Evaluation of a Meteorological Model For Use in Urban-Scale Monitoring of Carbon Dioxide Emissions in Portland, OR

Abstract:

Observation-modeling systems that monitor and verify greenhouse gas (GHG) inventories require accurate representation of atmospheric circulation and boundary layer dynamics which are often provided by numerical weather prediction models. Careful testing and validation of the model meteorology are required to ensure confidence in estimated emissions. As part of our effort to build a multi-sector carbon dioxide monitoring system (CO2) for Portland, OR, we evaluated the skill of the WRF model in retrodicting meteorological fields over a 2-week period using a number of model physics parameterizations and settings. Model data within the Portland metropolitan area (PMA) were evaluated against surface observations and upper air observations measured by commercial aircraft during takeoffs and landings at the Portland International Airport. We used 3 nested domains (12, 4, and 1.3-km) and an outer domain that covered the eastern Pacific Ocean and western U.S. Our model experiments were designed to test the skill of different boundary layer (BL) physics schemes, an urban canopy model (UCM), different meteorological input fields, and different types of data assimilation including grid analysis (or nudging) and observation nudging. As part of this experiment we tested separately the use of surface and upper air observations for both nudging and creation of objective analysis. Without data assimilation, simulated temperatures within the boundary layer were typically biased cold (2-4C) while wind speeds were overestimated (0.5-1.5 m/s). We found significant differences in model biases between BL physics schemes. Mean differences between observed and modeled hourly BL heights ranged between 3-300 m across model configurations. Most BL scheme correctly simulated the BL diurnal cycle. The use of the UCM did not affect surface temperature or wind speeds significantly but lowered BL heights. Upper air model biases were reduced significantly using both grid and observation nudging separately, but no additional improvement was seen when using both. With the use of nudging, we find the model demonstrates reasonable skill in reproducing the atmospheric circulation within the PMA. The expansive set of WRF configurations tested in this study should help inform similar air quality and GHG modeling studies in other urban areas.