Analysis of deforestation patterns and drivers in Swaziland using efficient Bayesian multivariate classifiers

Abstract

Deforestation is a global phenomenon that threatens not only biodiversity but also the livelihoods of people who depend on them. Recent global analyses indicates that this problem is growing, hence the growing need to identify the underlying drivers in order to develop more responsive policies. We present a machine learning-based method which automatically identifies key drivers and makes predictions from available spatial data in Swaziland during the post millennium period. The efficient Bayesian multivariate classifiers (EBMC) are used to learn feature-selected Bayesian network (BN) models of deforestation from multisource data. The EBMC models, learned using the K2 and BDeu algorithms, were also used to predict the probability or risk of deforestation in addition to providing a directed acyclic graphical view of the key interacting factors. These were compared with constraint and knowledge-based BNs developed using the common EBMC-selected variables. All the models performed consistently well (log loss <0.3, AUC >0.8) when evaluated against observed deforestation patterns. The knowledge-based and constrained-based BNs performed well highlighting the need for developing a causal structure of interactions between variables. The findings indicate that deforestation patterns are determined by an interaction of proximate and underlying factors primarily fuelwood use, human population density, human settlements, protection and land ownership status. The findings indicate that, unless robust conservation measures are put in place, deforestation is likely to continue as more areas become vulnerable. The models produced plausible results that can be used for preventive planning and policy making.