Authors: Md. Shahinoor Rahman*, Center for Spatial Information Science and Systems, George Mason University, VA, USA, Liping Di , Center for Spatial Information Science and Systems, George Mason University, VA, USA, Eugene Yu , Center for Spatial Information Science and Systems, George Mason University, VA, USA, Chen Zhang, Center for Spatial Information Science and Systems, George Mason University, VA, USA, Hossain Mohiuddin, School of Urban and Regional Planning, University of Iowa, USA
Topics: Agricultural Geography, Remote Sensing, Land Use and Land Cover Change
Keywords: CDL,Crop Data Layer, Landsat, Major Crop, USA
Session Type: Paper
Start / End Time: 1:10 PM / 2:50 PM
Room: Buchanan, Marriott, Mezzanine Level
Presentation File: No File Uploaded
Field level crop type information is vital for many research and application. USDA is providing crop types information for US cropland as cropland data layer (CDL). However, CDL is only available at the end of the year after the crop growing season. Therefore, CDL data cannot support many in-season research and decision making such as crop loss estimation, yield estimation, and grain pricing. USDA mostly relies on field survey and farmer’s report for the ground truth to train the classification model, which is one of the major reasons for the delayed release of CDL. This research aims to use trusted pixels as ground truth to train the model. Trusted pixels are these pixels which have been following a specific crop rotation pattern since 2007. These trusted pixels are used to train the image classification model for the classification of in-season Landsat images to identify major crop types. Six different classification algorithms are investigated and tested to select the best algorithms for this study. The Random Forest algorithm outstands among selected algorithms. This study classified Landsat scenes between May and mid-August for Iowa states. The result shows overall agreement with CDL are 84%, 94%, and 96% for May, June, July respectively. The accuracies have been assessed through more than 600 ground truth collected from the field. The overall accuracy is 85%, 90% and 94% for May, June, and July respectively. The classification accuracies of multi-date multi-band images are higher than single date multi-band images.