Conference Agenda

The emergence of carbon markets as a crucial tool for incentivizing greenhouse gas emissions reduction has necessitated the establishment of reliable and transparent monitoring mechanisms. This work highlights the potential of employing Earth Observation data for the Monitoring, Reporting & Verification of carbon projects within a carbon marketplace. Two effective methodologies have been proposed to estimate carbon stocks, crucial indicators of a vegetation ecosystem's carbon sequestration capacity, initially estimated as above-ground biomass (AGB) and then converted into carbon stocks using empirical rules.

The first method, "ReUse: REgressive Unet for Carbon Storage Estimation," employs deep learning to estimate global carbon sequestered by greenery. Utilizing publicly available AGB data from the European Space Agency's Climate Change Initiative Biomass project, a time series of Sentinel-2 images and a pixel-wise regressive U-Net are employed to estimate the carbon sequestration capacity of any land area. Incorporating Sentinel-1 satellite radar images and Digital Elevation Models enhances the model, enabling a more precise estimation of global carbon stocks. This tool offers quick estimates even in challenging conditions, such as after fires or hard-to-reach areas.

The second method, "Forest Carbon Stock Estimation Using Machine Learning Ensembles: Active Sampling Strategies for Model Transfer," targets localized regions instead of providing global estimations. The method uses active sampling and satellite imagery to pinpoint these cases' most significant data collection points. Based on Shannon's entropy for sample selection, the approach innovatively transfers a calibrated regression model between different areas using an active-learning methodology after initial calibration in a reference area. Various sampling methodologies and regression strategies have been explored to minimize fieldwork while maintaining estimation accuracy, which leads to a reduced set of points for new ground truth data retrieval, lessening the need for physical measurements. Experimental results show that this blend of regression ensembles and active learning techniques significantly reduces field sampling while providing carbon stock estimates equivalent to conventional methods.

The two methods provide complementary approaches to carbon stock estimation: a global approximation for difficult-to-access or rapidly changing areas and a focused, localized analysis with minimized field sampling.

Finally, a carbon marketplace, AICarbonHub, has been proposed. The market matches the compensation needs of businesses and citizens and the need to recover money to maintain and improve the quality of green spaces by property owners. Integrating the approaches above into the marketplace, we could monitor and check carbon stored continuously, ensuring the credibility and accuracy of the carbon credits sold.

Although gas exchange between the atmosphere and cropland can be measured with high precision in the field, it remains a challenge to assess it at the plot scale and over large areas due to the wide range of soil and climate conditions and crop management practices.

In this study, we aim to develop three remote sensing based approaches of increasing complexity to assess the carbon budget of cropland.

The first and simplest approach is based on the work of Ceschia et al., (2010) and relates the number of days of active vegetation (NDAV) to net ecosystem productivity (NEP). The NDAV is weighted by complementary variables such as temperature and radiation, allowing the effect of climate on CO2 fluxes to be taken into account.

Although it can be applied anywhere and very easily, this method remains inaccurate because it treats all crops as identical.

The second approach is based on the SAFYE-CO2 model (Pique et al., 2020b). It is a parsimonious agro-meteorological model, crop specific, operating on a daily basis that was obtained by coupling SAFY-CO2 (Pique et al., 2020a) and SAFY-WB (Duchemin et al., 2015). It allows estimation of the green area index, the production (biomass, yield) and the CO2 and water fluxes. The main advantage of this approach is that it requires few input data and no information on management practices, allowing the model to be applied to large areas. In addition, the effect of weeds, cover crops or regrowth on CO2 fluxes and hence the carbon budget can be taken into account by using remote sensing to drive the model.

Finally, a machine learning (ML) approach is also developed to estimate CO2 fluxes and hence carbon budget. Based on daily CO2 flux data from the Integrated Carbon Observation System (ICOS) a ML algorithm is designed considering features as remote sensing indices, meteorological data and soil maps.

However, it should be noted that all developed methods require plot scale data on carbon inputs from organic amendments and straw management in order to estimate field carbon budgets.

We believe here at DigiFarm that the future of a reliable carbon market in agriculture will be driven by extensive MRV initiatives, which includes continuous monitoring and change assessment of agricultural land. This is not an easy process and will require the entire ecosystem to contribute, we also see the increased focus on MRV and it's expected reliability will increase and come under scrutiny (also as a result of the outcome of early carbon market players such as Verra) we already see this change in increased monitoring of changes in agricultural land from CAP and IACS/LPIS regulations in the EU-region among national paying agencies. We believe the private sector will follow and adapt similar frameworks as the EU-commission has put in place in the public sector. Additionally, there will certainly be a need for both on-the-ground truthing and reference data along with remote sensing channels and as all in-field analytics, large or small, require both super-high resolution SatEO (this due to 83% of all the agricultural field in the world are smaller than 2 hectares) and an accurate way of remotely identifying agricultural parcels, the only way to do this effectively is automatically with sophisticated AI- models (as manually digitising field boundaries every year is too time-consuming and expensive). In order to develop a sustainable carbon market which will benefit all stakeholders, contribution from all angles is critical, and no one player will solve this challenge alone and we believe DigiFarm can play a role in this through providing the industry with the most fundamental data layers required to do this analysis: this starts with detecting field boundaries and seeded acres automatically using high-resolution SatEO (1m per pixel resolution, deeply resolved Sentinel-2) with the ability to recognise changes in activities and productivity on any particular parcel of agricultural land.

