MRV turns farm claims into evidence that can be inspected.

Measurement, Reporting, and Verification (MRV) is the evidence layer of the Kokonut Framework. Its purpose is not to make every farm claim automatically true. Its purpose is to make important claims traceable:

what was observed;
where and when it happened;
how it was measured or reported;
which record supports the claim;
whether the record has been reviewed;
whether it has been published or attested;
what uncertainty or limitations remain.

At Adelphi, MRV is already part of the operating model. Public farm records are exposed through the Kokonut Hub, while the broader MRV and attestation infrastructure continues to expand.

Inspect Adelphi's public records

Use the Hub for current published farm records, MRV events, harvest information, and project data.

Understand the data infrastructure

Kokonut Intelligence is the active data layer that stores, reviews, analyzes, publishes, and attests farm evidence.

MRV is a chain of evidence, not a magic proof mechanism. A sensor reading, a reviewed database record, an EAS attestation, and an impact conclusion are different things and should be described separately.

MRV in one sentence

Observe what happened → structure the record → review it → publish the evidence → attest when useful → interpret it with the appropriate methodology.

That sequence matters. A common MRV failure is to jump directly from “a record exists” to “the impact claim is proven.” Kokonut instead separates the evidence lifecycle into distinct layers. Each layer answers a different question.

Layer	Question
Observation	What was seen, measured, logged, or recorded?
Structured record	Can another system understand the event consistently?
Review	Has an authorized person or process checked the record?
Publication	Is the record approved for dashboards, APIs, reports, or downstream use?
Evidence package	Can the underlying material be traced and inspected?
Attestation	Who attested to this record, when, and against what evidence reference?
Interpretation	What can the evidence reasonably support?
Decision support	How should operators, contributors, governance, or partners use the result?

What MRV can support — and what it cannot

The strongest wording should match the strength of the evidence.

Question	Evidence that can support it	Important limitation
Is the farm active?	Geospatial records, field logs, infrastructure records, crop records, photos, harvest activity	A record can be incomplete or stale; publication date matters
What was planted or harvested?	Crop-cycle records, operator logs, GPS records, harvest entries, photos	Forecast quantities are not harvest actuals
What changed on the land?	Remote sensing, vegetation indices, soil readings, repeated field observations	Change does not automatically prove causation
Are ecological conditions improving?	Baselines plus repeated vegetation, soil, biodiversity, and water observations	Interpretation depends on methodology, season, sampling, and time period
Did a reported event have provenance?	Evidence CID, author/reviewer metadata, EAS attestation	Provenance does not guarantee scientific correctness
Can an impact claim be reported?	Reviewed records tied to a defined method such as EBF, SDG evidence, or another framework	A reporting framework is not the same as third-party certification
Can a farm forecast be improved?	Historical actuals, losses, harvests, prices, weather, crop-cycle records	Forecasts remain scenarios rather than guarantees
Can the model be compared across farms?	Shared schema, consistent time periods, comparable record types, documented methods	Local ecology and operating context still matter

Attested does not mean scientifically verified. An EAS attestation can make authorship, timing, and evidence references more inspectable. It does not by itself prove a carbon claim, certify ecological impact, guarantee a harvest, or create eligibility for a regulated environmental credit.

The five evidence classes

Farm documentation becomes clearer when every claim is identified by evidence class.

1. Direct observation

A field note, photo, GPS record, operator observation, or documented farm event.

2. Measurement

A sensor reading, area measurement, soil value, quantity, timestamped harvest entry, or remote-sensing output.

3. Derived indicator

A value calculated from source data, such as a vegetation index, loss rate, trend, score, or aggregation.

4. Forecast or model

A scenario based on assumptions: future yield, revenue, impact, carbon, risk, or another projected outcome.

5. Attestation or reviewed claim

A record whose provenance, review, or publication state has been formally captured. The attestation describes the claim; it does not replace the underlying evidence.

Interpretation

A conclusion produced by applying a defined method to evidence — for example EBF reporting, CRISP review, SDG mapping, or an agronomic assessment.

This vocabulary should be used throughout Kokonut farm pages. For example:

“110 hens documented” is a recorded state.
“48,450 lettuces per harvest” is a forecast when derived from planning assumptions.
“NDVI increased over a period” is a derived remote-sensing observation.
“The farm restored biodiversity” is a higher-level impact claim that requires a defined baseline, method, period, and supporting evidence.

Record lifecycle in Kokonut Intelligence

Kokonut Intelligence defines a review lifecycle before operational records become trusted enough for public use.

Status	Meaning
Draft	A record exists but is not ready for review.
Submitted	A field worker, integration, or supervisor has submitted the record for review.
Verified	An authorized reviewer or process has approved the record for the relevant workflow.
Published	The record can be used in dashboards, reports, APIs, or attestations.
Rejected	The record requires correction or additional evidence.

