Coastal Land Cover Mapping by High-Resolution Satellite Images and Airborne Lidar Point Clouds

To better estimate effects of relative sea level rising, it is essential to obtain accurate land cover mapping for coastal zones as possible. The main objectives are to identify tidal-creek systems, bare land, algal flats, marshes, upland grasses, shrubs, and trees for a coastal zone. We developed a novelty stacked classification approach to take advantages of high-resolution satellite imagery and airborne lidar point clouds. Here, a rule-based filter classifier is stacked on a Random Forest classifier for multispectral images and a filter classifier for lidar point clouds. Thus, it is not necessary to deal with the different data acquisition dates, and smaller training sets are needed. High-resolution satellite imagery such as WorldView 2 can provide 2m multiple spectral data, which are suitable for separating water, land, and vegetation. However, spectral data possibly misclassify algal flat and marsh due to their similar chlorophyll absorption. Apparently, elevation from lidar points is helpful to separate them. The airborne lidar can deliver land surface point clouds by average point density 4 lidar points per square meter or more. Water mislabeling could be produced by both spectral images and point clouds in coastal zones. For examples, a spectral classifier may classify spilling and plunging breakers to land due to the white foam of the breaking waves that is different to typical spectral response of water. A point cloud classifier may classify these waters to low vegetations due to elevation changes of the waves. Due to different physical reasons, the two misclassifications have different neighboring components. The rule-based filter classifier can correct the misclassifications very well. This classification approach shows promising results with the experiments of 26 wv2 images and 1568 lidar 1500m x 1500m tiles by Matagorda Bay region, Texas. The approach also is used to support our Effects of Sea Level Rise (ESLR) project for the six-county Coastal Bend of Texas.