An integrated framework for satellite-based flood mapping and socioeconomic risk analysis: A case of Thailand

Abstract

This study introduces a novel approach to monitoring floods and estimating socioeconomic impacts in Thailand. The approach leverages advancements in geospatial data, employing two web-based applications developed on the Google Earth Engine platform. These tools provide user-friendly access to a vast array of satellite data at the provincial level, including flooded areas, nighttime-light density, drought index, rainfall, cropland, and urban areas. The study also merges these satellite-based indices with official provincial GDP data from 2018 to 2022 to empirically analyze socioeconomic impacts using four machine learning algorithms. The result obtained from Random Forest (RF) demonstrates the highest predictive power for GDP forecasting (r-squared value of 0.912). Feature analysis methods identified the proportion of flooded urban areas as one of the most significant variables in predicting provincial GDP. The RF prediction model was also employed to conduct counterfactual simulations for the period 2018–2022, hypothesizing a scenario devoid of flood events. This approach facilitated the determination of a theoretical GDP value in the absence of floods, thereby enabling the calculation of flood-related economic losses, which averaged 0.945 % of GDP. The study's analytical framework, notable for its cost-effectiveness, leverages openly accessible data and open-source software packages, making it highly applicable to various developing countries.