In this session, we will explore how to model forest structure and quantify ecosystem services using Python and Earth observation data. Leveraging GEDI, Sentinel-1, Sentinel-2, and SRTM datasets, we’ll walk through data preprocessing, machine learning techniques (e.g., Random Forest), and interpretation using SHAP for understanding feature importance.
We will demonstrate how to estimate canopy height and aboveground biomass, and how to generate meaningful maps to support forest monitoring, ecosystem service assessment, and climate change research. Special emphasis will be placed on the role of forests in the carbon cycle and their significance in climate resilience strategies.
This session is ideal for students, researchers, and practitioners in forestry, ecology, climate science, and remote sensing. Whether you're new to forest modeling or looking to enhance your geospatial machine learning skills, this hands-on walkthrough will provide valuable insights and practical tools.
Background
Forest Modeling and Ecosystem Services Evaluation Through a Data-Driven Python Workflow
Outline

• Introduction to forest remote sensing datasets: GEDI, Sentinel-1, Sentinel-2, and SRTM
• Overview of forest structure, aboveground biomass (AGB), and ecosystem service indicators
• Preprocessing spatial and tabular data using Python (geopandas, rasterio, pandas)
• Machine learning approaches for forest modeling: Random Forest, feature selection, and model evaluation
• SHAP (SHapley Additive exPlanations) for model interpretation and understanding variable importance
• Estimating and mapping forest canopy height, biomass, and ecosystem services
• Case study: combining GEDI, Sentinel, and SRTM data for forest modeling
• Discussion and Q&A – Applications, challenges, and future directions

Prerequisites

• Basic Python programming knowledge

Resources

• will be shared