Evidence in action
A shared evidence approach. Four applications.
Four illustrative examples of how one digital twin and one evidence record could turn satellite observations, field records and laboratory results into evidence that independent verifiers can check. They describe a proposed service. None of them shows a real field, live monitoring or an issued credit.
Start with the land
Every example begins with a landscape divided into management zones. Evidence is kept zone by zone, so each record can be traced back to a place.
Gather observations, each with its own time
Satellite passes, field sensors, soil samples sent to a laboratory and photographs from the farmer arrive at different times. None of them is treated as live, or as if they happened together.
Keep every observation as a record
Each record holds the location, the time, the source and the type of evidence, and the checks it has passed. A separate status shows whether a person has reviewed it.
A digital twin, in this story, is a model tied to a real field or management zone, with its soil, water, crop, practice and history. Observations update it, models compare choices or flag changes, and people review the interpretation and decide what to do.
One loop, used in every example
Observe, check, model, review, act, follow up
- ObserveGather observations from space, the field and the laboratory.Measured
- CheckTest place, time, units and source; flag gaps and disagreements.
- ModelEstimate what cannot be measured directly, with its uncertainty.Modelled
- ReviewA person signs off the result and its exceptions.Reviewed
- ActAdvice goes back to the farmer or the programme.Estimated
- Follow upThe next reading returns to the same record.
Then the loop starts again with the next observation.
How to read the labels on this page
- MeasuredModelledEstimatedWhat kind of evidence a value is.
- ReviewedPending reviewWhere it is in the process. Reviewed is a process state, not verification.
- IllustrativeEvery example, figure and date on this page. Independent verification is done by accredited verifiers.
Explore what this could look like in practice.
Chapter 01 · Soil
01 / Soil
Can the field record support the soil-carbon claim?
Illustrative example · concept field
A field changes how it is farmed: a winter cover crop instead of bare ground, and less ploughing. The claim is that more carbon stays in the soil. The record below supports a soil-carbon model. It is not a direct measurement of the carbon itself.
Establish the field
Plot SC-0217 is a 2.4-hectare concept field, followed for one season of 300 days. Five soil cores, analysed in a laboratory, set the baseline. A field node logs soil moisture and temperature every hour.
Optical satellite passes estimate ground cover in 10-metre pixels, and the farmer adds photographs.
Build the record
Ground cover is recorded across the season. The cover crop establishes in autumn and is ended in February, and the spring crop follows in March.
Each farmer photograph sits at the date it was taken. This is an example record, not farming advice.
The record is complete. It is not perfect, and the evidence should say so.
Show the imperfect evidence
From mid-December, cloud blocks the optical satellite for 19 days. Values in that gap are estimated and drawn dashed, never smoothed into the measured line.
Radar passes see through cloud, but one of them disagrees with the farmer's photograph taken the next day. The disagreement waits for a person to review it before anything is counted. Here, the reviewer finds that surface water affected the radar reading.
Explain the estimate
A soil-carbon model compares the old practice with the new one. The modelled change is about 1.10 tonnes of CO₂e per hectare for the year, with an estimated uncertainty of ±52%.
A claim would use the conservative figure, 0.53 t CO₂e per hectare: about 1.3 t for this plot.
How these illustrative figures are built
The model estimates the change from the practice record, the baseline soil cores and local weather. The uncertainty combines model and sampling uncertainty. The conservative figure is the lower end of that range: 1.10 × (1 − 0.52) = 0.528, shown as 0.53. For 2.4 hectares, 0.528 × 2.4 = 1.27 t, shown as about 1.3 t.
These are illustrative figures for a concept field, not results.
Chapter 01 · Output
An evidence record, prepared for independent verification
- Plot
- SC-0217 · 2.4 ha · season September 2026 to June 2027 (illustrative)
- Inputs
- 14 clear optical passes · 5 radar passes · 6 farmer photographs · 5 soil cores (laboratory)
- Exceptions
- 19-day cloud gap, values estimated · 1 disagreement, reviewed
- Output
- Modelled change with its uncertainty, and the conservative figure
- Review
- Complete
- Independent verification
- Not started
An accredited verifier checks the record. Climate Ground prepares evidence and never verifies its own. See a specimen evidence pack and the label definitions
Chapter 02 · Agroforestry
02 / Agroforestry
What if we compare the choices before planting?
