Evaluation of optical remote sensing models for crop residue cover assessment

Measurement of crop residue cover over large areas is useful for monitoring conservation tillage adoption, assessing carbon sequestration potential and erosion modeling. This study was designed to test the accuracy of crop residue estimates in current Tillage Transect Surveys, and to test the feasibility of predicting crop residue cover based on data recorded by Landsat Enhanced Thematic Mapper Plus (ETM+) satellite scenes. A total of 468 corn and/or soybean fields in 11 Minnesota counties were characterized for residue cover in the course of three sampling campaigns coinciding in time with satellite scene acquisition. Results showed that Tillage Transect Survey estimates were correct for 49 percent to 74 percent of fields when either five or two categories were used in classification respectively. Regression analysis showed a strong positive relationship between percent soybean residue cover and ETM+ bands 1, 3, and 7 (r2 =0.66) and between percent corn residue and ETM+ bands 4, 5 and 7 (r2 = 0.44). Three additional indices based on satellite digital numbers, the Soil Tillage Index, Normalized Difference Index, and Normalized Difference Tillage Index had coefficients of determination between 0.02 and 0.56 for corn and soybean residues. The Crop Residue Index Multiband model, a more physically based model, correctly predicted residue cover categories for 30 to 64 percent of fields when five or two categories were used in classification respectively. We conclude that remote-sensing techniques had accuracy as good or better than Tillage Transect Surveys estimates when residue cover classifications were decreased to two categories (0 to 30 percent, and >30 percent). Since residue cover information is primarily needed to assess the extent of two categories, conservation and conventional tillage, remote sensing with Landsat imagery provides a means of sampling every field with an efficient, economical and uniform methodology.