Seagrass ecosystems are among the most important organic carbon sinks on Earth, playing a key role as climate change buffers. Posidonia oceanica is an endemic seagrass species from the Mediterranean Sea and has been observed to feature the highest carbon stock among all seagrasses. The Balearic Islands’ (Western Mediterranean) coastal waters host a vast area of seagrass meadows - around 633 km2 which can grow in depths of up to 45m among which 95% are P. oceanica meadows, approximately. These meadows have been declining in the region, threatened by anchoring, human pressure and an increase in water temperature, highlighting that continuous monitoring is needed. It is important to monitor its distribution by using field campaigns, however, it is expensive in resources and only covers a very limited area. In order to generate a synoptic tool for seagrass monitoring, we used the Copernicus Sentinel-2A/B satellite imagery at a 10-meter spatial resolution to generate a multi-temporal composite (2019-2021) of the Balearic Island coastal waters within the Google Earth Engine cloud-computing platform. Scalable machine learning algorithms have been applied to optimize sunglint, seagrass detection and benthic habitat classification, obtaining a spatial-explicit seagrass map of up to 30 m depth. After estimating the seagrass extent, regional in situ carbon stock data and other available sources were standardized and processed to estimate the blue carbon stock of the mapped seagrass extension. Moreover, we estimated the carbon stock in P. oceanica meadows by monetizing it in EUR per megatonne (Mg) of CO2. The generated data highlights the key role of the seagrass ecosystem in climate change mitigation in the Mediterranean Sea and the need to establish a baseline for seagrass ecosystem monitoring and management. This information aims to support the development of blue carbon strategies with scalable, time- and cost-efficient monitoring solutions within and beyond the Balearic Islands.

The need for information on the amount of carbon emission and removal in different regions is pervasive across various domains. Therefore, reliable and precise estimations of carbon emission and removal are required. In Thailand, Land Use, Land-Use Change, and Forestry (LULUCF) sector has significant contribution to greenhouse gas emission. These LULUCF activities contributed to a tremendous increase in net removal from the atmosphere since 2000 due to the net removal of rubber plantations. These rubber plantations have large potential to sequester atmospheric carbon into the biomass and soil. However, various measurement methods apply general emission factors and do not establish a clear baseline for rubber plantation. Traditional biomass assessment methods based on field survey to estimate carbon stock are difficult to conduct over large areas and are costly, time consuming, and labor intensive. Observation from space is now regarded as the best technical means for large scale monitoring and estimating of vegetation biomass and productivity. In this research, we used CO2 flux from Eddy Covariance technique (EC) and spectral information provided by satellite remote sensing from in combination with machine learning to estimate CO2 emission and removal rate from rubber ecosystem in Chachoengsao province, Eastern Thailand, between 2013-2019. The results show that (1) the combination of MODIS and Global OCO-2 SIF (GOSIF) images as predictor variables in the linear regression model provided the best CO2 emission and removal rate estimation, and (2) annual CO2 sequestration from 2013 to 2019 averaged 42.8 tons CO2 ha-1 yr-1. This research provides a new way of estimating CO2 emission and removal rate for the rubber plantation based on remote sensing data through the synergy of different sensors datasets and modeling algorithms. Additionally, rubber plantation can potentially support the achievement of Thailand’s emission reduction target from the LULUCF sector, which is expected to be a net sink of carbon by 2050.

This research analyses with a non-cooperative game under which conditions it becomes incentive compatible for the project owner and the buyer of offset credits to participate in the strategy in which Satellite EO-based Monitoring, Reporting & Verification (MRV) systems are used versus the strategy in which they are not used. Considering the voluntary carbon market, one concern relates

to the credibility and transparency of offset credits. Monitoring, reporting, and verification (MRV) systems based on satellite EO have the potential to increase transparency, accuracy, and reliability in the issuance of offset credits, thereby increasing convenience for both the buyer and the project owner. To determine the conditions under which both actors choose to participate in the voluntary

market with satellite EO-based monitoring, reporting, and verification (MRV) systems, the functions of the payoffs of the two actors in the two different strategies with and without satellite EO-based MRV systems are determined.

In order to solve the game theory model, the system of inequalities

with the payoff functions is solved by isolating the parameters and the conditions are identified for which the strategy with Satellite EO-based Monitoring, Reporting & Verification (MRV) systems is incentive compatible to both. The use of MRV systems based on satellite EO might lead to an increase in the total utility of project owners and buyers of offset credits under certain conditions.

This study emphasizes the need to adopt EO-based monitoring, reporting, and verification systems for voluntary carbon market growth

The expansion of agricultural land, particularly for the production of commodities such as soy, beef, palm oil, cocoa, coffee and wood, is the primary cause of deforestation and forest degradation. As a major economy and consumer of these commodities associated with deforestation and forest degradation, the EU acknowledges its partial responsibility for this problem. With this in mind, the EU has introduced a new regulation (2021/0366(COD)) aimed at minimising the consumption of products associated with deforestation or forest degradation. The regulation also aims at increasing demand for legal and ‘deforestation-free’ commodities and products. The regulation requires mandatory due diligences to be carried out by operators placing specific commodities on the EU market, including among others wood, as well as some derived products.

This study investigates the sustainability of supply chains within the corporate wood, food, fashion and metals industry, to better understand the potential need for EO-enabled solutions to improve business processes and transparency.

We assessed the impact these regulations may have on corporates in the wood, food, fashion and metals industries, as well as the role EU Space data and signals can play in monitoring the sustainability measures implemented by the regulations and the corporates, support the reduction of emissions or pollution by corporates, provide information on risk parameters and potential solutions, and help improve corporate processes to contribute to sustainability measures.

The adoption of regenerative agricultural practices is gaining traction as an approach to enhance soil health and sequester carbon to combat climate change. Different sustainability frameworks and programmes are now incentivizing producers to transition to regenerative farming. These initiatives need robust and transparent Measurement, Reporting and Verification (MRV) platforms to track cropland practices and outcomes. To help scale initiatives we use an automated approach currently quantifying rotations, tillage practices, cover crops, and soil carbon across European agri-landscapes.

Our approach leverages multi-source Earth Observations (largely Sentinel-1, Sentinel-2), process-based soil models, independent ground truth, and surveys. A current example is tracking tillage practices with tens of thousands of training data gathered across Europe. Independent training data is collected seasonally to ensure robust and transferability. Together, these multi-source data feed into machine learning models (LSTM, CNN, gradient boosting) to classify features such as tillage practices and cover crops. Withheld independent observations and data science best practices are used to gauge model performance and accuracy depending on regional schemes and landscape practice variability. With this approach, the Community of Practice can robustly track field conditions over seasons and feed downstream applications, such as estimating Soil Organic Carbon (SOC) and impacts of practices. Combining these tools with open operational data streams such as Copernicus, we look forward to helping grow regenerative agriculture impacts and carbon farming initiatives across Europe.

