The AI assistant called "PlantVillage Nuru" is being used across Africa to diagnose crop diseases.
The Penn State researchers rigorously tested the performance of their machine-learning models with locally sourced smartphones in the typical high light and temperature settings of an African farm.
In these tests, the app was shown to be twice as good as human experts at making accurate diagnoses, and it increased the ability of farmers to discover problems on their own farms.
"Our goal is to nudge behavioural changes that will help farmers prepare their farms to be climate ready," said David Hughes, associate professor of entomology and biology.
"There are proactive behaviours, such as planting for increased crop diversity, promoting soil moisture conservation and engaging in water harvesting, that are known to increase resiliency. Our AI tool is in the early stages, but it will get better over time and with more training," he added.
"PlantVillage Nuru" can draw in data from the United Nations' WaPOR (Water Productivity through Open access of Remotely sensed derived data) portal, a database that integrates 10 years' worth of satellite-derived data from NASA and computes relevant metrics for crop productivity given the available water.
Specifically, the AI assistant has the opportunity to integrate diverse data streams to provide information about drought tolerance of crops and which crops are suitable in which areas.
In addition, the app offers advice that could help farmers learn about crop varieties that are climate-resilient, affordable irrigation methods, and flood mitigation and soil conservation strategies, among other best practices.
Although the tool is smartphone based, it can be accessed through a webpage to inform diverse stakeholders.
( With inputs from IANS )