Mapping Rice Growth Stages Employing MODIS NDVI and ALOS AVNIR-2

Rice, a staple food of most Asian inhabitants, is broadly cultivated and has attracted substantial research interest in the past few decades. Monitoring rice production areas is important due to the growing global demand for the crop. Since cropping systems vary across time and space, frequent monitoring over broad areas is required. This research exploits Moderate Resolution Imaging Spectroradiometer (MODIS) data to identify the stage of rice growth, and Advanced Land Observing Satellite–Advanced Visible and Near Infrared Radiometer Type 2 (ALOS AVNIR-2) to map the stages. Exploiting X12-ARIMA to decompose Normalized Difference Vegetation Index (NDVI) time-series for growth-stage indication and five classifiers for mapping the growth stages, the framework was tested in Indonesia’s “paddy basket,” namely the North Coastal Region of West Java. A conventional classifier, Maximum Likelihood, was compared with some decision tree algorithms, namely Classification Rule with Unbiased Interaction Selection (CRUISE) and Quick, Unbiased, Efficient, Statistical Tree (QUEST); neural network (NN); and support vector machine (SVM) for mapping the growth stages. The seasonal component of time-series decomposition assisted in indicating the stages. Meanwhile, decision tree algorithms produced interpretable rules for rice growth stages while the spatial representation of SVM and NN was closer to the ground truth.