Determining input factor importance in a compartmental model using screening methods.

AbstractModels are used in a wide variety of scientific disciplines to mathematically and simply represent real phenomenon, to understand and explain observed behaviors, and to make predictions of future behaviors. A compartmental model is a type of model that is used to evaluate an experimental investigation concerning transport of material in a system of the real world. Sensitivity analysis (SA) of models is required to investigate the effects of changes in the inputs on the model output(s). When dealing with large, complex, computationally expensive models, screening methods are essential to reduce the dimensionality of the input space for a subsequent SA which can help to determine and quantify the change in model behavior as model factors change. In this study, our focus is on the application of screening designs to a 25-compartment global carbon cycle model, which help us identify the most important input factors that control most of the output uncertainty. We explore similarities and dissimilarities in the results from these screening methods. [ABSTRACT FROM AUTHOR]