Measuring carbon sequestration requires tracking how much carbon dioxide an ecosystem pulls from the atmosphere and stores over time. There’s no single method. The right approach depends on where the carbon is stored: in trees, soil, coastal sediments, or across an entire landscape. Most projects combine direct field measurements with remote sensing and modeling to get reliable estimates.
Measuring Carbon in Trees
Forests store the bulk of their carbon in aboveground biomass: trunks, branches, and leaves. The standard field method involves measuring individual trees and then plugging those measurements into mathematical formulas called allometric equations that convert tree dimensions into biomass estimates. The three key measurements are stem diameter (in centimeters), tree height (in meters), and crown diameter (in meters). Of these, stem diameter at breast height is the most commonly collected and the most predictive.
The widely used equation from Chave et al. (2014) estimates aboveground biomass using diameter, height, and wood density. You multiply the diameter squared by height and wood density, raise the result to the power of 0.976, and apply a correction factor. The output is dry biomass in kilograms per tree. Since roughly 47 to 50 percent of dry biomass is carbon, you multiply by that fraction to get the carbon stored in each tree. Scale that across a measured plot, then across the full forest area, and you have a carbon stock estimate.
For large areas where measuring every tree isn’t practical, newer equations let you estimate biomass from height and crown diameter alone, which are the two measurements that satellites and aerial sensors can capture. These equations account for the fact that flowering trees (angiosperms) and cone-bearing trees (gymnosperms) scale differently. This opens the door to estimating forest carbon from above without setting foot on the ground, though ground-truth plots remain essential for calibrating those estimates.
Measuring Carbon in Soil
Soil holds more carbon globally than the atmosphere and all plant life combined, but measuring it is slower, more expensive, and more variable. The USDA’s standard protocol calls for collecting soil cores at four depth increments: 0 to 15 cm, 15 to 30 cm, 30 to 60 cm, and 60 to 100 cm. The probe should be at least 100 cm long and 5 cm in diameter.
At each depth, you need two things: the concentration of organic carbon in the soil and the bulk density (how much a given volume of soil weighs when oven-dried). Labs determine carbon concentration through dry combustion, which burns the sample at high temperature and measures the CO2 released. The formula for soil organic carbon stock at each layer is:
SOC stock = bulk density × carbon concentration × layer thickness × (1 minus the fraction of rocks and gravel)
You then add up all four layers to get the total carbon stock per unit area, typically expressed in metric tons of carbon per hectare. If you’re excavating rather than coring, bulk density can be measured using a water or foam displacement method, but whatever technique you choose must stay consistent across all future sampling rounds. Consistency matters because the changes you’re trying to detect, often just a few tons per hectare per year, are small relative to the total stock. Replicated field experiments need to run for at least several years before meaningful differences emerge between management practices.
Measuring Carbon Exchange in Real Time
Eddy covariance flux towers measure the actual flow of CO2 between an ecosystem and the atmosphere, giving you a near-continuous reading of whether a landscape is gaining or losing carbon. The technique works by pairing a fast-response wind sensor with a CO2 analyzer mounted on a tower above the canopy. Together, these instruments track tiny updrafts and downdrafts of air and the CO2 concentration in each, calculating the net flux dozens of times per second.
This produces a measurement called net ecosystem exchange: the balance between carbon absorbed through photosynthesis and carbon released through respiration by plants, microbes, and animals. A negative value means the ecosystem is pulling in more carbon than it’s releasing. Flux towers are the gold standard for understanding how carbon uptake responds to weather, seasons, drought, and land management over time. Networks like FLUXNET operate hundreds of these towers worldwide, providing continuous data that helps validate satellite-based estimates.
Measuring Blue Carbon in Coastal Ecosystems
Mangroves, seagrass meadows, and salt marshes store enormous quantities of carbon in their sediments, often far more per unit area than terrestrial forests. Measuring this “blue carbon” requires sediment coring, similar in concept to soil sampling but with additional steps to account for the unique dynamics of coastal environments.
Accurate seagrass carbon burial rates, for example, require five steps: first, determine the sediment accumulation rate below the surface mixed layer using a naturally occurring radioisotope (lead-210) and porosity measurements. Second, measure the concentration of organic carbon in the buried sediment. Third, multiply the accumulation rate by the carbon concentration. Fourth, if you want to isolate carbon that came specifically from the seagrass itself rather than material that drifted in from elsewhere, use chemical biomarkers to determine that proportion. Fifth, account for any carbonate minerals formed within the meadow, which can actually release CO2 during formation and offset some of the sequestration benefit. Compaction during coring is a known source of error and must be carefully measured and reported.
Satellite and Airborne Remote Sensing
Ground measurements are precise but limited in scale. Remote sensing fills the gap by estimating carbon stocks across millions of hectares. The most important tool for forest carbon is LiDAR, which bounces laser pulses off the canopy and ground to build a three-dimensional picture of forest structure. NASA’s GEDI sensor, mounted on the International Space Station, collects LiDAR measurements globally.
GEDI’s canopy height measurements carry an error of about 5.6 meters on moderate slopes (20 to 30 degrees) and 6.5 meters on steeper terrain. When GEDI-derived carbon density estimates were compared against 220 ground-truth plots, the correlation was strong (R² of 0.69) with a relative error of about 24 percent. The sensor shows minimal saturation effects up to roughly 178 metric tons of carbon per hectare, meaning it remains reliable across most temperate forests. For tropical forests with extremely high biomass, accuracy can drop.
Optical satellite imagery from sensors like Landsat provides complementary data on vegetation greenness, texture, and terrain. On its own, spectral data yields moderate accuracy for carbon estimation. But when combined with vegetation indices, texture features, terrain data, and principal component analysis, machine learning algorithms can push overall classification accuracy above 84 percent. Random Forest models consistently outperform other algorithms, achieving R² values above 0.66 for predicting carbon stocks across different forest types. These models work by learning complex, nonlinear relationships between what the satellite sees and what ground plots actually contain.
Putting It Together for Carbon Credits
If your goal is to certify carbon sequestration for voluntary carbon markets, registries like Verra have specific methodological requirements. Verra’s VM0045 methodology for improved forest management, for instance, focuses on estimating carbon stock changes rather than absolute stocks. This emphasis on change over time, rather than a snapshot, improves precision.
The methodology requires a quasi-experimental design: you establish permanent measurement plots in both the project area and matched control sites, with control data sourced from existing national or regional forest inventories. This matched-pairs approach lets you isolate the carbon impact of your management practices from background trends. Projects must differentiate between emission reductions (preventing carbon loss) and carbon dioxide removals (pulling new carbon from the atmosphere). VM0045 has been approved by the Integrity Council for the Voluntary Carbon Market as meeting their Core Carbon Principles, which is currently the highest credibility benchmark in the voluntary market.
Cost Considerations
Monitoring costs vary enormously depending on scale and method. Variable costs for plot-based field sampling typically range from $100 to $1,500 per plot per monitoring cycle, depending on terrain accessibility, depth requirements, and lab analysis fees. Soil carbon is particularly expensive to measure directly because spatial variability means you need many samples to detect small changes with statistical confidence.
Larger projects can reduce per-hectare costs by integrating remote sensing with targeted ground-truth sampling. Satellite imagery covers vast areas cheaply, while a strategic network of field plots keeps the estimates honest. Machine learning models trained on this combined data can interpolate carbon stocks between plots with reasonable accuracy, reducing the total number of expensive ground measurements needed. For projects under a few hundred hectares, though, direct field measurement remains the most practical and credible approach.