ICESat-2 automatic water detection model

This repository contains a pipeline for classifying water surfaces in ICESat-2 ATL03 photon data using a Random Forest classifier. Developed by H.B. Rotteveel for his MSc thesis in Geomatics at Delft University of Technology (TU Delft). The pipeline downloads and processes raw ICESat-2 granules, computes photon-level features, labels points against a reference water polygon dataset, and trains a classifier to detect water bodies. It consists of seven scripts that are split up in two different sequences (/data_processing and /machine_learning).