Abstract
Understanding land surface morphology is critical for Earth system science, yet high-resolution global datasets for relief and landforms have been largely unavailable. This paper introduces the Global Relief Classification (GRC), a unique dataset derived from 1-arcsec (~30m) digital elevation models (DEMs). By integrating land surface ontologies and novel algorithms like the Surface Relief Index (SRI), the GRC provides a precise depiction of Earth's terrain. Comparative analyses show that the GRC captures finer spatial disparities than previous datasets, revealing significant topographic diversity in regions like Asia, Peru, and China. The dataset is open-access and designed for integration into hydrological, climatic, and ecological workflows.
Introduction: The Resolution Gap in Geomorphology
While high-resolution satellite imagery is now commonplace, global maps of landforms—the actual "shape" of the ground—have lagged behind. Most existing datasets operate at resolutions of 90m or 1km. At these scales, critical features such as narrow valleys, small ridges, and precise boundaries between plains and mountains are often lost.
Furthermore, traditional geomorphology relies on "moving window" analysis, which calculates relief within a fixed box (e.g., 3x3 pixels). This approach is inherently flawed: a small window cannot capture the full height of a large mountain, while a large window blurs distinct features together. The GRC addresses these issues by treating landforms as discrete objects with morphological integrity.
Methodology: A Hierarchical, Object-Based Approach
The GRC employs a hierarchical classification system divided into two levels:
- Level 1 (L1): A binary classification between Flat Terrain and Rugged Terrain.
- Level 2 (L2): Nine specific classes based on altitude (for flats) and relief intensity (for rugged areas), ranging from "Low-altitude flat terrain" to "Very-high-relief rugged terrain."
To achieve this, the researchers introduced two primary algorithmic innovations:
- Accumulated Slope (AS): Rather than relying on local slope alone, this "cost-distance" approach identifies the boundaries of flat areas by calculating the cumulative slope from a "core" flat region. This distinguishes true plains from small flat benches on mountain slopes.
- Surface Relief Index (SRI): This is the dataset's core innovation. Instead of a fixed window, the system constructs a "base terrain" (a simulated floor) for rugged areas. The SRI measures the height of a point relative to this base, effectively measuring a mountain from its actual foot rather than against its immediate neighbors.
Data Sources and Fusion
The GRC was constructed by fusing several high-resolution Digital Elevation Models (DEMs) to ensure global coverage and accuracy:
- FABDEM: The primary source, a "bare-earth" DEM that removes vegetation and building heights to reveal the true ground surface.
- AW3D30: Used to fill gaps in FABDEM coverage.
- REMA (Reference Elevation Model of Antarctica): Provides data for high-latitude southern polar regions.
- OpenStreetMap (OSM): Utilized as a land mask to separate terrestrial relief from marine environments.
Key Results and Global Insights
The transition from pixel-based to object-based analysis has yielded several significant findings:
- Asia’s Ruggedness: The GRC reveals that Asia is the most topographically diverse continent. Only 59% of Asia is classified as flat terrain—the lowest percentage of any continent—highlighting its massive, continuous mountain systems.
- Diversity Hotspots: Peru and China emerged as global leaders in relief diversity, containing nearly the full spectrum of the nine L2 relief classes within their borders.
- Hydrological Alignment: A striking correlation was found between GRC boundaries and surface runoff patterns, suggesting the dataset can serve as a functional map for predicting water movement across landscapes.
- Improved Precision: Validation against older 90m and 1km datasets shows that the GRC more accurately represents continuous valley systems and sharp mountain-plain transitions that were previously fragmented.
Limitations and Future Work
While the use of "bare-earth" DEMs like FABDEM is a major advancement, some noise from extremely dense forests or urban centers may persist. Additionally, defining boundaries in highly fragmented archipelagos currently requires some manual human-computer interaction. Future research will leverage this 30m baseline to explore the relationships between land morphology, biodiversity, and microclimates in greater detail.
Data Availability
The full GRC dataset is open-access and available in both raster (GeoTIFF) and vector formats via the Deep-time Digital Earth Geomorphology platform and Zenodo.
Summary updated