Ground Truth vs Satellite: Why Field Plots Still Matter

Updated 2026-11-02 · 4 min read · By the akar.earth specialist team

Ground truth vs satellite carbon data: what satellites can and cannot prove, why field plots still anchor credits, and how to combine both efficiently.

Satellites are excellent at telling you where forest cover is and when it changed, but they do not weigh trees, so field plots remain the anchor for carbon stocks and for validating every remote sensing product. The most credible projects use satellites for coverage and frequency, and ground truth for accuracy, with each checking the other.

What each source is good at

Neither column wins. The comparison of field plots and remote sensing goes deeper.

Why methodologies still require plots

Standards and integrity frameworks reward conservative, verifiable quantification. Remote sensing models are statistical predictions; their errors can be large and biased in particular forest types. Field plots supply:

• Calibration. Models need measured biomass to learn from. See remote sensing biomass.
• Validation. Independent plots show whether a map is biased.
• Carbon pools satellites miss. Below-ground biomass, dead wood, litter and soil.
• Species and wood density, which drive biomass as much as size does.

Verra’s VM0047 shows the current balance. Its area-based approach uses remotely sensed project and control plots to build a dynamic performance benchmark, then uses plot sampling to quantify stocks. See the VM0047 area-based approach.

How to combine them well

1. Map first. Use satellite data to stratify the area by cover, canopy height and change history.
2. Allocate plots by stratum, weighting toward variable and high-carbon strata. See plot sampling design.
3. Locate plots precisely with RTK GNSS so they match pixels.
4. Measure at the same time as imagery, or adjust for growth and disturbance between dates.
5. Train and validate models with separate plot sets.
6. Monitor change with satellites between field campaigns, and send crews to check flagged changes.
7. Reconcile. Field-observed losses should appear in alerts, and alerts should be checked on the ground.

Ground truth is more than biomass plots

Ground truth also includes:

• Land-cover reference points for classification accuracy.
• Visits to deforestation alerts to confirm cause and extent.
• Photos and interviews documenting land use history for land eligibility.
• Farmer-confirmed plot boundaries for smallholder projects and EUDR.

Budgeting the balance

Teams often ask how to split money between imagery and fieldwork. Free Sentinel and Landsat data make satellite coverage cheap; the costly parts are analyst time and field campaigns. A useful rule of thumb is to spend field effort where remote sensing is weakest: high-biomass strata, mixed agroforestry, peat and mangroves, and areas with frequent alerts. Use satellites to reduce the number of plots through stratification, not to remove them. Recheck the balance after each monitoring period, since better models may allow fewer plots in stable strata.

Common mistakes

• Satellite-only claims. Marketing “100% satellite MRV” usually means stocks rest on someone else’s global map with unknown local error.
• Plots only where access is easy, which biases the model.
• Using the same plots to train and validate.
• Ignoring time lag. A plot measured two years after the image in a fast-growing ARR site is not a match.
• Treating ground truth as a one-off. Models drift as vegetation changes; recalibrate each monitoring period.

What good evidence looks like

For a VVB, the strongest package shows how satellites and plots fit together: the stratification map, the plot design, the plot data with coordinates, the remote sensing inputs and code, the calibration and validation statistics, and the uncertainty budget. Field photos, tree tags and QA/QC remeasurements prove plots exist. Alert verification logs show that change detection was checked.

The Indonesian case

Cloud, haze, steep terrain and peat swamp forest make remote sensing harder in Indonesia than in many regions. Smallholder mosaics in Sumatra, Java and Sulawesi mix trees and crops at sub-pixel scale. These conditions make ground truth more valuable, not less. Field-verified data is also what links carbon work to EUDR evidence, so one dataset can serve both; see one dataset for carbon and EUDR.

We run RTK-surveyed plot campaigns and link them to satellite models through our field data campaigns service. For the full system, visit the dMRV hub.