Extracting tempo-spatial features of paddy rice using time-series MOD09A1-A case study in Hunan, China

Abstract

Time-series of 8-day composite MODIS surface reflectance product (MOD09A1) data were used to identify the spatial patterns of double-season early rice, single-season middle rice and double-season late rice of Hunan Province, China in 2010. Firstly, the available MODIS images of transplanting and heading stages were assured in accordance with the schedules of local traditional paddy fields tillage in 2010, and then time-series NDVI images were smoothed by Harmonic Analysis of Time Series (HANTS) algorithm to remove the noise and the atmospheric effects. Secondly, the spatial distribution and planting acreage of three types of rice were derived from combining the Enhanced Vegetation Index (EVI) and Land Surface Water Index (LSWI) according to the water background and variation characteristics of NDVI values at transplanting and heading stages. Finally, the accuracy evaluation was performed by the statistical data. The results showed that three types of rice were mainly distributed along the Dongting Lake Basin in Hunan Province and the relative errors were −10.99 %, 1.46 %, −5.87 %, respectively.