ad

AIにより過去の洪水リスクデータをデジタル化し将来予測を可能に(New Research Uses AI to Unlock Decades of Hidden Flood Risk Data)

ad

2026-04-29 ヒューストン大学(UH)

ヒューストン大学の研究は、AIを用いて洪水リスクマップを高精度に更新する手法を開発した。従来の地形データや降雨履歴に加え、機械学習により都市開発や土地利用の変化を反映した動的な予測が可能となり、洪水発生確率や浸水範囲をより現実的に評価できる。これにより、迅速な防災計画や都市インフラ設計の高度化が期待される。特に気候変動に伴う極端降雨の増加に対応するため、継続的なデータ更新と予測精度向上の重要性が示された。

<関連情報>

都市における洪水ハザードの長期評価:米国テキサス州ヒューストンの氾濫原記録の解明 The longitudinal assessment of flood hazard in cities: Unlocking the floodplain record of Houston, TX, USA

Francisco Haces-Garcia, Craig L. Glennie, Hanadi S. Rifai, Vedhus Hoskere

Journal of Hydrology: Regional Studies  Available online: 20 January 2026

DOI:https://doi.org/10.1016/j.ejrh.2026.103113

AIにより過去の洪水リスクデータをデジタル化し将来予測を可能に(New Research Uses AI to Unlock Decades of Hidden Flood Risk Data)

Highlights

  • Develops framework to extract flood hazard data from historical flood insurance maps.
  • Framework is systematically validated, with good data extraction for GIS analysis.
  • Applies framework to three case studies in Greater Houston, TX, USA.
  • Flood hazard in case studies significantly expanded over time.
  • Critical flood resilience consequences for expanding hazard were discovered.

Abstract

Study Region: Houston, TX, USA

Study Focus: The data-driven quantification of evolving urban flood hazard is challenging. Historical flooding data is readily available from the US National Flood Insurance Program, which has mapped Flood Hazard Areas (FHAs) since the 1970s. However, estimated FHAs are generally not used in modern flood studies due to the lack of georeferencing information. This poses a key impediment for fine-scale floodplain analysis, with critical implications for the study of urban flood change. This research develops a framework to automatically georeference historical Flood Insurance Rate Maps, and extract their floodplain data using photogrammetry, geomatics, and artificial intelligence. The registration framework is systematically validated to ensure the accurate extraction of longitudinal flood data. A median georeferencing residual of 23.1 m was obtained, which was smaller than the validation dataset accuracy. The framework provides an avenue towards the widespread assessment of longitudinal flood hazard, with significant implications for the study of urban flood resilience. Three flood-prone case studies are presented to exemplify the usefulness of the framework; Brays Bayou, Hunting Bayou, and Cypress Creek in Greater Houston.

New hydrological insights for the region: The case studies quantify the change of flood hazard within these watersheds. Floodplain expansion had significant flood resilience consequences. Population exposure was estimated to have risen by up to 635%, with a concurrent increase in the vulnerability of critical infrastructure.

ad
0904河川砂防及び海岸海洋
ad
ad


Follow
ad
ad
タイトルとURLをコピーしました