Carbon Stock (Trend) tracks the temporal evolution of carbon stored in ecosystems across a site over multiple years. The indicator measures how much carbon is currently sequestered, how much additional capacity remains unrealised, and what the future trajectory looks like under a high-emission scenario.
Carbon is stored in three pools: Above-Ground Biomass (AGB) — living plant material above the soil surface (trunks, branches, leaves); Below-Ground Biomass (BGB) — living root systems; and Soil Organic Carbon (SOC) — carbon bound in organic matter within the soil. The sum of all three gives the total carbon stock, expressed in tonnes of carbon per hectare (MgC/ha).
Unlike the snapshot indicator (Carbon Stock Distribution), this KPI shows change over time, allowing users to assess whether a site is acting as a carbon sink (accumulating carbon) or a carbon source (releasing it). It is directly relevant to ESRS E1 (Climate Change) reporting.
According to the Global Carbon Project (Friedlingstein et al., 2023), the terrestrial carbon sink absorbs approximately 3.1 +/- 0.6 Gt C per year globally, but this capacity is vulnerable to land-use change and rising temperatures.
Carbon stock values are estimated using a multi-temporal approach combining a 2016 baseline with machine-learning predictions for subsequent years.
For the base year 2016, data are sourced from the Global Potential Carbon dataset (Harvard Dataverse) at 500 m resolution, downscaled to approximately 167 m via bilinear interpolation. For subsequent years, Above-Ground Biomass (AGB) and Below-Ground Biomass (BGB) are estimated using a Random Forest model trained on 2016 biomass data combined with Sentinel-2 NDVI and Google Dynamic World land cover:
where RF = Random Forest model (100 trees), NDVI = Normalized Difference Vegetation Index from Sentinel-2, LandUse = land cover classification from Google Dynamic World
Soil Organic Carbon (SOC) is estimated via linear interpolation between the 2016 baseline and a 20-year delta:
where annual delta = (SOC current - SOC reference) / 20
Unrealised potential is the difference between maximum theoretical carbon storage capacity and current stock. Future projections (2050) use the RCP 8.5 scenario.
Data are sourced from the Global Potential Carbon dataset at 500 m resolution (2016 baseline), Sentinel-2 at 10 m (2017 onwards), and Google Dynamic World at 10 m (continuous).
Line Chart. A multi-year area chart displaying the temporal evolution of the carbon stock metric for the site.
Purpose: Is this site gaining or losing carbon storage capacity over time, and how does it compare to its surroundings?
Description: The chart is embedded in a full-width card within the Land Use assessment section. The X-axis shows years (one data point per available year, from 2016 onwards). The Y-axis shows the selected carbon metric value in MgC/ha or %. Each data point is a circle whose color reflects the quality grade. Solid line segments connect data points; future-dated points (beyond the current year) are shown with reduced opacity (50%) and connected by a dashed line to signal projected rather than observed data. Above the chart, a dropdown selector allows choosing the stock type (Current / Unrealised potential / Future forecast / Future variation) and a detail level toggle (Total / AGB+BGB / SOC / AGB / BGB). A Site/Control toggle switches between the site polygon (ROI) and the control area (CA).
How it's calculated: For the selected stock type and detail level, the platform retrieves the full trends time series from the map layer KPI API. The ROI trend uses values aggregated over the site polygon; the CA trend uses the surrounding control area. Point colors are assigned using the quality thresholds.
Legend: Point and line segment color follows the A-E quality scale for the selected sub-type.
Total Carbon Stock (MgC/ha):
| Level | Range (MgC/ha) | Color | Meaning |
|---|---|---|---|
| A (Excellent) | >= 255 | ■ #00A67A | Dense mature forest — very high carbon storage |
| B (Good) | 175 - 255 | ■ #00DF80 | Well-conserved temperate forests and natural ecosystems |
| C (Moderate) | 100 - 175 | ■ #FFD21E | Semi-natural areas, mixed land use |
| D (Poor) | 40 - 100 | ■ #FF8B16 | Grasslands, intensive agricultural land |
| E (Critical) | 0 - 40 | ■ #FF367F | Artificial surfaces, bare or severely degraded soil |
Interpretation example:
If the chart shows values rising from 48 MgC/ha (orange, D) in 2016 to 130 MgC/ha (yellow, C) by 2023, the site has undergone significant reforestation — a positive trend indicating growing carbon sequestration capacity that may eventually reach forest-level stocks.
Assessment Sidebar Row. A single row in the Land Use KPI panel within the Assessment sidebar, showing the current Carbon Stock quality grade.
Purpose: Provides a quick at-a-glance A-E rating for carbon stock alongside other Land Use KPIs.
Description: The row displays the label "Carbon Stock", the current numeric value (total current stock, ROI mean), and a colored grade badge (A-E). Clicking the row scrolls to the full Carbon Stock section.
How it's calculated: The grade is assigned by matching the site's total current carbon stock value against the quality thresholds: A >= 255, B 175-255, C 100-175, D 40-100, E < 40 MgC/ha.
Interpretation example:
If the sidebar shows a yellow "C" badge next to 142 MgC/ha, the site has moderate carbon storage — typical of semi-natural mixed landscapes where some natural vegetation is present but full forest cover has not yet developed.
