Research Proposal · Remote Sensing × Soil Science × Machine Learning

Predicting Soil Carbon Sequestration Under Regenerative Management Using Remote Sensing and Machine Learning by Aldi Airori

Integrating 3-year in situ SOC fractions, GHG flux, and soil nitrogen data with Sentinel time series to project long-term carbon accumulation under cover crop × cattle grazing rotations in Nebraska.

3 Years of field data

Nebraska ENREEC Study area

5+ Soil parameters

20 yr Projection horizon

Research Gap

Why This Project Matters

Current MRV protocols rely on process-based modeling, not measured SOC dynamics. There is no validated framework integrating SOC fractions, GHG flux, and RS time series for long-term carbon projection under regenerative management.

In Situ Data Assets · 3 Years

Field Measurements

SOC fractions (POM & MAOM)
GHG flux — CO₂, N₂O, CH₄
Total nitrogen & mineral N
Soil microbial biomass C & N
Bulk density, pH, texture, EC
Cover crop biomass & grazing records

Remote Sensing Inputs

Satellite Covariates

Sentinel-2 MSI

Sentinel-1 SAR

Landsat TIR

ERA5 Climate

Bare soil composites for SOC spectral signal
NDVI / NDRE / EVI phenology curves
SAR VV/VH for moisture & tillage

Methodology

RS + ML Data Fusion Pipeline

In Situ Data

SOC fractions · GHG · N · Microbial

›

RS Extraction

Google Earth Engine · Temporal alignment

›

ML Model
Random Forest · GPR · LSTM

›

Calibration

Fraction-level validation · Feature importance

›

Projection

5 · 10 · 20 yr scenarios + uncertainty bounds

Novel contribution: First framework to predict SOC fraction dynamics (POM vs MAOM) from multispectral RS time series under known management history — directly addressing the permanence assessment gap in carbon credit protocols.

Expected Outputs

Output 1 · Interactive Map

GitHub Pages Visualization

Animated SOC change layers (year-by-year)
GHG flux hotspot heatmap overlay
Treatment zone polygons (cover crop vs grazing)
Scenario toggle: 5 / 10 / 20 year projections
Built with Python Folium → hosted free

Output 2 · Research Article

Full Research Paper (6–8k words)

RS + ML fusion methodology (novel framework)
SOC fraction-level predictions vs bulk SOC
GHG flux patterns under regenerative management
Long-term carbon accumulation projections
Implications for MRV protocol design

Geoderma

SOIL

AEE

RSE

Project Timeline · 8 Months

Mo 1–2

GEE RS extraction & temporal alignment

Mo 2–3

Exploratory analysis & feature engineering

Mo 3–4

ML model training & validation

Mo 4–5

Projection scenarios & uncertainty

Mo 5–6

Folium map build & GitHub deployment

Mo 6–7

Paper writing & figure prep

Mo 7–8

Review & journal submission

→ View Supporting Literatures