Skip to main content
ARS Home » Plains Area » Bushland, Texas » Conservation and Production Research Laboratory » Soil and Water Management Research » Research » Publications at this Location » Publication #266570

Title: Statistical learning algorithms for identifying contrasting tillage practices with landsat thematic mapper data

item SAMUI, PIJUSH - Vellore Institute Of Technology, Vit
item Gowda, Prasanna
item OOMMEN, THOMAS - Michigan Technological University
item Howell, Terry
item MAREK, THOMAS - Texas Agrilife Research
item PORTER, DANA - Texas Agrilife Extension

Submitted to: International Journal of Remote Sensing
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 8/29/2011
Publication Date: 3/17/2012
Citation: Samui, P., Gowda, P., Oommen, T., Howell, T.A., Marek, T.H., Porter, D.O. 2012. Statistical learning algorithms for identifying contrasting tillage practices with landsat thematic mapper data. International Journal of Remote Sensing. 33(18):5732-5745.

Interpretive Summary: Tillage practices affect evaporation and soil erosion from agricultural fields due to wind and water erosion. Consequently, models that simulate agricultural systems require tillage as input. Collecting tillage data over large areas is a time consuming and costly task. In this study, we evaluated two new statistical techniques for identifying and mapping tillage practices, using Landsat satellite data. Comparison of the models indicated that relevance vector machine-based models were found to be superior in identifying contrasting tillage practices.

Technical Abstract: Tillage management practices have direct impact on water holding capacity, evaporation, carbon sequestration, and water quality. This study examines the feasibility of two statistical learning algorithms, such as Least Square Support Vector Machine (LSSVM) and Relevance Vector Machine (RVM), for classifying and mapping two contrasting tillage management practices using remote sensing data. The LSSVM is firmly based on the statistical learning theory, whereas the RVM is a probabilistic model where the training takes place in a Bayesian framework. Input to the LSSVM and RVM algorithms were reflectance values at different bandwidths and indices, derived from Landsat Thematic Mapper (TM) data. Ground truth data for this study were collected from 72 commercial production fields in two counties, located in the Texas High Plains of the south-central United States. Numerous LSSVM and RVM-based tillage models were developed and evaluated for tillage classification accuracy. Percent correct and kappa statistics were used for this purpose. Results showed that the best LSSVM and RVM models included the use of TM band 5 or vegetation indices that involved TM band 5, indicating sensitivity of near infrared reflectance of crop residue cover on the surface. This is consistent with other remote sensing models reported in the literature. Overall classification accuracy of the best LSSVM and RVM models were 87.8% and 90.2%, respectively. The corresponding kappa statistic for those models was 0.75 and 0.8, respectively. Further comparison of the best LSSVM and RVM models with published logistic regression based tillage models developed with the same data, indicated a superiority of the RVM over LSSVM and logistic regression models in determining contrasting tillage practices with Landsat TM data.