2.2.1. MODIS NDVI Data

Time-series MODIS NDVI data sets were used as classification inputs in this study. The MODIS-based vegetation index has been widely used for large-scale crop classification because it provides time-series information at a 250 m spatial resolution [1,29,30]. Similar to Kim and Park [31], we aimed for early crop map production, prior to the release of the CDL 2015 data, as part of crop acreage estimation. Thus, a total of 13 MODIS 16-day composite NDVI data sets from January to July 2015 were experimentally used to account for time-series variations of various crops in the study area. To minimize the effects of clouds in the 16-day composite data sets, the Savitzky-Golay filter [32] was applied prior to classification (Figure 1a). The study area consisted of a 400 by 400-pixel array with a spatial resolution of 250 m.
