Blending multi-temporal SPOT and MODIS imageries for rice crop phenology detection

Information on rice crop phenology is important for crop management. The main objective of this study was to generate the higher spatiotemporal resolution data from SPOT and MODIS data for rice crop phenology detection in Taiwan. The data were preprocessed using the spatial and temporal adaptive fusion model (STARFM) and the methodology comprises four main steps: (1) constructing the time-series Normalized Difference Vegetation Index (NDVI) data from STARFM MODIS-SPOT synthetic images, (2) filtering noise from the time-series data, (3) detecting rice crop phenology, and (4) validating the results of rice crop phenology detection. The preliminary results obtained from comparisons between the SPOT NDVI and the synthetic NDVI image indicated a close correlation between the two datasets (R² > 0.8). The comparisons between the estimated phenological dates (sowing and harvesting dates) and the field survey data confirmed the satisfactory results. The RMSE values were respectively 10.7 and 15.3 days for the sowing and harvesting dates, which were smaller than those achieved from the MODIS data (i.e., 21.0 and 20.5 days for the sowing and harvesting dates). The results obtained from this study led to a realization that the synthetic MODIS-SPOT data could be used for monitoring rice crop phenology of small and fragmental rice fields in Taiwan.