Reducing agricultural emissions and creating conditions to financially reward sustainable farming practices are the goals of carbon farming, a concept gaining traction worldwide. Artificial intelligence plays a crucial role in quantifying the benefits and the amount of CO2 captured.
Greenhouse gas emissions are on the rise in agriculture, which is responsible for 11% of global emissions [source: World Resource Institute]. Reducing all climate-altering gases, particularly the most significant one, CO2, is imperative.
In agricultural practice, carbon farming is a method of growing interest for capturing and sequestering carbon dioxide. It involves practices that enhance the capture and storage of CO2 in soil and vegetation, preventing its re-release. Examples include reforestation and the management of peatlands and wetlands.
Agriculture urgently needs to reduce its impact, including greenhouse gas emissions. The European Union, for example, is exploring the most effective ways to achieve this. As part of this effort, the EU has developed a certification framework for CO2 removal (carbon dioxide removal – CDR).
TAKEAWAYS
Europe’s plans on carbon farming
In February 2024, the European Council and the European Parliament decided to establish an EU certification framework for CO2 absorption, implementing a regulation that offers an open definition of carbon removal, in line with IPCC (Intergovernmental Panel on Climate Change) guidelines, which covers only carbon dioxide removal interventions. Among the measures included in the regulation is the temporary storage (lasting no less than five years) of carbon dioxide through carbon farming.
In April 2024, the European Parliament formally approved the provisional agreement reached with the Council of the EU in February 2024 on a regulation establishing the world’s first voluntary certification framework for carbon absorption. Named the Carbon Removals and Carbon Farming Regulation (CRCF), this regulation not only sets EU quality criteria and processes for monitoring and reporting but also aims to facilitate investments in innovative carbon removal technologies and sustainable carbon farming solutions, while simultaneously avoiding greenwashing. Through this regulation, those who remove and store CO2 could potentially receive adequate incentives in the future.
However, understanding how to calculate the amount of carbon dioxide removed and stored through carbon farming remains a challenge. This is where artificial intelligence techniques can be utilised. Various research projects have focused on AI to better manage the quantity of CO2 emitted and stored, as well as to improve agricultural practices and contribute to their sustainability.
What is carbon farming?
In the quest for ways to remove carbon dioxide from the atmosphere through capture and storage in the soil,carbon farming has increasingly garnered interest. The term refers to practices aimed at enhancing the sequestration and storage of CO2 in forests and soils. These practices include a range of activities, from agroforestry to mixed farming that integrates trees or shrubs with crop or livestock management, and soil protection measures such as cover crops and intercropping, reforestation, and peatland restoration [source: EU Commission].
According to the Carbon Cycle Institute, the Natural Resource Conservation Service (NRCS) identifies at least 35 carbon farming practices that improve soil health and sequester carbon, producing significant benefits including increased soil water retention, hydrological function, biodiversity, and resilience.
The study “Carbon Farming – Making Agriculture Fit for 2030,” commissioned by the European Parliament, defines carbon farming as the activity of reducing greenhouse gas emissions at the enterprise level. The term also refers to agricultural management practices that can promote climate mitigation:
«This involves the management of land and livestock, all carbon reserves in soil, materials, and vegetation, as well as CO2, methane, and nitrous oxide flows».
It includes the permanent sequestration and storage of CO2 in soil and biomass, the avoidance of emissions by preventing the loss of already stored carbon, and the reduction of climate-altering emissions. For the US Department of Agriculture, carbon farming means adopting specific on-farm practices to sequester carbon dioxide from the air and store it in soils and vegetation.
Beyond its environmental goals, the practice of capturing and storing carbon in soil can become a highly profitable economic activity. It is estimated that in North America alone, the potential annual supply of carbon credits in the agricultural sector is 326 million tonnes of CO2, compared to the potential demand from corporate buyers, which amounts to 190 million tonnes of CO2 depending on the region. The market is valued at $5.2 billion [source: S&P Global].
Quantifying soil CO2: the role of AI
Given the potential significance of carbon farming in agriculture, understanding how to accurately assess the amount of CO2 sequestered in the soil, including the quantification of potential credits, remains essential. This challenging task can be facilitated by the adoption of artificial intelligence techniques. This was highlighted in an article published in Molecular Plant, titled “Going deep: Roots, carbon, and analyzing subsoil carbon dynamics,” by experts from the Alliance of Bioversity International and the International Center for Tropical Agriculture.
The article explains that the adoption of machine learning models, trained using data from various sensors including electromagnetic induction and gamma-ray spectrometers, has the potential to provide valuable insights into the content and distribution of soil organic carbon (SOC). The study, initially aimed at estimating SOC concentration in Canada at various depths up to one metre, demonstrated that a specific random forest algorithm successfully explained 83% of the variation in the spatial and vertical distribution of SOC.
