Seasonal change detection of riparian zones with remote sensing images and genetic programming in a semi-arid watershed.

Riparian zones are deemed significant due to their interception capability of non-point source impacts and the maintenance of ecosystem integrity region wide. To improve classification and change detection of riparian buffers, this paper developed an evolutionary computational, supervised classification method--the RIparian Classification Algorithm (RICAL)--to conduct the seasonal change detection of riparian zones in a vast semi-arid watershed, South Texas. RICAL uniquely demonstrates an integrative effort to incorporate both vegetation indices and soil moisture images derived from LANDSAT 5 TM and RADARSAT-1 satellite images, respectively. First, an estimation of soil moisture based on RADARSAT-1 Synthetic Aperture Radar (SAR) images was conducted via the first-stage genetic programming (GP) practice. Second, for the statistical analyses and image classification, eight vegetation indices were prepared based on reflectance factors that were calculated as the response of the instrument on LANDSAT. These spectral vegetation indices were then independently used for discriminate analysis along with soil moisture images to classify the riparian zones via the second-stage GP practice. The practical implementation was assessed by a case study in the Choke Canyon Reservoir Watershed (CCRW), South Texas, which is mostly agricultural and range land in a semi-arid coastal environment. To enhance the application potential, a combination of Iterative Self-Organizing Data Analysis Techniques (ISODATA) and maximum likelihood supervised classification was also performed for spectral discrimination and classification of riparian varieties comparatively. Research findings show that the RICAL algorithm may yield around 90% accuracy based on the unseen ground data. But using different vegetation indices would not significantly improve the final quality of the spectral discrimination and classification. Such practices may lead to the formulation of more effective management strategies for the handling of non-point source pollution, bird habitat monitoring, and grazing and live stock management in the future.