Selection of multi-temporal scenes for the Mississippi Cropland data layer, 2004

Abstract

This paper describes the use of the National Agricultural Statistics Services remote sensing methodology to compare three multi-temporal scene-processing scenarios of the Mississippi Delta Region to produce the Mississippi Cropland Data Layer for 2004. The Mississippi Cropland Data Layer Program, initially started in 1999, uses medium resolution satellite imagery, random sample selection of fields, and field data to produce crop acreage estimates and a categorized data layer of the state showing the land use. The focus is on the major field crops and allows a visualization of the entire State of Mississippi land use by year when cloud-free imagery is available. The results for the three multi-temporal strategies show use of 2004 Landsat 5 scenes moderately early (5/6/04) and late (9/27/04) in the season give the superior classification for the Mississippi Delta and the 2004 Cropland Data Layer based on multi-temporal classification statistics and observation of the classified images. On the Mississippi Department of Agriculture and Commerce web site (http://www.mdac.state.ms.us/) you can currently see the changes in Mississippi farming per year from 1999 to 2003. In July 2005, the 2004 data will be added. The Cropland Data Layer results compare favorably with the official NASS Mississippi crop estimates for the major field crops in 19992004. Keywords-component; Landsat 5; Cropland Data Layer; Mississippi Delta; multi-temporal; crop acreage estimate