Feed the Future
This project is part of the U.S. Government's global hunger and food security initiative.

Digital Development for Feed the Future: Building an Innovative Community of Practice to Respond to Smallholder Farmers' Needs

This post is the first in a two-part series focusing on data-driven agricultural development and was authored by Karina Lundahl, Facilitator for USAID’s Innovation for Data-Driven Agriculture Convening on April 27–28, 2017.

With the continued global proliferation of smartphones, sensors and advanced analytics, opportunities and challenges relevant to smallholder agriculture in emerging economies are increasing. In the focus countries of the U.S. Government’s Feed the Future initiative, smartphone adoption increased an incredible 800 percent between 2010–2015 according to data from GSMA Intelligence. And in 2017, the combined processing power of global smartphones will surpass the processing capacity of all servers worldwide. To create an agile and informed response to technological opportunities addressing the context-specific problems faced in developing regions, a diverse group of thinkers and innovators is required.
 
Addressing this, USAID, in collaboration with the Sustainability Innovation Lab at the University of Colorado, Boulder (SILC), hosted its second convening focused on building a cross-industry community of practice in data-driven agricultural development. Representatives from the U.S. Global Development Lab and Bureau for Food Security at USAID joined a group of researchers, tech innovators, funders and development practitioners to discuss the state of the industry as well as paths forward for data-driven approaches to agricultural development. Through a series of presentations, panels and workshop activities, three major themes emerged:
 
1. Opportunities and challenges in the data landscape: collection, analysis, open sharing and distribution
Growing opportunities for data-driven agricultural innovation include the use of smartphones, low cost ground sensors, weather stations and remote sensing to gather on-farm data and landscape information. Ever-improving machine learning and predictive analytics, which capture existing information and generate predictions where gaps in site-specific information exist, can also be employed to model data for further insights.
 
Some promising examples of current data collection and analysis presented included:

  • Stantec Consulting is using soil moisture sensors, combined with real-time weather data and on-demand water delivery, to provide optimal irrigation levels in water-scarce conditions in California. These systems facilitate up to 30 percent increases in crop yields while also reducing water and energy use by up to 30 percent.
  • The Geospatial and Farming Systems Research Consortium (GFSRC) is conducting quantitative monitoring of crop trial experiments through unmanned aerial vehicles (UAVs), or drones, equipped with sensors.
  • The Land Potential Knowledge System (LandPKS) is a mobile application that performs digital assessment of site vegetation and obtains on-farm location and soil data, cross-referenced with global soil databases, to produce a high resolution estimate of local soil characteristics.
  • DigitalGlobe is collecting and analyzing satellite imagery aided by spectral analysis. Their data offers insights into population distribution, land use, crop yields, crop health and key vulnerabilities such as food security in areas otherwise inaccessible due to conflict or other crises. DigitalGlobe also employs crowdsourcing to identify items of interest within imagery. Crowdsourcing results are then used to train machine learning algorithms to improve the accuracy of automated object identification.
  • aWhere is applying weather modeling and prediction, along with agronomic modeling, to offer recommendations on potential pest and disease crop stress, production forecasting and more.

 
For all the opportunities, challenges remain in data collection and sharing for agricultural development. Some of the challenges discussed include:
 

  • Sparse or low quality on-farm data

Some data — such as observed weather data — is not available at the resolution needed. Other data — including on-farm management activities — is variable, inconsistent or of poor quality in many locations. On-farm observations of crop selection, timing of planting, use of fertilizers and other key decision points provide the most direct insight into the successes and failures of crops over time and across space.
 
Since technologies such as machine learning, simulation models and predictive analytics require site-specific information to generate successful estimates and predictions, there remains a need to collect accurate on-site information to provide ground-truthing and locally-specific knowledge for farmers. This need will likely increase as new analytical capabilities come online.

  • Uncertain relevance or metadata

Collected data, once shared, often lacks relevance or metadata. Metadata provides the key information on how data is organized and its relevance. It is critical for easily finding and using data for analysis. Without it, shared data may ultimately not be of use to stakeholders or researchers.

  • Increasing weather variability

There are unique agricultural and developing economy risks that innovators must take into account. Due to planting and harvest cycles, farmers’ economic concerns change throughout the year and growing season and are highly dependent on weather. And, as weather becomes more variable, farmers will experience extreme episodic events and periods of acute risk and vulnerability more frequently.
 
2. How to better incorporate smallholder farmer concerns during design and implementation
Convening discussions also focused on the need to frame research, programs and ventures within a firm understanding of smallholder needs and concerns. Not only will farmers in different contexts access data and technology differently, data and technology may be irrelevant if a farmer lacks reliable access to market and has no power to set prices for goods. 


Some methods discussed for understanding and engaging farmers were:

  • Human-centered design to understand smallholder concerns
    Adam Reineck, Design Director and Studio Lead with IDEO.org, demonstrated a human-centered design (HCD) approach at the convening when sketching a conceptual framework of smallholder farmers’ information ecosystem in East Africa (Figure 1). This included different farmer groups, communication channels they access and other influential actors, such as extension agents, agrovets and middlemen. HCD is a rigorous and iterative design process employed by IDEO.org that centers on the needs of the end-user from the beginning of every project to arrive at a contextually-relevant and actionable solution. 

East Africa
Figure 1: A Sketch of Smallholder Information Ecosystem: East Africa.
 

  • Personalizing communication with farmers
    Communication channels offer potential for personalization at scale as farmers’ needs change over time. Information desired will likely change depending on location and crops grown, for instance. By applying multiple channels of contact, perhaps radio advertising followed by interactive voice response (IVR) for instance, one can tailor information to farmers’ specific needs and keep them engaged longer. There was also discussion of how channels might be used to collect more information on the usefulness of information and/or programs for farmers in order to improve information over time.

3. Engaging a diverse community of practice
As we move toward building a dynamic community of practice, workshop attendees noted industry-specific strengths and possible gaps in expertise. Some of the gaps were:

  • A research to application gap for some academic researchers and technology developers. For instance, often promising tools are developed without a clear path to markets.
  • A lack of shared risk between implementing programs and farmers. Program success is often measured by the number of farmers reached instead of positive livelihood impacts for farmers.
  • A lack of incentive for funders to engage in small-scale or high-risk ventures that may be promising for smallholder farmers. The funders’ incentive is to invest in the fastest and largest return on investment, but this may be in conflict with how capital ebbs and flows throughout agricultural growing seasons.

Taking into consideration these possible gaps, workshop attendees made commitments as to how they will move forward with building a community of practice for data-driven agricultural development. This included activities such as co-writing concept notes, hosting follow-up workshops and exploring collaborative projects. A follow-up event to learn about the outcomes of these commitments and actions would be beneficial to supporting this innovative community of practice.

Digital Development for Feed the Future is a collaboration between USAID’s Global Development Lab and the Bureau for Food Security and is focused on integrating a suite of coordinated digital tools and technologies into Feed the Future activities to accelerate agriculture-led economic growth and improved nutrition.  

For more information on Digital Development for Feed the Future's work on data-driven agricultural development, please read the Key Findings Report from the Innovation for Data-Driven Agriculture convening in April 2017.