Land cover modelling for tropical forest vulnerability prediction in Kalimantan, Indonesia

Abstract

This study aimed to estimate the vulnerability of tropical forest in Kalimantan, Indonesia, from simulated land cover changes using the Weighted Normalised Likelihood–Markov Chain model (WNL-MC) for the period 2018–2050 under a forest conservation and a no-conservation scenario. Predictions of future vulnerability and forest change were based on land cover maps for 2010 and 2014 developed using the integration of Landsat, ALOS PALSAR, and Sentinel-1 data. The four main land cover classes in the region were oil palm and rubber plantations, native forests, and non-forested areas. The performance of the WNL–MC model was evaluated by comparing the simulated and actual land cover maps for 2018 and validating them against high-resolution images. Kappa accuracy for the simulated maps was >85%, and overall accuracy was >90%. Under the scenario of no forest conservation, native forest showed approximately 50% loss over the next 30 years, while the forest conservation scenario showed a slowing of current deforestation by 18% from 2018 to 2030 and by 5.95% from 2030 to 2050. Native forest outside forest conservation zones was most vulnerable to conversion to other land uses, particularly in lowland areas close to settlements and roads. Estimations of native forest loss, particularly under the no-conservation scenario, emphasise the need for policies to preserve and conserve remaining native forest areas.