High-Resolution Flood Probability Mapping Using Generative Machine Learning with Large-Scale Synthetic Precipitation and Inundation Data

2024-09-20

arXiv.org

AI Summary

High-resolution probabilistic flood mapping is hindered by sparse historical observations and the prohibitive computational cost of physics-based models. To address this, we propose Flood-Precip GAN—a novel precipitation-driven, unit-level deep estimator integrated with a physics-constrained generative adversarial network (GAN). Our framework incorporates a synthetic-event calibration mechanism combining threshold-based filtering and K-nearest-neighbor (KNN) smoothing. Leveraging knowledge distillation from high-fidelity hydraulic simulations and rigorous multi-scale similarity and correlation validation, we generate 10,000 high-fidelity precipitation–inundation paired samples. These enable construction of probabilistic flood maps stratified by water depth classes. Validated against ground-truth measurements, synthesized inundation distributions achieve structural similarity >0.92 and Pearson correlation >0.89. The approach effectively overcomes data scarcity and computational bottlenecks, establishing a scalable, high-accuracy paradigm for large-scale flood risk assessment.

Technology Category

• Computer Vision: Generative Adversarial Networks (GANs) for Vision
• Machine Learning: Calibration & Uncertainty Quantification
• Natural Language Processing: Generation

Application Category

• Social Networks and Social Media: Generative AI / large language models and their impact on social systems
• Economics, Online Markets and Human Computation: Economic ramifications for generative AI infrastructure and applications
• Web Mining and Content Analysis: Web data generation and simulation

Abstract

High-resolution flood probability maps are essential for addressing the limitations of existing flood risk assessment approaches but are often limited by the availability of historical event data. Also, producing simulated data needed for creating probabilistic flood maps using physics-based models involves significant computation and time effort inhibiting the feasibility. To address this gap, this study introduces Flood-Precip GAN (Flood-Precipitation Generative Adversarial Network), a novel methodology that leverages generative machine learning to simulate large-scale synthetic inundation data to produce probabilistic flood maps. With a focus on Harris County, Texas, Flood-Precip GAN begins with training a cell-wise depth estimator using a limited number of physics-based model-generated precipitation-flood events. This model, which emphasizes precipitation-based features, outperforms universal models. Subsequently, a Generative Adversarial Network (GAN) with constraints is employed to conditionally generate synthetic precipitation records. Strategic thresholds are established to filter these records, ensuring close alignment with true precipitation patterns. For each cell, synthetic events are smoothed using a K-nearest neighbors algorithm and processed through the depth estimator to derive synthetic depth distributions. By iterating this procedure and after generating 10,000 synthetic precipitation-flood events, we construct flood probability maps in various formats, considering different inundation depths. Validation through similarity and correlation metrics confirms the fidelity of the synthetic depth distributions relative to true data. Flood-Precip GAN provides a scalable solution for generating synthetic flood depth data needed to create high-resolution flood probability maps, significantly enhancing flood preparedness and mitigation efforts.

Problem

• Generates synthetic flood data for high-resolution maps
• Reduces computation time for flood probability mapping
• Enhances flood risk assessment with machine learning

Innovation

• Generative machine learning for synthetic flood data
• Conditional Generative Adversarial Network for precipitation
• Scalable pipeline for high-resolution flood probability maps