Development Of Visualization And Analysis Tool For Air Quality Decision Support Based On Response Surface Model

Combined air pollution has become a serious worldwide problem with the rapid pace ofindustrialization and urbanization. The third generation Community Multi-scale Air Qualitymodeling system (Model3-CMAQ) is developed to approach air quality as a whole byincluding state-of-the-science capabilities for modeling multiple air quality issues. However,CMAQ is a time-consuming and Linux-based tool for most of the policy makers, which limitsits application to a small group of professional users and scientists.To address these problems, a CMAQ-based Response Surface Model (RSM) is proposedin this dissertation, to incorporate the statistical relationships between model inputs andoutputs to provide real-time estimate of the air quality changes. The RSM is based on an airquality meta-modeling approach that aggregates numerous pre-specified individual CMAQmodeling simulations into a multi-dimensional air quality "response surface". Once the RSMis simulated, this meta-modeling technique, which is a "model of the model", can be shown toreproduce the results from an individual modeling simulation with little bias or error. Basedon the created RSM method, a visualization and analysis tool (RSM-VAT) is developed for airquality control decision support. First of all, the RSM is validated by two different cases, aPM2.5case of America, and an ozone case of China. The ability of RSM to reproduce CMAQin various scenarios is discussed by comparing the performance of six statistics, whichprovides a quantitative measurement of the agreement between RSM prediction and CMAQ"true" values. The bias and error between RSM prediction values and CMAQ simulationvalues show good agreement in the PM2.5, US case, with MB≤0.196μg m3, NME≤3.890%and R≥0.997. For the China case, results also show that the created RSM can reproduceCMAQ simulation with little bias and error. When the validation statistics are withinacceptable range, this created RSM can be applied to estimate air pollution status over8typical cities of America and4megalopolises in China respectively. The pollutantconcentration varying with different emission control scenarios can be analyzed by thefunctional modules of "Visualization and Analysis" in RSM-VAT. The relationship betweenemission control factors and ambient concentration of both the PM2.5case and ozone case ispresented and analyzed in the form of real-time response concentration map, visualization display graph, and data analysis charts.Results in this dissertation show that based on well-designed experimental matrix, theRSM-VAT is a modeling surrogate tool that can real-time simulate pollutant impacts.RSM-VAT has various user-friendly functions, such as generating estimates of the benefits ofreductions in precursors and providing screening level estimates of the impacts of controlstrategies, etc. It is highly useful in the context of providing illustrative control scenarios forselected areas, and understanding the contribution of different source categories, sourceregions and pollutant emissions to air quality. The RSM-VAT is intended to provide areal-time decision support tool for regulation, air pollution control and management.