Atmospheric Environment 34 (2000) 4889}4906 Improving ozone modeling in regions of complex terrain using observational nudging in a prognostic meteorological model Mike Barna, Brian Lamb* Laboratory for Atmospheric Research, Department of Civil and Environmental Engineering, Washington State University, Pullman, WA 99164-2910, USA Received 8 November 1999; accepted 3 April 2000 Abstract Three air quality simulations are presented for an ozone episode that occurred in the Cascadia region of the Paci"c Northwest during 11}14 July 1996. These three scenarios were the result of three di!erent "ne-scale (5 km grid) wind "eld predictions that were input into the air quality model. Wind "eld modeling in the Cascadia region is di$cult due to the complex terrain that exists within the domain. Wind "elds were developed using the prognostic model MM5 and the diagnostic model CALMET. The "rst wind "eld was created by MM5 and did not employ observational nudging in the data assimilation process. Results were often poor, especially in mountainous regions. The second wind "eld com- bined the "rst MM5 solution with CALMET's objective analysis feature to adjust the MM5 predictions toward the observed winds. This was an iterative approach that yielded acceptable results both in terms of the wind "eld prediction and subsequent air quality prediction, but was very time consuming to apply. The "nal wind "eld was an MM5 simulation that incorporated observational nudging, using hourly surface wind measurements. This approach yielded better wind "eld predictions relative to the MM5 simulation that did not employ observational nudging. In addition, ozone predictions were improved when this wind "eld was input into the air quality model, relative to the other two ozone simulations. These results illustrate the importance of applying observational nudging in a prognostic meteoro- logical model when simulating air quality in regions of complex terrain.  2000 Elsevier Science Ltd. All rights reserved. Keywords: Ozone modeling; Observational nudging; Cascadia; Complex terrain; MM5; CALMET; CALGRID 1. Introduction A critical component of any air pollution modeling study is the representation of the meteorology within the model domain, since an accurate air quality simulation requires an accurate portrayal of the three-dimensional wind "elds. Two types of meteorological models com- monly used for computing wind "elds are diagnostic models and prognostic models. A diagnostic model ex- trapolates existing surface and upper air observations to develop a mass consistent #ow "eld within the model * Corresponding author. Tel.: #1-509-335-5702; fax: #1- 509-335-7632. E-mail address: blamb@wsu.edu (B. Lamb). domain. Prognostic models develop a full numerical solution of the mass, energy, and momentum conserva- tion equations that govern the atmospheric #ow. Both diagnostic models and prognostic models have limita- tions. Diagnostic models contain limited physics in their formulation, and critically depend upon the density, fre- quency, and accuracy of the observations used as input. Prognostic models are very sensitive to initial and boundary conditions, and require much greater comput- ing resources than diagnostic models. Diagnostic meteorological models, such as CALMET (Scire et al., 1999) or the Diagnostic Wind Model (Doug- las et al., 1990), are typically used to generate the requi- site wind "elds for an air quality simulation. These wind "elds are then input into Eulerian photochemical models such as CALGRID (Yamartino et al., 1992) or the Urban 1352-2310/00/$ - see front matter  2000 Elsevier Science Ltd. All rights reserved. PII: S 1 3 5 2 - 2 3 1 0 ( 0 0 ) 0 0 2 3 1 - 4