Network analysis reveals where to act effectively in seed systems

Photo credit: Paul Rachkara.

An innovative approach to seed systems, spearheaded by scientists working with the CGIAR Research Program on Roots, Tubers and Bananas (RTB), offers new ways to enhance the adoption of novel technologies and limit the spread of pests and diseases. By focusing on the network that links farmers, traders and others, it identifies the best places to influence how things – seeds, pathogens, knowledge, technologies – travel through the network, suggesting efficient ways in which farmers can be helped to improve their productivity.

Most of the world’s poorer farmers do not buy high-quality seed. They typically plant material they have saved themselves, or that was saved by relatives and neighbors. Some may buy seed from traders, but often that seed was similarly saved by an unknown farmer, with no guarantees of quality. One problem inherent to this kind of informal seed system – especially for roots, tubers and bananas, in which the “seed” is vegetatively propagated rather than a true, botanical seed – is that many pests and diseases are transmitted along with the seed. The result is often seed degeneration, and with it declining yields, because even seeds that start as high-quality accumulate pathogens and pests over cycles of growth and reproduction.

Network architecture

A formal network analysis of the seed system considers the nodes (which might be farms, markets, agricultural stores, extension workers, male and female farmers) and the links among them. Where does a farmer’s seed or information come from? Where does it go?

In a highly-connected network, seed can travel easily from farmer to farmer. A centralized network may allow seed to travel from a single node, perhaps a market, to all the other farmers, but not from farmer to farmer. In a more decentralized network, there may be clusters where seed travels easily within the cluster, but not often between clusters.

“The network architecture can tell us where to monitor for diseases, where to introduce new varieties, how to develop capacity and much more,” said Jorge Andrade-Piedra, plant pathologist at the International Potato Center in Lima, Peru.

Where to look for disease

One recent study, on which Andrade-Piedra worked, looked at 48 potato farmers in Ecuador, all of the members of a regional consortium called Consorcio de Productores de Papa (CONPAPA). The consortium includes some specialist farmers who produce higher quality seed potatoes. It provides training and expertise and also processes and sells ware (eating) potatoes on behalf of members. Network analysis revealed that the system is quite sparse, with many links between farmers and both the CONPAPA team and the market in their nearest town, but few links among different CONPAPA-member farmers.

Knowing the structure of the network, the team wanted to know which might be the best locations to detect a newly introduced disease. In a simulation, they introduced the disease on one farm node at random and asked how many other farms it would spread to before being detected at another randomly-selected farm. CONPAPA itself is clearly the most important node for monitoring because it is the source of most of the seed coming from off the farm. By the same token, the most effective route for introducing new varieties and other beneficial technologies to these farmers would be through CONPAPA. While individual farms are less critical for monitoring than CONPAPA itself, the network analysis also shows which farms would be most important for additional analyses of the potential introduction of a pathogen.

“It is clear that some spread can always occur via the seed system,” said Andrade-Piedra. “Knowing where to concentrate monitoring efforts is a good way to reduce the potential spread of pests and diseases.”

The same goes for information. The network for sources of information about integrated pest management (IPM) was similar to the seed network, with most farmers getting their information from the CONPAPA management team. Farmers also often reported that they got information from agrichemical stores. Farmers who reported getting their information about IPM from less reliable sources might be worth monitoring and training, because their farms could have additional risk factors for the introduction of diseases.

Training store owners to share information about pest management with farmers might seem a good idea. The problem is that farmers don’t trust store owners as much as they trust CONPAPA, and that might be justified. Store owners face a conflict between supplying information that they know to be correct and recommending the use of products that they happen to have in stock.

“In addition to understanding network architecture,” Andrade-Piedra added, “we need socioeconomic data to make sure we understand the role of information sources for farmer decision-making.”

Controlling sweetpotato diseases

A related study of sweetpotato in Northern Uganda, conducted with RTB and colleagues by Kelsey Andersen of the University of Florida, found a different network structure. The network is less centralized than the CONPAPA network, with 27 sellers and almost 1000 buyers throughout the network, organized by the village. Most nodes have few links while a few nodes have many links.

As with potato, viral diseases transmitted through seed vines can put a major brake on sweetpotato production. The researchers used this network to study, among other things, how to control an epidemic if a pathogen should appear. Quarantining villages or managing the disease intensively, so that it could not spread, would obviously help to contain a disease, and the more villages are managed, the slower the disease spread. But which villages? The simplest metric, how many incoming links does it have, proved almost as effective as more complex criteria, and much more effective than just managing villages at random. However, management was effective only if 15 or more villages were knocked out of the network.

“The invasion may be able to overcome management of a small number of nodes, even if those are the nodes with the most links,” said Kelsey Andersen. In fact, although such management slows a disease down, it does not actually stop the spread. The network analysis suggests that it may be better to focus on seed health and disease resistance, to prevent diseases in the first place. The paper by Andersen and her colleagues gained a lot of attention and has already been downloaded more than 1,500 times.

Useful methods

Researchers recognize that each seed system is likely to have different network characteristics, but in every case understanding the structure of the network is an important first step in deciding where to allocate resources.

“We’re working to develop methods for the RTB toolbox that will allow scientists to perform what we call an ‘impact network analysis’ – evaluating regional management strategies in linked networks of communication, trade, and potentially spread of pathogens and pests—quickly and effectively,” said Professor Karen Garrett of the University of Florida, who is working with RTB researchers on seed systems. The RTB Toolbox is a collection of tools that allow practitioners to understand and systematically diagnose seed systems and determine how to most effectively intervene in them. The network analysis is an important component of the Toolbox, and Garrett is also working to adapt network analysis approaches for decision-makers.

Mainland Southeast Asia, she points out, is grappling with an outbreak of cassava mosaic disease. Plant health regulations are in place, but they are not being applied to cassava, where the seed system is largely outside the formal sector. Traders regularly transport cassava material for long distances and across international borders. Rather than stricter policing, which might not work, analysis by RTB suggests that it might be better to encourage traders to get seed from regions without reports of the disease, such as the highlands of North Vietnam, and make use of existing trade routes to spread healthier seed.

“Systems analysis that integrates socio-economic and epidemic networks, including seed trade, offers insights that can improve food security,” Garrett says.