The researchers assert that if soil carbon could be measured more swiftly, accurately, and over large areas, it would facilitate easier assessment of its contribution. Additionally, farmers could more readily participate in carbon markets.
Active projects worldwide: including an italian initiative
A study conducted by a team from Wageningen University & Research (Netherlands), detailed in “Enabling soil carbon farming: presentation of a robust, affordable, and scalable method for soil carbon stock assessment,” focuses on the importance of increasing soil organic matter to mitigate climate change by sequestering atmospheric CO2. The scientists developed the Wageningen Soil Carbon Stock Protocol (SoilCASTOR), which integrates satellite data, direct proximal soil measurements, and machine learning to estimate soil carbon stocks. Leveraging over 17,000 samples analysed at AgroCares’ internal laboratory and the latest AI developments, significant results were achieved.
Soil carbon stocks are dynamically modelled using machine learning algorithms, achieving high precision estimates with a margin of error below 5%. Moreover, the method requires just 0.5 samples per hectare for farms ranging from 20 to 150 hectares. Tested on various soil types in the United States, it has demonstrated its versatility and applicability to different agricultural contexts.
An international team of scientists coordinated by the Centre for Planetary Health and Food Security in Brisbane, Australia, discussed in the scientific article “Preliminary Results in Innovative Solutions for Soil Carbon Estimation: Integrating Remote Sensing, Machine Learning, and Proximal Sensing Spectroscopy” the potential of integrating remote sensing with machine learning and mid-infrared (MIR) spectroscopy for estimating soil organic carbon. The team conducted a comparative analysis between laboratory-measured SOC values, using 36 soil samples, and found a high degree of correlation, underscoring the potential of this integrated approach.
This approach offers a wide range of applications, including comprehensive soil health mapping and carbon credit assessment.
Additionally, the AI-GROUNDS project, coordinated by Francesco Trovò, a professor in the Department of Electronics, Information and Bioengineering at the Politecnico di Milano, and developed in collaboration with CREA, employs AI for similar purposes. Specifically, it aims to analyse the relationships between soil characteristics, climatological data, and agronomic practices to develop an AI-based decision support system. This system will help agronomists recommend the most appropriate practices for managing soil organic carbon. As explained in a statement, this project: «will enhance the role of agricultural activities in soil carbon sequestration and create new opportunities for the development of carbon farming».
Satellite data and Machine Learning: research prospects
The focus of current research is increasingly on the use of satellite data combined with artificial intelligence techniques. The CERES project, for instance, has been dedicated to developing tools based on AI algorithms aimed at supporting agricultural production and carbon farming practices that facilitate carbon sequestration in the soil, addressing the climate crisis. Specifically designed for agricultural activities in Serbia, CERES holds strategic importance for the national economy.
The project’s concept involves creating models that utilise large geospatial data sets from various sources, including optical and radar satellite images (Copernicus missions), soil data, accurate meteorological data, and textual information available on agricultural internet portals. These models will generate new information automatically, aiding in timely and correct decision-making in agriculture.
Such models enable early identification of changes in crop growth, automated interpretation of the causes of these changes, yield estimation, organic carbon content in the soil, and identification of soil tillage activities.
Boomitra aims to quantify soil carbon increase using satellite and AI technologies to generate verified carbon removal credits. This cleantech startup provides data and forecasts related to carbon credits worldwide, as evidenced by projects in Africa, Mexico, and Mongolia. In Mongolia, it has also signed a collaboration agreement with the local government to develop regenerative agriculture practices. Boomitra specialises in developing machine learning techniques from extensive soil sample databases and satellite images. By carefully selecting satellite images correlated with soil samples, it monitors various useful parameters (such as soil carbon, nitrogen, phosphorus, potassium levels, and moisture) globally, without the need for further soil sampling.
Another intriguing trend in research is the use of deep learning combined with satellite data. A recent example is described in the article “Soil Organic Carbon Mapping Utilizing Convolutional Neural Networks and Earth Observation Data: A Case Study in Bavaria, Germany.” This research utilised Copernicus Sentinel-2 multispectral imagery, appropriately aggregated to extract useful information for soil mapping applications.
The study, conducted in the German federal state of Bavaria, aimed to provide estimates of soil organic carbon. Various time intervals were considered for generating composites, including multi-annual and seasonal, and a convolutional neural network was developed. By leveraging deep learning techniques and complementary information from different spectral pre-processing methods, the research achieved improved overall prediction performance.