The present piece provides insights into an exploratory study focused on the development of an advanced pipeline for land use and land cover (LULC) high-resolution classification and subsequent estimation of blue carbon stocks and sequestration rates for two Areas of Interest (AoI) that include a spectrum of environments: Ria Formosa National Park and Ria de Aveiro Natural Reserve. The initial phase of the pipeline focused on LULC classification of GEOSAT-2 imagery in the RGB and NIR bands, in orthorectified standard (3m) and resampled (1.6m) pixel size formats. Sentinel-2 and Landsat 8 imagery from the same period was also retrieved and collocated with the geospatial high-resolution layers from the Copernicus Land Service (i.e., coastal zones, tree cover density and type, imperviousness), for training data labelling purposes. Throughout this investigative study, several supervised classifications were evaluated—both pixel and object-based— by using several machine learning (ML) algorithms (e.g., SVM and RF). The aim was to achieve an optimal overall accuracy of, at least, 80%, while maintaining the highest resolution possible for the LULC classifications and devising a scalable solution. Subsequent phases involved the estimation of carbon stocks in two distinct pools: above-ground biomass (AGB) and soil organic carbon (SOC). Given the limited information on carbon values for specific locations and LULC classes, the construction of a database of carbon densities was initiated, encompassing all available in-situ measurement data from existing literature. This resource enabled more informed decisions regarding the values used per LULC class present in the AoI. From a broader perspective, the pipeline's objective is to facilitate the monitoring of carbon stocks and sequestration over time by incorporating changes in LULC, carbon stock fluctuations, and ecosystem dynamics. This aspect of the pipeline will employ a time-series approach, offering an upgrade from the static nature of previous models and enabling more dynamic and precise evaluations. However, to counteract potential inaccuracies and enhance precision, future iterations should test the integration of alternative data sources (e.g LiDAR) for extrapolating biomass information.

Ultimately, this research aims to enrich the fields of environmental monitoring, carbon accounting, and climate change mitigation strategies. Moreover, it is poised to become an indispensable tool for policymakers, environmental managers, and researchers who need accurate, up-to-date data on LULC change and carbon dynamics across multiple scales.

Keywords: High-resolution satellite imagery; LULC classification; Machine learning; Carbon stocks.

Credible monitoring of forest carbon requires availability of reliable information on forest biomass. The most exact way to get this information is to make measurements of forest attributes on ground and transform these measurements to biomass and carbon. However, such approach is often too expensive or unfeasible as an exclusive approach.

In this paper we present a concept, in which the reference data are collected with limited amount of ground measurements or even without any ground observations. The latter approach is based on conducting two independent surveys whose results are compared. The result of the survey can be considered reliable, if the two independent approaches give similar enough results.

We demonstrated the concept by developing a method for the assessment of forest area and structural variables in the cases when the availability of representative ground reference data is poor, and these data are not collected from the whole area of interest. The two tested independent approaches for the estimation of forest variables of European boreal forest were: i) computation of wall-to-wall estimates using moderate-low resolution VIIRS imagery of Suomi NPP mission; and ii) visual interpretation of plots of samples from Very High Resolution (VHR) satellite data obtained with a two-stage design.

The forest area prediction from VIIRS for the whole study area was 1.2% higher than the VHR-based result. All other structural variable predictions using VIIRS fitted within the 95% confidence intervals computed from the VHR sample except estimates of the main tree species groups were outside the limits. Comparison of VIIRS based forest area estimates with Finnish and Swedish NFI data indicated 10.0% points and 4.6% points overestimation, whereas the total growing stock volumes were overestimated by 8% and underestimated by 3.4%, respectively. Our concept using VIIRS data with coarse resolution was applicable for the estimation and overall mapping of the forest area and central structural variables at regional to national levels.

The same approach can be applied for higher resolution wall-to-wall imagery, such as Sentinel-2. If it is possible to measure part of the VHR image plots on ground, reliable information on the structural variables and carbon of the forest can be produced with the same two-phase sampling of the entire area of interest. The sample data can then be used to correct the bias of wall-to-wall mapping. The approach is further developed in the ESA-funded project https://www.forestcarbonplatform.org/ .

Ref. Tuomas Häme, Heikki Astola, Jorma Kilpi, Yrjö Rauste, Laura Sirro, Teemu Mutanen, Eija Parmes, Jussi Rasinmäki, Mohammad Imangholiloo 2023. Forest Area and Structural Variable Estimation in Boreal Forest Using Suomi NPP VIIRS Data and a Sample from VHR Imagery. Subm. to Remote Sensing.

The Cerrado formations (Brazilian Savanna) represent a diverse and unique tropical savanna biome. Despite its ecological significance and rich biodiversity, the Cerrado remains one of the most threatened biomes globally, primarily due to agricultural expansion and land-use changes. Understanding the biomass dynamics within these formations is crucial for effective conservation. Challenges for biomass estimates in the Cerrado include the heterogeneity of vegetation types, complex vertical structure, and the influence of anthropogenic activities and climate change on wildfire regimes and vegetation dynamics. Precisely estimating carbon stocks is a crucial undertaking for both climate change mitigation and conservation initiatives, particularly in the context of REDD+ projects. Traditional approaches, relying on costly field-based carbon measurement methods, encounter constraints related to their spatial coverage, impeding comprehensive and accurate assessments across expansive forested areas. To address this challenge, remote sensing with UAV-LiDAR emerges as a valuable solution that has the potential to overcome these limitations. By harnessing remote sensing technologies, carbon stock estimates can be scaled up, fostering transparency in carbon reporting. However, achieving dependable outcomes mandates the development of precise regional models and the adoption of high-accuracy techniques to ensure reliable and robust results. In this sense, we assess the capability of high-density UAV-lidar to estimate and map the aboveground carbon density (AGCt) of live trees in the Brazilian Savanna (Cerrado) using 45 field-based sample plots (30 m × 30 m). Three generalized least square regression models were tested, and we considered the presence/absence of palm trees as variable to improve precision in biomass estimates. We hypothesize that the presence of palm trees is an indirect indicator of higher soil moisture, occurrence of trees with higher canopy cover and biomass. We found that considering the presence of palm trees as a biological index of areas with higher carbon content improves the model accuracy by 27%. Comparing the LiDAR estimates and field-based control plots, we found a global uncertainty of 2.1% and sub estimates of 1.6% when applied to 1000 iterations of the resampling procedure. These results demonstrate the feasibility and potential of UAV-lidar technology in accurately estimating and mapping AGCt in the Cerrado. The findings of this study serve as a benchmark for future research endeavors aimed at generating accurate carbon maps and providing baseline data for the efficient management of fire and predicted climate change impacts on tropical savanna ecosystems.

