Spatio-Temporal data modeling in response to deforestation monitoring (a case study of small region in Riau Province, Indonesia)

Abstract

Indonesia with large amount of area covered by tropical forest faces a critical problem of deforestation. A lot of forested areas were converted into other coverage influenced by human activities. Therefore, deforestation monitoring and forest prediction have to be done in order to manage the sustainability of forest. To monitor deforestation, this research has analyzed the trend of forest cover in the study area by combining NDVI differencing and image classification to describe the forest cover change. In order to do that, Landsat images acquired in different time (1996, 2000, and 2005) have been chosen as input. NDVI differencing has been conducted by doing normalization of one image to another image initially. Subsequently, thresholds to identify the change and no change have been carried out separately for decrease and increase part. Apart from that, image classification was applied using supervised classification. Eventually, land cover change detection has been performed by combining NDVI differencing and image classification. It has been proved by the research that forest in study area has decreased by 6% during 1996-2005. In order to forecast future forest cover, three models were chosen to get the best model for prediction. These models are Stochastic Markov Modal, Cellular Automata Markov (CA_Markov) Model, and GEOMOD. To measure the best model among them, Kappa index was employed to validate the simulation. As the result, GEOMOD performed the highest Kappa. Therefore, GEOMOD was implemented to model forest cover in 2015. The result of GEOMOD implementation revealed that forest cover will be decreased by 12% during 2005-2015.