Back to overview

How to use Carbon-Eye

From creating an account to reading your first validated result - everything below matches the real dashboard, not a simplified version of it.

Satellite view of mixed farmland and woodlots near Kabarak
Farmland around Kabarak, Nakuru CountyContains modified Copernicus Sentinel data (2024-2026), processed by Carbon-Eye in Google Earth Engine.
A forest path and wooden railing beside a red-brown river under tall trees
Karura Forest, Nairobi's urban forest, along the Karura RiverPhoto: Kaethe17, CC BY 4.0, via Wikimedia Commons (resized)

Six steps, start to finish

  1. 1

    Create an account

    Sign up with an email address (Clerk-hosted, no special setup) and you're taken straight to the dashboard.

  2. 2

    Choose your area

    Pick one or more of Kenya's 47 counties - search by name, select a whole region in one tap, or switch to “Draw area” to outline a custom boundary directly on the map.

  3. 3

    Set the accuracy preset, year and units

    Quick preview samples fewer pixels for a fast look; High accuracy samples more for a more reliable estimate. Pick a reference year (or average a range of years), and whether results default to carbon, biomass, or both.

  4. 4

    Run the analysis

    This kicks off a real Earth Engine pipeline - not a lookup against a pre-computed map. Random Forest, Gradient Tree Boosting and an SVM each train and validate against your exact selection, usually finishing within a few minutes.

  5. 5

    Ask the guide if anything's unclear

    Every run has its own “Ask the Environmental Guide” chat, plus a one-click AI-written executive summary - both grounded in that run's actual numbers, not generic answers.

  6. 6

    Save, export or share

    Save a run to your portfolio to compare it with others later, download a PDF report or the county CSV, or generate a read-only link to share results without giving someone dashboard access.

Carbon-Eye's carbon map of Kabarak University campus drawn in red to yellow over satellite imagery of the surrounding farms
The carbon map over real satellite imagery - Kabarak University campus, 94 ha: 6.7 t C/ha on average, 626 t C in all.

The twelve results tabs, at a glance

Every finished run opens the same twelve tabs. You don't need to read them in order — Map and Briefing cover the headline story; the rest are there when you need more depth.

Map

The default landing tab: a plain-language explanation of the run, its headline numbers, the biodiversity co-benefit card, and the interactive carbon/AGB map over satellite imagery, with a model picker and overlay layers.

Validation

Actual-vs-predicted scatter plots and an RMSE / MAE / bias / R² table, one row per model with its weight in the ensemble - a model that does no better than the average gets no weight and is marked "excluded this run".

Model Comparison

Where the three models agree and disagree, spatially: an RF-vs-GTB scatter plot alongside |RF−GTB| and three-model-spread maps.

Data Quality

An explainable 0–100 screening score built from sample volume, validation strength, model agreement and geographic coverage - a quick read on whether to trust this particular run.

Zonal Stats

Per-county table of mean / min / max / total carbon and tree-cover-loss percentage, sortable and exportable as CSV.

Importance

A ranked bar chart of which satellite, climate and terrain predictors the Random Forest and boosting models actually relied on.

Forest Change

Recent tree-cover alerts (Landsat GLAD-L and Sentinel-1 radar RADD), annual loss, hotspots to check first and the carbon at the alerts - plus a dynamic baseline: forest loss in matched forest outside the area, year by year beside the area's own.

Priorities

A conservation-priority map that weighs carbon, pressure, intactness and restoration potential - set the weights to match your mandate and see which places rise to the top.

Restoration

A growth-curve projection for a restoration scenario, a restoration-suitability map, and methodology screening: land that could qualify under Verra VM0047 (ARR) and VM0048 (REDD+), with the reason for every excluded class.

Decision Tools

Two groups. Planning: biomass history, a carbon-change explorer, a restoration-project revenue calculator, climate exposure and watch-list alerts. Field work: plot planner, field checks, the field-data loop (ODK / Kobo plot form, plot upload and recalibration) and drone or lidar canopy upload. Share a run from the Share button in its header.

Briefing

An AI-generated (or deterministic fallback) executive summary of the whole run - strongest model, uncertainty, county hotspots and cautions - downloadable as text.

Report

Download a shareable PDF summary, and manage your saved-project portfolio with a combined PDF across runs.

First time on the dashboard itself?

The New Analysis form has its own short, interactive tour built in the first time you sign in — look for the “Take the tour” link on the New Analysis card if you want to replay it.