I recently returned from Zambia, where Our Sci is supporting research efforts by the International Maize and Wheat Improvement Center (CIMMYT) into the delivery of extension services to smallholder farmers. Specifically, CIMMYT wants to know if using technology to deliver site-specific extension advice changes farmer adoption of good agronomic practices and soil health building activities compared to general extension advice.

Since founding Our Sci LLC, our goal has been to support community driven science using open-source tools. This collaboration with CIMMYT provides an excellent opportunity to test two of our tools, which were central to the research, in difficult environments:

1. Our Sci Reflectometer: a low cost, hand-held spectroscopy tool (left, above).
2. SurveyStack: a flexible data collection and management platform (center and right, above).

My primary task was to train local extension agents on the use of the Reflectometer and SurveyStack. But first, some background…

What is CIMMYT doing?

CIMMYT seeks to improve the livelihood of resource-poor farmers by improving production systems of basic grain crops (maize, wheat, millet, etc) and building more resilient agrifood systems. Achieving this requires research into both what practices increase food production and resilience and how best to deliver that knowledge to farmers.

For this research project, CIMMYT partnered with the Zambia Agriculture Research Institute and trained local extension agents on two extension advice delivery methods:

• General extension – farmer interactions. Each extension agent visits households in their area, discusses the challenges that farmers face and provides information about soil health building practices (legume rotation and intercropping, using compost, etc) and good agronomic practices.
• Site-specific recommendations based on soil health measurements. Each extension agent uses the Our Sci Reflectometer to measure and classify soil health in multiple fields managed by the household. The Reflectometer is calibrated to predict degraded soils, and targeted recommendations were developed by regional experts from CIMMYT based on the Reflectometer outputs. I have outlined how we generate meaningful soil classifications and prediction models in a previous post if you want to know more about the process.

What kinds of site-specific recommendations are being generated?

When a soil is classified as degraded, the model recommends practices that build soil health and suggests farmers not use inorganic fertilizer, since degraded soils tend to have low fertilizer use efficiency. A few examples of soil health building practices include adding organic amendments (compost, manure, green manures, etc) or integrating legume crops in rotation or as intercrops.

Zambia has a fertilizer subsidy program, so each household gets access to a limited amount of subsidized fertilizer, and utilizing this fertilizer effectively is important for resource-poor farmers. Soils classified as not being degraded are more likely to respond well to fertilizer, and the extension advice focuses in part on proper fertilizer use (application timing, placement, etc) to ensure farmers benefit from fertilizer.

How is SurveyStack used in this project?

First, project managers created a “Group” in SurveyStack, inviting all of the participating extension agents to the group and created a survey that walks the extension agents through the process. The survey leveraged numerous SurveyStack features to simplify the user experience and protect sensitive data:

• Uploading ontology lists of local districts so extensions agents could quickly select the correct area from auto-select dropdown lists.
• Check the “Private” box for all questions that might include sensitive household data. SurveyStack allows privacy settings on a question-by-question basis, ensuring that private data is protected, while still maximizing community-driven learning.
• Survey logic to only show extension agents the questions that are appropriate to their situation. This allows them to avoid scrolling through irrelevant questions when they are in the field.

SurveyStack also has unique scripting and hardware integration features, allowing users to take Reflectometer measurements in the app and then run custom prediction models to use both responses to survey questions and measurements to output soil health classifications and related recommendations. Developing custom scripts can require knowledge of javascript and significant time and effort to code and test. However, SurveyStack has a “Question Set Library” that allows users to create and share groups of questions easily. For this project, staff added questions sets relating to hand-texturing of soil, calibrating the Reflectometer and taking soil scans, allowing staff to leverage previous Reflectometer research efforts to add validated methods and scripts to the survey quickly. 

The surveys are accessed through the SurveyStack progressive web app, which is fully offline compatible. When extension agents arrive at a new household to generate and discuss the site-specific recommendations, they open the survey in the app and walk through the entire workflow, answering questions, taking measurements and generating site-specific recommendations seamlessly. Once complete, they can submit the results to the database. If there is no internet connection, the results will be queued to automatically upload to the database the next time the phone or tablet is connected to the internet. Once the response is submitted, it is available in real-time to all project managers. Project managers also have the ability to edit the responses as needed. For example, if an extension agent forgets to answer a question, either the extension agent or project staff can answer the question later.

Next Steps

It’ll take some time to ascertain if using the Reflectometer to generate site-specific recommendations changes farmer behavior, or if those behavior changes lead to better outcomes for resource-poor households. 

For now, though, we can learn a lot about how well (or not) our tools function in different environments. Today, the extension agents are actively engaging with participating households, trying to provide meaningful advice before the rainy season arrives and farmers start planting their crops.

If you think SurveyStack or the Reflectometer might be the right tools for your community research, contact us at info@our-sci.net