Mapping tillage practices with landstat thematic mapper based logistic regression models

Tillage information is crucial in environmental modeling as it has a direct impact on sediment delivery, phosphorus loss, and water holding capacity o f agricultural soils. Remote sensing techniques can provide information abo ut tillage practices over large areas. In this study, six Thematic Mapper ( TM)-based logistic regression models proposed by van Deventer et dl. (1997) were used to distinguish conventional and conservation tillage practices i n the Lower Minnesota River watershed located in southern Minnesota. Accura cy assessments of tillage maps derived from Landsat TM data were made using field data collected by the Natural Resources Conservation Service (NRCS). Regression models were easy to use, cost and time effective and produced r easonably accurate tillage maps. The "percentage correct" and kappa (k) val ues varied from 42-77% and 0.03-0.51, respectively, with best values for lo gistic regression models that use TM band 5 or the difference between TM ba nds 3 and 5 images. This approach is promising for the rapid collection of tillage information on individual fields over large areas.