California, San Diego’s RSM; the National Center for Atmospheric Research’s
WRF-RUC; and the Lawrence Berkeley National Laboratory/University of California,
Berkeley’s WRF-CLM3) and one statistical downscaling approach (the University of California,
San Diego’s CANA) were used to downscale 10 years of historical climate in California. To
isolate possible limitations of the downscaling methods, initial and lateral boundary conditions
from the National Centers for Environmental Prediction global reanalysis were used. Results of
this downscaling were compared to observations and to an independent, fine-resolution
reanalysis (the North American Regional Reanalysis). This evaluation is preparation for
simulations of future-climate scenarios, the second phase of this California Energy Commission
climate projections project, which will lead to probabilistic scenarios. Each model has its own
strengths and weaknesses, which are summarized here. In general, the dynamic models
perform as well as other state-of-the-art dynamical regional climate models, and the statistical
model has comparable or superior skill, although for a very limited set of meteorological
variables. As is typical of dynamical climate models, there remain uncertainties in simulating
clouds, precipitation, and snow accumulation and depletion rates. Hence, the weakest aspects
of the dynamical models are parameterized processes, while the weakest aspect of the statistical
downscaling procedure is the limitation in predictive variables. However, the resulting
simulations yield a better understanding of model spread and bias and will be used as part of
the California probabilistic scenarios and impacts.