Verifying Haze Dispersion from Peatland Fires using Classified Landsat 8

Haze dispersion from peatland fire that influence human health can be recognized by analyzing haze trajectory pattern. Previous studies have successfully analyzed haze dispersion as the result of the simulation using HYSPLIT. However the haze dispersion have not verified and compared with the real situation. This study aims to verify haze dispersion from peatland fires in Riau Province using Landsat 8. The Landsat 8 was classified using the C5.0 decision tree algorithm. There were eight classes used in the classification which are cloud, smoke/haze, water body, shadow, built up area, burning area, burned area, and vegetation. The best classification model has the accuracy of 97.58% and kappa coefficient of 0.97. This study has successfully verified 58 haze locations in the study area. As many 98.28% of those locations were found on smoke/haze and cloud. In addition, about 41.38% of 58 haze points are located at height and pressure that are harmful for human health.