Paddy and non-paddy crops mapping using multi-temporal data of Sentinel-1A in part of Bantul Regency

摘要

Monitoring of rice field, as a place for producing paddy is very important to realize one aspect of food security, namelyfood availability. Modern agriculture has been widely utilize remote sensing data, especially optical images formonitoring agricultural land in various aspects of land management. However, the use of optical images is hampered bycloud cover when monitoring rice fields because most of them located in tropical countries, so there is an alternative tousing SAR imagery that has ability to penetrate clouds. One of the SAR image products is Sentinel-1A with band C onits sensors which was launched in 2014 and the data can be utilized by the wider community for free. The purpose of thisstudy was to determine the ability of multitemporal Sentinel-1A SAR imagery in identifying paddy and non-paddy inBantul Regency’s agriculture field which was measured through its mapping accuracy. Sentinel-1A multi-temporalimages with ten recording dates from February to May 2018 were used as the main data for this study. The method usedis a digital classification with two approaches i.e. parametric with MLC algorithm and non-parametric with k-NNalgorithm. In addition, the Sentinel-1A, which consists of VV and VH polarization, performed in three classificationschemes (VV multi-temporal, VH multi-temporal, and VV&VH multi-temporal). The classification results show thatmulti-temporal Sentinel-1A can be used to identify paddy and non-paddy crops with an accuracy of 77.69% (VV multitemporal-MLC), 82.15% (VH multi-temporal-MLC), 88.45% (VV&VH multi-temporal-MLC), 76.64% (VV multitemporal-kNN), 78.47% (VH multi-temporal-kNN) and 79.52% (VV&VH multi-temporal-kNN).