Agriculture can make a significant contribution to reducing atmospheric CO2 by reducing GHG emissions through appropriate farming practices and by actively sequestering atmospheric CO2 in the soil through the targeted build-up of humus/organic carbon in the soil. Furthermore, more resilient farming practices such as crop rotation and planting catch crops make a significant difference regarding a more sustainable food production.

To harvest the power of this potential, it is necessary to be able to monitor, report and verify (MRV) the measures to conserve and bind CO2 in the soil as a part of Carbon Farming. Earth Observation data, being available globally in high spatial resolution, containing relevant information about the vegetation and soil, and being available long-term – both in hindsight and in the future - is a uniquely qualified data source to develop a tool fulfilling these needs.

Within the ESA Artes Business Application project “EO4CarbonFarming”, we aim at developing such an Earth Observation-based MRV tool that will be able to monitor catch crops, report on farming measures for higher carbon sequestration and verify organic carbon build-up in the soil. For this, we utilize high resolution optical and radar data (Sentinel-1, Sentinel-2), sophisticated pre-processing methods, radiative transfer modelling and newly developed dedicated algorithms for soil organic carbon estimation.

The MRV-tool indicators carbon stock in biomass and soils, catch crop detection and proof of crop rotation will be visualized in the ‘EO4CF Cockpit’ and also provided via API to the documentation system of the service customers and users.

At the workshop, the user needs and requirements from different sectors such as carbon farming sponsorships or machinery manufacturers as well as the ways how such an MRV can fulfil them will be shown. This elaborates one way how a more transparent and reliable performance assessment of sustainability measures and more resilient farming practices and their impact can be achieved.

Reliable determination of field carbon stocks is key to monitoring the climate impacts of carbon farming and to developing business models.

This project is funded within ESA Contract Number 4000134139/21/NL/MM/mr.

The organic carbon (OC) stock in the soil is a crucial parameter for mitigating the effects of global warming due to its potential as a carbon sequestration indicator. Efficient carbon sequestration requires careful consideration of field practices and cropland areas, as these factors significantly influence the carbon status in the soil. Soils have recently gained importance in the global carbon agenda for climate change mitigation as part of the CO2 sink. An increase of only 4% in global OC stocks within the top 35 cm layer of non-permafrost soils would be equivalent to annual atmospheric CO2 sequestration. Therefore, monitoring the OC stocks in the soil can serve as an indicator of the climate change phenomenon. To model the carbon stock using remote sensing (RS) techniques, we employed a hyperspectral sensor in the Campania region of southern Italy. This involved utilizing a soil spectral library (SSL) based on the Campania region in conjunction with an aerial hyperspectral image obtained from the hyperspectral airborne AVIRIS NG sensor from NASA JPL. The study area selected for this investigation was an agricultural field situated in the Sele-River area, which is part of Campania region of southern Italy. The outcome of this study includes four raster layers with a very high spatial resolution (1m) map, representing the OC stocks, as well as three other relevant soil attributes: OC content, clay content and bulk density (BD) value (0-30cm). We observed that the spectral absorption of clay minerals at 2200 nm, associated with OH in clay minerals, significantly influenced the prediction of the examined soil attributes. By applying the model on a pixel-by-pixel basis to the reflectance cube of AVIRIS-NG, we generated a quantitative map that was validated through in-situ observations. The results demonstrated the potential of combining SSL datasets with high spectral/spatial resolution RS data to estimate various essential soil attributes, particularly carbon stock.

As carbon project consultants, working from initiation to credit sale and managing projects from tree to customer, we at FORLIANCE have realized that bridging the communication gap between project developers and corporations is critical to the success of carbon projects.

To increase communication options between project developers and customers, we have developed a customizable solution as a digital monitoring and verification platform. Using project information, processed and analyzed field data, and satellite imagery from Sentinel-1 and Sentinel-2, we present our clients with a package that allows them to ensure the transparency of their projects on various levels.

We perceive the need for more transparency and credibility as one of the VCM’s biggest challenges. What makes our solution unique is that we digitally monitor the progress of projects and enable communication of project-linked information using one easy-to-use dashboard infrastructure. We believe this is essential to improve trust while implementing carbon projects.

Our poster will highlight our digital Monitoring and Communication Platform.

The Monitoring and Communication Platform comprises our in-house developed modules and sub-modules, such as above and below-ground biomass estimation models, volume calculations, carbon projections, stored CO2 evaluations, and other customized key performance indicators, for example, land use change and cover and planting areas.

This enables project developers to monitor projects and communicate their impact to clients and stakeholders.

The Monitoring and Communication Platform is a web and mobile-based application visualizing data on the modules and sub-modules. In addition to visualizing and communicating, our Platform enables logged-in users to analyze the data for monitoring information on the specific project site.

In our poster, we will share experiences of implementing this platform which was designed using open-source geospatial infrastructures and is currently in a beta phase, aimed to be upscaled by the end of this year.

Our unique aspects are customizability and strong project focus, which blend in with our company’s unique portfolio as we offer consultancy at every level of carbon project management. With this expertise, we can understand the variety of needs in the carbon sector and provide practical solutions.

