Sensitivity Analysis. Could Better Methods Be Used?

Some recent articles, all in volume 102 of JGR, Atmosphere, use sensitivity analysis (SA) for mechanism identification and/or model selection. All authors implement SA via either an elementary One factor At a Time (OAT) approach or via a derivative-based method. Both approaches have important limitations: * Only a reduced portion of the space of the input factors is explored * The possibility that factors might interact is discounted. * These methods do not allow self-verification. Given that all models involved are highly nonlinear and potentially non-additive, these approaches might fail to provide the full effect of any given factor on the output, thus deceiving the analyst. Different methods are suggested, even to make efficient use of the computations involved. These are a rationalised OAT screening test, a regression-based method, and two implementations of global quantitative sensitivity analysis measures. Costs, advantages and limitations of the proposed strategies are discussed.JRC.(ISIS)-Institute For Systems, Informatics And Safet