AI for Sustainable Farming: Tackling Greenhouse Gas Emissions and Empowering Responsible Finance

Omdena and Agreed Earth built an AI model to estimate farm emissions using real and simulated data. The system helps banks track and reduce agricultural greenhouse gases, supporting sustainable finance and global net-zero goals.

The Problem

Modern farming, especially on a large scale, is not as environmentally friendly as it may appear. Activities such as crop production, livestock farming, and land-use change contribute significantly to greenhouse gas (GHG) emissions and other environmental impacts.

As the world faces the urgent challenge of climate change, financial institutions particularly banks play a growing role in promoting sustainability. Through sustainable finance, banks use their resources to support and reward environmentally responsible practices. This approach is becoming an essential part of global efforts toward sustainable development.

However, in sustainable finance, accurately estimating and reporting emissions from financed agricultural activities remains a major challenge. For banks, understanding these emissions is key to assessing their environmental footprint, guiding investment decisions, and promoting positive change within the agricultural sector.

Despite its importance, this process is complex. Gathering, analyzing, and verifying data on agricultural emissions involves numerous technical and operational obstacles. Overcoming these challenges requires innovative, data-driven solutions that combine technology, science, and financial insight.

The Background

Agriculture is one of the largest contributors to global greenhouse gas emissions. It accounts for nearly 30 percent of the world’s total emissions, driven mainly by the use of chemical fertilizers, pesticides, and animal waste.

This figure is expected to rise as global demand for food continues to increase. A growing population, greater consumption of dairy and meat products, and the intensification of farming practices all contribute to this upward trend. Another major factor is the conversion of non-agricultural land, such as forests, into farmland.

Among agricultural emissions, nitrous oxide and methane are the most significant, together representing more than half of total agricultural greenhouse gas output. These gases have a strong warming effect and are more harmful than carbon dioxide in the long term.

Addressing these emissions is essential. Without intervention, agricultural expansion and unsustainable land use will continue to intensify the effects of climate change. To counter this, innovative methods and sustainable practices are needed to support both productivity and environmental health.

The Goal

This article explores the Agreed Earth Omdena AI Innovation Challenge, where our team developed a machine learning model to estimate greenhouse gas emissions from farming. The model combines synthetic data generated through biochemical simulations with ground-truth data from actual emission measurements.

By improving the availability and accuracy of these emission estimates, the project aims to help farmers adopt sustainable practices and assist banks in advancing their sustainable finance commitments. With reliable data, financial institutions can make informed decisions, support eco-friendly projects, and guide the agricultural sector toward a low-carbon future.

Challenges in Estimating and Reporting Agricultural Emissions

The agricultural sector is complex, and its diverse processes make emission estimation difficult. Collecting accurate data, quantifying emissions, and reporting results all present significant challenges. To reduce environmental harm, there is a strong need to promote regenerative agriculture and lessen the effects of synthetic fertilizers on soil health, water quality, and the nitrogen cycle.

Based on research and reports from the Principles for Responsible Investment (PRI), Ceres, and the Task Force on Climate-related Financial Disclosures (TCFD), we identified several major challenges.

1. Data Availability and Quality

Banks often struggle to obtain complete and accurate emission data from all parts of the agricultural supply chain. Reliable estimation depends on consistent, transparent, and standardized data collection methods.

2. Scope and Boundaries

It is essential to clearly define the scope of emissions to be measured and reported. Collaboration among stakeholders helps establish consistent methodologies and ensures that each actor understands their role in the process.

3. Measurement and Verification

Accurate measurement requires suitable methods and consistent application. Verification of data increases both the credibility and the reliability of emission reports, helping build trust with stakeholders.

4. Supply Chain Complexity and Transparency

Agricultural supply chains involve many actors, intermediaries, and stages. Banks need to navigate this complexity by promoting transparency, collaboration, and information sharing. This helps track emissions more effectively across the chain.

5. Integration of Emerging Technologies

New technologies such as remote sensing, satellite imagery, and IoT sensors can improve emission estimates. However, banks must also manage challenges such as cost, data compatibility, and the technical expertise needed to use these tools effectively.

