Abstract
This paper explores how machine learning can quantify the economic value of ecosystems and nature-based solutions in reducing flood damage across South Florida. By applying decision tree–based models, including Random Forest and XGBoost, the study integrates hydrological, environmental, and socioeconomic data with flood insurance claims to assess the protective role of wetlands, mangroves, and related ecosystems. Ablation studies and quantile regression are used to evaluate predictive performance across a range of flood severities, while Kalman filters are employed to reduce noise in hydrological time series. Results show that machine learning models can capture both typical and extreme flood events, highlighting the cost-saving benefits of ecosystem preservation and restoration. These findings provide actionable insights for policymakers, insurers, and planners by linking ecological resilience to economic risk reduction, and they demonstrate how data-driven methods can inform sustainable strategies for climate adaptation.
Presenters
Jorge HernandezPost-Doc, Institute of Environment, Florida International University, Florida, United States Jamal Julien
Akorede Oluwo
Details
Presentation Type
Paper Presentation in a Themed Session
Theme
Economic, Social, and Cultural Context
KEYWORDS
MACHINE LEARNING, FLOOD RISK, ECOSYSTEM SERVICES, NATURE-BASED SOLUTIONS, ECONOMIC VALUATION
