State-level Crop Mapping in the U.s. Central Great Plains Agroecosystem Using Modis 250-meter Ndvi Data

Improved and up-to-date land use/land cover (LULC) datasets are needed for intensively cropped regions such as the U.S. Central Great Plains, in order to support a variety of science and policy applications focused on understanding the role and response of the agricultural sector to environmental change issues. The Moderate Resolution Imaging Spectroradiometer (MODIS) holds considerable promise for detailed crop-related LULC mapping in this region, given its global coverage and unique combination of spatial (250-meter), spectral, temporal (16-day composites), and radiometric (12-bit) resolutions. In this study, a hierarchical MODIS-based crop mapping protocol, which incorporates a time-series of MODIS 250-m NDVI data and a decision tree classifier, was tested for the state of Kansas. The protocol produced a series of three crop-related LULC maps that classified: 1) general crop types (alfalfa, summer crops, winter wheat, and fallow), 2) summer crop types (corn, sorghum, and soybeans), and 3) irrigated/non-irrigated crops. A statistical accuracy assessment and state and sub-state areal comparisons with USDA crop acreage information were conducted for each map to assess its overall quality and highlight any major areas of regional misclassification. The MODIS NDVI-derived maps generally had overall and class-specific accuracies of greater than 80%. Overall accuracies ranged from 84% (summer crop map) to 94% (general crop map). The classified crop areas were within 1-5% of USDA reported crop areas for most classes at the state level. Sub-state comparisons found that for most classes the areal discrepancies were relatively minor throughout the state. The largest areal differences occurred in eastern Kansas due to the omission of many small cropland areas that were not resolvable at the 250-m resolution. Regional areal differences were also found for selected classes that were the result of localized precipitation patterns and specific crop practices (i.e. double cropping). These results illustrate the considerable potential of the MODIS-based mapping protocol for spatially and thematically detailed regionalscale crop classification.