Shared read-only analysis
AGB reference year 2020
Mean density
17.5 t C/ha
Above-ground carbon
4.5M t C
Total area
257.4K ha
Tree-cover loss since 2001
13.1K ha
Land & soil context
What kind of land this area covers - useful for checking that a drawn boundary really is the farm or forest it is meant to be.
Cropland
36%
Land cover · Dynamic World
Topsoil organic carbon
21.3 g/kg
Average in the top 20 cm of soil. Soil carbon is reported for context and is not included in the above-ground carbon totals.
Land cover change 2017-2025 · Esri 10 m
Trees: Dynamic World 40%, Esri 21%. The two maps define trees differently - Dynamic World counts scattered trees, hedges and tea or coffee shade that Esri files under crops or rangeland, and Dynamic World is a recent-months average while Esri is one label per year. Neither tree share feeds the carbon figures (Dynamic World only masks out water, built-up land and snow); the carbon map measures vegetation directly.
Sources: RCMRD Kenya Farming Systems (Sentinel-2 2021-2023, 10 m); Google Dynamic World V1 (10 m, mode over the imagery window); iSDAsoil v1 (30 m, 0-20 cm); Esri / Impact Observatory Sentinel-2 10 m land cover (annual) (CC BY 4.0). The cropland sources can disagree - RCMRD can miss flooded rice paddies, and Dynamic World and Esri often read dryland farms as shrub or rangeland - so a low-cropland warning only appears when every source agrees.
Above + below ground
Roots hold carbon too - this adds below-ground carbon to the map's above-ground estimate.
Above ground
17.5 t C/ha
4.5M t C in total
Below ground
6.3 t C/ha
1.6M t C in total
Total
23.7 t C/ha
6.1M t C in total
Root-to-shoot ratio 0.48. Below-ground carbon = this run's above-ground map x root-to-shoot ratios from NASA/ORNL harmonised biomass carbon (Spawn et al. 2020, 300 m)
How precise is this?
90% ranges, measured on parts of the area the model never trained on.
Area average
16.5 - 17.9 t C/ha
map says 17.5 t C/ha · based on ~35 test blocks
±4% at 90%: within the ±10% that crediting standards ask of field inventories.
Any single 100 m pixel
± 11.3 t C/ha
held 92% of the time on test blocks (target 90%)
Ranges are relative to the ESA CCI reference biomass map (its own error is not included). Read them as "how closely the map reproduces the reference" - field measurements are needed to pin down the absolute error.
Independent lidar check
This run's map compared with NASA GEDI spaceborne lidar measurements it never learned from.
Lidar footprints
1.4K
Correlation
0.55
Map vs lidar
-2%
Typical difference
20.1 t C/ha
Median at the lidar spots: lidar 9.1 vs this map 14.5 t C/ha. The map and the lidar broadly agree on the level of carbon here.
NASA GEDI L4A v2.1 lidar footprints (25 m, 2019-2023), held out from training. GEDI is least reliable in sparse dryland shrub and on steep slopes.
Forest structure (lidar)
How tall and how closed the vegetation is, measured by NASA GEDI's laser at 71.8K spots, 2019–2025.
Canopy height
7.5 m
middle half 3.4–15.5 m
Canopy cover
8%
middle half 3–32%
Plant area index
0.15
leaf + branch area per m² of ground
Layering
1.79
higher = more canopy layers
| Year | Height | Cover | Footprints |
|---|---|---|---|
| 2019 | 4.7 m | 5% | 7.6K |
| 2020 | 7.2 m | 9% | 20.1K |
| 2021 | 7.2 m | 10% | 11.5K |
| 2022 | 9.5 m | 10% | 16K |
| 2023 | 6.2 m | 7% | 6.1K |
| 2024 | 6.3 m | 7% | 7.4K |
| 2025 | 7.9 m | 8% | 4K |
Lidar samples, not a wall-to-wall map. Each year's footprints fall in different places, so year-to-year changes are indicative.
NASA GEDI L2A canopy height (rh98) and L2B canopy cover, plant area index and foliage height diversity; good-quality footprints only
Risk screen
Would carbon stored here stay stored? The questions a carbon buyer asks before the numbers.
Deforestation
medium1.4% of the area lost tree cover in 2021-2025; 0.8% within 5 km around it.
Fire
low0% burned at least once in the last 5 years.
Protected areas
15%Overlaps Dagoretti, Embakasi.
Sources: Hansen Global Forest Change (global_forest_change_2025_v1_13); MODIS MCD64A1 burned area (500 m); WDPA (UNEP-WCMC / IUCN). A screening, not a due-diligence assessment.
Biomass history 2003-2019
An independent satellite record of woody biomass here, from before our own map's period.
2003
32.6 t C/ha
2019
37.4 t C/ha
Trend
+0.31 t C/ha/yr
90% range 0.31 to 0.32
+14.9% overall
This run's map: 17.5 t C/ha (dashed line). Chloris reads 2.1× this run's level. Both are above-ground carbon (Chloris biomass × 0.47), but they are different products: Chloris counts woody biomass in ~4.6 km cells with its own calibration, while this map follows the ESA reference at 100 m or finer. Use Chloris for the direction of change, and this run for today's level.
Chloris Global Biomass 2003-2019 (Chloris Geospatial, via Microsoft Planetary Computer) · ~4.6 km cells · Chloris Global Biomass is licensed CC BY-NC-SA 4.0: free for non-commercial use with attribution. Shown here as an independent reference only.
| County | Mean (t C/ha) | Sum (t C) | Tree-cover loss |
|---|---|---|---|
| Kiambu | 17.5 | 4.5M | 5.1% |
Generated with Carbon-Eye — learn more