Comparison of WRF/CAMx and MM5/CAMx simulations for an ozone episode in California

Ozone concentrations frequently exceed federal and California State standards in central and northern California. Over the last twenty years, a number of comprehensive field studies were carried out and data analysis and photochemical modeling were conducted to better understand the chemical and physical dynamics of ozone formation in the region. One of the modeling systems that has been used extensively relied upon MM5 to simulate meteorological conditions and CAMx to simulate the photochemical production of ozone. Recently, MM5 has been applied to simulate episodic conditions captured during the 2000 Central California Ozone Study (CCOS) and extensively evaluated with the CCOS meteorological measurements at surface and aloft layers. However, NCAR has decided to discontinue future development of the MM5 in favor of a new model, the Weather Research and Forecast (WRF) model, thereby prompting our interest in the application of WRF in Central California. In this study, both MM5 and WRF models were used to prepare meteorological inputs to CAMx. Simulated meteorological fields from both models as well as the pollutant concentrations simulated using each set of meteorological inputs were compared to each other as well as to CCOS observations.

Application of the MM5/CAMx models to the ozone episode of July 30 - August 2, 2000, successfully replicated observed ozone in the San Francisco Bay area and the Sacramento area. In the San Joaquin Valley, however, ozone concentrations were generally underestimated. Detailed examination of the simulated meteorological fields by MM5 revealed several problems thought to influence simulated ozone:

1. The afternoon temperatures in the Livermore Valley, located in the eastern Bay Area and where the highest ozone in the Bay Area occurred, were several degrees cooler than the observed values. 2. The mountain blocking effect of terrain in the eastern Bay Area was too weak, causing earlier than observed air flow over the mountains. 3. The sea breeze in the San Francisco Bay Area in general was too strong. 4. The planetary boundary layer (PBL) simulated in the San Joaquin Valley was too deep.

These problems further prompted interest in exploring the properties of the new WRF meteorological model. Some of the problems may be attributed to the dynamic framework of MM5, especially the second-order advection scheme, which tends to produce spurious oscillations and requires numerical smoothing. The WRF model has a choice of a third- or fifth-order advection scheme and is able to run without numerical smoothing to give improved results.

We have made several WRF model runs using different PBL schemes for the July-August 2000 CCOS episode. We present comparisons of the WRF and MM5 model results, including detailed model-to-model comparisons and model-to-observation comparisons. Our focus is on features that may affect air quality simulations, such as temperature distribution, mountain blocking, transport properties, wind convergence, turbulence strength, and PBL heights. In general, the WRF model produced wind fields that are similar to those generated using MM5. There are, however, some differences between the two models thought to be important to successful ozone simulations. The WRF model produced a better defined convergence line in the Livermore Valley than the MM5. Our investigations have shown that the location of this convergence zone is critical to successfully predicting the ozone peak at Livermore. However, the predicted surface-level temperatures from the WRF model are several degrees cooler during the day and several degrees warmer at night than the corresponding MM5 predictions and do not agree as well with observed temperatures.

As part of this study, we developed a WRF to CAMx model interface (WRFCAMx), which converts the WRF outputs to CAMx ready meteorological inputs. The resultant fields from WRFCAMx are dynamically consistent, and comparable to those generated using the previously available MM5CAMx. We present results from CAMx runs using both WRF and MM5 fields to illustrate the effect of different meteorological fields on ozone formation. Model verification will be presented for several regions of interest, including the San Francisco Bay area, the Sacramento area, and the north, central and south San Joaquin Valley.