“Verified” here is a workflow state. It means the record passed the applicable Kokonut review process. It does not automatically mean external scientific certification, regulatory approval, or independent audit.

The observation stack at Adelphi

Adelphi combines several evidence sources because no single instrument can describe a whole farm.

Ground sensing
Remote sensing
Community and field observations
Per-plant and geospatial records

Ground observations describe conditions close to the crop and soil.Documented variables include:

volumetric water content;
electrical conductivity;
soil temperature;
soil and crop condition;
water observations;
operator field notes.

These measurements can support irrigation, soil, nutrient, and crop-condition analysis when collected with consistent units, locations, and time periods.

Documented tools and methods

The public repository currently references the following tools and evidence surfaces for Adelphi.

Tool / method	Role in the evidence stack	Typical output
Silvi	Per-plant GPS and health records	Location, status, phenology, plant-level records
Atlantis App	Field observations and project evidence	Reports, photos, timestamps, field context
QGIS	Geospatial and vegetation analysis	Raster analysis, derived vegetation indicators
Pix4Dcapture / drone workflow	Flight planning and imagery	Georeferenced imagery and orthomosaics
Landsat / Sentinel	Landscape-scale remote sensing	Repeated vegetation and land observations
Kokonut Hub	Public farm-data surface	Published farm records, milestones, MRV and harvest information
Kokonut Intelligence	Canonical data, review, analytics, and evidence infrastructure	Governed records, metrics, reports, APIs, attestation workflows
EAS on Celo	Public attestation layer	Attestation records tied to evidence and claims

Tools can change. The durable part of the methodology is the evidence requirement: source, timestamp, location or farm context, payload, provenance, review state, and limitations.

The current MRV payload

The builder documentation defines a shared payload shape for ground, remote, and community observations.

interface KokonutMRVPayload {
  farm_id: string;
  timestamp: string;
  measurement_type: "ground" | "remote" | "community";

  ground?: {
    volumetric_water_content?: number;
    electrical_conductivity?: number;
    soil_temperature?: number;
  };

  remote?: {
    ndvi?: number;
    reci?: number;
    ndre?: number;
    msavi?: number;
    source: "landsat_8" | "sentinel" | "drone";
    image_url?: string;
  };

  community?: {
    crop_cycle_stage?: string;
    plant_health_notes?: string;
    water_analysis?: string;
    soil_texture?: string;
    disease_flags?: string[];
  };

  attested_by?: string;
  attestor_address?: string;
}

This is a shared interface for interoperability. It should not be mistaken for a complete scientific methodology. A serious measurement can also require:

units;
instrument or source metadata;
calibration information;
sampling method;
coordinates or plot reference;
observation period;
baseline;
reviewer;
evidence files;
uncertainty or quality flags.

Changes to the shared MRV payload should be treated as compatibility-sensitive and reviewed before implementation. See the builder primitives →

From record to public evidence

Not every observation needs to be put on-chain. A practical evidence path is:

1. Capture the source

Record the field event, measurement, imagery, harvest, GPS point, photo, or other primary evidence with enough context to understand it later.

2. Structure the record

Associate the observation with a farm, timestamp, measurement type, source, payload, and any relevant crop cycle, plot, operator, or evidence reference.

3. Review the record

Check completeness, units, plausibility, provenance, duplication, and whether the record supports the claim being made.

4. Publish approved records

Published records can feed the Kokonut Hub, APIs, governed metrics, reports, and downstream analysis.

5. Package durable evidence when needed

Important evidence can be referenced through content-addressed storage such as IPFS/Filecoin, preserving a stable pointer to the evidence package.

6. Attest significant claims when useful

Kokonut Intelligence supports EAS on Celo for farm-data attestations. The attestation can link a public claim to evidence, an attestor, and a timestamp.

7. Interpret with the right method

Use the evidence for agronomic analysis, forecast reconciliation, EBF reporting, CRISP review, SDG evidence, governance review, or another defined purpose.

Gnosis and Celo have different roles. Kokonut governance and treasury execution are documented on Gnosis Chain. Farm-data attestations are implemented through EAS on Celo. Governance capital and farm evidence can be connected without placing them on the same chain.

What should be attested

On-chain attestations are most useful when a record has lasting coordination value. Examples can include:

a significant MRV submission;
a harvest milestone;
a farm-data claim referenced by governance;
an impact-report evidence anchor;
a funding or milestone record;
a compliance or partner record where public provenance is useful.

Not every sensor reading needs its own transaction. A high-frequency soil stream can remain in the governed data layer while a reviewed period summary or significant claim receives the durable attestation.

Attest the claim that needs public provenance — preserve the detailed source data where it can be reviewed appropriately.