6. Alignment with Reporting Standards

Finally, aligning emission reporting with recognized standards like TCFD enhances transparency and comparability. Consistent financial disclosures help investors, regulators, and stakeholders make informed decisions and assess climate-related risks more accurately.

Our Approach

The Omdena Challenge project aimed to solve several key challenges by creating a machine learning system to estimate nitrous oxide (N₂O) emissions using input data such as soil properties, weather conditions, and crop details. The first phase of the project involved exploring and analyzing available datasets, especially satellite data. The team identified APIs that provide access to satellite imagery and summarized their functionalities. They also collected information about satellite characteristics, including spatial resolution, temporal resolution, and available data bands.

This research created a strong foundation for developing a reliable model capable of generating accurate emission estimates based on environmental and agricultural data

Available satellite data APIs

Characteristics of available satellites

Knowledge-Guided Machine Learning (KGML)

The project employed the Knowledge-Guided Machine Learning (KGML) framework to enhance N2O emission prediction, crucial for advancing sustainable farming practices. By blending synthetic data with scientific models, KGML aimed for more accurate forecasts by combining scientific principles with data-driven methods. It addressed limitations of existing systems like Ecosys and DNDC, known for their complexity and outdated code bases, thus enabling more effective agricultural practices.

KGML revolutionized model training by first learning from synthetic data generated through process-based simulations and then fine-tuning using ground-truth emissions data. Despite initial challenges in model architecture selection and data availability, the project adopted a sophisticated approach prioritizing direct mapping of relevant variables to N2O emissions. Through meticulous dataset preparation and architectural modifications, the project not only overcame data dependencies but also enhanced predictive accuracy, establishing a robust methodology for N2O flux prediction in agriculture.

Figure 1: Two KGML architectures. The left architecture stacks layers of GRU units and directly maps fertilizer rate, soil and crop properties, weather conditions, and IMVs to the N2O flux. The right architecture contains two independent modules of GRU layers, one for predicting IMVs and the other for predicting the N2O flux (reproduced and adapted from Ref. [6]).

Data Collection

Developing the KGML model for predicting N₂O emissions required two main types of data:

1. Input data for the DNDC model to generate synthetic data for pre-training.
2. Ground-truth data for fine-tuning the model.

DNDC Input Data for Pre-Training

To run DNDC to generate synthetic data, we used the GUI interface of the DNDC software to  supply location-specific input data on climate, soil characteristics, vegetation, and management practice. A tabular dataset was created including daily and annual climate data for the selected years, various soil properties for the specific locations, and crop and management practices for each year. Vegetation was assumed to be crops, and management practices (tillage and fertilizer application) were set specifically for the tests/experiments considered. Multiple DNDC runs were performed for a range of configurations to capture various scenarios.

Running DNDC simulations with these input data generated output files containing daily values for each variable for the selected site, including soil temperature, moisture, oxygen content, microbial activity, pools and fluxes of elements (carbon, nitrogen, phosphorus), soil water, field management, crop information, and grazing. These outputs were essential for further analysis, ensuring a robust dataset for training the KGML model to predict N2O emissions.

Ground-Truth Data for Fine-Tuning

We obtained the ground-truth N2O flux measurements for real UK sites from GHG Nitrous Oxide Datasets in the Agricultural and Environmental Data Archive (AEDA). The raw files from these datasets included most of the required input variables, such as geographical coordinates and daily values for soil moisture, soil mineral nitrogen, rainfall, and air temperature. However, they lacked some crucial variables, which we obtained from other sources. For example, we obtained wind and humidity data from NASA’s data access viewer for the specific sites of the experiments where the N2O flux was measured.

To prepare the dataset for fine-tuning, the raw data was split into time series samples based on location, block number, and treatment. The Harmonized World Soil Database (HWSD) provided sand and silt content, which was used for selecting, renaming, unit-converting, and calculating the variables needed for the model. The datasets were then arranged as a full-year time series for each input variable, with missing time steps generated and filled with known constants and values from the weather data sources. The missing NH4 flux values were filled using interpolation, and the missing N2O flux values were imputed with values predicted by DNDC.

