Using Evidence Theory in Land Cover Change Prediction to Model Imperfection Propagation with Correlated Inputs Parameters

Abstract

The identification and the propagation of imperfection are important. In general, imperfection in land cover change (LCC) prediction process can be categorized as both aleatory and epistemic. This imperfection, which can be subdivided into parameter and structural model imperfection, is recognized to have an important impact on results. On the other hand, correlation of input system parameters is often neglected when modeling this system. However, correlation of parameters often blurs the model imperfection and makes it difficult to determine parameter imperfection. Several studies in literature depicts that evidence theory can be applied to model aleatory and epistemic imperfection and to solve multidimensional problems, with consideration of the correlation among parameters. The effective contribution of this paper is to propagate the imperfection associated with both correlated input parameters and LCC prediction model itself using the evidence theory. The proposed approach is divided into two main steps: 1) imperfection identification step is used to identify the types of imperfection (aleatory and/or epistemic), the sources of imperfections, and the correlations of the uncertain input parameters and the used LCC prediction model, and 2) imperfection propagation step is used to propagate aleatory and epistemic imperfection of correlated input parameters and model structure using the evidence theory. The results show the importance to propagate both parameter and model structure imperfection and to consider correlation among input parameters in LCC prediction model. In this study, the changes prediction of land cover in Saint-Denis City, Reunion Island of next 5 years (2016) was anticipated using multi-temporal Spot-4 satellite images in 2006 and 2011. Results show good performances of the proposed approach in improving prediction of the LCC of the Saint-Denis City on Reunion Island.