Spatio-temporal reverse semantic kriging

Abstract

Spatio-temporal prediction and forecasting of the terrestrial land-use/land-cover (LULC) distribution of a RoI facilitates their management, city planning, mitigation of adverse climate change, etc. However, the spatio-temporal change in LULC distribution is not a trivial phenomena to be modeled as it often shows nonlinear behavior in the presence of different factors such as human impact to the ecosystem (e.g., anthropogenic activities, urbanization), change in meteorological parameters, etc. Therefore, for the prediction and forecasting of LULC distribution, incorporation of the meteorological knowledge into the prediction process may facilitate us to develop an advanced model. This work aims to model the behavioral change of interannual LULC pattern of a region by analyzing different related meteorological parameters. This study also attempts to forecast the future LULC distribution using the basic idea of semantic kriging. Here, SemK approach is extended for the spatio-temporal analysis and a new variant is proposed, which is referred to as ST-RevSemK. It captures the semantic relationships among different related meteorological parameters and use this relation to forecast the semantic terrestrial distribution pattern. From the empirical performance evaluation of this framework, it is found that the spatio-temporal modeling of meteorological parameters facilitates improved prediction of LULC pattern.