Cutting Blue Carbon MRV Costs at Scale:
The WeForest Sine Saloum Case
Large scale mangrove restoration leads to climate, biodiversity and social benefits for thousands -
but monitoring isn’t easy
Nature pioneers already restoring with us
7,000 hectares. 700+ polygons. 200km of coastline.
One question: how do you monitor all of this affordably?
The WeForest blue carbon project in Senegal covers over 7000ha of mangroves across the UNESCO heritage Sine Saloum Delta and Casamance river. These regions represent crucial coastal and estuarine habitats for local fish, as well as important wintering sites for many species of birds on the East Atlantic Flyway. However, almost 25% of the country's mangroves have been lost since satellite monitoring became available in the 1970s, due to factors including droughts and unsustainable wood harvesting practices. To reverse this trend, the WeForest project is protecting and restoring many of these crucial mangrove habitats.
The project generates multiple environmental and socio-economic benefits beyond carbon sequestration. Between 2020 and 2022, 7,019 hectares of degraded mangroves were restored, improving coastal ecological stability, reducing erosion, and enhancing natural protection for coastal communities. From a social perspective, more than 4,400 community members participated in training, awareness raising, and livelihood support activities. Livelihood diversification initiatives—such as beekeeping, oyster processing, salt production, aquaculture, and agriculture—have supported over 4000 beneficiaries, many of them women involved in local value chains. In addition, more than 8,900 community members benefited from awareness campaigns, restoration participation, and community strengthening activities, contributing to improved well-being across villages in the project areas.
Ecologically, the restoration efforts support improved habitat quality and ecological connectivity, benefiting several vulnerable fish and bird species associated with mangrove ecosystems. The project also strengthens local monitoring capacity and long-term scientific knowledge, with trained community monitoring teams and regular biodiversity monitoring campaigns.
Figure above shows the scale of just half of the project, indicating the large geographical spread of many small polygons along the Casamance river in Senegal.
This large scale project (consisting of over 700 polygons spread out across 200km) is the result of a collaborative effort of many stakeholders and partners, from government and local communities to local technical contractors. Crucial for project financing and effective management however is efficient monitoring over this large scale and difficult to access ecosystem, where Inverto Earth has contributed their technical assistance in the analysis of UAV data collected by the local Senegalese drone service provider Earth Geomatique.
From forest to finance: the MRV chain that unlocks blue carbon credits
As a brief background to the importance of digital MRV to the project, a large source of the funding that enables this climate and biodiversity action is generated through voluntary carbon credits. The first step in generating these carbon credits is measuring the biomass within the project boundary at different points in time, and from this the equivalent tons of CO2 sequestered by the project. This measured value of sequestered CO2 is what is used to calculate the amount of carbon credits the project can issue, after being confirmed through an audit from an independent Validation and Verification Body (VVB). Crucially, the biomass of the mangrove forest must be known with confidence intervals of 10% or smaller, which requires a large amount of ground sample plots across the project area.
Measuring the biomass at a single sample plot (in this case sample plots are circles of 10m radius) requires taking measurements from every tree meeting the inclusion criteria within the sample plot boundary, as well as the GPS location of the centre of the plot. The plot biomass is then calculated by applying a species and region appropriate allometric equation to these measurements. In this project, the exact tree measurements taken include the height of each tree and two canopy diameters, which are then used to calculate biomass based on published allometric equations from other projects within the same delta.
_edited.jpg)
Image showing team on site taking tree measurements from within a sample plot
This alone is a significant task, as the sample plots are typically difficult to access, underwater for half the day, and even walking through them can become extremely difficult as the mangroves grow denser. For this reason, reducing project monitoring costs is dependent on minimising the number of sample plots required to meet the target confidence intervals. For this, optimisation of the sampling strategy is required, and currently the project is using a stratified sampling method to ensure the sample plots provide as much information as possible about the total project biomass.
The objectives of the drone campaign went beyond simply estimating tree biomass. Specifically, they included:
-
Improving stratification to better account for mortality and plantation success.
-
Categorizing planted areas by actual density and biomass for more accurate carbon accounting.
-
Exploring how future monitoring could integrate drones or high-resolution imagery to reduce reliance on labor-intensive biomass measurements, potentially incorporating radar or LiDAR as trees grow larger. This also includes correlating biomass plots with imagery to reduce the number of plots needed or eventually replace them.
-
Mapping pre-existing vegetation to refine the planted area and enable more precise carbon accounting.
Fewer plots, same confidence: the case for smarter sampling
Standard error of a study variable, Soil organic matter in g/kg vs number of samples for an example case taken from here.
