UJICOBA MODEL PEMETAAN LAHAN SAWAH BERBASIS PERUBAHAN PENUTUP LAHAN CITA LANDSAT MOSAIK TAHUNAN DI JAWA BARAT

Abstract

Land cover changes of bare land, water and vegetation can be used as a basis for paddy field mapping by using probability theory approach, that is, the probability of one area can be determined as paddy field if the changes of water, bare and vegetation in multi time series can be detected. The results of preliminary studies that have been done on Tenggamus region – Lampung showed that probability theory approach produces a mapping accuracy reaches 91.2%. Based on this results, it has been carried out the model of validation for the wide region for some districts in Province West Java. The data used in this study are multitemporal Landsat 2000-2009. Data processing methods include: 1. Unsupervised digital classification of global land cover to map the bare land, vegetation and water from Landsat images, 2. Merger of each two multitemporal land cover so that the three spatial information obtained: bare land, vegetation and water 2000-2009. The validation of land cover changes made by overlaying the three spatial information. The evaluation results conducted by the confusion matrix (error matrix) by using reference paddy field 1:50,000 scale in 2010. Results of the testing showed that the average mapping accuracy of this probability model reaches 65.5%.