Identification of windbreaks in Kansas using object-based image analysis, GIS techniques and field survey

摘要

Windbreaks are valuable resources in conserving soils and providing crop protection in Great Plains states in the US. Currently, Kansas has no up-to date inventory of windbreaks. The goal of this project was to assist foresters with future windbreak renovation planning and reporting, by outlining a series of semi-automated digital image processing methods that rapidly identify windbreak locations. There were two specific objectives of this research. First, to develop semi-automated methods to identify the location of windbreaks in Kansas, this can be applied to other regions in Kansas and the Great Plains. We used a remote sensing technique known as object-based image analysis (OBIA) to classify windbreaks visible in the color aerial imagery of National Agriculture Imagery Program. We also combined GIS techniques and field survey to complement OBIA in generating windbreak inventory. The techniques successfully located more than 4500, windbreaks covering an approximate area of 2500, hectares in 14 Kansas counties. The second purpose of this research is to determine how well the results of the automated classification schemes match with other available windbreak data and the selected sample collected in the field. The overall accuracy of OBIA method was 58.97 %. OBIA combined with 'heads up' digitizing and field survey method yielded better result in identifying and locating windbreaks in the studied counties with overall accuracy of 96 %.