Abstract

Improved High-Resolution Land-cover and Impervious Surface Mapping Using Advanced Modeling
Track: Imagery and Raster
Authors: Philipp Nagel

Highly accurate land cover and impervious surface maps are often required by local governments and other agencies to aid in decision-making. Regional datasets with acceptable accuracy and resolution are frequently lacking. Typically, regional land cover and impervious surface data was derived from 30-m Landsat imagery using various digital techniques. The recently available 1-m, four-band National Agricultural Imagery Program (NAIP) orthoimagery and the Light Detection and Ranging (LIDAR) data for Minnesota pose an opportunity to update and improve existing datasets. It also creates a challenge in developing methods to process the new data. In this study, a new approach was developed to integrate object-based classification and regression tree techniques. The method was applied to the 1-m resolution datasets in Minnesota to extract highly accurate land cover and impervious surface maps. The resulting data can be utilized by local governments. The proposed method can be calibrated to be used in other regions.