An automated cropland classification algorithm (ACCA) that is rule-based is illustrated here for the state of California, USA. The goal of the ACCA is to automatically compute cropland characteristics such as: (a) cropland extent\area; (b) crop type, (c) cropping intensity, and (d) irrigated versus rainfed. However, ACCA here is focused on automatically determining cropland extent using multi-sensor remote sensing and secondary data for the state of California. First, a Mega-file data cube (MFDC) (see section 2.0) was created using Moderate Resolution Imaging Spectroradiometer (MODIS) for year 2010 monthly maximum value composite (MVC) normalized difference vegetation index (NDVI) time-series and Landsat TM5 July 2010 surface reflectance (one time) data. Secondary data (e.g., precipitation, elevation, temperature) did not add additional value in cropland classification and hence were dropped. Second, ACCA algorithm is developed to accurately determine cropland extent. The ACCA algorithm is developed to replicate a truth layer (e.g., a nationally derived map or rigorously interpreted map of cropland that is quite accurate). Third, the goal of ACCA is to automatically determine the cropland characteristics once the MFDC of the area is ready. This should allow computation of cropland characteristics, for which ACCA is developed, year after year and should have the ability to hindcast and nowcast. An additional benefit of ACCA is that once the croplands are determined, the rest of the croplands should be fallow croplands. ACCA is described in detail below using the State of California. You can also read more on ACCA in a paper (Thenkabail and Wu, 2012).
