Skip to content

OpenClimateScience/M4-Open-Science-for-Soil-Health

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

20 Commits
 
 
 
 
 
 
 
 

Repository files navigation

DOI

OpenClimateScience M4 – Open Science for Soil Health

High-Resolution Soil Organic Carbon Mapping (Algeria)

Project Overview

Global soil products such as SoilGrids provide invaluable information, but their spatial resolution (≈250 m) is often too coarse for local agricultural decision-making. Farmers and land managers operate at much finer spatial scales.

This project demonstrates an open-science, reproducible workflow to generate high-resolution (30 m) Soil Organic Carbon (SOC) maps by downscaling coarse global soil data using high-resolution satellite imagery and terrain information.

The workflow is implemented as a Google Colab notebook and is designed for:

  • teaching and training
  • rapid prototyping
  • transparent, cloud-native geospatial analysis

Objectives

  • Access cloud-hosted geospatial data (Landsat, NASADEM, SoilGrids)
  • Engineer environmental predictors relevant to soil processes
  • Downscale coarse SOC data using machine learning
  • Produce a fine-resolution SOC map with uncertainty estimates
  • Demonstrate open, reproducible geospatial science

Study Area

  • Region: Tell Atlas, Northern Algeria
  • Extent: ~10 × 10 km agricultural window
  • Spatial Resolution: 30 m (output)

Methodology (Conceptual)

  1. Target Variable (Teacher)

    • SoilGrids SOC stock (0–30 cm depth, ~250 m)
  2. Predictors (Students)

    • Vegetation indices from Landsat:
      • MSAVI (Modified Soil-Adjusted Vegetation Index)
      • Bare Soil Index (BSI)
    • Topographic variables:
      • Elevation (NASADEM)
      • Slope
  3. Downscaling Strategy

    • Resample coarse SOC to the 30 m grid
    • Learn relationships between SOC and environmental predictors
    • Predict SOC continuously at high resolution
  4. Model

    • Random Forest Regressor (scikit-learn)
  5. Uncertainty Quantification

    • Ensemble approach (multiple RF models)
    • Pixel-wise standard deviation as uncertainty proxy

Notebook

The full workflow is implemented in the notebook: https://github.com/OpenClimateScience/Module_4_Open_Science_for_Soil_Health_FR/blob/main/Module_4_Open_Science_for_Soil_Health_FR.ipynb

About

Lesson materials for Module 4, "Open Science for Soil Health"

Resources

License

Stars

Watchers

Forks

Packages

 
 
 

Contributors