A multi-scale accuracy assessment of the MODIS irrigated agriculture data-set (MIrAD) for the state of Nebraska, USA

Accurate and timely information about the geographic distribution of irrigated cropland is important for a range of applications including crop assessments, water resources management, drought monitoring, and environmental modeling. In the United States, a consistent, seamless irrigated agricultural lands map was not available until the development of the 250-m moderate resolution imaging spectroradiometer (MODIS) irrigated agriculture data-set (MIrAD), which was developed from time-series MODIS normalized difference vegetation index, county-level crop area statistics, and land use/land cover (LULC) data using an automated county-level classification approach. This study performed a detailed multi-scale assessment of the MIrAD over the state of Nebraska, which is extensively irrigated, to determine the thematic accuracy of classified irrigation patterns at various spatial scales (i.e., landscape, field, and subfield) and over various crop types. The MIrAD was found to map comparable irrigated cropland patterns to a Landsat-derived LULC map (82% pixel-level thematic agreement) over Nebraska, with most areas of disagreement occurring in transitional mixed pixel locations between adjacent LULC types. Classification accuracy was found to have little variation by general field size and crop type, but accuracies were lower for pixel locations at the fields edges. Overall, this study found the MIrAD to be a relatively accurate and consistent irrigated cropland classification within the US Central Great Plains region that can be used for local-and regional-scale applications.