5.3 Evaluation of a WRF dynamical downscaling simulation over California (2008

5.3

Evaluation of a WRF dynamical downscaling simulation over California

Peter Caldwell, LLNL, Livermore, CA; and D. C. Bader, H. N. S. Chin, and G. Bala

Impacts of global warming will largely be felt on scales too small to

be resolved by today's general circulation models (GCMs). One popular

approach for obtaining climate change predictions at these smaller

scales is dynamical downscaling, whereby a regional model is run for a

number of years based on lateral boundary conditions imposed by a

global model. To be more useful in regional climate change studies,

the accuracy of this approach should be better established. With this

aim, a rigorous test of the ability of a new WRF-based dynamical

downscaling simulation to reproduce the current climate of California

is presented in this study. To improve statistical robustness and

regional detail, this GCM-forced simulation is both run for a longer

time (40 yrs) and at a higher resolution (12 km) than previous

simulations over North America

Spatial representation of modeled precipitation and surface

temperature are found to agree much better with observations in the

downscaled run than in the forcing GCM, though the climatological

magnitude of WRF precipitation is substantially overestimated along

windward slopes. This is due to strong overprediction of

precipitation intensity; precipitation frequency is actually

underpredicted by the model. Potential sources for this bias

are investigated. Regional model surface temperatures agree well with

observations in most regions and in most seasons, though the model

inherits a domain-wide warm bias of several degrees from the GCM in

summer and is warmer than observed along the coast in all seasons due

to overpredicted near-land GCM SST. Modeled snowfall/snowmelt

agrees quite well with observations, but snow water equivalent is

found to be much too low due to monthly reinitialization of all

regional model fields from GCM values.