Delineating rice cropping activities from MODIS data using wavelet transform and artificial neural networks in the Lower Mekong countries

Delineating rice cropping activities is important for crop management and crop production estimation. This study used time-series MODIS data (2000, 2005, and 2010) to delineate rice cropping activities in the Lower Mekong countries. The data were processed using the wavelet transform and artificial neural networks (ANNs). The classification results assessed using the ground reference data indicated overall accuracy and Kappa coefficients of 83.1% and 0.77 for 2000, 84.7% and 0.8 for 2005, and 84.9% and 0.8 for 2010, respectively. Comparisons between MODIS-derived rice area and rice area statistics at the provincial level also reaffirmed close agreement between the two datasets (R2 ≥ 0.8). An examination of relative changes in harvested area revealed that from 2000 to 2010 the area of single-cropped rice increased 46.1%, while those of double- and triple-cropped rice were 20.1% and 25%, respectively.

► Study delineated farming practices from MODIS data using wavelet transform and ANNs.
► The results were verified using ground reference data and rice area statistics.
► The overall accuracy and Kappa coefficients were in turn greater than 80% and 0.7.
► There were close agreement between MODIS-based rice area and statistics (R2 ≥ 0.8).
► Remarkable changes in rice cropping activities were observed during 2000–2010.