At the end of 2016, the government of Azerbaijan wishes to provide to their citizens independently the majority of their wheat requirements, in order to reduce their dependence on import. Wheat is one of the main staples of the Azeri national nutrition. The policy leaders believe that Azerbaijan has enough land, rain, and water to grow most of the national consumption of wheat and barley.
Azerbaijan has suitable agricultural land occupying about 50 thousands square kilometers. But despite the government investments of millions of Dollars in agricultural subsidies, procurement of agricultural machinery, and impressive irrigation infrastructures, the amount of wheat yield has not met the expected targets. Therefore they still have to import almost half of the country’s wheat requirements.
The governmental team who were assigned to determining the country’s economic policy asked for information that might spread a light on many unanswered questions that they had:
Why, despite the considerable financial investment, we have been unable to grow sufficient amount wheat?
Are the subsidized farmers are growing wheat or something else?
How much of the land seeded with wheat has actually yield? And in what productivity?
Ofek experts started investigating the issue of wheat cultivation in Azerbaijan, with main challenge: How to distinguish between wheat, barley and the other seasonal grains?
Main task: How to differ between wheat, barley, and other grains within a short season in enormous areas?
Azerbaijan has a variety of climates and topography, from the mountains that reach 4,000 meters in height, to the flat terrain that descends to sea level. Every area and valley grows different varieties of wheat and other grains using different techniques, in hundreds of small fields which are often adjacent and mixed with each other.
The solution was found in the development of a special technique for interpretation the satellite images. The inspiring idea was to try to separate the wheat from the barley and the other grains by examination of their unique way of growing during the season (Crop Phenology). This method, which has been formulated for the first time, based on the assumption that if we successfully manage to monitor the behavior of every field crop in Azerbaijan during its growing season, we might be able to identify its unique development patterns. Knowing that, we will manage to delineate the fields we are looking for.
The first stage was to separate the wheat and barley from the other plantations by identification of the features that are unique to these crops. For this purpose, we made field surveys and collected thousands of samples by a combined team of Israeli experts and local Azerbaijani surveyors.
For the second stage, we recruited the best brains from the Israeli academic community in the fields of Artificial Intelligence (AI), image processing and computerized interpretation of agricultural data. Their challenge was to develop a complex and unique algorithm that can find and download the relevant images that were captured by a variety of satellites over Azerbaijan during a given period, cataloging them to the enable season oriented process, removing the cloud and shadow effect, and dispersing the fog to make the hidden visible. The result surprised us with its accuracy: The wheat and barley fields were very finely and clearly identified from multilayered images.
Following the success of separation between wheat and Barley to other grains we now faced the main challenge.
The impossible task: how to differ between wheat and barley?
In Ofek’s initial studies during the preparation for the project, we found that up to date, no solution was proved to successfully separate wheat and barley crops. Furthermore, it appears that even when an expert stands in a field in the beginning of the season, he will find it difficult to distinguish of one grain from the other. If that is the case, how can we trust separation made by images that were taken by satellite from a height of hundreds of kilometers?
The solution found: A field visit to meet local farmers and obtain their help in learning about the minor differences in cultivation practice, which we can train the computer to identify.
After overcoming the natural suspicion towards uninvited foreigners that come around asking questions, something that doesn’t happen often in these areas and thanks to the assistance of local team that speaks the language and was trained by Ofek staff to ask the professional questions, the farmers in the field were convinced to cooperate. With their help, we had the full picture relating to the wheat and barley growing procedure and realized which kind of differences our algorithm should be trained to look for.
Information worth its weight in gold
Agricultural mapping is being conducted for decades worldwide. Currently, higher accuracy can be achieved due to the ability to download satellite images taken every 10 days, to process it automatically and create a clear picture of the seasonal status of crops. Following the adjustment of the special algorithms to identify the changes in the field, we succeeded in separating the wheat and barley crops from the grass and other grains and even from each other.
This exercise enabled not only to quantify all the wheat and barley areas seeded all over Azerbaijan, but also to learn how much of these areas yield by the end of the season.
Based on this we could also identify how much area has been seeded, the state of crops in the planted area, identify irrigation problems, understand which areas are not productive and estimate how much wheat and barley are being grown in the country at a given moment.
After physical verification of the results by another visit to the cultivated areas, we were able to present to the government our conclusions from the project:
1.Less than 70% of the areas seeded with wheat reached with the required yield for harvesting (!).
2.The study discovered many areas that had been seeded with wheat and barley but the farmers lost hope to see yield at the end of the season and replaced the wheat with another crop, in order to try to save their season.
3. An important observation was that in areas smaller than 5 hectares, which are cultivated with a lot of effort and hard work by thousands of farmers, the yield that is harvested at the end of the season is less than 50% of the areas that are planted.
So…how can wheat yield be improved?
The secret of success in the agricultural practice relates to precise activities and excellent timing. Monitoring growth, early identification of irrigation and drainage problems, pests and weeds and other cultivation faults, enables alerting farmer’s attention to those problems, taking immediate operative actions and gain hundreds of thousands of tons of wheat.
The success in analyzing the behavior of the grain crops across the last season now allows Ofek to propose actions that will help to improve the yield of wheat during the next season.
Ofek continues to work closely with the Azerbaijan government in pursuit of its objectives. We hope to expand the lesson learned and the use of the developed tools to other countries and organizations that are facing similar challenges.