Effective carbon sequestration assessment in forests demands precise data regarding the age and above-ground biomass (AGB) of constituent species. This study presents an innovative approach to estimate these parameters in Para rubber (Hevea brasiliensis) forests using satellite imagery and terrestrial laser scanner (TLS) technology. Our methodology capitalizes on the unique electromagnetic interaction signatures reflected in satellite imagery, attributable to varying canopy characteristics at different Para rubber growth stages. This information provides a basis for rubber age classification, supplemented with field surveys and other trustworthy data sources. This remote sensing technique provides a non-invasive, comprehensive, and timely means of assessing forest age distribution. Simultaneously, we employed TLS technology for an exhaustive field survey to estimate carbon sequestration. The high-resolution, 3D point cloud data from TLS effectively captures spatial variations in biomass density across different forest types, enabling a detailed analysis of forest carbon dynamics. The location for sampling was guided by previous AGB map data, using a line transect sampling method within a 20x20m plot with an inter-plot distance of 200-500 meters. The gathered data were processed using the R program, elucidating a significant correlation between the rubber trees' age and carbon sequestration potential. Specifically, our results highlighted an increase in carbon sequestration from 10.07 tons per rai for trees aged 7-14 years, to 40.26 tons per rai for trees older than 20 years. This finding underscores the value of mature rubber forests in carbon management strategies. By harnessing the potential of remote sensing and TLS technologies, we provide an efficient, scalable solution for monitoring forest carbon dynamics. Our study paves the way for improved forest management strategies, contributing to sustainable carbon sequestration efforts crucial in combating climate change.

Keywords: Carbon Sequestration, Para Rubber, Terrestrial Laser Scanners, Aboveground Biomass (AGB), Satellite Imagery.

Soil organic carbon (SOC) sequestration is considered an efficient and actionable solution for reducing atmospheric CO2, whilst at the same time improving chemical, physical, and biological soil characteristics. By adopting practises that capture carbon into the soil, farmers significantly contribute to addressing this global challenge. Carbon Credits provide an incentive to help farmers overcome the perceived risks of adopting SOC improving practices. Today, measuring and monitoring SOC requires thorough soil sampling and analysis which account for a significant proportion of the carbon credit generation cost, limiting the return on investment for companies generating these credits.

constellr develops a constellation of state-of-the-art high-resolution visible, near and thermal infrared satellites, planned for launch in 2024, to measure carbon, water and temperature. Leveraging its proprietary data, imagery from public missions and strong remote-sensing analytics we provide a fast and highly scalable solution to optimize soil sampling and estimation for SOC monitoring. This solution combines a stratification approach with an Organic Carbon Stock (OCS) value range measurement.

On the one hand, the parcel stratification approach combines vegetation patterns with bare soil thermal characteristics and topographic information to generate strata of consistent soil OCS at the field level. This mapping of homogeneous areas within a field provides the base for an optimized stratified soil sampling and enbales reducing the number of samples required to characterize field’s SOC variability. Hence, significantly lowering field data acquisition costs and outperforming simple random sampling.

On the other hand, constellr’s SOC product estimates, from the soil reflectance spectra and local soil profiles, the corresponding range of OCS (t/ha) in the top 0-30 cm of the soil for each stratum, at pixels and field level. Such estimates of SOC stock support efficient baseline mapping and enables the prioritization of fields to implement carbon projects.

Accurate estimations of carbon stocks in tropical forests are a critical challenge for climate change mitigation and conservation projects. Conventional field-based carbon measurement methods are expensive and limited spatial coverage, hampering comprehensive and precise assessments in vast forested regions like the Brazilian Amazon. Remote sensing can overcome these limitations, potentially scaling up carbon stock estimates and enhancing transparency in carbon reporting. However, achieving reliable results necessitates the development of accurate regional models and the use of high-accuracy techniques. LiDAR (Light Detection and Ranging) remote sensing has emerged as a powerful tool for tropical forest monitoring. By generating detailed three-dimensional forest models, LiDAR complements satellite image-based models, enabling more precise and comprehensive estimates of carbon stocks while facilitating the monitoring of forest degradation and carbon fluxes. This technology enhances the accuracy, transparency, and reliability of carbon credit calculations. The Brazilian company BrCarbon has successfully employed LiDAR remote sensing and embarked on Unoccupied Aerial Vehicle (UAV) in two REDD+ projects in the Brazilian Amazon. The second project, the Brazilian Grouped Project (Verra ID 2551), encompasses 27,182.20 hectares of avoided deforestation, totaling 18,497,463 VCUs (2021-2030 para Brazilian Amazon, e 2022-2031 para o Cauaxi / baseline period). AUV-LiDAR provides highly accurate and detailed data acquisition capabilities, facilitating identifying subtle changes in forest structure indicative of degradation, such as biomass loss. Furthermore, AUVs enhance monitoring efficiency, allowing for broader and more frequent coverage of forested areas. BrCarbon's implementation of this technology in REDD+ projects in the Brazilian Amazon demonstrates promising results in carbon estimation and ongoing forest degradation monitoring.

Integrating LiDAR and drone-based remote sensing offers a comprehensive solution for obtaining precise and timely information on carbon stocks and forest degradation in tropical regions. This approach can potentially revolutionize forest monitoring practices, contributing to improved conservation and climate change mitigation efforts in the Brazilian Amazon and beyond.

Acorn is an initiative of Rabobank, which supports small holder farmers in their transition to sustainable agroforestry practice by giving them access to the voluntary carbon market. Agroforestry is the intentional combination of agriculture with forestry, such as planting trees and bushes on pastureland. This traditional farming principle offers farmers significant ecological and economic benefits and makes an important contribution to climate change mitigation.

To quantify the carbon sequestration of trees on agricultural land, we use remote sensing data to estimate above ground biomass in agroforestry. The values are then converted to carbon removal units, following a certified ACORN methodology, which ensures that uncertainty, leakage, additionality is accounted for. Strict eligibility criteria are followed where each farmer plot is checked for illegal deforestation or non-sustainable use of additives.

