Visibility Estimation Due to Forest Fire Smoke Using Backward Elimination Multiple Regression of Himawari-8 Satellite Data over Sumatera and Borneo Island Indonesia

Abstract

Land and forest fires smoke has been regularly revealed over Sumatera and Borneo Islands for past two decades. Various sector have been severely impacted, including air transportation. Air obscurity caused by smoke reduce horizontal visibility and profoundly affects aviation regularity, efficiency and safety. The study aimed to develop a visibility estimation of an orbital satellite using remote sensing and observation data. Orbital satellite is one of the technological advances that have potentials for supporting and improving the ability of meteorological service provider to meet user's needs related to adverse weather. The Automatic smoke detection using classification tree analysis was used as the first step. Then backward elimination for multiple regression was used to produce visibility estimation based on Himawari-8 satellite-smoke-detection and actual observation of visibility in airports. The results showed that the correlation between the estimated visibility in satellite-detected smoke and observed visibility was high (R2=0.68). This approach very sensitive to thick smoke with horizontal visibility of shorter than 2000 meters.