Abstract
Most of the research on sensitivity analysis focus on models having only uncorrelated input parameters. In addition, the very few research that focus on models having correlated input parameters presume several assumptions about the parameters.
This paper presents a sensitivity analysis approach that addresses the imperfections in a model for Land Cover Change (LCC) prediction. The main objective of this paper is to improve the decisions about the forthcoming LCC. The proposed approach addresses two main problems: (1) different types of imperfection related to the LCC prediction model (aleatory imperfection and epistemic imperfection) and (2) correlation between the input parameters.
Results on real world data show that taking into account both types of imperfection in the process of LCC prediction and the correlation between parameters improve the resulting predictions about the land changes. Comparisons with existing methods are also discussed in terms of accuracy and computation time and they show that the proposed approach outperforms existing methods. (C) 2017 Elsevier B.V. All rights reserved.