A map of stored carbon is a snapshot. A map of annual growth is a test of change.
A forest is a balance sheet—but climate policy needs a cash-flow statement
A forest is not a warehouse of carbon that stays unchanged until somebody cuts it down. Trees grow, compete, die and regenerate. Windstorms alter the canopy. Logging transfers wood out of the landscape. Soil microbes release carbon even as living trunks add more. The crucial question for climate policy is therefore not merely how much carbon a forest contains today but how that amount changes over time.
At an October 20 online technology showcase organized by the Japan Science and Technology Agency and Hokkaido University, Professor Tomomichi Kato of the university’s Graduate School of Agriculture presented a method designed to estimate annual growth in above-ground forest biomass. The published program describes combining airborne laser surveys, satellite observations and machine-learning models, with a stated ground resolution of 10 by 10 meters per pixel.
What is actually new?
Satellite biomass estimation is not new. Satellites have long helped scientists classify forest cover, infer canopy properties and estimate standing timber or carbon stocks. Kato’s stated innovation is to estimate a rate of growth—a change per year—rather than only a stock at a single point in time. That distinction is consequential. A mature forest can store considerable carbon without necessarily recording rapid annual increases in above-ground woody biomass.
JST identifies carbon-credit calculations, timber-volume estimation and regional carbon accounting as possible applications. They are use cases, not evidence that a nationwide annual-growth product has already been independently validated or that carbon credits have been issued from it.
Why a satellite cannot simply photograph carbon
Optical imagery observes reflected light, which depends on tree species, season, leaf condition and illumination. Radar measures returned microwave signals whose behavior also reflects structure and moisture. Neither produces a direct weight reading for the wood in a tree. Scientists need reference observations to connect remote-sensing signals with material on the ground. Airborne lidar, which uses laser ranging to describe forest height and three-dimensional structure, helps supply that bridge.
A machine-learning model then searches for relationships between the reference measurements and wider satellite datasets. Used carefully, such a model can provide coverage far beyond the places that crews have surveyed. Used carelessly, it can produce confident-looking maps in unfamiliar ecosystems with unreliable results. A model trained extensively on one set of forest types is not automatically accurate in every Japanese prefecture.
The earlier national map that makes this credible
The Japan Aerospace Exploration Agency has already published a national high-resolution dataset of above-ground forest carbon stocks, biomass, timber volume and tree height. JAXA attributes the work to joint research led by Kato. Its published methodology cites the Japanese ALOS-2 satellite, the European Sentinel-2 satellite, airborne lidar measurements and machine learning. That established mapping work supplies important historical context for the new announcement.
But it would be incorrect to treat the older stock map as proof that the new annual-growth estimates are accurate everywhere. Estimating a quantity once and detecting a relatively small year-to-year change are different statistical challenges. JAXA’s already available map and JST’s October 20 presentation should not be conflated.
Ten-meter pixels—and the accuracy trap
A ten-by-ten-meter pixel covers 100 square meters. Such detail could be useful for spotting variation within a managed forest, comparing stands or examining recovery after disturbance. Yet spatial resolution is not the same as measurement accuracy. The position of a pixel, mixed land cover at forest edges, weather conditions, tree species and errors in the training data all affect estimates.
Annual change can be especially difficult. If the uncertainty surrounding biomass measurements is large compared with a year’s real growth, a visually precise map may still struggle to distinguish biological change from noise. Publication of independent error statistics, validation locations and performance across multiple years will be essential.
Carbon uptake is not identical to tree growth
Growing trunks, branches and foliage generally indicate that living vegetation has stored more carbon. That is valuable information. But a rise in above-ground live biomass is not automatically a measurement of net ecosystem carbon sequestration. Below-ground roots, soils, decomposition, dead wood, timber removals and disturbances all influence the larger balance. The research description does not establish that every component has been measured.
Nor does a forest growth map by itself establish the eligibility of a carbon credit. Credit programs may demand additionality, a credible baseline, durability, protection against leakage, monitoring and independent verification. The capability to estimate a biophysical quantity is only one link in that chain.
The longer Japanese history behind the measurement
Japan has accumulated a formidable remote-sensing infrastructure through successive Earth-observation satellites, airborne surveys and forestry data. JAXA’s ALOS-2 radar and optical data from Sentinel-2 underpin earlier carbon-stock mapping. ALOS-4, now operating, carries more advanced radar capabilities: JAXA says its improved observation swath supports more extensive monitoring, including forestry. However, the October presentation does not establish that Kato’s new model specifically used ALOS-4 imagery.
Japan’s satellite ambitions also intersect with global needs. In April 2026, JAXA and the UN Food and Agriculture Organization extended their partnership for forest monitoring. Countries with vast forests and sparse ground-survey coverage have particular incentives to develop credible satellite-based observation, but local field measurements remain indispensable.
What forest owners and businesses might gain
For Japanese municipalities managing plantations, a defensible annual-growth layer could help prioritize field visits, identify poorly performing stands, monitor replanting and inform timber estimates. For companies, greater spatial and temporal detail could improve risk assessments and the scrutiny of nature-related claims. For investors considering forest-carbon projects, better data could reduce one source of uncertainty.
Yet the technology’s economics cannot be inferred from a description of free satellite input data. Acquiring or licensing suitable airborne lidar, developing reference plots, running models, updating estimates and commissioning independent audits all carry potential costs. The full workflow, rather than the price of a satellite image, determines the practical business proposition.
Who is accountable when the map is wrong?
A forest carbon estimate can influence a public grant, a land-management decision or a credit claim with real financial consequences. That raises questions beyond computer vision. Which institutions sign off on model error? Can a landowner challenge an erroneous classification? Will smaller rural communities have access to the results? How will a disturbance be reflected in the next reporting period?
JST’s technology introduction does not answer those governance questions. It indicates possibilities for commercial partnerships and future applications rather than a fully deployed, audited forest-carbon accounting service.
What readers should watch next
The published material does not disclose a comprehensive independent validation dataset, national annual-growth accuracy, species-by-species error rates, commercialization terms or the launch date for a public annual-growth map. These are unanswered questions, not shortcomings that can be filled with assumed numbers. The most revealing next step would be an openly described comparison of predicted annual gains with repeated ground and airborne measurements in varied forests.
The intriguing achievement is conceptual as well as technical. Mapping where trees stand is a portrait. Mapping how their biomass changes is a moving picture. Kato’s research seeks to take Japan from one to the other—while reminding us that the most convincing account of a forest still depends on connecting space-based observations to reality on the forest floor.
Sources and further reading
Based on the October 20 JST presentation and JAXA primary-source material. No independent researcher interview or accuracy replication was conducted. Annual growth is not equated with complete ecosystem carbon accounting.