Five years ago, monitoring a 10,000-hectare REDD+ project at quarterly cadence required a budget that priced out most jurisdictional programmes. Today the data layer is structurally different. Sentinel-2 (10 m optical, 5-day revisit at the equator), Sentinel-1 (C-band SAR, all-weather, ~6-day revisit), and Landsat 8/9 (30 m, 16-day revisit) are free under Copernicus and USGS open-data policies. ESA's Copernicus Browser and AWS Open Data make full-archive access routine; cloud-native COGs and STAC catalogs have removed most of the data-engineering tax.
What costs money now
The cost stack for production-grade satellite MRV has inverted. Imagery is approximately zero marginal cost. Compute on Sentinel Hub, Microsoft Planetary Computer, or Google Earth Engine sits in the cents-per-scene range. What costs money is:
Methodology calibration. Translating a wall-to-wall classification into a creditable activity-data layer requires methodology-aligned definitions of forest, deforestation, degradation, and stratification. Under VCS VM0048 (the consolidated REDD methodology released in 2023) and the associated VMD0055 module for unplanned deforestation, allocated baselines are recalculated every six years against a jurisdictional or regional reference, which requires sustained calibration work, not a one-time map.
Ground-truth networks. Sample plots, biomass equations, and field validation are still the dominant unit-cost line in nature-based MRV. The Sustainability Atlas 2026 cost benchmark places professional EO platform fees in the order of USD 1–3/ha/year for large landscapes, but the all-in MRV cost — including ground plots, allometric equations, and verification fieldwork — typically remains an order of magnitude higher [1]. The Carbon Balance and Management study on REDD+ in Tanzania remains a useful baseline: monitoring costs scale strongly with required uncertainty bounds, not project area [2].
Model-ops engineering. An MRV pipeline that must run for a 10–30-year crediting period needs the same operational discipline as any production data system: schema versioning, reproducible training datasets, drift monitoring, and verifier-grade audit logs. This is the bucket most underestimated by first-time developers.
Implications for project developers and hosts
The economic question is no longer 'can we afford satellite MRV?' but 'do we have the institutional capacity to operate it credibly over a multi-decade crediting period?' Jurisdictional programmes that build internal capacity around free Copernicus and Landsat data are running monitoring at a fraction of the cost of pre-2020 vendor contracts; programmes that outsource the entire pipeline to a single vendor remain exposed to platform-lock-in and methodology-version risk.
Mozambique's 2025 REDD+ Technical Annex 2 submission, covering results-based payments for 2017–2020, illustrates the institutional bar: a national MRV system with documented activity data, emission factors, and uncertainty quantification submitted to the UNFCCC Technical Assessment process [3].
Where CAS fits in
CAS designs and implements the engineering layer that turns open EO data into validator-ready MRV. Specifically: (a) Sentinel-1/-2 and Landsat pipelines aligned to VCS VM0048 / VMD0055, ART TREES, or jurisdictional REDD+ requirements; (b) stratification and sampling designs that meet the uncertainty deductions in the chosen methodology; (c) model-ops infrastructure with methodology-version control and reproducible re-runs across the crediting period; and (d) handover packages so the host country or project operator can run the system in-house. We do not act as a Validation and Verification Body and we do not issue credits.
Sources
[1] Sustainability Atlas, 'Earth observation for climate analytics costs in 2026'.
[2] Bottcher et al., 'An assessment of monitoring requirements and costs of Reduced Emissions from Deforestation and Degradation', Carbon Balance and Management.
[3] Republic of Mozambique, 'REDD+ Technical Annex 2 — Modified Submission', UNFCCC, May 2025.