Our approach currently includes 10 project areas in 17 EcoRegions across 8 countries in Latin America and Africa.

We use a global machine learning model, trained and validated on 16,884 plots referred to as ground truth data. All trees in each sample plot are hand measured for diameter at breast height, and tree height. Tree height measurements are further enhanced with airborne Lidar data collected from 5 of the projects. Species inventory is then used to convert the measurements into biomass values representative for the plot area using a generic allometric equation. The model is built to establish the relationship between in-situ data and remote sensing imagery. It combines machine learning method, physical modelling and crop recognition, developed for biomass estimation. Sentinel 1, Sentinel 2 and ERA5 data were fused at a feature level for training a light gradient-boosting machine (LightGBM) model. Various features including spectral and texture features were extracted from these datasets. The model is currently used on approximately 300,000 farmers in the 10 projects.

Our study shows how remote sensing data can be successfully used for above ground biomass estimation in agroforestry.

The REDD related poster that SarVision would like present is about the question how to facilitate smaller organizations that would like to join a REDD initiative to generate carbon credits. Often complexity and lack of capacity and finance make organizations hesitant to join a REDD program. RS systems can help them. Traditional forest carbon assessments require a lot of biomass related field sampling to meet strict standard (e.g. VERRA) requirements and are therefore difficult to implement and costly. RS based assessments can reduce the number of required field samples. With SarVision's SarCarbon system it is possible to produce detailed biomass baseline maps and to monitor biomass related loss, gain and leakage.

SarCarbon not only includes biomass losses as a result of deforestation, but also losses caused by forest degradation. SarVision uses its unique SarSentry forest monitoring system to map and quantify biomass loss caused by forest degradation. At present SarSentry is the only wide scale RS forest monitoring system that includes forest degradation e.g., as a result of (illegal) selective logging, and we believe it will play an important role in the development of a new standard for the Estimation of Emissions Reductions from Avoiding Unplanned Forest Degradation.

During the poster sessions SarVision will present some practical examples and implementation of its innovative multiple-sensor based SarCarbon system that is used to assess and monitor biomass and carbon for forests, and for individual trees in a tree-scarce environment.

SarVision invites organizations that are interested to join the carbon market to come to the poster exhibition. We would also like to discuss with other companies, that offer REDD and carbon related services, how these services can be offered best to meet the requirements of both the clients and selected standard.

Blue carbon ecosystems—seagrasses, mangroves and tidal flats—provide globally significant yet vastly

underestimated and impacted ecosystem services to humans, biodiversity and economies like carbon

sequestration, coastal protection and biodiversity maintenance—the so-called natural climate

solutions. Accelerating climate change, biodiversity loss, uneven levels of protection, and infancy in

pertinent spatially explicit accounts and frameworks are all significantly stressing these physical and

financial benefits of coastal ecosystems, necessitating cost- and time-effective contemporary

technologies.

Here, we present the novel coastal ecosystem accounting framework of the Global Seagrass Watch

service. Our developed scalable ecosystem accounting framework blends modern Earth Observation

advances—cloud computing, artificial intelligence, big satellite data analytics— with high-quality field

data collections and other public geospatial datasets, across multi-national and multi-annual scales.

We showcase the scalability, effectiveness, and confidence of our Earth Observation technological

framework through its recent applications across both tropical and temperate coastal biomes.

Leveraging our cloud-native framework within the Google Earth Engine cloud platform, we nationally

aggregate high-resolution analysis-ready mosaics using the open Sentinel-2 and NICFI PlanetScope

image archives. These analysis-ready image pixels at 5 and 10 m resolution are then transformed into

physical seagrass ecosystem extent, condition (e.g., bathymetry, water quality) and service (e.g., blue

carbon stock and sequestration rate) accounts. Our showcased national spatial seagrass accounts to

date cover more than 76,000 km2 of areal extent across around 74,000 km of coastline in 28 countries and three seagrass bioregions, encompassing the Western Indian Ocean, the entire Mediterranean,

and the Caribbean.

We discuss the real-world impact of our ecosystem accounting technology towards climate change

mitigation in a recent blueprint project for the Seychelles. We also articulate current technological

challenges and respective research and development solutions, and near-future opportunities and

applications towards the full operationalization of Earth Observation for transparent and effective

coastal ecosystem accounting, decision making, financing and resilience—within and beyond the 21st

century.

High quality estimations of carbon removals together with certification frameworks that avoid greenwashing are two major priorities of the European Union (EU). The achievement of these objectives requires comprehensive approaches that integrate positive and valuable features of innovative technologies. The assimilation of Earth Observation (EO) data into process-based models leveraging machine learning, geostatistics, and cloud computing is a game changer for the development of operational and cost-effective Monitoring Reporting and Verification (MRV) approaches. However, heterogeneous environmental and regulatory contexts have hindered the implementation of such frameworks for the EU land use, land-use change and forestry sector (LULUCF). To overcome these challenges, MRV4SOC assimilates high quality in-situ and EO data into statistical and process-based models in 14 Demo Sites (DS) covering 9 Land Use and Land Cover (LULC) classes i.e., croplands, grasslands, pasture, agroforestry, forests, peatlands, wetlands, paludiculture, and peri-urban areas subject to conversion. DS are located in 5 European countries and 1 Associated Country with heterogeneous pedoclimatic conditions. MRV4SOC accounts for changes in as many carbon pools as possible to estimate greenhouse gas (GHG) and full carbon budgets, couple carbon and nitrogen cycles, quantify Soil Organic Carbon (SOC) accumulation, and assess the results of traditional management practices and carbon farming. To successfully achieve these objectives, MRV4SOC proposes a Tier 3 methodology applied to different spatial levels (i.e., sub-landscape and landscape) and temporal scales (i.e., long-term experiments, new observations, and future climate change and land use change scenarios) to assess robustness, transparency, scalability, standardisation, and cost-effectiveness towards a EU MRV framework for the LULUCF sector. MRV4SOC will target 6 specific objectives: i) to measure long-term SOC accumulation in 9 EU representative LULC classes, ii) to assess how carbon farming practices drive carbon flux dynamics in the 9 LULC classes, iii) to assess the impact of climate change on SOC accumulation, iv) to develop a robust, transparent, standard, and cost-effective MRV to facilitate results-based payments associated with carbon farming practices, v) to seek out revenue opportunities to unlock results-based payments, vi) to increase stakeholders’ faith in Voluntary Carbon Markets. MRV4SOC identifies barriers and enablers to adopt carbon farming practices in each region and integrate outcomes towards policy recommendations. The results of this 3-year project will be shared with local stakeholders and key actors to increase their faith in Voluntary Carbon Markets and leverage results-based payments.

