Evaluating the effects of preprocessing, method selection, and hyperparameter tuning on SAR-based flood mapping and water depth estimation

TL;DR

The paper tackles SAR-based flood mapping and water-depth estimation by systematically varying preprocessing, flood-mapping, and depth-estimation methods and their hyperparameters across two Garonne River flood events. It demonstrates that speckle-filter choice and associated hyperparameters strongly influence flood extents, while supervised flood-mapping methods typically outperform unsupervised ones, though tuned unsupervised approaches can approach their performance. Water-depth estimates inherit substantial variability from both input flood maps and depth-estimation parameters, with Fw-DET and FLEXTH showing comparable median accuracy but differing sensitivity. The study argues for ensemble processing and acknowledging pipeline-related uncertainty to improve operational reliability and hydraulic-model calibration, and it provides open-source resources to enable reproducibility and further exploration.

Abstract

Flood mapping and water depth estimation from Synthetic Aperture Radar (SAR) imagery are crucial for calibrating and validating hydraulic models. This study uses SAR imagery to evaluate various preprocessing (especially speckle noise reduction), flood mapping, and water depth estimation methods. The impact of the choice of method at different steps and its hyperparameters is studied by considering an ensemble of preprocessed images, flood maps, and water depth fields. The evaluation is conducted for two flood events on the Garonne River (France) in 2019 and 2021, using hydrodynamic simulations and in-situ observations as reference data. Results show that the choice of speckle filter alters flood extent estimations with variations of several square kilometers. Furthermore, the selection and tuning of flood mapping methods also affect performance. While supervised methods outperformed unsupervised ones, tuned unsupervised approaches (such as local thresholding or change detection) can achieve comparable results. The compounded uncertainty from preprocessing and flood mapping steps also introduces high variability in the water depth field estimates. This study highlights the importance of considering the entire processing pipeline, encompassing preprocessing, flood mapping, and water depth estimation methods and their associated hyperparameters. Rather than relying on a single configuration, adopting an ensemble approach and accounting for methodological uncertainty should be privileged. For flood mapping, the method choice has the most influence. For water depth estimation, the most influential processing step was the flood map input resulting from the flood mapping step and the hyperparameters of the methods.

Figures (23)

• Figure 1: Workflow for processing Sentinel-1 (SAR) images to derive flood maps and water depth fields. Raw images are preprocessed using 26 hyperparameter configurations, then each image is turned into 48 flood maps (or 432 if using morphological operations). Each flood map is transformed into water depth fields using 19 configurations. The configurations are described in Table \ref{['table:all_methods']}.
• Figure 2: Visualization of the study area on the Garonne River in France and the extent of the Sentinel-1 A (S1A) and Sentinel-1 B (S1B) acquisitions during 2019 and 2021 flood events.
• Figure 3: Measured discharge at gauging stations (Tonneins, Marmande, and La Réole) during the (a) December 2019 and (b) February 2021 flood event, along with (c) spatially distributed Strickler values used in the study area.
• Figure 4:
• Figure 5: Backscatter intensity results from applying various speckle filters on a Sentinel-1 image in VH polarization.