Claim standard for Kokonut farm pages

A material farm claim should answer as many of these questions as applicable:

Field	Question
Claim	What exactly are we saying happened?
Evidence class	Observation, measurement, derived indicator, forecast/model, attestation, or interpretation?
Farm / location	Where does the claim apply?
Time period	When was the evidence collected or the outcome recorded?
Source	Operator log, sensor, satellite, drone, GPS, harvest record, financial record, etc.?
Method	How was the value measured, calculated, or interpreted?
Record status	Draft, submitted, verified, published, or rejected?
Evidence pointer	Hub record, file, CID, photo, dataset, transaction, or attestation?
Reviewer / attestor	Who reviewed or attested the record, when applicable?
Limitations	What uncertainty, missing data, assumptions, or scope constraints remain?

This standard is especially important for claims about:

harvests;
revenue;
jobs;
carbon;
biodiversity;
soil improvement;
water;
SDGs;
public-goods allocation;
certification;
community benefit.

Forecasts belong in MRV — but as forecasts

MRV is not only about recording achieved outcomes. It also makes planning assumptions testable. The Adelphi Crops & Harvest Forecast contains projected yields and revenues based on explicit assumptions. MRV closes the loop by comparing those projections against:

planted quantities;
survival and loss rates;
actual harvest volume;
realized prices;
crop-cycle timing;
operational interruptions;
field conditions.

The goal is not to defend the original forecast. The goal is to make the next forecast better because actual farm performance is increasingly documented.

EBF, CRISP, and SDGs are interpretation layers

Raw evidence and impact interpretation should remain separate.

MRV — evidence

What happened, what was measured, where the record came from, and what supports it.

EBF — value interpretation

Organizes ecological, economic, social, and sustainability evidence into a broader impact profile.

CRISP — risk interpretation

Surfaces risks that can weaken carbon or ecological delivery claims.

SDG reporting similarly maps documented activities and outcomes to Sustainable Development Goal categories. None of these interpretation layers should create impact simply by being calculated. They are only as credible as:

the underlying evidence;
the baseline;
the reporting period;
the method;
the review;
the disclosed uncertainty.

AI agents can prepare evidence, not self-certify it

Kokonut × AI Agents describes developing automation around MRV and farm data. Useful agent workflows can include:

checking payload completeness;
calculating derived vegetation indicators;
comparing forecasts with actuals;
identifying stale or missing records;
preparing draft MRV summaries;
formatting evidence packages;
generating draft reports;
routing records for review.

Agents should not independently:

certify an impact claim;
declare a farm compliant;
guarantee yield or revenue;
change the Common Data Schema silently;
publish high-impact conclusions without the appropriate reviewer.

The Agentic Marketplace remains in development. Agent-assisted MRV should be described as automation around the evidence workflow, not as an autonomous verification authority.

Data quality matters more than data volume

More records do not automatically create better MRV. A smaller dataset with clear provenance can be more useful than a large dataset with unclear units, missing timestamps, inconsistent sampling, or no review.

Minimum quality questions

Before using a record in a public claim, ask:

Is the farm and location clear?
Is the timestamp or reporting period clear?
Are units defined?
Is the source known?
Is the method reproducible or documented?
Is there a baseline when the claim depends on change?
Is the record complete enough for the intended use?
Has it passed the appropriate review?
Are assumptions separated from observations?
Are limitations disclosed?

Known limitations and failure modes

Credible MRV includes the reasons a conclusion may be weak.

Risk	Why it matters
Missing baseline	Change cannot be interpreted confidently without knowing the starting condition
Inconsistent sampling	Different timing, locations, or methods can create false trends
Sensor or instrument issues	Calibration, placement, failure, and unit errors can contaminate measurements
Remote-sensing limits	Clouds, resolution, canopy stage, sensor differences, and season can affect derived indices
Self-reporting bias	Human observations are valuable but should preserve authorship and review context
Attribution error	A measured change does not necessarily prove that one intervention caused it
Forecast confusion	Planned yield, revenue, carbon, or impact can be mistakenly reported as an actual outcome
Attestation overclaim	On-chain provenance can be mistaken for external scientific validation
Privacy exposure	Not every raw record should be made public when it contains sensitive personal, partner, or operational data
Method drift	Comparing periods can become misleading if formulas, schemas, or measurement procedures change silently

Private or sensitive evidence can remain off-chain while public claims expose only the evidence reference, attestation, or appropriately redacted record needed for verification.

Current maturity

The MRV stack combines live components with infrastructure that is still expanding.

