Autologistic regression and multicriteria evaluation models for the prediction of forest expansion

Abstract

Land use changes are complex ecological processes driven by the interaction of biophysical and human related factors. The prediction of forest land use changes is important for sustainable forest management and biodiversity conservation. This study investigates the modelling process of the spatial dynamics of a forest ecosystem in north eastern Greece. For the prediction of forest expansion, based on land use data of the study area, a deterministic approach using logistic regression and heuristic methods of multi-criteria evaluation is adopted. The set of factors driving forest expansion are: the slope, the distance to roads, the distance to urban areas, the distance to forest, the soil depth, the soil erosion and the influence from the land uses of the neighbourhood. The spatial autocorrelation of driving factors is addressed using an autologistic regression model. The multicriteria evaluation approach is developed using weighted linear combination (WLC) and ordered weighted averaging (OWA) methods. In WLC method the relative importance of each factor was estimated using the analytical hierarchy process. In the OWA method, decision strategies are generated using a selection of relative linguistic quantifiers, which allow different Risk in decisions. The accuracy of the models produced was tested with real data for the year 2001 using the ROC validation method. All the methods produced satisfactory results. Autologistic regression showed slightly better performance than multicriteria evaluation methods due to higher degree of objectivity in defining the importance of driving factors for forest expansion.