VELSTROM Datasets
Proprietary datasets produced by VELSTROM's research team. These datasets combine laboratory measurements with satellite-derived environmental features for carbon science and environmental intelligence.
Soil Carbon
VELSTROM's proprietary soil carbon dataset combining field laboratory analysis results with satellite-derived environmental features. Designed for soil organic carbon (SOC) estimation, carbon stock mapping, and MRV (Monitoring, Reporting, Verification) workflows.
Coverage
India — phased expansion to South and Southeast Asia
Version
v0.2-alpha
Schema
Lab: SOC (%), bulk density, texture, pH, EC | Environmental: NDVI, EVI, LST, CHIRPS precipitation, DEM-derived terrain variables
Methodology
Stratified random sampling with soil cores at 0–30 cm and 30–100 cm depths. Lab analysis via dry combustion (Walkley-Black). Environmental features extracted at point locations from Sentinel-2, ERA5, and SRTM.
Citation
VELSTROM Research (2026). Soil Carbon Dataset. Internal publication.
Carbon Benchmarks
Curated benchmark dataset for evaluating soil and ecosystem carbon models. Includes standardised train/test splits and evaluation metrics for reproducible comparison of SOC prediction methods.
Coverage
Multi-regional benchmarks — expanding
Version
v0.1-draft
Schema
Benchmark splits, ground truth labels, model evaluation metrics (RMSE, R², MAE, bias)
Methodology
Working towards a standardised benchmark protocol. Will follow ML-ready conventions with spatial cross-validation to prevent data leakage.
Citation
VELSTROM Research (2026). Carbon Benchmarks. Forthcoming.
Environmental Intelligence
Composite dataset combining multiple environmental indices (NDVI, NDWI, SAVI, EVI, NBR2, etc.) with climate and terrain variables into a unified, harmonised feature set for machine learning and analytics.
Coverage
India — select pilot regions
Version
v0.1-alpha
Schema
Multi-temporal index composites, percentile features, seasonal statistics, terrain derivatives
Methodology
Temporal compositing of Sentinel-2 indices with cloud masking. Statistical aggregation over user-defined time windows. Normalisation and quality flagging.
Citation
VELSTROM Research (2026). Environmental Intelligence Dataset. Internal publication.