Measuring soil carbon: which methods really work
Satellites promise global soil carbon measurement. Physics says: impossible. This analysis shows which methods really work, where their limits lie — and why carbon farming credits are on thin ice. Based on more than 40 scientific studies.
The fundamental measurement problem
Soil carbon is invisible, unevenly distributed and changes slowly. This creates a signal-to-noise problem that pushes most measurement methods to their limits.
“You can’t manage what you can’t measure. And measuring soil carbon is one of the most difficult tasks in environmental science.”
Why satellites cannot measure soil carbon
The electromagnetic radiation that satellites receive penetrates no more than 175 micrometres into the soil. But the carbon lies up to 100 cm deep. This gap cannot be bridged by physics.
“A satellite measuring soil carbon at a depth of 30 cm is like a microphone trying to capture the colour of a painting — the wrong instrument for the wrong question.”
Laboratory methods: The only reliable foundation
Direct soil sampling with laboratory analysis is the gold standard. But even here there are differences in accuracy, cost and feasibility.
Dry combustion (DC)
The sample is combusted at 950–1350 °C. CO₂ is quantified by infrared detector or thermal conductivity. DIN EN 15936 / ISO 10694.
Walkley-Black (WB)
Chromic acid oxidation. Captures only 60–86 % of organic carbon. The correction factor of 1.32 is soil-type-specific and leads to systematic errors.
Loss on ignition (LOI)
The sample is ignited at 400–550 °C. Mass loss ≈ organic matter. Conversion to C uses a factor of 1.724 (= 58 % C content), which varies greatly depending on soil type.
The same soil sample, analysed in different laboratories, can give results that differ by a factor of 2.
Proximal sensing: Fast, but with limits
Technologies used directly on the soil or in the field. Faster than laboratory analysis, but less accurate — and highly dependent on local calibration.
VNIR/MIR spectroscopy
Visible–near-infrared (VNIR, 350–2500 nm) and mid-infrared (MIR, 2500–25000 nm) reflectance analyses the molecular composition of the soil. MIR is more accurate than VNIR.
LIBS — laser-induced breakdown spectroscopy
A high-energy laser pulse vaporises a tiny amount of soil. The resulting plasma emits element-specific light that quantifies C directly.
INS — inelastic neutron scattering
A neutron source irradiates the soil. Carbon nuclei emit characteristic gamma radiation at 4.44 MeV. Measures total C in the 0–30 cm soil volume without destroying the sample.
Process models: Calculating instead of measuring
Mathematical models simulate the carbon cycle in the soil. They are the backbone of many MRV systems — but only as good as their input data and calibration.
RothC
Monthly time steps. 5 C pools: DPM, RPM, BIO, HUM, IOM. Requires weather, clay content, land use and C input.
DNDC
Daily time steps. Models the C and N cycles in coupled form. Designed specifically for N₂O and CH₄ emissions from agriculture.
Century / DayCent
Long-term C dynamics in grassland and arable land. Basis for COMET-Farm (USA). Monthly to daily steps (DayCent).
Models calculate carbon changes — they do not measure them. The results depend critically on the input data: C input, weather, soil parameters. Small errors in the inputs lead to large errors in the output.
Dupla et al. (2024): Model-based SOC credits are “empty promises” if the models are not regularly calibrated and validated against real measurements.
Eddy covariance: Highly precise, but impractical
Measures the actual CO₂ exchange between ecosystem and atmosphere in real time. Highest accuracy — but suitable for research purposes only.
Operating principle
An ultrasonic anemometer and an open-path infrared gas analyser measure vertical wind speed and CO₂ concentration 10 times per second. The covariance gives the net flux.
Why it does not scale
Autonomous robots: Sampling without people
Robots automate the bottleneck of soil sampling. They take GPS-referenced soil cores — but laboratory analysis is still needed.
ROGO
Autonomous soil sampling robot. RTK-GPS, 24/7 operation, 200+ samples/day. First commercial provider in the USA.
SoilSense 360
Integrates VNIR spectroscopy directly into the drilling process. Combines sampling with in-situ measurement.
MoistureMapper
Measures soil moisture and conductivity. Serves as a proxy for SOC distribution and optimises sampling points.
Important: Robots solve the sampling problem (cost, speed, reproducibility). They do not solve the analysis problem — the samples still have to be analysed in the laboratory (DC). The robot is a logistics tool, not a measuring instrument.
LiDAR & drones: Surveying trees, calculating carbon
For tree biomass, remote sensing really does work — because height and structure can be measured directly. From there, carbon content is calculated using allometric equations.
UAV LiDAR (drone)
A drone-mounted LiDAR scanner creates 3D point clouds of forest stands. Tree height, crown diameter and stand density are derived from these.
Allometric equations
Biomass is calculated from tree height, DBH (diameter at breast height), tree species and wood density. Basis: statistically derived equations from destructive harvests.
Drone — multispectral/hyperspectral
Camera-based drones capture spectral reflectance in 4–200+ bands. Vegetation indices correlate with leaf area, health and, indirectly, biomass.
TLS & satellite LiDAR: From the millimetre to the globe
Terrestrial laser scanning (TLS) delivers millimetre accuracy for individual trees. GEDI and ICESat-2 survey global forest biomass from orbit.
TLS — terrestrial laser scanning
A stationary scanner creates 3D models of individual trees accurate to the millimetre. Stem volume is calculated directly — without allometric equations.
GEDI / ICESat-2
Full-waveform LiDAR from orbit. 3.77 billion laser measurements. Measures canopy height and vertical structure globally. Basis for global biomass maps.
Biochar analysis: carbon, permanence & stability
Biochar quality is not determined by carbon content alone. For carbon credits, only carbon that demonstrably remains stable for centuries counts. Permanence is decisive — and this is exactly where Prof. Hamid Sanei’s inertinite benchmark comes in.
A biochar with 80% carbon, produced at low temperature (350°C) may contain unstable carbon that breaks down within a few decades. Conversely, a biochar with 55% carbon produced at 650°C can consist almost entirely of inertinite.
H/Corg molar ratio
The most widely used proxy worldwide. Lower value = higher degree of aromatisation = greater stability.
Random reflectance & inertinite benchmark (IBRo2%) — gold standard
Prof. Hamed Sanei (Aarhus University, Director of the LOC laboratory) has developed what is currently the most scientifically robust method for assessing permanence:
Production temperature has the greatest influence on biochar stability:
✅ How to measure biochar correctly
❌ Common mistakes
All methods in direct comparison
Interactive accuracy comparison of all measurement methods. The bars show R² values — the longer, the more accurate.
What works — and what is greenwashing
The scientific evidence is clear: there is no single miracle method. But there is a clear hierarchy — and a pragmatic hybrid approach.
Works
Usable with limitations
Does not work
This tiered approach combines the strengths of each method: DC ensures accuracy, proximal sensing increases data density, models fill temporal gaps, UAV LiDAR captures above-ground biomass, and the inertinite benchmark (IBRo2%) secures biochar permanence. Satellites provide context (land use, vegetation), but no soil C values.
“There is no shortcut. If you want to measure carbon in soil, you have to go into the soil. If you want to measure carbon in trees, you need 3D data. Satellites alone are not enough for either.”
Satellites measure surfaces — not soil carbon.
Direct sampling remains indispensable for MRV.
For trees, LiDAR works.
For soil, there is no shortcut.
Biochar permanence needs more than just carbon content —
the inertinite benchmark is the gold standard.