Certified carbon credits have become an essential tool for financing ecosystem restoration programs, agro-forestry development projects and cities wishing to invest for a more resilient environment. But obtaining these credits requires a reliable identification and validation and a precise monitoring and evaluation over time. Various models based on satellite data are already used to assess the impact of these projects on carbon sequestration and in monitoring green gas emission reductions as well. Everimpact and VisioTerra have collaborated on various projects to develop or enhance algorithms using hybrid approaches combining ground and satellite sensing.

In Carbon tracker, a Connect by CNES sponsored projects, EVERIMPACT has deployed 75 ground sensors over the area of Dijon Metropole and assimilated measures in a satellite imagery derived urban structure model and CFD to account in a reliable manner the performance of Dijon climate plan implementation. Furthermore, in cooperation with VISIO TERRA, a sequestration monitoring component was developed to provide a compete carbon neutrality monitoring solution to the client city.

EVERIMPACT and VISIOTERRA have developed at the call of a major Japan-based multinational corporation to feed their new service offering to Japanese community forest owners to evaluation forest carbon sequestration performance, guide forest management enhancement and facilitate certification of additional carbon sequestration achieved.

At the call of another Japanese company, EVERIMPACT and VISIOTERRA are developing a satellite imagery-based detector of Moringa oleifera fields and carbon sequestration performance.

In support to a major international research institute and a renowned bilateral development agency, EVERIMPACT and VISIOTERRA are developing a landscape sustainability enhancing tool dedicated to guide coffee expansion to provide additional ecosystemic services including landslide risk reduction, biodiversity recovery and carbon sequestration.

Mitigating the effects of climate change requires combining projects of various sizes, on the scale of cities, countries or regions. For example, Everimpact, VisioTerra and their partners have offered to provide the traceability solution of two major African initiatives: the President Buhari's Pledge which provides for the restoration of 4Mha of degraded land in Nigeria and the Great Green Wall which aims to restore 100Mha in 11 countries from Senegal to Djibouti, to sequester 250 million tons of carbon and create 10 million green jobs. These two projects must be completed in 2030 thanks to public and private financing, including carbon credits. Thanks to satellite sensing, other initiatives of the kind will multiply on all continents in the years to come, using the tools developed.

One of the most significant human-caused sources of methane emissions is rice cultivation. Changing rice-growing practices may reduce greenhouse gas (GHG) emissions. "Alternative wetting-drying" (AWD) is one method of growing rice that is environmentally friendly. AWD aims to use less water in rice cultivation, which alters the soil chemistry and reduces methane production without impacting rice yield. As Thailand's most widely grown crop, rice takes up 40% of all cultivated land. It can generate significant tradable carbon credits with the GHG emission countries. However, assessing AWD using Internet of Things (IoT) devices may be impossible to implement in all of Thailand’s paddy fields. As a result, GISTDA proposed a new validation technique using satellite-derived parameters to assess the AWD method and increase the feasibility of selling carbon credits from rice farming in Thailand. Our research in central Thailand's Ayuthaya province has shown that the Normalized Difference Moisture Index (NDMI), Normalized Difference Water Index (NDWI), and Band 9 (water evaporation) from Sentinel-2 are potential parameters to assess the adoption of the AWD method in paddy fields. These parameters have lower values in areas of AWD practice when compared to non-AWD. Also, rice farms tend to practice AWD when the NDWI and NDMI are lower than 0.2 from day 15 to day 60 of rice cultivation. Based on these preliminary results, the following study will expand to incorporate Sentinel-1a data into the evaluation method. In addition, the eddy covariance flux tower will be used to compare these satellite parameters with direct measurements of methane exchange in AWD and non-AWD paddy fields. This new method will enhance the feasibility of trading carbon credits from rice farming and increase farmers' income. Furthermore, it will also widely promote the use of the AWD method in paddy fields, which will significantly reduce GHG emissions in the future.

The goal of the FAO OpenForis Initiative (openforis.org) and SEPAL (sepal.io) is to remove the barriers to earth observatobservation and to allowoin data access and cutting-edge processing methods so anyone, anywhere can produce sophisticated, useful, and actionable results, especially where these results inform locally relevant decisions. By enabling better decision making and targeted action, OpenForis and SEPAL will likely improve natural resource management and facilitate increased ecosystem and human resilience to otherwise high-impact environmental perturbations.

To ameliorate, manage, and potentially reverse the negative impacts on human life and livelihoods caused by climate change, accurate information on natural resources is required to catalyse good decision-making. This information, data, and methods need to be both consistent across large geographies and available to people in situ, at the point where decisions are being made. In many cases people most vulnerable to changing environmental conditions and emergencies are also located in areas traditionally underserved by technology capable of increasing resilience and informing positive actions.

All OpenForis software developed by FAO Forestry is free and open-source, enabling autonomous processing and analysis of geospatial data for customized land and forest monitoring by a broad spectrum of stakeholders. The tools and platforms available through OpenForis empower users to process satellite data, create customized maps, and detect land cover and land-use change. They also provide many other functions critical to effective land management without the need of coding skills. The tools work seamlessly with modern geospatial data infrastructures, such as Google Earth Engine, further driving the generation of high-integrity forest and land-use information that help users answer complex questions about any area of interest, land managers make more informed decisions, and countries to attract finance for forest-related climate action.

