Adapting remote-sensing data for an array of applications in agriculture


Adapting remote-sensing data for an array of applications in agriculture

CIP scientists made progress in 2016 on forecasting tuber yields in potato over large areas by linking modeling and remote-sensing data. Using satellite images of a potato-farming area in the US state of Idaho, researchers parameterized physiological and crop growth models that were then used to forecast yield. According to Roberto Quiroz, leader of CIP’s Crops and Systems Sciences Division, the parameterized model prediction was compared with crop statistics from the area in the satellite image. The comparison showed that combining modeling and remote-sensing data can produce excellent results.

The experiment was the latest on a growing list of CIP’s applications of remote-sensing data to agricultural research. CIP began using satellite images to estimate sweetpotato crop areas almost two decades ago, but those images were less useful for studying potato since potato-farming areas are often obscured by clouds. CIP scientists consequently began to explore other remote-sensing options, such as attaching cameras to balloons or miniature planes. In 2012, they tried an unmanned aerial vehicle (UAV)—commonly known as a drone—which soon proved to be the best tool for gathering high-resolution images.

Quiroz subsequently led the creation of a UAV-based, agricultural remote-sensing integrated platform that has validated the use of drones for analyzing crop area in the field in Peru and Tanzania. For more accurate assessments, algorithms were developed for geometric and radiometric correction of data from sensors in order to enhance their applicability to agricultural research.

Technological advances, coupled with the decreasing cost of hardware, have opened the door for a rapid expansion of applying this technology to agriculture. Yet CIP researchers have also assembled drones from parts in African countries, purchased sensors instead of cameras, and developed open-access software to make the technology as affordable as possible.

One such open-access program allows researchers to stitch together many different images to study large areas. Researchers in Tanzania used that program to stitch together more than 600 images taken during one UAV flight over Kilosa District. They produced an image that depicts about 100 ha, in which they identified more than 20 crops with less than 20% error. Quiroz noted that taking more photos would have reduced the error. He explained that his team is trying to develop a tool to correlate UAV remote-sensing data with satellite data in order to improve the accuracy of such large area analyses.

CIP and partners formed a ‘UAV for Agriculture’ community of practice (CoP) in Africa to share knowledge and software, but it has quickly expanded to include members outside Africa. The CoP, which is hosted by the Technical Centre for Agricultural and Rural Cooperation at Wageningen University & Research, the Netherlands, had almost 1,000 members by the end of 2016.

While remote sensing has primarily been used to assess crop area, CIP researchers are exploring other applications such as yield prediction and detecting the onset of diseases or pest infestations, or the effects of climate change. Remote sensing has the potential to improve the accuracy of government agricultural statistics.

The ability to detect crop pests or diseases early, or forecast low yields, can help governments predict and prepare for food shortages.

CIP researchers have used remote-sensing images to detect potato yellow vein virus before pathologists could detect it in individual plants, and the technology has implications for an array of other crops too.

In another innovative development, CIP programmers produced software in the R language that can extract information on individual plants in a field from a stitched, 120-megapixel image with a resolution of less than 5 cm. This non-intrusive technology has the potential to replace visual inspection for a more accurate, high-throughput phenotyping, since it would reduce human error. “This is a way of generating a public good that breeders around the world could use to improve the efficiency of their phenotyping,” said Quiroz.

“Technological advances, coupled with the decreasing cost of hardware, have opened the door for a rapid expansion of applying this technology to agriculture.”Roberto Quiroz, leader of CIP’s Crops and Systems Sciences Division

Photo: Testing a drone in sub-Saharan Africa. CIP

Share this

Related Publications