WRF 2011a Meteorology Results
The 3SAQS used the Weather Research Forecast (WRF Version 3.5.1) Advanced Research WRF (WRF-ARW) to simulate meteorology for the 2011 calendar year with a 36 km continental U.S. (CONUS), 12 km western U.S. (WESTUS), and 4 km three-state domain that covered the states of Colorado, Wyoming, and Utah and neighboring areas. The 3SAQS 2011 WRF 36/12/4 km simulation was run in 5-day segments with a half day initialization. The annual WRF simulation used 16 processors per run segment. Each run segment took on average two wall clock days to complete.
The 2011 annual WRF simulations were evaluated for surface wind, temperature, mixing ratio observations, and monthly accumulated precipitation from the PRISM analysis fields based on observations. In addition to the standard meteorological metrics for air quality, WRF performance includes targeted episodic evaluation for wintertime ozone episodes. There were numerous WRF sensitivity simulations performed to help identify an optimal WRF configuration for the annual 2011 simulation. Sensitivity simulations were tested for wintertime (January) and summertime (July) periods.
The 2011 final WRF configuration model performance for surface winds, temperature and mixing ratios was evaluated across the states of Colorado, Wyoming, and Utah as well as across the 4 km three-state and 12 km WESTUS domains. The WRF model performance statistics were compared against meteorological model performance benchmarks that have been developed by analyzing meteorological model performance of 30 “good” performing prognostic model simulations conducted to support air quality modeling.
Quantitative evaluation of surface winds, temperature, and water mixing ratio showed that WRF’s representation of these fields in the 4 km simulation is generally very good. The monthly bias and error statistics rarely exceeded the model performance benchmarks when the fields were evaluated on a domain-wide and state-by-state basis. This evaluation revealed a number of minor biases in the WRF simulation that are worth noting. WRF appears to have some difficulty simulating the nighttime temperature inversion commonly observed in regions with mountainous terrain.This leads to too warm temperatures at night in Utah during the winter months and a cool bias during nighttime hours in other areas. It was also found that modeled temperatures are too warm immediately following sunrise in many locations. WRF consistently under-predicts wind speed by approximately 0.5 m/s throughout the entire year across much of the modeling domain. We also observed a distinct seasonal pattern in mixing ratio bias. Mixing ratio is generally over-predicted during the cooler months, peaking during the daytime hours. Conversely, mixing ratio is under-predicted across much of the 4 km domain during the warmer months. The 3SAQS evaluation for the 4 km three-state and 12 km WESTUS domains also including precipitation. In general the WRF did a very good job in reproducing the spatial distribution and magnitudes of the PRISM monthly precipitation analysis fields during the winter, early spring and late fall. However, during the summer monsoon conditions, the WRF monthly precipitations exhibited more differences with the PRISM analysis fields with WRF estimating higher and more wide-spread convective precipitation events in the summer than indicated by PRISM analysis fields. This is likely due in part to the difficulty in capturing spotty transitional localized precipitation from convective cells in the precipitation observational network and son that they won’t be fully represented in the PRISM analysis fields.
The WRF 2011 36/12/4 km simulation was also evaluated for winter ozone event periods using routine and special field study data from the Upper Green River Winter Ozone Study (UGRWOS). Although the initial 2011 WRF simulation was not configured to simulate meteorological conditions associated with winter ozone events (e.g., cold pooling), the evaluation for the winter high ozone periods will help setup a more focused winter ozone modeling analysis planned for later stages in the 3SAQS. WRF did reproduce the slow wind speeds, temperature inversions and dry conditions associated with winter ozone events, but exhibited high wind speed and direction error.
The 3SAQS 2011 WRF application exhibited reasonably good model performance that was as good or better than other recent prognostic model applications used in air quality planning and it was therefore reasonable to proceed with its use as input for the 3SAQS regional photochemical grid modeling. Additional WRF sensitivity simulations with alternative physics, input, and data assimilation options should be evaluated before the WRF results are used to simulate winter ozone exceedanceexceedence episodes.