12.5
National irrigated lands mapping via an automated remote sensing-based methodology
Jesslyn Brown, SAIC, Sioux Falls, SD; and B. D. Wardlow, S. Maxwell, S. Pervez, and K. Callahan
According to the U.S. Department of Agriculture (USDA) 2002 Census of Agriculture (published every five years by the National Agricultural Statistics Service), more than 50 million acres in the United States are irrigated. However, there is no detailed national-level map of irrigated lands. Knowledge of the distribution and location of irrigated lands can play a role in multiple water-resource management decisions. This information can factor into hydrologic research questions and operational decisions related to crop water use, strategic planning, water rights, irrigation system performance diagnostics, and water supply impact assessment. Mapping all of the irrigated areas across the conterminous United States is challenging due to the geographic extent of the study area (over 2.9 million square miles) and the heterogeneity of the agricultural landscape. Some spatially-specific irrigation information is available, but it is not of adequate spatial detail, not up-to-date, not consistently available across all of the conterminous states, or a combination of these factors. Although several state-level mapping activities have delineated irrigated areas, they were not consistent in their classification methods or production dates. Remote sensing offers an objective synoptic source from which to compile geospatial information related to water use and management, including irrigated areas. The goal of this project is to develop a robust, consistent, and efficient methodology to identify irrigated lands across the United States based on remotely sensed and other geospatial input data generated in an automated and repeatable fashion. This methodology incorporates three data sources: satellite-derived vegetation index (VI) data, USDA county-level irrigation summary statistics for 2002, and general land cover information. The satellite data, collected by the Moderate Resolution Imaging Spectroradiometer (MODIS) aboard the Earth Observing System Terra platform, provides a measure of the annual peak growing season productivity at a 250-m spatial resolution. In an automated classification environment (ArcView Avenue), the 2002 county statistics (i.e., number of irrigated acres) set the criteria for dynamically identifying and selecting the number of cells with the highest annual peak MODIS VI within the appropriate land cover classes. The land cover information is contributed by the National Land Cover Database. A comprehensive evaluation of the initial irrigated lands classification is underway and will incorporate several comparison data sources because there is no single national irrigated lands map available at this scale. A recent evaluation conducted over Nebraska utilized detailed Landsat-derived irrigated cropland data sets [prepared for the Cooperative Hydrology Study (COHYST)] in the central Platte River basin for a time period comparable to the MODIS-derived results. Initial comparisons found a moderate level of areal agreement at a 250-m resolution between the two maps, both in total acreage and geographic correspondence. Further evaluations of the MODIS-derived irrigated lands map will be carried out during the next six months and will include comparisons with existing state or regional land use and land cover maps containing irrigated crop classes and comparisons with geospatial irrigation well location data.
Session 12, Advances in Remote Sensing and Data Assimilation in Hydrology, Part IV
Thursday, 24 January 2008, 3:30 PM-5:00 PM, 223
Previous paper
Extended Abstract (1.3M)
Recorded presentation