Boomitra Grassland Restoration in East Africa Through Soil Enrichment

Audit Analysis

This Kenyan grassland soil-carbon project (VCS-3340) uses a novel remote-sensing AI model for soil carbon measurement, which introduces significant methodological uncertainty. While additionality is VVB-confirmed and a 15% buffer pool is in place, the validation process surfaced 34 clarifications, 28 corrective actions, and three Forward Action Requests questioning the AI model's accuracy, bias correction, and temporal alignment of satellite data. The project's claim safety is further weakened by a project-specific baseline, an unquantified CORSIA/CCP status, and a contradiction in the claimed ERR figure across documents.

Project Description

Red Flags

• ⚠FAR#03 (validation report) specifically questions the accuracy, robustness, and uncertainty quantification of the remote-sensing AI model used to estimate soil carbon — a core methodological risk for a soil-carbon project
• ⚠Corrective actions require a detailed bias-correction plan for ERR accounting (RothC vs. remote sensing) and a revised model plan to mitigate overestimation from temporal misalignment of satellite data, suggesting the 93,371 tCO₂e figure may be inflated
• ⚠The claimed ERR of 93,371 tCO₂e in the validation report contradicts a figure of 109,885 tCO₂e in an undated, unidentified document, and the source of the lower figure is untraceable
• ⚠Low extraction confidence across the document set means at least one key document was poorly readable, limiting the reliability of the extracted record
• ⚠CORSIA eligibility and CCP status are not stated in any available document, leaving dual-channel risk unassessed

Credit Vintages

No issuances recorded on the registry.