Advancing Crop Analytics for Smallholder Farmers
The Enabling Crop Analytics at Scale (ECAAS) team has been diligently working to address unprecedented challenges to the agricultural sector resulting from the interconnected impacts of climate change, water scarcity and food security. As detailed in an earlier Agrilinks post on strengthening the collection and sharing of agricultural data, the Bill and Melinda Gates Foundation-funded ECAAS initiative is scaling the development and use of advanced analytics in smallholder farming communities to increase agricultural productivity and resilience against shocks and stressors.
The ECAAS team is expanding our community of practice, refining and promoting the use of agricultural data collection, and developing new tools and resources, including public training datasets and a podcast! To expand visibility, we attended the successful Cracking the Nut and Market Systems Symposium virtual events. Digging deeper, we sponsored and presented at two conferences this past spring — ICTforAg and World Agri-Tech Summit. While the participants varied widely between the ICTforAg and World Agri-Tech conferences, several overlapping themes emerged, which are summarized below and contextualized through the ECAAS frame.
Recurring Themes from Ag-Tech Conferences
- High-quality ground truth data gaps persist across applications and use cases. Participants across different use cases continued to highlight the need for high-quality, ground truth training data with proper spatial, temporal and metadata characteristics. Collecting the type, quality and quantity of data needed to develop trustworthy applications or services requires significant time and resources. In addition to filling specific use case needs, the community continues to request benchmark datasets to benchmark model performance and create a more transparent analytics ecosystem.
At the ICTforAg session on Data-Informed Decision-Making, Dr. Catherine Lilian Nakalembe, associate research professor at the University of Maryland, talked about current limitations to the use of machine learning, noting that, “Fundamentally, ground data collection is critical. Having a good representative dataset is important. Both across space and time. From a data science perspective, getting a ground truth point is really hard.” David Guerena, an agronomist with QED, reiterated the difficulty, saying it would take one person 170 years to delineate cropland by hand across the African continent. With a well-trained model based on ground truth training data, a cropland detection model could process the same amount of data in a number of hours or days, bringing insights and analysis to researchers, governments and businesses, including farms.
The increased access to and use of high-quality data remains a core ECAAS objective. In coordination with Radiant Earth, ECAAS has piloted various techniques to drive down the time and cost of data collection and increase data sharing by developing replicable standards and repositories where others can access data. In addition, we are exploring how to bridge these and other repositories through a common search function to make finding a geo-specific dataset as easy as possible. For now, existing machine learning-ready data repositories are listed on our website.
- Farmer-centered approaches support solution permanence. Participants referenced “Ag-tech burnout” at several conferences, which often results from a mismatch between current market solutions and farmers’ needs. Burnout lowers farmers’ willingness and capacity to welcome new technologies and services. This is a crucial moment for ag-tech, where market solutions should be tailored to solve specific, clearly identified problems.
At the World Agri-Tech conference, participants discussed the need to overcome ag-tech burnout through very intentional, user-centric design. Specifically, services should be built in direct partnership with agricultural users and designed to work within existing day-to-day operations. At World Agri-Tech, the Farmer Business Network exemplified farmer-first approaches to product design by curating a farmer-to-farmer agronomic information network that increases supply chain transparency.
Participants in the ICTforAg session reiterated the importance of contextually specific solutions for local farmers. ICTforAg speaker Claire Rhodes from Producers Direct specifically recommended, “Design in-person and digital solutions from the foundation of farmers’ reality, knowledge and experience with the ultimate aim of improving farmer livelihoods, financials and resilience.” Solutions should not be proposed without consultation with an agricultural community.
Within the context of ECAAS, the standards and products that we emphasize must reflect the reality of how farmers do business, and put farmer autonomy and needs first instead of imposing new ways of working. We continuously challenge our assumptions by asking key questions like, “Are the benefits lasting, viable and/or sustainable?” to ensure that long-term sustainability plans fit farmers’ needs.
- Emerging importance of carbon markets and their role in climate change mitigation. Carbon markets and ecosystem service programs provide crucial pathways to reward farmers for improved agricultural management practices, which some studies say could increase annual carbon storage in soil by up to 21% from current levels. Carbon market applications rely on accurate monitoring, reporting and verification (MRV) systems to guarantee transparency, precision, comparability and compensation for carbon information — but reliable data are often lacking. Crop analytics realized through remote sensing and machine learning approaches can help overcome this challenge.
In the rapidly emerging domain of carbon markets, advanced analytics can help estimate and inform both carbon capture, related offsets and MRV for environmental social governance (ESG) goals. Experts from World Agri-Tech explained that since carbon is not a physical commodity, tracking efforts require accurate digital records, or “digital twins,” as noted by Chris Harbourt from Indigo Ag. These systems are rapidly expanding and include core ground truth data, such as field sizes, boundaries and yields.
While much of the carbon discussion at World Agri-Tech focused on markets in developed countries and the connection to corporate ESG goals, organizations at ICTforAg were generally more focused on smallholder farmers around the world. Existing carbon marketplaces are often set up with the “off-set” buyer or corporations as the primary target and beneficiary. To scale these applications to the level required to meet the enormous carbon challenge, we need to flip the script to boost farmer participation in carbon marketplaces by placing smallholder farmers at the center of the equation.