Glimpses of Futures
Although several research projects have been initiated to accurately measure the quantities of CO2 captured and stored in the soil through carbon farming, and to make the carbon credit market viable and profitable – thereby promoting sustainable agricultural practices – there are still few methodologies that rely solely on remote sensing and artificial intelligence.
To better understand the prospects opened by this approach, let’s anticipate possible future scenarios by analysing the impacts of the evolution of carbon capture and storage techniques in the soil from multiple perspectives using the STEPS matrix.
S – SOCIAL: Carbon farming involves adopting specific agricultural activities to capture and store carbon dioxide from the atmosphere in the soil and plants. This not only helps improve soil health but also increases crop yields and combats climate change by reducing greenhouse gases. The economic benefits derived from carbon credits and the related market, in addition to environmental benefits, will help agricultural businesses improve their management and operations. As illustrated by the Department of Primary Industries and Regions of the Government of South Australia, carbon farming practices are already helping landowners enhance agricultural resilience and productivity through activities such as agroforestry. Moreover, it allows income diversification through CO2 stored in forest plantations and promotes sustainable landscape management.
T – TECHNOLOGICAL: the use of artificial intelligence techniques, combined with satellite data and remote sensing, is a factor that, as we have seen, allows for improved precision in information regarding the amount of CO2 sequestered in the soil. Some have already adopted carbon farming practices to reduce emissions from non-agricultural activities. An example is the telecommunications company Telstra, which in 2022 launched a land-based business model, utilising internet-connected drones to plant and maintain approximately 158,000 native trees on a 240-hectare experimental site in New South Wales, Australia. This project aims to eliminate carbon dioxide emissions from the atmosphere (Telstra itself aims to store about 160,000 tonnes of carbon dioxide by 2050) while also employing IT technologies. Telstra has partnered with a specialised company to develop drone seeding and use IoT sensors to monitor environmental conditions and weather forecasts. As the company stated, it plans to use robotics and artificial intelligence in the future to improve the management of pests and weeds, alongside drones and sensors to monitor tree health and calculate stored carbon.
E – ECONOMIC: the removal of CO2 from the atmosphere and its capture and storage in the soil can provide substantial economic benefits. Carbon dioxide removal could generate a $1.2 trillion industry [source: McKinsey]. Specifically, carbon farming has significant potential. For instance, in the state of Queensland, Australia, under the federally implemented Emissions Reduction Fund (ERF), 137 projects have been contracted with an expected revenue of $835.3 million in local economy investments over 16 years. If well managed, carbon farming has the potential to provide up to $8 billion by 2030 [source: Department of Environment and Heritage Protection Queensland Government].
P – POLITICAL: within the European Union, as mentioned earlier, the European Parliament adopted the Carbon Removals and Carbon Farming (CRCF) Regulation in April 2024. This regulation aims to establish an EU-wide certification framework for three practices: carbon farming, carbon removal (via direct air capture and storage – DACCS – or combined with bioenergy – BECCS), and carbon storage in products (such as wood products and construction materials). In the USA, the Growing Climate Solutions Act was approved in 2022, facilitating farmers, ranchers, and foresters in participating in carbon markets and adopting climate-friendly practices. Specifically, it aims to help generate and sell carbon credits by establishing a third-party certification process through the United States Department of Agriculture. In 2011, the Australian Government enacted the Carbon Farming Act to establish a voluntary carbon offset programme known as the Carbon Farming Initiative.
S – SUSTAINABILITY: carbon farming is poised to play a key role in the EU’s goal of achieving net-zero emissions. A report by Ecologic and IEEP, titled “Carbon farming co-benefits: Approaches to enhance and safeguard biodiversity,” highlights the need to reconcile biodiversity restoration with climate change mitigation and adaptation through dedicated practices. The authors emphasise that integrating biodiversity safeguards into carbon farming standards «creates opportunities to design standards that can produce biodiversity benefits while minimising risks». For farmers, «including biodiversity co-benefits in a carbon removal certification mechanism offers the chance to receive higher payments for adopting practices beneficial to both climate and biodiversity. Even if farmers opt not to integrate biodiversity-enhancing practices, the design of a certification mechanism could still facilitate funding that benefits biodiversity». The reconciliation of soil carbon capture and storage practices with biodiversity protection is crucial. Soils are currently the largest carbon reservoir, containing 1700 Gt of sequestered organic carbon in the top metre. This amount is four times greater than that in global vegetation, double that in the atmosphere, and 160 times the current annual rate of anthropogenic CO2 emissions, as noted by scientists from Wageningen University in the article “Carbon for soils, not soils for carbon.” They stress the vital role of soils in carbon removal but caution that carbon farming is not beneficial in all conditions and should be adopted on a case-by-case basis.
