You are on page 1of 1

H43A-1174

Evaluation of a WRF downscaling simulation over Western Montana


N. Silverman and M. Maneta University of Montana Watershed Hydrology Lab Contact: nicholas.silverman@umontana.edu

Introduction
Dynamic downscalling of global circulation models (GCMS) provides a mean to interpret large scale data at a resolution more relevant to policy and locality. Regional climate models (RCMs), such as the Weather and Research Forecasting (WRF) model, can take advantage of finer resolution topography, water-land boundaries, and land use delineation to provide a better estimate of precipitation and temperature at the regional scale. High resolution WRF simulations for climate studies in the Western U.S. have primarily focused in areas largely influenced by coastal characteristics such as California and the Pacific Northwest. These areas are largely affected by sea surface temperatures, coastal weather patterns, and unique moisture fluxes around land and sea boundaries. Since model performance may vary with geographic location it is important to evaluate the model in a variety of areas and topographic terrain.

WRF

PRISM

WRF

PRISM

Probability Distributions
Monthly Precipitation (mm/month) Average Temperature oC

Purpose
This is the first of three WRF downscaling simulations that we will run for Western Montana and our preliminary evaluation of the output. Our focus for this project is on identifying and understanding any biases that may be present in the temperature and precipitation data. Ultimatetly, we will use this data as forcings for a newly developed ecohydrologic watershed model.
Difference Difference

Variables vs. Elevation


Winter Summer

Winter Precipitation (mm)


Monthly winter precipitation was calculated using hourly WRF output data. Hourly totals were accumulated for each month and then averaged across the winter season. The winter season was defined as December, January, February, and March. PRISM data is calculated by interpolating observational station data at various temporal resolutions and then averaging for each month. The largest difference in precipitation values is seen over areas of large topographic variation. In high elevations WRF predicts significantly greater precipitation than PRISM. In valleys WRF tends to predict lower values than PRISM.

Summer Precipitation (mm)


Monthly summer precipitation was calculated using hourly WRF output data. Hourly totals were accumulated for each month and then averaged across the summer season. The summer season was defined as June, July, August, and September. PRISM data is calculated by interpolating observational station data at various temporal resolutions and then averaging for each month. The largest difference in precipitation values is seen over areas of large topographic variation. Differences are smaller than those calculated in the winter
Monthly Precipitation (mm/month)

Methods
In this study, we present the results of a downscaled reanalysis of climate over Western Montana for the years 2000-2006. We used the WRF model to resolve onedegree Global Forecast System (GFS) data to 4-km grid spacing. To evaluate the model we compared monthly precipitation and temperature for winter and summer months to observational analysis datasets from PRISM (Parameter-elevation Regressions on Independent Slopes Model).
WRF Domain DEM

Missoula

Elevation (m)

Monthly Precipitation (mm/month)

Elevation (m)

Winter

Summer

Temperature

Winter Temperature (oC)


Winter min and max temperatures were calculated by averaging daily values for each month. The winter season is defined as December, January, February, and March. PRISM values are calculated by interpolating observational station data using a range of temporal resolutions. WRF tends to predict lower max. temperatures in comparison to PRISM. Differences in min. temperature seem to be correlated with elevation.

Summer Temperature (oC)


Summer min and max temperatures were calculated by averaging daily values for each month. The summer season is defined as June, July, August, and September. PRISM values are calculated by interpolating observational station data using a range of temporal resolutions. WRF tends to predict higher minimum temperatures and lower maximum temperatures than PRISM.

WRF Domain

Average Temperature (oC)

Average Temperature (oC)

Regional Downscaling Domain

WRF-Min

PRISM-Min

Difference-Min

WRF-Min

PRISM-Min

Difference-Min

Elevation (m)

Elevation (m)

Discussion
As expected, precipitation differences between WRF and PRISM are largely correlated with variable terrain and they appear to be at a maximum in higher elevations. WRF predicts more precipitation on mountain tops and predicts less precipitation in valley bottoms. In flat terrain, WRF and PRISM report very similar values for precipitation. Winter precipitation differences are greater than summer. Temperature differences are more challenging to summarize and do not appear to be consistent between min and max values. This may be attributed to different temporal resolutions and/or land surface moisture fluxes. Temperature averages show less variation between the two models in both summer and winter. Since, this is a preliminary evaluation more analysis is needed to identify trends and sources of these biases. Three sources are possible: (1) Global forecast model, (2) WRF model, or (3) PRISM. Understanding these biases and their sources will be imperative to moving forward with our two other simulations. We hope to use this GFS-WRF simulation to help calibrate a similar PCM-WRF run over the same time-period so that we can more accurately account for a final midcentury PCM-WRF simulation. We will then use this data as forcings for a newly developed ecohydrologic watershed model.
References
Caldwell, P., H. N. S. Chin, et al. (2009). "Evaluation of a WRF dynamical downscaling simulation over California." Climate Change 95(3-4): 499-521. Daly, C. (2002). "Variable influecne of terrain on precipitation patterns: Delineation and use of effective terrain height in PRISM." http://www.ocs.orst.edu/prism/effter.pdf. Pan, L.-L., s.-H. Chen, et al. (2010). "Influences of climate change on California and Nevada regions revealed by a high-resolution dynamical downscaling study." Climate Dynamics. Qian, Y., S. J. Ghan, et al. (2010). "Downscaling hydrolclimatic changes over the Western US based on CAM subgrid scheme and WRF regional climate simulations". International Journal of Climatology 30(5): 675-693.

WRF Settings
Three 2-way nested domains (36km, 12km, 4km grid spacing). Three-minute timestep reporting every hour. Micro-physics scheme = WRF single moment 3-class. Longwave radiation scheme = rapid radiative transfer model (RRTM) Shortwave radiation scheme = Duhdia, downward integration Surface layer scheme = Monin-Obukhov Land surface model = Unified Noah Cumulus scheme = Kain-Fritsch for domains 1 and 2 Landuse and soil data from WPS/Geogrid

WRF-Max

PRISM-Max

Difference-Max

WRF-Max

PRISM-Max

Difference-Max

Acknowledgements
We would like to acknowledge the Montana Water Center, NASA Montana Space Grant Consortium, and the Environmental Protection Agency (EPA) for their contributions and help funding this project.

You might also like