Uncertainty in forecasting land cover land use at a watershed scale: Towards enhanced sustainable land management

Human-induced environmental stressors represent a major source of recent global changes that are reflected through alterations in the land cover land use (LCLU) that are concomitantly driven by natural and socioeconomic factors. While the Cellular Automata Markov Chain (CA-MC) process has been advocated as a promising approach in LCLU predictions, this process excludes the underlying processes that drive behavior in the context of LCLU change. Behavioral analyses can uncover motivations by considering a wider range of factors and shedding light on potential uncertainties associated with the CA-MC process. Therefore, the main objective of this study is to incorporate social behavior in the process of LCLU predictions through a novel comparative framework that allows for uncertainty assessment in predicting LCLU changes. We propose a novel comparative framework to assess uncertainty in predicting LCLU changes. The framework compares a behavioral mental model with a Cellular Automata Markov Chain (CA-MC) model expanded through its multi-criteria decision analyses (MCDA) to cover natural and socio-economic factors under a GIS platform. Both models rely on field survey responses of farmers under climate change stress in the same test area as well as historical data for validation and verification with the aim to improve the LCLU prediction ability. Using past land cover maps, the CA-MC approach exhibited a good similarity between simulated and past observed results (71 %) and predicted future urbanization expansion (93 % in 15 years) being the driving force behind the loss of grasslands (73 %) and forests (5 %). However, when both models are compared, agricultural and urban areas exhibited 17 % and 47 % differences reflecting an appreciable level of uncertainty in LCLU predictions. The outcomes highlight the importance of considering behavioral processes in LCLU predictions towards sustainable land management.