Proposed extension · illustrative landscape
Trees and crops share the field
Rows of productive trees follow the contours, with crop alleys between them, so the field keeps producing while the trees establish.
Connect habitat and protect water
A habitat corridor links the existing woodland to the stream, and planted buffers protect the water's edge.
Keep the field workable
A track keeps access open for planting, harvest and monitoring. The geography underneath stays fixed, so every change can be seen.
A plan is only a proposal. The next questions are which version to plant, and how to look after it.
Compare layouts before planting
The twin compares layouts before anything is planted, all against the same baseline: the farmland as it is today. Option A keeps wide crop alleys between tree rows. Option B plants more trees and connects habitat to the stream.
Trees compete with crops for light and water as well as helping them, so more trees should not automatically score better. The scores stay empty until a real scenario run, with a common baseline, weather and time horizon.
Record establishment
An approved design becomes planting lines and seasonal tasks. Each tree is recorded as it is planted, with its species and who did the work.
A scheduled survival check finds one tree missing. Its replacement is logged against the same tree, with the reason, and the zone record keeps the history. The dates are left unset in this example.
Chapter 02 · Outcomes
Each outcome kept separate, with its own evidence
There is no combined impact score. Each result keeps its own baseline, method, period, evidence type and uncertainty. None is assessed in this example.
- 1CarbonNet t CO₂e against a defined baseline and period.Needs inventories, sampling and models.Not yet assessed
- 2BiodiversityHabitat condition and selected species indicators.Needs ecological surveys; canopy cover alone is not a gain.Not yet assessed
- 3ResilienceLoss and recovery during a defined shock.Needs a stated shock, a comparison and a recovery period.Not yet assessed
- 4AdaptationPerformance under specified future conditions.Needs defined scenarios; it stays a projection.Not yet assessed
- 5LivelihoodsProduction, gross margin, labour and income diversity.Needs farm records.Not yet assessed
Why keep them apart?
A single score would hide trade-offs. Denser planting might store more carbon while reducing crop space and changing labour and income. Habitat maps can guide ecological surveys, but canopy cover alone does not show a biodiversity gain.
Test the design against different seasons
A design that works in an average season may not work in heat, drought or intense rain. The comparison would vary survival, yield, input costs and prices, show ranges rather than one future, and revise the design where it is weak.
These pictures explain the questions being tested. They are not forecasts, and the hazards and time horizon are pilot parameters to agree.
Chapter 02 · Output
A proposed design, its establishment history and the evidence still needed
- Design
- Today, option A and option B compared on a common baseline; the choice is made with farmers and local specialists
- Establishment
- Recorded tree by tree · 1 tree not found, replaced and logged with the reason
- Outcomes
- Carbon, biodiversity, resilience, adaptation and livelihoods, each with its own evidence · none assessed yet
- Conditions
- Average, hotter and drier, and wetter seasons · parameters to agree in a pilot
- Decisions supported
- What to plant where, what to replace, and what to measure next
- Status
- Proposed extension · no gains claimed
Chapter 03 · Farmer alerts
03 / Farmer alerts
When something changes, what should someone check?
Illustrative service concept · AG-003, Zone 03
A new example, separate from the soil field and the agroforestry design: one management zone, a change in the evidence, and the human actions that follow. The question is how a change in the evidence leads to a useful action.
Gather observations
AG-003 is an illustrative farm with six management zones. In Zone 03, a soil-moisture sensor, a weather station and a drone survey each report at their own time.
The latest imagery is two days older than the sensor reading. The record keeps each source's age rather than treating them as simultaneous or live.
Flag a difference
The drone survey's reflectance for Zone 03 differs from a reference canopy, mostly in the near infrared. The curves shown are schematic and synthetic.