Training and Results

Once the datasets were prepared, the model was trained in two main steps. First, it learned from the synthetic data generated using DNDC, and then it was fine-tuned with the ground-truth data collected from UK sites.

The results showed that the machine learning approach can overcome many limitations of traditional process-based models. By combining both real-world and simulated data, the KGML model successfully reduced the challenge of limited data availability in certain regions.

The findings aligned with scientific research and followed the Intergovernmental Panel on Climate Change (IPCC) guidelines. This confirmed that the model could reliably estimate N₂O emissions in diverse agricultural settings.

Using this system, banks can now improve their estimation and reporting of emissions from financed farming activities. The KGML model enables financial institutions to combine scientific data with AI, allowing for more accurate sustainability assessments and better-informed investment strategies.

Results from pre-training (top) and fine-tuning (bottom) after 1000 epochs.

Outcome

The integration of AI technologies, such as the developed KGML model as a B2B SaaS platform, can empower banks to facilitate sustainable farming practices and support the transition to a low-carbon economy.

The solution offers remote sensing insights on farm-level N2O emissions. Through the utilization of satellite imagery, drones, and other remote sensing tools, the platform collects comprehensive data on agricultural activities, enabling banks to gain valuable insights into emissions hotspots and identify opportunities for emission reduction. The AI-powered analysis and visualization capabilities of the solution empower banks to navigate the complexities of sustainable farming by providing them with actionable information to support decision-making and risk assessment.

The KGML model and the B2B SaaS platform would work in tandem to enhance the accuracy of emissions estimation, improve risk assessment, and promote environmentally conscious lending practices. The use of AI in enabling sustainable farming can also extend beyond emissions estimation and risk assessment. AI technologies can be leveraged to optimize resource management, improve crop yield predictions, and support precision agriculture practices. By analyzing vast amounts of data and generating actionable insights, AI empowers farmers to make data-driven decisions, maximize resource efficiency, and minimize environmental impact. For real-world examples of controlled-environment agriculture using data and automation, explore the top companies in vertical farming.

Real World Applications

Risk Assessment

Understanding the emissions profile of agricultural supply chains can help banks identify potential risks from regulatory changes, market shifts, and climate change impacts. By implementing proactive risk mitigation strategies and supporting the transition to low-carbon agricultural systems, banks can contribute to a more resilient and sustainable agricultural sector.

Market Incentives

Transparent information on emissions intensity incentivizes practices that reduce carbon footprints and promote regenerative agriculture. Banks can play a crucial role in driving the shift towards a sustainable and low-carbon agricultural sector by encouraging farmers and agribusinesses to adopt sustainable practices, invest in renewable energy, and implement climate-smart technologies, 

Standardization and Transparency

Addressing the challenges of estimating and reporting agricultural emissions requires collaboration, knowledge sharing, and the integration of emerging technologies. The KGML model can be used to promote transparency in emissions estimates, enabling banks to work together with financial institutions, agricultural stakeholders, and scientific communities to develop standardized methodologies and share best practices. 

Influencing Policy and Regulations

Banks can leverage their insights and data to support the development of policies that incentivize regenerative agricultural practices, promote carbon pricing mechanisms, and facilitate the transition to a low-carbon economy. Through policy influence and advocacy, banks can create an enabling environment for sustainable finance and drive systemic changes in the agricultural sector.

Other Applications of This Model

1. Transportation

Businesses can optimize routes, minimize emissions, and improve public health by knowing more accurate emission levels in logistics and transit planning.

2. Environmental Monitoring

Environmental Protection organizations can use this model to predict air and water quality, track biodiversity changes, and forecast environmental impacts. It can also analyze sensor data to provide insights into pollution levels, habitat degradation, and climate change trends.

3. Energy Production

The model can be used by energy companies to optimize energy generation processes, improve renewable energy system efficiency, and minimize environmental impacts. 

4. Manufacturing

Manufacturers can optimize manufacturing processes, reduce waste, and minimize pollution.

5. Urban Planning

The model can support sustainable urban planning efforts by analyzing population growth, land use, and infrastructure data. It can also predict urbanization’s environmental impact, assess planning policies, and inform decision-making for sustainable development.

Want to work with us too?

Let’s see if we are a good fit