Knowledge of where, when, and how much paddy rice is planted is crucial information for understating of regional food security, freshwater use, climate change, and transmission of avian influenza virus. We developed seasonal paddy rice maps at high resolution (10 m) for Bangladesh and Northeast India, typical cloud-prone regions in South Asia, using cloud-free Synthetic Aperture Radar (SAR) images from Sentinel-1 satellite, the Random Forest classifier, and the Google Earth Engine (GEE) cloud computing platform. The maps were provided for all the three distinct rice growing seasons of the region: Boro, Aus and Aman. The paddy rice maps were evaluated against the independent validation samples, and compared with the existing products from the International Rice Research Institute (IRRI) and the analysis of Moderate Resolution Imaging Spectroradiometer (MODIS) data. The generated paddy rice maps were spatially consistent with the compared maps and had a satisfactory accuracy over 90%. This study showed the potential of Sentinel-1 data and GEE on large scale paddy rice mapping in cloud-prone regions like tropical Asia.

No abstract

Abstract:Rice is the staple food for half of the world's population. Therefore, accurate information of rice area is vital for food security. This study investigates the effect of phenology for rice mapping using an object-based image analysis (OBIA) approach. Crop phenology is combined with high spatial resolution multispectral data to accurately classify the rice. Phenology was used to capture the seasonal dynamics of the crops, while multispectral data provided the spatial variation patterns. Phenology was extracted from MODIS NDVI time series, and the distribution of rice was mapped from China's Environmental Satellite (HJ-1A/B) data. Classification results were evaluated by a confusion matrix using 100 sample points. The overall accuracy of the resulting map of rice area generated by both spectral and phenology is 93%. The results indicate that the use of phenology improved the overall classification accuracy from 2%-4%. The comparison between the estimated rice areas and the State's statistics shows underestimated values with a percentage difference of´34.53%. The results highlight the potential of the combined use of crop phenology and multispectral satellite data for accurate rice classification in a large area.