Detection of yearly change in farming systems in the Vietnamese Mekong Delta from MODIS time-series imagery

Abstract

This paper briefly introduces an integrated remote-sensing methodology based on MODIS time-series imagery for detecting spatiotemporal changes in Vietnamese Mekong Delta (VMD) farming systems. The integrated methodology consists of six parts, and uses a wavelet-based filter to smooth MODIS-derived time-series indexes (EVI, LSWI, and DVEL) from 2000 to 2008. Through a proposed decision tree using the smoothed indexes, we classified several farming systems, viz., triple rice cropping, two types of double rice cropping, shrimp-rice farming (rotational cropping), and inland aquaculture (monoculture). The MODIS-derived estimate of the total rice-planted area, shrimp- rice farming area, and inland aquaculture area agreed well with statistical data at the province level (R2 ≥ 0.96). However, in some provinces, the estimate has a large margin of error, probably because of the mixed-pixel effect due to the moderate spatial resolution of MODIS (250 m). According to the estimated spatial pattern of the farming systems in the whole VMD, inland aquaculture and shrimp- rice farming areas are distributed mostly in the coastal provinces. The areas of the farming systems steadily expanded until 2007, and double rice cropping systems in both upper and coastal regions were replaced by triple rice cropping because of infrastructure improvements. The proportion of the triple rice cropping area peaked in 2005 and then declined steadily over the next 2 years. We discuss the advantage of the proposed methodology for detecting the spatiotemporal changes of land use patterns, especially of farming systems in a regional area.