The ECAAS team will continue to explore the many opportunities to help redirect carbon markets to focus more intentionally on smallholder farmers as both primary actors in the equation and as important data contributors to refine and improve MRV modeling. Successful approaches often simplify entry requirements, create systems that are easy to navigate and comply with, and bundle carbon market services with other extension or advisory services. Accurate ground truth data will continue to play a critical role in producing accurate carbon measurements and could be a potential use case for ECAAS’ future ag-data accelerator, should the community continue to prioritize this technical domain.
- Data sovereignty often determines who can access and benefit from data. Data sovereignty, also synonymous with data governance, refers to the degree of control an individual or organization has over the data they generate and use. In agriculture, farm data can be generated by growers, farm agents or via remote sensors, resulting in data with various levels of identifiability, marketability and commercial value. Farmers have different levels of ownership and control over the different data sources, resulting in potential exploitation.
ICTforAg speaker and data protection expert, Foteini Zampati, advised that more legal frameworks be developed around agricultural data privacy in low- and middle-income countries, similar to data rules and protections already in place in developed countries. Data policies can protect farmers against data asymmetries and knowledge imbalances between stakeholders because some farmers have limited access to technology and the data they generate.
At World Agri-Tech, participants highlighted the importance of data privacy and consent. However, some companies spoke in favor of bypassing farmer consent practices for data collection campaigns, as consent was viewed to inhibit the collection of data and subsequent scaling of new technologies. This concept goes against many data principles and local regulations and highlights the diversity of thought regarding data ownership. The conversation also highlighted the urgency of discussions about data privacy and ownership that are needed to avoid exclusionary and harmful practices.
As operating principles, ECAAS prioritizes trust, transparency and privacy of farmer data. By detailing the anticipated use of data and seeking explicit consent, ECAAS aims to follow good data practices. This idea includes understanding more from the farmers’ perspective about the benefits of data sharing, as arrangements with service providers should be, but are not always, mutually beneficial. All of ECAAS’ use cases are centered around delivering useful information back to the smallholder farmer providing data.
- Technology solutions risk exacerbating the gender digital divide. The gender digital divide reflects the inequalities between men and women regarding digital technology access and use, especially in rural smallholder farm contexts. When created without a critical lens on the issue or a clear path to address the unequal access and use of technology, digital solutions can easily exacerbate the already-existing gaps and risk worsening gender inequality for women worldwide.
Unfortunately, the intersection of ag-tech with gender was left largely undiscussed at World Agri-Tech, as is often true of similar tech-centric conversations and workshops. At ICTforAg, Alfred Yeboah, regional director for Africa at the Grameen Foundation, shared, “Access to technologies is unequal [...] Women, in particular, have limited access to smartphones. The use of digital agricultural services could potentially increase gender inequalities instead of reducing them. And, therefore, do more harm to women farmers and ag businesses.” To reduce existing gender divides, the digital agriculture community needs to better understand women’s capabilities, adoption and use of digital ag services.
One of ECAAS’ governing principles states that fundamentally, good ground truth data collection needs to be representative not only in size and crop standards, but in terms of who is farming and how advanced analytics can meet the needs of individuals and vulnerable groups. ECAAS incorporates local knowledge and gender-transformative approaches in our work by engaging a network of local stakeholders that drive informed and gender-inclusive innovations.
The ICTforAg and World Agri-Tech conferences reinforced areas for continued focus within the advanced crop analytics space. Geeta Sethi of the World Bank noted at World Agri-Tech that as a relatively high-risk and low-return sector, smallholder agriculture needs support from governments and philanthropy to help finance innovations where financial returns may otherwise be difficult to achieve. These groups must partner with private actors to collaboratively transition to a more inclusive and digitally informed food system.
Future Ag-Data Planning
Building from these takeaways and momentum around new connections, the ECAAS team is ramping up efforts to help the Bill and Melinda Gates Foundation prepare for the initiative’s future. The next phase involves formalizing a cohesive data-sharing network of public and private sector actors, with long-term data-sharing infrastructure, data exchange arrangements and a standard set of data practices and norms, as prioritized by our community-driven Innovation Agenda.
ECAAS will continue to codevelop future services that meet the stated needs of advanced analytic providers, such as data and model benchmarking, while supporting more coordinated efforts to collect and share quality ground truth data to unlock the potential of advanced analytics for smallholder farmers.
Tools and Resources
- Access Training Datasets. Organizations are deploying the ECAAS Field Mapper Tool for ODK for ag-data collection campaigns in Tanzania, Zambia, Niger and other countries. Practitioners can access the resulting agricultural training datasets generated by these (and other collection) efforts at Radiant Earth MLHub. The Field Mapper tool is now available with French language support; Swahili support is currently in development. We strongly encourage more organizations to utilize these resources when planning data collection campaigns. Together, we can collectively advance information and insights available for smallholder farmers in a harmonized way.
- Listen to our Podcast. Our recently launched podcast, The Dirt on Data, builds awareness about the challenges and innovations at the intersection of big data and small-scale agriculture for the global development community.
- Connect with the Community. To learn more about the ECAAS community of practice and how you can get involved, visit https://cropanalytics.net/.