EO and in-situ based information framework to support generating Carbon Credits in forestry

Project description

The overall objective of the ForestCO2 project is to develop methodologies and produce a set of geospatial information mostly by using Earth Observation data and machine learning algorithms, that will support Carbon Credit projects based on afforestation. This will enable boosting afforestation projects that will result in the decrease of carbon footprint, capture CO2 in biomass and soil and creation of new economic models. The results would support information-driven decisions in forestry management, increase the profitability of the forestry sector and significantly remove carbon from the atmosphere.

ForestCO2 project aims at leveraging free and open data from the Copernicus missions, together with soil observations, ground truth (in-situ) observations from forest inventory as well as the available other data like meteorological, climate, land use, land cover data and similar. Such a huge amount of freely available data allowed for the design of advanced machine learning techniques able to leverage both the spatial and temporal features of the available data thus providing the reliable information necessary for making the right decision at the right time. The main focus of this project will be related to forest monitoring aiming to support Carbon credit projects, including: finding optimal sites for afforestation projects, providing information for applicability for Carbon credit projects, supporting Monitoring Reporting Verification (MRV) process, soil organic carbon modeling and measuring.

The products designed to fit the requirements of Carbon Credits methodologies will include: suitable locations for afforestation, forest biomass, wetlands data, forest degradation from fires, disease, or cutting, optimal locations for soil sampling and Soil Organic Carbon (SOC) estimation from soil spectroscopy. The products will significantly facilitate Carbon Credit projects implementation and make afforestation economically viable. To train the robust models, comprehensive in situ data on forest biomass, forest disturbances, and SOC content will be used, either from available sources or collected within the scope of the ForestCO2 project.


The main focus of this project will be related to forest monitoring aiming to support Carbon credit projects, including: finding optimal sites for afforestation projects, providing information for applicability for Carbon credit projects, supporting Monitoring Reporting Verification (MRV) process, soil organic carbon modeling and measuring.

Objective 1

Supporting Carbon Credit projects in forestry - applicability conditions and MRV

In the context of carbon credits certification by Carbon offset standards (e.g. Verra, CDM), projects can be registered and implemented if designed in accordance with an existing methodology. Algorithms for the generation of data products based on EO data and collection of the existing geospatial data products from open access will be defined to support the requirements of the Carbon Credit methodologies. These requirements are related to checking if the applicability conditions are satisfied (e.g. project can not be established on wetlands and organic soils) and Monitoring Reporting and Verification process (MRV). We will create our solutions for required data provision to fit VERRA Methodology for Afforestation, Reforestation and Revegetation (ARR) Projects 2 that is based on AR-ACM0003 methodology 3 of CDM.

The project will develop a methodology for assessing baseline and project scenario and additionality. For the MRV process a substantial amount of data has to be collected and processed regularly. We will create the models based on ML and EO data to generate information products that will provide inputs for MRV. The products will include: high resolution land use/land cover classification maps, estimation of live aboveground biomass stocks, identification of deforested areas (due to fires, tree cutting or disease)

Objective 2

Identifying suitable sites for afforestation projects

Finding suitable sites is essential for the success of any afforestation program as site conditions define productivity and resilience of the newly established forest ecosystem, as well as the challenges in the future. To meet this challenge, ForestCO2 will implement Multi-Criteria Land Suitability Analysis (LSA) methods to identify the major limiting factors for different vegetation species. Spatial Decision Support System (SDSS) will be applied to support identification of the most suitable sites for afforestation. SDSS will use a variety of spatial and nonspatial information and at least two known points of land use in history to investigate the effects of different scenarios (i.e., afforestation and alternative land use). To maximize biomass production and ecosystem resilience, ForestCO2 will implement Species Site Matching Tools (SSMT) for selection of the most suitable tree species for afforestation at given sites. 

Objective 3

Developing methodology for high resolution gridded SOC data in forest area in Serbia and optimal soil sampling strategy

Increase of Soil Organic Carbon (SOC) is another result of deforestation projects whose quantification is required for generating Carbon Credits.. SOC is a major source of terrestrial carbon and also a key element for soil quality and fertility representing an important element of terrestrial ecosystems, due to its great potential to affect the climate, food security, forest and agricultural sustainability. The surface reflectance and vegetation indices derived from the optical satellite imaging (Sentinel 2), soil texture derived from radar imaging (Sentinel 1), climatic variables, and terrain factors will be used as covariates in order to build a predictive model for SOC. For that purpose, the performance of several different machine learning algorithms based on in situ measured data values will be tested with the objective to design the best predictive model for a particular site. Soil sampling and analysis is needed to assess and monitor the SOC content, however, it could be a time-consuming and expensive task. To overcome the problem, an optimal sampling algorithm will be developed and validated to determine where to take soil samples on a forest/region. This will significantly lower the costs and effort of verification of C sequestration. 

Objective 4

Developing methodology for SOC estimation based on soil spectroscopy data

Finding inexpensive but reliable ways to measure SOC is essential for maintaining costs of Carbon Credit project implementation low. Reflectance spectroscopy, the measurement of light adsorption at different wavelengths, has emerged as an important rapid and low-cost complement to traditional wet chemical analysis. 

The VNIR spectrum of a soil sample portrays the physical and chemical chromophores present within; in other words, machine learning techniques can infer from a VNIR spectrum key physical and chemical parameters with good levels of accuracy which are instrumental for the health of the soil ecosystem, including: soil organic carbon (SOC), particle size distribution (sand/silt/clay), electrical conductivity, pH, total nitrogen and more. The main objective is to fit calibration models and/or adopt the standard international models to assist with the conversion of soil spectroscopy data to analysis-ready soil data.