ABSTRACT

Sawgrass marsh occupies about 70% of the Everglades and distributes from the north to south Everglades. The Everglades is a peatland with significant carbon storage coming from on-site plant production. The emergent plant marshes such as sawgrass are particularly productive. The chapter presents a methodology is presented for sawgrass biomass estimation using Landsat data. Field sawgrass biomass data were collected in the plan to identify baseline conditions and evaluate potential impacts of the uprate. The biomass model can be applied to any objects generated from different scale parameters based on what level of detail is needed and is adaptable to different applications. Based on the object-based matched dataset, four machine learning regression algorithms, Support Vector Machine (SVM), Random Forest, k-Nearest Neighbor, and Artificial Neural Network were evaluated and compared with the parametric regression approach Multiple Linear Regression. Kernel based SVMs are commonly used and several parameters need to be tuned, including kernel to be used, precision, and penalty parameters.