EarthScape: A Multimodal Dataset for Surficial Geologic Mapping and Earth Surface Analysis

Matthew A. Massey, Nusrat Munia, Abdullah Imran

Keywords: Datasets and Benchmarks, Multimodal Learning, Multimodal Fusion, Remote Sensing, Geospatial Analysis, Earth Science, Surficial Geology

Abstract: Surficial geologic (SG) maps are essential for understanding surface processes and supporting infrastructure planning, but current workflows are labor-intensive and difficult to scale. We introduce EarthScape, an AI-ready multimodal dataset for SG mapping that integrates digital elevation models, aerial imagery, multi-scale terrain features, and hydrologic and infrastructure vector data within a unified, reproducible pipeline. We report baseline benchmarks across single-modality, multi-scale, and multimodal configurations. In our experiments, terrain-derived features provide the most reliable predictive signal, while spectral inputs and raw elevation degrade substantially under cross-region evaluation. Cross-generalization and multimodal fusion remain challenging, underscoring the need for models that capture shape-driven surface processes. EarthScape offers a geographically compact but modality-rich benchmark for multimodal fusion, domain adaptation, and surface-process modeling.

Primary Area: datasets and benchmarks

Submission Number: 13767