Stratified random sampling is a common method used in forestry inventories to improve the statistical efficiency of sampling. This is performed by grouping the area to be measured into regions of different strata, based on characteristics that correlate strongly with biomass such as planting age, NDVI from satellite data, or canopy coverage. The variation of biomass within each strata is smaller than the variability in biomass over the entire area. As an example of how such a strategy can improve sampling results, the plot above shows the standard error of a study variable (soil organic matter in g/kg in this case) vs the number of samples. The points labelled SI are the results from simple random sampling (no stratification), STSI(Neyman) are the results using stratification with Neyman allocation, and STSI(prop) use proportional allocation. In all cases, stratified sampling methods significantly improve the standard error at the same number of sample sizes, or allow a target standard error to be reached with much fewer samples. For example, if a study was targeting a standard error of 8 g/kg in this case, using a stratified sampling method would allow this error to be met with over 25% fewer samples (26 samples vs 35), which is a significant reduction in sampling costs. Or, if a study used 25 sample points, a stratified sampling
method would increase the accuracy from an error of 11g/kg to 7.5g/kg, an improvement of 30%. For stratification to be effective however, strata need to be divided based on meaningful characteristics or covariates that relate to biomass.
One option for stratification over large areas is to use satellite data. This can be very effective at massive scales and for established forests, but in the initial years of plantation small saplings are not visible at the 10m resolution of most freely available satellite data. For mangroves, in the first one to two years after planting, the sapling canopy diameter can be in the order of 10cm or less, which requires very high resolution imagery to be detected. To meet this, drone imagery has been used.
Commoditised tech and local experts
Drones are no longer the “high tech” and complicated instruments they used to be only a decade ago, their costs have dropped dramatically, their usability and reliability has increased, and they have become commoditised almost to the level of smart phones. They are being used in every country in the world, and one can find proficient and experienced local drone service providers for any location on the planet. Collecting drone data does not require importing expensive foreign experts. Networks of drone service providers can be found on multiple platforms, services such as Globhe or networks such as Flying Labs make this expertise local. For this project, the local Senegalese company Earth Geomatique was contracted to collect 7000 hectares of high resolution RGB imagery.
Why mangroves are the hardest ecosystem to monitor - and how local teams delivered
The scale of this project, and the difficult environment of mangrove forests, make drone data collection from these environments non-trivial, which again highlights the importance of working with local experts such as EGEO. Data needs to be captured at low tides, avoiding solar noon (to minimise glare in the images), and many of the survey regions are remote and difficult to access. The tidal requirements mean the data collection windows can be quite short, only a few hours per day, and some areas may always have some presence of water affecting the data collected. EGEO has done an incredible job in collecting this aerial imagery at this massive scale, in difficult conditions, all at a ground resolution of below 3cm per pixel, ten times higher than the best commercial satellite imagery.
From raw imagery to stratification-ready data: delivered remotely
With this enormous amount of data (over 4 terabytes), the next step was to turn this into usable data for sampling stratification, which was the role of Inverto Earth. We received the high resolution imagery collected by EGEO remotely, without needing to travel to the site. The key metrics analysed were sapling counts (with detection rates of at least 80% or higher), sapling canopy coverage and the canopy coverage of pre-existing (non-planted) mangroves. The sapling counts also serve the purpose of estimating sapling mortality in the different regions of the project.
The imagery captured was extremely varied over the project, with different lighting conditions, presence of water and algae, and the roots and washed up branches of dead trees all providing additional challenges to the task of identifying and counting saplings. Through a combination of our algorithm, using both thresholding and machine learning methods, we were able to meet the requirement of 80% sapling identification accuracy across the surveyed area, as well as measure the sapling canopy area and pre-existing canopy coverage.
The image above shows a planted region with the small saplings identified and positioned with red points, and the larger pre-existing vegetation segmented in light blue. The small saplings have also been segmented, in green in the image below, which is only possible due to the resolution of this imagery.
This segmentation of the smallest saplings provides estimates of canopy size, which is strongly correlated with biomass and growth. These small saplings would be barely visible in high resolution satellite imagery, and it might be another year until they reach a size at which they could be measured using even high resolution commercial satellite imagery.
Pre-existing trees are segmented to ensure that they are excluded from the carbon accounting calculations, as they cannot be considered additional. The locations and coverage of pre-existing trees is recorded so they can continue to be monitored in the future as well.
In many cases the young saplings are even smaller than shown in the image above, without significant canopy that can be detected. In these cases features such as their shadows and stems are crucial to detecting them, which wouldn’t be possible in satellite imagery.
These results have been provided to WeForest, who have finalised the stratification and have begun the auditing process in February 2026. Working with local community members they have collected data from 481 ground sample plots which will be stratified into 5 strata based on 5 different levels of canopy coverage. WeForest keeps a track of annual mortality in the entire project area, and plans to do biomass measurements in these specific plots every 3 to 4 years for carbon accounting.
Mangroves are some of the most challenging ecosystems to access and measure, with very short windows for image acquisition due to the limited times of appropriate tides, the presence of near-permanent water in certain areas causing reflections, and sun glare. Sampling methods that can reduce the MRV effort while maintaining accuracy targets can quickly result in significant project cost savings, particularly for large scale projects. Drone and satellite data are very powerful tools for this, and can be applied either for high resolution stratification techniques or even for advanced model based approaches, which can be used to map out biomass at a high resolution over the entire surveyed area. These biomass maps can lead to crucial insights on growth and highlight regions with potential hydrology issues or other risks in time to take preventative actions. The tools to do this are available, already in the hands of local experts, and can reduce everyone’s time spent wading through the mud.
If you’re interested in learning more about advanced mangrove MRV methods for carbon or other restoration projects, get in touch.