Debasish PaiMazumder, David Henderson, Nicole Molders
University of Alaska
Fairbanks, College of Natural Science and Mathematics, Department of
Atmospheric Sciences, and Geophysical Institute, 903 Koyukuk Drive, P.O.
Box 757320, Fairbanks, AK 99775-7320, USA.
The Weather Research and Forecasting (WRF) model was run as a regional model without data assimilation or
nudging (31 36h-simulations) for July and December 2005 over a limited area domain covering Siberia to examine
weather formation in an air-mass source region. The WRF-results were compared to NCEP1/NCAR-reanalysis,
International Satellite Cloud Climatology Project, Global Precipitation Climatology Centre and Canadian Meteorological
Centre data to assess model performance and identify shortcomings. WRF is capable of predicting air-mass formation.
Simulation errors are within the error range of other models. The timing of best/worst agreement differs among quantities
depending on their sensitivity to systematic (model deficiencies) and/or unsystematic errors (e.g. initial conditions).
Overall, the WRF-results agree better with reanalysis for July than December. WRF-results and reanalysis agree best
under persistent high pressure and worst during frontal passages and transition from one pressure regime to another. In
July, WRF provides smaller diurnal amplitudes of 2m-temperature with up to 5.4 K lower, and 3.5 K higher values at
0000 and 1200 UTC than the reanalysis. In December, WRF overestimates 2m-temperature by 1.4 K. WRF-temperatures
excellently agree with the reanalysis from 700 hPa to 300 hPa. Except during frontal passages, wind-speed shows positive
bias. Typically root-mean-square errors and standard deviation of errors of wind-speed (temperature) increase (decrease)
with height. In December, WRF has difficulty predicting the position and strength of the polar jet. WRF underestimates
cloudiness and snow-depth, but overestimates precipitation. In July, predicted convective precipitation is related strongly
to boundaries between different land-cover. WRF-predicted snow-depth strongly correlates with terrain and misses the
observed fine features.