Component	Status	What to assume
Adelphi public farm data	Live	Use the Kokonut Hub as the current public farm-data surface
MRV methodology	Live and evolving	The evidence model is in active use and continues to improve
Ground / field observations	Active	Records should preserve source, context, units, and review state
Remote sensing	Active methodology	Vegetation indicators support monitoring but require interpretation
Kokonut Intelligence	Active infrastructure	Canonical records, review states, analytics, reports, and attestation workflows
EAS on Celo	Live infrastructure	Farm-data schemas and attestation capability exist; check individual records rather than assuming everything is attested
IPFS / Filecoin evidence references	Supported evidence layer	Useful for durable evidence packages; not every record needs content-addressed storage
EBF / CRISP interpretation	Framework tooling	Decision support and reporting methods, not external certification
AI-agent automation	Developing / expanding	Agents assist workflows but should remain review-gated
Multi-farm comparison	Next major test	Shared records must prove they remain useful across additional local contexts

How different readers should use MRV

Farm operators

Use observations and trends to support field decisions while keeping agronomic judgment close to the farm.

Impact and research contributors

Review methodology, baselines, evidence quality, uncertainty, and whether claims match the data.

Governance and capital participants

Use published evidence to evaluate milestones, funding requests, risks, and proposal assumptions without treating dashboards as automatic decisions.

Partners and grant reviewers

Trace material claims back to reviewed records, evidence packages, reporting periods, and documented methodology.

Developers

Preserve schema compatibility, provenance, lifecycle state, and evidence references when building farm-data integrations.

Agent builders

Build bounded automation that prepares, checks, or analyzes evidence while leaving high-impact approval to authorized reviewers.

Contributing to MRV

MRV changes can affect live farm records and downstream systems, so material changes should be reviewed before implementation. Good contribution areas include:

payload validation;
schema versioning;
evidence-quality rules;
field forms;
sensor metadata;
remote-sensing methodology;
forecast-vs-actual reporting;
Hub visualization;
attestation schemas;
privacy and selective disclosure;
EBF / CRISP evidence requirements;
agent-assisted QA;
multi-farm comparison methods.

Open an issue before changing the shared MRV payload, attestation format, evidence lifecycle, or interpretation methodology. These are compatibility-sensitive primitives consumed by farm, reporting, governance, and developer workflows.

Continue through Kokonut Farms

Adelphi — Farm Overview

See how MRV fits into Kokonut’s first live reference farm.

Crops & Harvest Forecast

See how forecast assumptions are designed to be reconciled against actual farm records.

Ecological Impact Frameworks

Understand how EBF and CRISP interpret evidence without replacing the MRV layer.

Kokonut Intelligence

Explore the canonical data, review, analytics, attestation, and agent-access infrastructure behind MRV.

Build with Kokonut

Review the public payload structures, developer primitives, repositories, and compatibility requirements.

Open Collaboration

Contribute to data quality, MRV methodology, farm records, evidence review, or tooling.

Good MRV does not eliminate uncertainty. It makes uncertainty, evidence, provenance, and interpretation visible enough that better decisions can be made.

​MRV turns farm claims into evidence that can be inspected.

Inspect Adelphi's public records

Understand the data infrastructure

​MRV in one sentence

​What MRV can support — and what it cannot

​The five evidence classes

1. Direct observation

2. Measurement

3. Derived indicator

4. Forecast or model

5. Attestation or reviewed claim

Interpretation

​Record lifecycle in Kokonut Intelligence

​The observation stack at Adelphi

​Documented tools and methods

​The current MRV payload

​From record to public evidence

​What should be attested

​Claim standard for Kokonut farm pages

​Forecasts belong in MRV — but as forecasts

​EBF, CRISP, and SDGs are interpretation layers

MRV — evidence

EBF — value interpretation

CRISP — risk interpretation

​AI agents can prepare evidence, not self-certify it

​Data quality matters more than data volume

​Minimum quality questions

​Known limitations and failure modes

​Current maturity

​How different readers should use MRV

Farm operators

Impact and research contributors

Governance and capital participants

Partners and grant reviewers

Developers

Agent builders

​Contributing to MRV

​Continue through Kokonut Farms

Adelphi — Farm Overview

Crops & Harvest Forecast

Ecological Impact Frameworks

Kokonut Intelligence

Build with Kokonut

Open Collaboration

MRV turns farm claims into evidence that can be inspected.

MRV in one sentence

What MRV can support — and what it cannot

The five evidence classes

Record lifecycle in Kokonut Intelligence

The observation stack at Adelphi

Documented tools and methods

The current MRV payload

From record to public evidence

What should be attested

Claim standard for Kokonut farm pages

Forecasts belong in MRV — but as forecasts

EBF, CRISP, and SDGs are interpretation layers

AI agents can prepare evidence, not self-certify it

Data quality matters more than data volume

Minimum quality questions

Known limitations and failure modes

Current maturity

How different readers should use MRV

Contributing to MRV

Continue through Kokonut Farms