Skip to main content
ARS Home » Plains Area » El Reno, Oklahoma » Oklahoma and Central Plains Agricultural Research Center » Agroclimate and Hydraulics Research Unit » Research » Publications at this Location » Publication #414359
Research Project: Development of a Monitoring Network, Engineering Tools, and Guidelines for the Design, Analysis, and Rehabilitation of Embankment Dams, Hydraulic Structures, and Channels

Location: Agroclimate and Hydraulics Research Unit

Title: Integrating LANDSAT imagery and HAND-FIM data with machine learning for enhanced flood extent prediction

Author
item ABEYSINGHE, UMANDA - University Of Missouri
item WEISS, DAVID - University Of Missouri
item ALOYSIUS, NOEL - University Of Missouri
item Hunt, Sherry

Submitted to: Meeting Abstract
Publication Type: Abstract Only
Publication Acceptance Date: 3/29/2024
Publication Date: 4/9/2024
Citation: Abeysinghe, U., Weiss, D., Aloysius, N., Hunt, S. 2024. Integrating LANDSAT imagery and HAND-FIM data with machine learning for enhanced flood extent prediction. Poster. 2024 Show Me Research Week - University of Missouri Research Forum, April 8-12, 2024, Columbia, Missouri.

Interpretive Summary:

Technical Abstract: Predicting the extent of flooding with high accuracy and efficiency remains a critical challenge in flood management and disaster response. Traditional hydrodynamic models, while comprehensive, often demand extensive computational resources and detailed input data that may not be readily available for all regions. This study introduces a novel surrogate model that integrates Height Above Nearest Drainage (HAND) Flood Inundation Mapping (FIM) data with LANDSAT imagery to predict flood extents for specific events. By leveraging machine learning techniques, our model utilizes the spatial and temporal data derived from HAND-FIM and satellite observations, enhancing the detection of water bodies and flood-affected areas through indices such as the Normalized Difference Water Index (NDWI). The model extracts key features associated with flood susceptibility, including elevation, slope, and changes in land cover, to inform its predictions. Employing a variety of machine learning algorithms, such as Random Forest and Neural Networks, the surrogate model is trained and validated against LANDSAT-derived flood extents, ensuring the alignment of datasets both geographically and temporally. Our results demonstrate the model's capability to accurately predict flood extents, as validated by comparative analyses against observed LANDSAT imagery. This approach not only reduces reliance on detailed hydrodynamic simulations but also offers a scalable and efficient tool for flood risk assessment and planning in both urban and rural settings. The development of this surrogate model represents a significant step forward in the utilization of available remote sensing data and machine learning for enhanced flood management practices. USDA is an equal opportunity provider and employer.