Efficient techniques for sensitivity and uncertainty analysis of multiscale air quality models
S. Isukapalli, S.W. Wang, N. Lahoti, P.G. Georgopoulos
Environmental and Occupational Health Sciences Institute, UMDNJ - R.W. Johnson Medical School and Rutgers University, Piscataway, NJ
Ambient air quality modeling, using numerical simulation systems such as the Community Multiscale Air Quality model (CMAQ), involves uncertainties of multiple origins, including uncertainties in the modeling of meteorology, emissions, and chemical interactions. Therefore, uncertainty propagation and identification of contributions of input uncertainties to output uncertainties are important tasks. Traditional techniques such as Monte Carlo and Latin Hypercube Sampling impose significant computational demands, typically requiring several hundreds to thousands of simulations to obtain combined sensitivity and uncertainty information. Recently developed, computationally efficient, alternative techniques include the Stochastic Response Surface Method (SRSM) and the High Dimensional Model Response (HDMR), which have been implemented as modules of the MENTOR system (Modeling ENvironment for TOtal Risk studies). SRSM approximates model outputs as probabilistic response surfaces involving orthogonal polynomials of random variables ("polynomial chaos"), whereas HDMR approximates model responses in terms of orthogonal polynomials in the deterministic space. This presentation focuses on the implementation and evaluation of the SRSM and HDMR to the assessment of the impacts of uncertainties in emissions estimates on different CMAQ model outputs of concern, as well as on other important metrics such as population-weighted mean concentrations, and number of exceedances of ambient concentration standards. The SRSM and HDMR are evaluated from the perspective of (a) estimating output uncertainties, (b) assessing contributions of individual input uncertainties, (c) computational efficiency, and (d) ability to provide an approximate model.
This work is funded in part by a University Partnership Agreement between USEPA and EOHSI. Viewpoints expressed here do not necessarily reflect the views of USEPA or its contractors.