Rice is grown from the ground up, but more and more, advances to grow the crop are happening in the cloud.
LSU AgCenter precision agriculture specialist Luciano Shiratsuchi and AgCenter statistician Thanos Gentimis are working to help rice farmers grow their crop more efficiently using large datasets collected on the ground and from satellites.
Shiratsuchi is using ground truth data — collecting sensor readings from more than 50 rice fields across the state so he can have an initial measurement of what may be inferred from satellites images.
“The satellite is saying I have this much biomass, but we have to have calibration to say, OK, yes, that makes sense,” Shiratsuchi said.
As the crop grows, Shiratsuchi’s sensors can tell him things like leaf area index or canopy temperature for water stress.
“Once you have that, the satellite’s imagery can extrapolate to a bigger area,” he said. “The handheld sensors are collecting data for us the same way the satellites are at the same time.”
He is working on collecting data from not only his sensors, but also from farmers because the more data he can take in, the better his software can become. His plan is to transform that data into models to predict input needs and best production practices for individual fields.
Currently farmers mainly use uniform nitrogen rate applications, but based on results from Shiratsuchi’s research, farmers will be able to use maps to support variable rate applications. Specifically for rice farmers, the data and models generated also can pinpoint timing for applying the first flood.
None of the data he collects is made public. It is only used to make predictive models.
“The idea is to provide farmers with unbiased precision ag applications,” he said.
He said farmers could use the platform he is developing to draw a boundary, download satellite data for that area and then connect to the server to generate prescription maps.
Some big farms pay for monitoring and data collection, but Shiratsuchi envisions his platform would be available to any farmer interested in using it.
“The intelligence will handle the data, crunch this information and come up with some recommendation. It seems like magic, but it’s not magic,” he said.
If there is any magic, it is in the math, and that is where Gentimis comes in. A mathematician by training, he is working with a large dataset that researchers at the Rice Research Station had been collecting for years. He worked with a graduate student to untangle all the information in it, clean it up and establish protocols for adding to it.
With the clean data, his student was able to look at optimal planting dates for rice.
“The AgCenter comes up with an optimal window for planting, and it’s based on science, but the science was about 20 years old, and things have shifted,” Gentimis said.
Once the dataset was in place, Gentimis said it was easy to search by variety, planting times and locations.
“The information was always there, but it was so broken apart that you couldn’t put it together,” he said.
Combining data collected into something useful to growers is Gentimis’ aim. He said researchers are independently collecting data to answer a specific question, and once that research is done, the data are left. By putting it all together, valuable information may be extracted. And more data may follow.
“When you start publishing the utilization of datasets that already exists, stakeholders will give you more data on the off chance you can find something. That is a pillar of digital ag,” he said.
Gentimis also worked with a graduate student to fly drones over fields to replicate handheld GreenSeeker technology. They were trying to determine if they could use a drone to figure out how much pest pressure was occurring in the field.
The student would take the large image the drone created and use software to create a program that would segment the image.
“If you don’t have an automated system, it is time-consuming work analyzing the photos,” he said.
Gentimis said they encountered a problem in flooded rice fields. The background the water provided in the images made it difficult for the software to see whether one area looked more damaged than others. He said this technology may not work best for rice.
Even with mixed results, the statistician is helping his colleagues explore techniques they don’t normally use.
“A lot of people are trying to get into drones, but they don’t know how to use the output, so I help them extract information and do the appropriate analysis and help teach a lot of the students on how to do the work,” he said.
For Gentimis, digital ag is a huge umbrella. He said no projects are alike, and he is seeing digital agriculture techniques becoming more mainstream.
“It’s already being used by big companies, and they are already creating algorithms to determine the best seed, the best planting time,” he said. “For big applications in Brazil no one does inspection of the field anymore. They all fly drones and use image analysis.”