Mapping Prominent Cash Crops Employing ALOS PALSAR-2 and Selected Machine Learners

Monitoring crops area is essential in achieving food security. The pro- duction coverage, crop types, and their growth phases are the key for monitoring food supply. Remote sensing plays a critical role to provide reliable data on regional basis supporting food production monitoring. In this research, we evaluated the use of Phased Array-type L-band Synthetic Aperture Radar (PALSAR-2), coupling with selected machine learners to map crop areas in the South Burnett, Queensland, Australia. Feature amendments onto dual polarimetric of ALOS PALSAR-2 were then assessed by means of variable importance to improve classifcation perfor- mance. Four machine learners were selected based on previous research and evalu- ated through classifcation accuracy. The best performer was Random Forest followed by C5.0, which generated accuracy at 82% and 81%, respectively. The response of data amendment varied over different classifers. Random Forest and C5.0 seem to produce the highest accuracy at the best data-subset, while additional