Restoration projects generate data everywhere. Decisions get made nowhere.
A satellite image says the land is degraded pasture. A drone orthophoto reveals a gully that the satellite missed. A soil sample shows pH 4.3 and severe compaction at 30 centimeters. A climate model projects declining suitability for half the planned species by 2070. Each source tells part of the story. None tells enough to act on alone.
The problem is not a lack of data. It is the gap between having data and having a basis for decision.
What each source sees - and what it misses
Satellite imagery provides the macro view. Slope gradients across the entire property. Vegetation cover classified by type - pasture, degraded, secondary growth, native forest fragments. Fire history over the past decade. Seasonal rainfall patterns. Elevation models. All of this at 10-meter resolution, updated every few days, covering any site on Earth.
What satellite cannot see: the pH of the soil. Whether the ground is compacted at 30 centimeters. Which invasive species are present and how deep their roots go. Whether the existing vegetation fragments contain species worth preserving or are dominated by exotics. The satellite sees the surface. The decisions that determine success or failure happen below it.
Drone orthophotos close the resolution gap. At 1-3 centimeter resolution, a drone survey captures erosion channels, individual tree crowns, standing water, access paths, and micro-terrain features that satellites cannot resolve. Multispectral sensors add NDVI and NDRE indices at plant-level precision.
What drones miss: they capture a single moment. A drone flight in the dry season shows one reality; the same area in the wet season may be impassable. Drones also cannot penetrate the canopy or the soil surface. And drone surveys are expensive enough that you cannot fly every site every month - you fly when you need a high-resolution snapshot to validate or contradict what the satellite suggested.
Field surveys deliver the ground truth that no remote sensor can provide. Soil chemistry from laboratory analysis: pH, nitrogen, phosphorus, potassium, cation exchange capacity, organic matter content. Penetrometer readings that reveal compaction layers invisible from above. Floristic inventories that identify every species present, native or exotic, and assess their condition. Soil moisture profiles. Root depth observations.
What field data misses: coverage. A field team can visit 12-15 measurement plots across an 8,000-hectare site in a two-week campaign. That gives precise data at each plot, but the space between plots is interpolated, not measured. Field surveys are slow, labor-intensive, and weather-dependent. They are essential precisely because they answer questions that nothing else can - but they cannot cover everything.
Models project what has not yet happened. Climate trajectories under different emission scenarios. Carbon sequestration curves based on allometric equations and species growth rates. Species suitability ranges for 2040, 2070, 2100. Revenue projections under different carbon pricing assumptions.
What models miss: they are hypotheses, not measurements. Every model rests on assumptions - about growth rates, mortality, climate sensitivity, market conditions. A carbon projection is only as good as the baseline it starts from and the field data that validates its assumptions. Models without ground truth are speculation.
When sources contradict each other
The most valuable moment in a diagnostic is when two sources disagree.
Satellite analysis classifies a zone as operable - moderate slope, adequate rainfall, no legal constraints. Then the field team arrives and finds laterite compaction at 25 centimeters that makes conventional soil preparation impossible without deep ripping. The satellite was right about the surface. The field data reveals a subsurface reality that changes the entire treatment prescription.
Or: a drone orthophoto shows what appears to be healthy vegetation cover across a riparian buffer. But the floristic inventory identifies that 80% of the cover is Leucaena leucocephala - an aggressive exotic species that will outcompete any planted natives unless removed first. The drone saw green. The botanist saw a problem.
"Each data source has a confidence interval. Satellite gives you the big picture with known blind spots. Drone gives you precision at a single point in time. Field gives you depth at sample points. Models give you the future with stated assumptions. The value is not in any single source - it is in knowing exactly what each one can and cannot tell you." - Jérémy Giral, Platform & Data Engineer, MORFO
These contradictions are not failures. They are the diagnostic working as intended. A system that only confirms what the satellite already showed adds no information. The entire point of layering four sources is to catch what any single source would miss.
The carbon market is learning this the hard way. Canopy maps built on unprocessed satellite data. Height models riddled with cloud artifacts. Biomass layers that have nothing to do with what is actually growing on the ground. The interfaces look credible - polished dashboards, impressive maps. But when the underlying data has not been validated against field reality, the entire crediting chain inherits the error. Beautiful maps are not the same as good science.
Confidence, not certainty
Every data point in a restoration plan carries a source. Satellite-derived estimates are tagged as such. Field-validated measurements carry a different confidence level. Modeled projections are explicit about which assumptions they rest on.
This matters because different stakeholders need different levels of assurance. A field team needs to know which zones are ready for soil preparation - that requires field-validated compaction data, not satellite estimates. An investor evaluating carbon projections needs to know whether the baseline was measured in the field or extrapolated from regional averages. An auditor preparing a verification needs traceable evidence: what was measured, by whom, when, with what method.
"The terrain always has the last word. We can model, we can fly drones, we can analyze satellite time series. But the soil sample, the penetrometer reading, the species count in the field plot - that is what converts a hypothesis into a prescription." - Rebecca Montemagni Almeida, Senior Forest Ecosystem Engineer, MORFO
The goal is not to eliminate uncertainty. It is to make uncertainty visible, traceable, and bounded. When someone asks "how do you know this zone is operable?" the answer should not be "we think so." It should be: "satellite slope analysis (SAT, 10m resolution, March 2026) shows 78% below 15 degrees. Field penetrometer readings (FLD, 14 plots, December 2025) confirm workable compaction above 30cm. Drone orthophoto (DRN, 2cm resolution, November 2025) shows no erosion channels in the mechanization corridors."
Four sources. One decision. Every link in the chain visible.
