Using AI to learn and generate physically consistent and realistic landscape topography and fluvial river bathymetry
PI: Xiaofeng Liu Ph.D., P.E., Associate Professor, Department of Civil and Environmental Engineering, ICDS co-hire
Other senior team member: Roman DiBiase, Ph.D., Associate Professor, Department of Geosciences
The goal of this project is to use AI and ML tools to learn and generate physically consistent and realistic landscape topography and fluvial river bathymetry. Topography and bathymetry are important for both scientific researches such as landscape evolution and engineering applications such as flood modeling. This research hypothesizes that landscape topography and fluvial river bathymetry result from a complex co-evolutionary process involving interactions among terrestrial landscape, river morphology, hydrological processes, and climate factors. Advanced artificial intelligence (AI) and machine learning (ML) methodologies, particularly generative models like diffusion models, can learn the inherent structural relationships embedded in available data and thereby generate synthetic yet physically consistent and realistic topographic and bathymetric datasets. The objectives of the project are: (1) to investigate the inherent structural relationships between topography, river bathymetry, physiography, climate, precipitation, and river discharge. (2) to develop AI and ML models capable of generating synthetic, physically realistic landscape topography and river bathymetry. In time permits, the third objective is to validate synthetic datasets against existing measured data and physically-based hydrological and geomorphological models. The project will employ generative AI models, with particular focus on diffusion models, leveraging their strengths in capturing complex spatial correlations and patterns. The methodology will include: 1. Data Collection and Preprocessing: Compile high-resolution topography data from USGS LiDAR datasets and river bathymetry from existing limited datasets. Supplement these with additional physiographic, climatic, precipitation, and discharge data. 2. Model Training: Train diffusion models on the assembled multi-variable dataset, capturing underlying spatial and temporal dependencies. 3. Synthetic Data Generation: Use the trained models to generate new topographic and bathymetric datasets, conditioned on various climatic, hydrological, and physiographic scenarios. 4. Validation and Physical Consistency Checks: Evaluate generated synthetic datasets through statistical comparison with measured data, along with simulations using established hydrological and geomorphological numerical models. Available datasets include the following:
- Topography: USGS LiDAR digital elevation models (DEMs).- River Bathymetry: Available bathymetric survey data from selected rivers and stream reaches.
- Physiographic Data: Land cover, soil type, geology maps.
- Climatic Data: Historical and projected climate data including temperature, evapotranspiration, and precipitation.
- Hydrological Data: Historical river discharge records and simulated runoff data.
The anticipated outcomes of this research include the development of robust generative AI/ML tools capable of producing realistic and physically consistent synthetic topographic and bathymetric datasets. It will also enhance our understanding of landscape and riverine system evolution influenced by integrated hydrological and geomorphological processes. The AI/ML model can produce comprehensive synthetic datasets to support improved modeling and decision-making in both scientific research and engineering practices, especially in regions lacking detailed measurements. Ultimately, this research is expected to significantly advance predictive capabilities in landscape and river channel morphology, informing better resource management, flood risk assessment, habitat conservation, and infrastructure development.
A list of specific areas of computational and/or data science expertise or skills that the current team is particularly interested in recruiting to support the project: Diffusion models, multimodal model building and training, and general background in AI/ML.
Any other requirements or expectations of potential ICDS Junior Researchers: currently a post-comps graduate student in a related field, such as computer science, IST, or math.
A list of specific objectives for work supported by this call: generating preliminary data to support a NSF proposal
At least one medium to long-term goal: A collaborative proposal to NSF Collaborations in Artificial Intelligence and Geosciences (CAIG) program, which is due February 4, 2026
A short statement (1 sentence to 1 paragraph) explaining the connection of the project to ICDS’s mission: The project falls right into the ICDS mission of advancing AI in research and education.
A paragraph summarizing team member’s recent and/or planned engagement with ICDS: Xiaofeng Liu is a co-hire of ICDS and regularly participate in ICDS activities and committees, such as monthly lunches, faculty search committee, and annual ICDS symposium.