The twin raises a modelled flag: change flagged, cause unconfirmed. Water, nutrition, crop stage, shade or damage could explain it, so it prompts an inspection, not a diagnosis.
Ask for a field check
The farmer receives a message tied to one zone: what changed, and what to check. They can add a photo, record observations or ask for local advice.
The field report comes back: leaf damage observed, pest not yet identified. A pilot would test local-language text or voice on the channel farmers already use.
Close the record loop
The detection, the farmer's observation, an adviser's review and a scheduled follow-up all land in the same zone record.
The follow-up stays pending. Reviewed records can become evidence for testing the next model version; they do not retrain the model automatically. The times explain the sequence; they are not service levels.
Chapter 03 · Output
A zone record that holds the signal, the check and what happens next
- Zone
- AG-003 / Zone 03 (illustrative)
- Inputs
- Soil moisture 22 Oct 08:10 · weather 08:00 · drone imagery 20 Oct 10:35, two days older
- Flag
- Modelled · cause unconfirmed
- Field report
- Leaf damage observed · pest not yet identified
- Adviser review
- Reviewed · follow-up scouting planned and recorded
- Status
- Follow-up pending
- Pilot measures
- Alert-to-check time, confirmed alerts, response completion, crop outcomes across seasons
No recovery, yield gain or verified outcome is shown, and no treatment is advised.
Chapter 04 · Government reporting
04 / Government reporting
How can checked field records inform a whole programme?
Illustrative state programme · abstract districts
Check before counting
This is an illustrative state programme on an abstract map of eight districts, A to H. Each symbol stands for a group of checked field records, not a set number of farms or hectares.
Records that fail a check are flagged and held for review. They stay outside the totals until a person has reviewed them.
Aggregate without losing meaning
Records roll up into district summaries: the share of cropland under the recorded practice, a winter cover crop with less ploughing, from 48% to 72%.
Across the programme, 60,000 of 100,000 hectares are under the practice in one season. Units, period and source travel with every total.
Show the result and its uncertainty together
Using the soil example's figures, the modelled change is about 66,000 t CO₂e for the season, with the estimated uncertainty of ±52% carried through. The conservative figure is about 32,000 t.
This is an illustrative calculation, not issued credits, an authorised transfer or an independently verified result.
How these illustrative figures are built
60,000 ha × 1.10 t = 66,000 t CO₂e, modelled. 60,000 ha × 0.528 t = 31,680 t, shown as about 32,000 t, the conservative figure. The ±52% from the field estimate is carried through unchanged, a deliberately cautious simplification.
What a government could use it for
The same evidence could support national inventory work, moving from default factors to recorded practice under the Paris Agreement's Enhanced Transparency Framework; Article 6 administration, where the government authorises transfers and tracks corresponding adjustments; and farm-programme monitoring.
Farm-level records stay in the country, in the evidence record. Only district and state summaries are shared with the state system. Climate Ground does not grant authorisations or approvals.
Chapter 04 · How a programme runs
Set up, run every season, then hand over
- 1Set up
- Baseline for the programme area
- Data-sharing agreements
- Link to the national registry
- 2Run
- Monitoring every season
- Exception review
- Annual report
- 3Hand over
- Training for government teams
- Open methods
- A system the government can run itself
We prepare the evidence. Accredited verifiers and the government's own review sign it off.
Pilot
Prove it in a bounded pilot, then decide whether to expand
- 1BaselineAgree the baseline, the decisions, the farmers and the delivery partners.Data access agreed
- 2TestData quality, useful advice and practical recording.Records usable
- 3DeliverResponse times, false alerts, workload and completion.Advice useful
- 4EvaluateEnvironmental and resilience outcomes, over the right periods.Expand?
DecisionExpand when the evidence and the operating model support the next step.
Proposed pilot evaluation. No quantified improvement is asserted. Results would be compared with the previous way of working or a suitable comparison group, and resilience assessed through loss and recovery under a defined shock.
Next step
Bring your evidence question
Tell us about the decision or the report the evidence needs to support, and what evidence you hold today.