Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

HOTA: Hierarchical Overlap-Tiling Aggregation for Large-Area 3D Flood Mapping

Wenfeng Jia, Bin Liang, Yuxi Lu, Attavit Wilaiwongsakul, Muhammad Arif Khan, Lihong Zheng

TL;DR

This work tackles the need for accurate large-area 3D flood maps by introducing HOTA, a plug‑and‑play multi‑scale inference strategy that aggregates overlapping tile predictions during inference. When combined with SegFormer for 2D flood segmentation and a dual-constraint DEM differencing depth module, the approach yields accurate 3D flood surfaces without retraining the backbone. In a Kempsey, Australia case study (March 2021), SegFormer with HOTA improved IoU from 73.05% (U‑Net baseline) to 83.97%, and the resulting 3D flood surface achieved mean boundary errors below 0.5 m, demonstrating rapid, large-area flood mapping capability for disaster response. The method’s reliance on DEM quality and single-date imagery is noted, with future work proposed on multi-source data, multi-temporal analysis, and uncertainty quantification to enhance robustness and applicability across diverse terrains.

Abstract

Floods are among the most frequent natural hazards and cause significant social and economic damage. Timely, large-scale information on flood extent and depth is essential for disaster response; however, existing products often trade spatial detail for coverage or ignore flood depth altogether. To bridge this gap, this work presents HOTA: Hierarchical Overlap-Tiling Aggregation, a plug-and-play, multi-scale inference strategy. When combined with SegFormer and a dual-constraint depth estimation module, this approach forms a complete 3D flood-mapping pipeline. HOTA applies overlapping tiles of different sizes to multispectral Sentinel-2 images only during inference, enabling the SegFormer model to capture both local features and kilometre-scale inundation without changing the network weights or retraining. The subsequent depth module is based on a digital elevation model (DEM) differencing method, which refines the 2D mask and estimates flood depth by enforcing (i) zero depth along the flood boundary and (ii) near-constant flood volume with respect to the DEM. A case study on the March 2021 Kempsey (Australia) flood shows that HOTA, when coupled with SegFormer, improves IoU from 73\% (U-Net baseline) to 84\%. The resulting 3D surface achieves a mean absolute boundary error of less than 0.5 m. These results demonstrate that HOTA can produce accurate, large-area 3D flood maps suitable for rapid disaster response.

HOTA: Hierarchical Overlap-Tiling Aggregation for Large-Area 3D Flood Mapping

TL;DR

This work tackles the need for accurate large-area 3D flood maps by introducing HOTA, a plug‑and‑play multi‑scale inference strategy that aggregates overlapping tile predictions during inference. When combined with SegFormer for 2D flood segmentation and a dual-constraint DEM differencing depth module, the approach yields accurate 3D flood surfaces without retraining the backbone. In a Kempsey, Australia case study (March 2021), SegFormer with HOTA improved IoU from 73.05% (U‑Net baseline) to 83.97%, and the resulting 3D flood surface achieved mean boundary errors below 0.5 m, demonstrating rapid, large-area flood mapping capability for disaster response. The method’s reliance on DEM quality and single-date imagery is noted, with future work proposed on multi-source data, multi-temporal analysis, and uncertainty quantification to enhance robustness and applicability across diverse terrains.

Abstract

Floods are among the most frequent natural hazards and cause significant social and economic damage. Timely, large-scale information on flood extent and depth is essential for disaster response; however, existing products often trade spatial detail for coverage or ignore flood depth altogether. To bridge this gap, this work presents HOTA: Hierarchical Overlap-Tiling Aggregation, a plug-and-play, multi-scale inference strategy. When combined with SegFormer and a dual-constraint depth estimation module, this approach forms a complete 3D flood-mapping pipeline. HOTA applies overlapping tiles of different sizes to multispectral Sentinel-2 images only during inference, enabling the SegFormer model to capture both local features and kilometre-scale inundation without changing the network weights or retraining. The subsequent depth module is based on a digital elevation model (DEM) differencing method, which refines the 2D mask and estimates flood depth by enforcing (i) zero depth along the flood boundary and (ii) near-constant flood volume with respect to the DEM. A case study on the March 2021 Kempsey (Australia) flood shows that HOTA, when coupled with SegFormer, improves IoU from 73\% (U-Net baseline) to 84\%. The resulting 3D surface achieves a mean absolute boundary error of less than 0.5 m. These results demonstrate that HOTA can produce accurate, large-area 3D flood maps suitable for rapid disaster response.

Paper Structure

This paper contains 16 sections, 2 equations, 11 figures, 2 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Related Work
  3. Research States for Flood Mapping
  4. Deep Learning Models& Flood Depth Estimation Methods
  5. Methodology
  6. Research Area & Datasets
  7. SegFormer and U-Net Building & Training
  8. Hierarchical Overlap-Tiling Aggregation
  9. Dual-Constraint Flood Depth and Boundary Joint Optimization
  10. Results and Analysis
  11. Evaluation Metrics
  12. 2D Flood Mapping Results
  13. 3D Flood Mapping Results
  14. Discussion
  15. Conclusion
  16. ...and 1 more sections

Figures (11)

  • Figure 1: Flood events from 1994 to 2024.
  • Figure 2: Difference method for flood depth estimation.
  • Figure 3: Complete technical workflow.
  • Figure 4: Overview of the Kempsey flood dataset.
  • Figure 5: SegFormer model.
  • ...and 6 more figures