Mapping Earth’s Hidden Scars: My Journey Contributing to a Global Gully Erosion Dataset

Over the past several months, I’ve been part of something bigger than any single project I’ve worked on—a global, scientist-led push to map how landscapes are eroding. Under the Global Soil Erosion Mapping Initiative, coordinated by Dr. Pasquale Borrelli and the EUSO Soil Erosion Working Group, nearly 300 researchers came together to digitize the world’s hidden geomorphic features: gullies, terraces, and badlands.

I contributed hands-on to Phase 1, focusing on gully mapping using high-resolution imagery and a harmonized protocol designed for consistency at continental scales. Our collective dataset just crossed 515,709 mapped features, and I’m incredibly proud that my work is part of this milestone.

Why Gully Erosion and Why Now?

Gullies carve deep, lasting scars into croplands and rangelands. They accelerate soil loss, reduce productivity, damage infrastructure, and alter hydrology and sediment connectivity. Yet globally, consistent, high-quality gully inventories remain rare, fragmented, or isolated to small study areas.

This initiative changes that by building a standardized, high-coverage, global gully dataset that researchers and policymakers can use to understand where, when, and why erosion happens, and how to respond.

What We Achieved in Phase 1

Between late 2024 and early 2025, the consortium received 165 mapping contributions. After harmonization, duplicate removal, and quality checks, the Phase 1 compilation contains:

  • 482,545 gully/gully-head points
  • 28,087 terrace mapping boxes
  • 5,077 badland mapping boxes
  • 515,709 total features so far

The descriptive stats per dataset are revealing: mean ~3,346, median ~2,631, with a min–max of 24–14,847 and IQR ~2,363. Some small submissions likely reflect limited study extents or partial sets; extremely large counts (>10,000 gullies) are under secondary review. A statistical validation step is now underway; each of us will map 50 further sites for quality assessment and calibration against initial patterns.

My Contributions

Here’s how I contributed and collaborated within the consortium:

  • Digitization & Feature Consistency
    I digitized gully heads and channels across assigned tiles using a consistent 250–300 m site protocol. I prioritized clean geometry, precise feature placement, and meticulous attribute coding (gully type/width/depth when observable).

  • Quality & Harmonization Mindset
    I adhered to the consortium’s harmonized schema, field names, units, codes, and spatial precision so my contribution could be instantly merged without downstream friction.

  • Methodological Rigor from UAV/SAR Experience
    Drawing on my background in multispectral/hyperspectral/LiDAR/thermal UAV campaigns and Sentinel‑1 SAR water/land segmentation, I cross-referenced micro‑topographic cues, shadow geometry, texture, and vegetation breaks in imagery to reduce false positives, especially in drylands and post-tillage landscapes where linear features can mislead.

  • Collaborative Feedback Loop
    I participated as a constructive reviewer, flagging edge cases, submitting notes on ambiguous terrace–gully transitions, and advocating for context-specific imagery checks (e.g., seasonal stacks or slope context at 10–30 m DEM) in tricky terrain.

  • Transparent Metadata & Reproducibility
    I documented source imagery, dates, visibility conditions, and disambiguation choices (e.g., ephemeral vs. permanent indicators) so my tiles can be retraced and verified by future analysts or ML training pipelines.

(If you want, I can insert the exact countries/regions you mapped, your total site count, and a map figure—just share the specifics or tiles list.)

How I Mapped (My Workflow)

  1. Tiling & Tasking
    I worked on assigned tiles aligned with cropland/grassland target areas, ensuring even spatial representation.

  2. Imagery & Context
    I used high‑resolution orthoimagery as the primary source. Where needed, I reviewed DEM‑derived slope/curvature and multitemporal scenes to distinguish ephemeral rills from stable channels.

  3. Feature Identification
    I prioritized gully heads and continuous channels, avoiding ephemeral surface marks from tillage or drainage channels unless they showed persistent incision and banks.

  4. Attribute Consistency
    I followed the common attributes for direction, mean length, width, and depth, which were always visible within the 250–300 m site logic.

  5. Error Awareness
    Guided by lessons from GE‑LUCAS 2018 & 2022, I kept an eye out for common pitfalls (commission/omission risks, landscape context, seasonality), noting uncertain cases for follow‑up validation.

Collaboration That Scales

One of the most inspiring aspects of this initiative is how standardized micro‑tasks add up to macro‑scale insight. Every contributor’s adoption of shared symbology, coding, and QC means we can build a global‑ready, ML‑ready dataset. That’s how we’ll enable the next steps:

  • Paper #1 (planned early 2026): A global database describing the spatial pattern of mapped gully/terrace/badland features
  • Paper #2 (planned early 2026): Machine‑learning interpolation to predict erosion features beyond mapped sites

I’m excited to continue through the 50‑site validation round, help debug outliers, and strengthen the interpretability of the forthcoming models.

Why This Matters Beyond Academia

This isn’t just a map; it’s actionable intelligence. With reliable inventories, we can:

  • Target conservation where gullies actively expand into cropland
  • Plan nature‑based solutions in vulnerable basins
  • Quantify policy impact (e.g., tillage practices, buffer strips, controlled drainage)
  • Improve flood and sediment risk models for downstream communities

For me, this work bridges my academic research in river processes and remote sensing with real-world soil and water conservation from watershed management to resilient agriculture.

Acknowledgments

Deep gratitude to Dr. Pasquale Borrelli and the EUSO Soil Erosion Working Group for spearheading a complex, inclusive, and rigorous collaboration. And to the 295+ scientists whose contributions, like mine, are helping turn scattered traces into a coherent, global picture of geomorphic change.

Comments

Post a Comment

Popular Posts