Fluvial Erosion–Deposition Generative Model (FED-GM): A Process-Driven Framework for Three-Dimensional Terrain Generation Through Sediment Transport and Geomorphic Feedback
Abstract
Three-dimensional terrain generation has been investigated through procedural modeling, physically based simulation, neural representations, and generative learning. However, the representation of a landscape and the process through which that landscape forms are conceptually distinct. This paper proposes the Fluvial Erosion–Deposition Generative Model (FED-GM), a process-driven framework for generating three-dimensional terrain through the coupled evolution of water flow, sediment transport, spatial erosion, sediment deposition, and terrain-dependent feedback. Rather than treating a final terrain as a static geometric object, FED-GM treats the terrain as an evolving spatial state whose geometry influences material transport and is subsequently modified by that transport. The proposed framework combines a terrain representation, a hydrological state, a sediment state, a transport-and-redistribution operator, and a feedback-conditioned generation mechanism. Material removed from one region is tracked as transported sediment and may be deposited elsewhere, subject to explicit mass-balance constraints and boundary conditions. The resulting evolution can produce candidate representations of river valleys, incised channels, alluvial fans, sandbars, floodplains, and depositional surfaces. The framework can be implemented using height fields, voxel grids, implicit surfaces, point clouds, or neural three-dimensional representations, and can be combined with learned generative priors when training data are available. This paper establishes the conceptual architecture, governing equations, optimization objectives, and an evaluation protocol for FED-GM. It does not claim completed experiments or demonstrated superiority over existing terrain-generation methods. The principal research hypothesis is that explicitly coupling material provenance, spatial redistribution, and terrain-dependent feedback within a controllable generation process can improve the physical consistency, controllability, and interpretability of generated fluvial landscapes.
Related
- Layerwise Deposition-Field Generative Modeling: A Process-Conditioned Framework for Three-Dimensional Generation Through Thin-Film Deposition and Surface Growth
- Deposition-Based 3D Generative Modeling: A Framework for Three-Dimensional Object Generation Through Gas-Phase Particle Deposition, Local Adsorption, and Branching Growth
- Granular Flow and Deposition Generative Modeling: A Physics-Driven Framework for Three-Dimensional Shape Generation Through Discrete Particle Dynamic
- Reaction-Driven Three-Dimensional Generative Computing: A Framework for Constructing 3D Objects Through Local Chemical Transformation, Deposition, and Structural Growth
- Collective Biological Construction: A Multi-Unit Generative Framework for Three-Dimensional Shape Formation Through Local Deposition and Collective Structural Evolution
BibTeX
@misc{zhang2026fluvial,
title = {{Fluvial Erosion–Deposition Generative Model (FED-GM): A Process-Driven Framework for Three-Dimensional Terrain Generation Through Sediment Transport and Geomorphic Feedback}},
author = {Jincheng Zhang},
howpublished = {Zenodo},
year = {2026},
month = oct,
doi = {10.5281/zenodo.23635945},
url = {https://zenodo.org/records/23635945},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - Fluvial Erosion–Deposition Generative Model (FED-GM): A Process-Driven Framework for Three-Dimensional Terrain Generation Through Sediment Transport and Geomorphic Feedback
AU - Jincheng Zhang
PY - 2026
DA - 2026/10/25
PB - Zenodo
DO - 10.5281/zenodo.23635945
UR - https://zenodo.org/records/23635945
AB - Three-dimensional terrain generation has been investigated through procedural modeling, physically based simulation, neural representations, and generative learning. However, the representation of a landscape and the process through which that landscape forms are conceptually distinct. This paper proposes the Fluvial Erosion–Deposition Generative Model (FED-GM), a process-driven framework for generating three-dimensional terrain through the coupled evolution of water flow, sediment transport, spatial erosion, sediment deposition, and terrain-dependent feedback. Rather than treating a final terrain as a static geometric object, FED-GM treats the terrain as an evolving spatial state whose geometry influences material transport and is subsequently modified by that transport. The proposed framework combines a terrain representation, a hydrological state, a sediment state, a transport-and-redistribution operator, and a feedback-conditioned generation mechanism. Material removed from one region is tracked as transported sediment and may be deposited elsewhere, subject to explicit mass-balance constraints and boundary conditions. The resulting evolution can produce candidate representations of river valleys, incised channels, alluvial fans, sandbars, floodplains, and depositional surfaces. The framework can be implemented using height fields, voxel grids, implicit surfaces, point clouds, or neural three-dimensional representations, and can be combined with learned generative priors when training data are available. This paper establishes the conceptual architecture, governing equations, optimization objectives, and an evaluation protocol for FED-GM. It does not claim completed experiments or demonstrated superiority over existing terrain-generation methods. The principal research hypothesis is that explicitly coupling material provenance, spatial redistribution, and terrain-dependent feedback within a controllable generation process can improve the physical consistency, controllability, and interpretability of generated fluvial landscapes.
ER -