As a global digital public good, the OpenForis initiative promotes and strengthens the collaboration between space agencies, intergovernmental and non-governmental organizations, research institutions, and civil society to effectively deliver the services promised by Big Data analytics, including the dissemination and exchange of results, which can improve the accessibility and transparency of data.

Various applications of the OpenForis and SEPAL will be presented in the framework of capacity building and use in developing countries for national forest and land monitoring including carbon accounting.

The voluntary carbon market is a vital mechanism for businesses and individuals to voluntarily compensate for their carbon emissions. Forestry projects, including forest conservation, planting and restoration, and improved management, are responsible for the majority of carbon credits issued from nature based solutions. BeZero assesses the efficacy of carbon credits, including forestry, by measuring if carbon projects deliver on their climate promises. Satellite imagery, along with machine learning, offers innovative ways to monitor changes in forest extent, health, and carbon stocks at scale.

Multiple global forest monitoring products are available from space agencies, academic labs, and research institutes, but often these consider forest or non-forest as binaries. For carbon credits issued by projects located in open forest systems, assessing quality requires that we monitor forests across a wide spectrum of tree cover. Regional differences in forest definitions and underestimates of tree cover in dryland ecosystems from global monitoring approaches pose further challenges for assessing ecosystem change and carbon stocks of seasonal open forests.

At BeZero, we are developing the targeted tools required to monitor forest dynamics in the most challenging of settings. For dry forests, our analysis begins with characterising vegetation phenology, to understand the times of year when tree crowns are most distinct. Sentinel-2 provides over seventy multispectral images per year for all global land areas at 10-m spatial resolution, but cloud cover can limit data availability during growing seasons. The algorithms we are developing optimise for cloud-free imagery, while tuned to the specific phenology of the project’s forests.

Information on tree height is derived from spaceborne LiDAR, and overlaid with high-resolution (e.g., 50 cm) commercial satellite data in samples across the project landscape. Using these data, we label individual tree crowns with high spatial precision. Our machine learning models use the labels for training and quality assessment, and extrapolate to the wider landscape based on the seasonal mosaics from Sentinel-2. Accurate forest monitoring provides novel insights on the quality of forest carbon offsets and transparency in the voluntary carbon market.

A carbon offset credit is a transferable instrument that represents a reduction in emissions of one metric tonne of CO2 or an equivalent amount of other greenhouse gases. The carbon offset credit can be purchased to claim emission reduction in relation to an organization’s own unmet targets. Earth observation actors could participate directly in the implementation of emission reduction projects. Calculating and monitoring avoided emissions thanks to satellite data would allow programs such as Copernicus to claim a share of these benefits. The paper seeks to assess the potential feasibility for Earth observation actors to issue offset credits by analysing regulatory, technical, and economic dimensions. Studies have shown that Satellite EO-based Monitoring, Reporting & Verification (MRV) systems have the potential to enhance transparency, accuracy, and reliability regarding offset credits, in addition that the programs themselves play an active role in reducing emissions. An analysis of the literature on voluntary carbon markets together with the literature on the distinctive characteristics of earth observation entities is presented. The Copernicus space component is used as a single case study to present the positive impact on emission reduction of earth observation satellites. Copernicus Space Component is the data collection element of the EU Space Programme Copernicus, an earth observation programme that generates and uses global satellite data to help various actors in civil security, environmental management and climate change comprehension and mitigation. Further research might identify a methodology for measuring avoided emissions using satellites and demonstrate the feasibility for earth observation actors to issue offset credits with a dedicated use case.

Carbon accounting is a continuing effort to combine sustainability with market forces. Merged disciplines, improved measurement, and regulatory tailwinds are supporting a new era of applications with carbon as the currency. The market, however, has not been without problems. Driven by the commodification of carbon credits and a lack of monitoring, reporting, and verification, many credit purchasers are unable to measure progress on their sustainability goals. In the worst cases, the carbon credit is invalid and does not reduce CO2e at all.

Companies buy a carbon credit for its impact. For some, this is a simple accounting of emissions with the aim of claiming that they are carbon neutral. But this is changing. The carbon market is already seeing buyers who target specific outcomes. They want to reduce emissions within their own supply chain, conserve land in areas close to their activities, and regenerate the natural systems that produce their raw materials. Measuring tons of CO2e reduced through credit purchase on an open market is no longer adequate for modern corporate sustainability goals. Sustainability is more than just carbon.

With the current carbon market, the deforestation of monocropped new growth timber is treated as equivalent to an old growth rainforest. From only a carbon credit perspective, these actions are the same. Critically, biodiversity and all of its value is not considered. Essential ecosystem services, genetic information, and natural resilience need to be included in the financial mechanisms of the carbon market. In 2023, the first commercial sale of biodiversity credits marks the start of the next era of green finance. In the new carbon market, targeted impacts and holistic projects will better include all aspects of natural welfare. To avoid the previous pitfalls of carbon credits, technology must be applied from the start for more transparency and better verification.

Earth observation (EO), especially Copernicus, provides low-cost global coverage and near real-time monitoring. EO approaches for Essential Biodiversity Variables (EBVs) must be combined with in-situ measurements, verification methodologies, and new field collection technology to define these emerging biodiversity credits. This talk will explore the overlaps between EO state-of-the-art for carbon accounting and the current status of biodiversity credits. Using a real-world project example, it will present a roadmap for project development and outline the challenges faced for the future.

Adding the value of biodiversity into carbon markets will require a comprehensive approach that combines in-situ and remote sensing data. Through a real-world example of a nature-based carbon offsetting project, Sylva will demonstrate the role of Copernicus in a biodiversity credit's evaluation, investment case, in situ collection planning, and MRV (monitoring, reporting, and verification). Essential Biodiversity Variables and their measurement are considered within the context of Earth observation capabilities and the cost of field data collection. Sylva also explores the potential contribution of future missions such as ROSE-L and CHIME using current missions with similar hyperspectral and SAR products.