Use of artificial intelligence in Agrosavia

Soil analysis is one of the most important activities for crop production in the country. For more than a year, the Colombian Agricultural Research Corporation, Agrosavia, has implemented artificial intelligence in this process to make it more efficient and provide more precise recommendations.

Through soil analysis, an agronomist can know the physical characteristics and chemical properties that it has and, based on the results, recommend to the farmer, according to what his crop needs, supplements and nutrients that allow him to achieve his production objectives. .

Agrosavia has provided this service to the country’s producers since 2007, but it has gained greater relevance and scope since 2015 thanks to the “Before sowing, the soil must be analyzed” program, which in agreement with the courier company Servientrega allows farmers to of more than 900 municipalities in Colombia to access the analysis of their soils so that they have planned production processes and better results.

The prediction made by the AI ​​model is of the nutrients required by the analyzed soil.

In order to contribute more and more to the development and strengthening of this sector, and given the success of the program, in 2018 Agrosavia joined forces with the Ministry of Information and Communication Technologies, Mintic, to implement artificial intelligence in this process and acquire a system  that facilitate the management of information in the laboratory.

According to Rafael Pedraza Rute, Laboratory Information Management Coordinator, for this program they acquired a LIMS system that has allowed them to improve management and communication with farmers who request the study of their soils.

And the implementation of artificial intelligence was possible because they had a large amount of historical data from the results of soil analysis and the fertilization recommendations that agronomists make.  With IBM Watson they developed the algorithm and the AI ​​model that would allow them to make predictions about the nutrients that the soil needs, for this they used nearly 10,000 fertilization recommendations.

How does your AI model work?

  • The LIMS system sends the data that is valued in the laboratory, for example, nitrogen, phosphorus, potassium, sulfur, iron, magnesium, calcium, among others, to the AI ​​model, which based on these and historical information makes a prediction amount of nutrients that a farmer should apply to his soil to improve its characteristics.
  • The agronomist in charge checks if this prediction is correct or if an adjustment needs to be made. If it needs to be done, he does so and the AI ​​model saves it and learns from this new information.
  • With the correction and validation of the agronomist, the model delivers a PDF file that will be uploaded to a portal so that the farmer can see the results of the study of his soil and the respective recommendations.

Main achievements

Initially, the AI ​​model they developed was trained on information from 200 types of crops. Today the best predictions that it makes are those for which there is more data: coffee, citrus, avocado, blackberry, grass … of the others, for example, custard apple or other fruits, more information is required so that you are increasingly more expert in these and make fewer mistakes.

From the implementation of this technology, Rafael highlights that  they have been able to speed up and be more efficient in predictions and recommendations. Before the team made 18 recommendations a day, now they make 36.  “We have increased our capacity for action, but we know that it will get better every time because the system continues to learn, we must reduce the errors that the agronomist must correct.”

Likewise, he says that they have improved the response time to farmers and the presentation with the recommendations that they deliver to them.

Without a doubt, “it is a technology that has served us a lot and we want to continue advancing in its implementation in other processes to be increasingly efficient and precise with what we do and recommend,” concludes Rafael.



Leave a Reply

Your email address will not be published. Required fields are marked *