Line Chart. A yearly trend line in the Multi-Year Trends section, showing how the selected carbon stock metric has evolved across all generated land cover versions.
Purpose: Compare carbon stock trajectories across multiple years and assess long-term ecosystem carbon dynamics.
Description: This chart appears in the Multi-Year Trend page alongside a donut gauge. The right panel shows the line chart; the left panel shows the donut gauge summarising the current period's ROI vs CA comparison. The X-axis shows selected years; the Y-axis shows MgC/ha. A year-filter toolbar at the top allows selecting specific years. Future-projected data points are displayed with dashed line segments and reduced opacity. The Site/Control toggle switches between ROI and CA series.
How it's calculated: Same data source as the Land Use section, but filtered by years selected in the Multi-Year Trends context. When SiteSpecific functionality is not active, only the CA trend is displayed.
Legend: Same quality-scale color coding as the Land Use line chart above.
Interpretation example:
If the Multi-Year Trend chart shows stable carbon stock around 200 MgC/ha (green, B) across 2019-2023 with a projected dip to 180 MgC/ha by 2050 (shown as a dashed line), the site is maintaining good carbon storage capacity in the near term, with a modest projected decline under the RCP 8.5 scenario.
Gauge. A semicircular dual-arc gauge in the Multi-Year Trends section comparing the site's carbon stock to the control area for the selected period.
Purpose: Visually compare the site's absolute carbon stock value against its surrounding control area.
Description: The outer arc represents the site (ROI) value; the inner arc represents the control area (CA). Both arcs are colored according to the A-E quality scale. The center displays the ROI mean value, the CA mean, a benchmark value, and a delta badge. This gauge is shown only when SiteSpecific functionality is active for the user account.
175Carbon Stock (MgC/ha)
How it's calculated: Values come from the ROI, CA, and benchmark fields for the selected carbon stock item. Arc colors are determined by the quality thresholds of the active sub-type.
Legend: Same A-E quality scale as the line chart.
Interpretation example:
If the gauge shows ROI = 175 MgC/ha (grade B, green) and CA = 80 MgC/ha (grade D, orange) with a delta of +95, the site stores significantly more carbon than its surrounding landscape — a strong positive signal for the site's ecosystem restoration efforts.
Map Layer. An interactive raster overlay on the site map showing the spatial distribution of the selected carbon stock metric at pixel level.
Purpose: Reveals which parts of the site store the most (or least) carbon, supporting targeted land management decisions.
Description: This is a complex spatial element. The layer is selected via the Land Use map section, then refined with two additional controls: the stock type selector (Current / Unrealised Potential / Future Unrealised / Future Change) and the detail toggle (Total / AGB+BGB / SOC / AGB / BGB). Each combination corresponds to a separate pre-generated raster. A color legend bar is displayed alongside the map showing the gradient scale for the active layer. An information panel describes the selected sub-type.
How it's calculated: Raster tiles are served from pre-generated files. Current stock uses the 2016 baseline dataset (500 m, downscaled) plus Random Forest predictions for later years. Future layers use RCP 8.5 projections. Coloring uses a red-to-green 11-stop gradient for Current/Future Change layers, and a black-to-white 9-stop gradient for Unrealised Potential layers.
Legend: Color gradient varies by sub-type. For Current Stock and Future Change: red (low/negative) to green (high/positive). For Unrealised Potential: black (fully realised) to white (large potential).
Interpretation example:
If the map shows a dense green area covering the central forest patch and orange/red pixels on the agricultural margins, it means the forested core has high current carbon storage (likely B-A grade) while the degraded agricultural zones have very low stocks (D-E grade) — identifying priority zones for restoration investment.
| Source | Provider | Coverage | Resolution | Period |
|---|---|---|---|---|
| Global Potential Carbon | Harvard Dataverse | Global | 500 m | 2016 baseline |
| Google Dynamic World | Global | 10 m | Continuous | |
| ESA Sentinel-2 L2A | ESA / Copernicus | Global | 10 m | 2017 - present |
The raw raster tile is loaded from pre-computed files on S3 for the 2016 baseline. A scale factor is applied to convert raw pixel values to the carbon stock ratio. The data undergoes bilinear resampling (3x interpolation from 500 m to approximately 167 m) to the output projection (EPSG:3857).
For years after 2016, a Random Forest model (100 trees) is trained per-request on 2016 AGB data, 2017 Sentinel-2 NDVI, and 2016 Google Dynamic World land use. The trained model then predicts current-year AGB+BGB values using the latest available NDVI and land cover data.
The SOC component uses a simplified 20-year linear interpolation between the 2016 baseline and a projected delta, not a machine learning model — SOC projections assume a steady rate of change.
For the map layer, coloring uses a red-to-green 11-stop gradient for Current Stock and Future Change layers, and a black-to-white 9-stop gradient for Unrealised Potential layers. Area for each class is calculated in metric projection for accurate surface measurement.
Future projections use RCP 8.5 scenario raster files. The future change indicator shows the percentage variation between the projected 2050 unrealised potential and the current value: