March 4, 2025.6 minutes

From credit to credibility: The importance of data verification

Katja Ravn van Overeem, Blayne Lees and Ben Smith

Measurement, reporting and verification (MRV) is a crucial process in the production of soil carbon solutions, monitoring farmers’ transition to regenerative agriculture. The final component, verification, is a key step to ensure the continued credibility of carbon credits. As highlighted in the first blog from this series, the integrity of the voluntary carbon market has recently been under increased scrutiny, with MRV technologies emerging as the linchpin on which the integrity of nature-based carbon solutions hinges.

Agreena is working with more than 2,300 farmers to transition 4.5m hectares of cropland to regenerative agriculture across Europe. This scale highlights the importance of having and evolving a strong verification approach to ensure carbon credits represent real, measurable, and permanent carbon reductions or removals. Over time, our internal quality assurance and quality control processes (QA/QC) have evolved to verify farmer data on a field level at scale. Farmer-provided data is a key input source for carbon calculations, so it is imperative this data is verified to a high standard.

When it comes to verifying farmer data, it’s really rather simple; the farmers provide us with their field data and we verify it. In this blog post we’re going to focus on the first stage of verification – our field boundary verification process. Field boundary verification is a crucial first step in determining if the field adheres to set standards and can proceed for further farm practice assessments. Furthermore, since our remote sensing approach verifying historical and current agricultural practices is dependent on the field area, it is essential that we use an accurate and verified field boundary.

As well as a detailed review of stage one, a high level overview is likewise provided for stages two and three, and future blog posts will dive into the details of how we verify historical and current practice data.

1. 1

   Verifying the field boundary Farmers provide shapefiles and we verify the following:

   Is the field unique and is its area valid; does it overlap with any other field boundaries within our programme; is the field situated within arable cropland; has the field been deforested in the last 20 years; and has the field undergone transition from a natural ecosystem within the last 10 years?

2. 2

   Verifying historical practices (i.e. historic tillage, cover cropping, fuel usage etc.)

   A process to ensure the farmer’s historic practice data is valid and correct using remote sensing, logic based checks, and farmer consultation.

3. 3

   Verifying harvest and practice data from the current year (i.e. current tillage, cover cropping, fuel usage etc.)

   A process to ensure the farmer’s current practice data is valid and correct using remote sensing, logic based checks, and farmer consultation.

Digital measurement, reporting and verification (dMRV) approach is central to the AgreenaCarbon project as well as additional verification methods. dMRV utilises satellite imagery, feeding trained remote sensing models, which can capture what practices are happening, and have historically happened, on the field.

Below, we’ll take a deep dive look into our field boundary detection (FBD) model and explore how it’s used to verify the specific boundary provided by the farmer.

The in-house FBD model utilises remote sensing data, in this case satellite imagery, and a branch of machine learning techniques, known as deep learning, in order to estimate field boundaries. Our FBD model has been trained on millions of data points across Europe over different growing seasons. This provides us with a robust model that can process thousands of fields in a short amount of time at high accuracy. After each growing season we run our FBD model over focal areas and generate all the boundaries for these. This data is then stored within our internal database. When a farmer joins the programme and uploads their field boundaries, we use our internal database of field boundaries to compare the farmer-uploaded boundary to our data. As detailed below, our main comparison is to use a metric called intersection over union. Essentially, this allows us to check that the field boundary submitted occupies the same area both geographically and shape-wise as the detected one. This QA process combines model outputs with real-time human evaluation to ensure that the farmer boundaries represent a real field – and one that we are able to accommodate in our programme.

To run the boundary verification we utilise in-house developed proprietary software displaying all fields in our program. It’s designed to make it easy to narrow down the display to a farm or field level, to directly focus on each farmer's cases.

Different flag colours are displayed to differentiate between the severity and type of cases. Two checks are run – the first is intersection over union.

This is the check where the FBD model comes into play. IOU is a metric used to measure the degree of overlap between the farmer-reported boundary and the FBD model boundary.

It ranges from 0 to 1, with 1 being two boundaries that perfectly overlap with the exact size and shape. This check is done to ensure that the boundary provided by the farmer is aligning with what we are detecting through remote sensing. In the cases where the deviations between the two boundaries are of a certain degree, the fields are flagged and assessed by one of our specialist geospatial quality team. This ensures that we get our eyes on fields that could potentially have an issue and which we are thereby able to correct.

The team will assess each flagged boundary and determine the appropriate solution which can include;
- Using the FBD boundary as the correct area
- Editing the farmer-provided boundary to create a valid field more in line with the model results
- Excluding the field from the programme
- Following up with the farmer to find the correct field area over video call
- Overriding the flag and accepting the original field

An example could be this one below, where we see the FBD boundary (pink) looks more accurate than the farmer provided boundary (blue outline), which seems to be slightly skewed. So, in this situation we would either accept the FBD or follow up with the farmer to confirm which is correct.

Overlapping boundaries are cases where the newly added boundaries overlap with other fields in our program. It’s crucial that these cases are caught and understood to ensure accurate field data and calculations. Fields overlapping with other boundaries are flagged and will not continue on to the next verification steps before the overlap has been resolved – the tolerance for overlap is 0 so any overlap flags are not dismissed. In cases where a solution is not straightforward, a consultation will be taken with the relevant farmers, to understand how the issue has occurred and can be addressed.

Below, we see a case of two fields overlapping slightly. Similarly to the IOU check, the team will assess whether contact needs to be made or if the issue can be addressed as it is minor. Regardless of the outcome, the field isn’t able to proceed further through the verification process without having this overlap resolved.

Following these two checks (IoU and overlapping), all eligible fields are passed on to the next stage of the process, which uses other remote sensing models for historical deforestation, natural ecosystem change, and land usage. This is carried out to ensure that only arable land is considered, and any roads, buildings etc. are disregarded. With this done, the field becomes eligible for Agreena carbon solutions, pending further verification checks – the field's historical and current agricultural practices also undergo assessment to establish their 10 year baseline, before being ready to start issuing credits.

Field boundary verification is a crucial foundation for Agreena’s broader data verification process. Accurate boundaries underpin the validation of historical and current farming activities, ensuring consistency and reliability throughout the programme. By getting this first step right, we strengthen the integrity of our entire verification framework and enhance the credibility of AgreenaCarbon credits, pending verification by Verra.

Orders for Verra-certified carbon credits are open.

Learn more

From credit to credibility: Explore the series

In this series, we're sharing an overview of what it's taken to build a carbon project that meets the highest of standards – Verra’s VM0042 methodology. This series of articles introduces the Agreena team that have worked closest to the project and continue to support the next steps to verification and issuance.

In this series you hear from:

1. Karolina Kenney, Senior Standards Specialist; outlining the processes of building a project with the highest integrity for the market;

2. Katja van Overeem, Blayne Lees and Ben Smith of the Data teams; explaining Agreena's work to ensure integrity of on-farm data capture at scale.

3. Dr Petros Georgiadis, Dr Marcos Alves and Dr Andrew Manderson from the Agreena Science and Statistical teams; unravelling the methodology to calculate soil carbon credits through sampling and modelling together;

Roberta McDonald, Head of Programme

From credit to credibility: Building a high-integrity soil carbon project

Read the full article

From credit to credibility: Quantifying and creating high-integrity carbon credits

Read the full article