Why AgTech is hot, really hot, you don’t know how hot
Per year more than 10 million hectares of agricultural land are lost to usage other than producing agricultural goods. In addition already more than 25% global surface contains less humus and nutrients than 25 years ago. According to the German environmental agency, this is mainly due to land production by deforestation, fire-drifting and intensive, non-locally adjusted agriculture.
In addition, the declining agricultural areas face a growing world population, which must be fed by the remaining usable area. Today, approximately 11% of the conquest area is agricultural, a total of around 5.6 billion hectares. The loss of 10 million hectares per year, on the other hand, then appears to be almost negligible at 0.18%.
If one analyzes the growth of world population, on the other hand, you may get a different angle of view: We are currently about 7.55 billion people on Earth. That means 0.74 hectares agricultural land per human being. According to the UN in 2030 world population will have grown to approx. 8.5 billion. The factor of cultivation area per world inhabitant then falls to 0,64 hectares. A decline of 13%.
Many start-up teams are working intensively to solve this problem, which is like a train approaching us in the tunnel at high speed, also with the help of AI. The aim is to increase efficiency of the cultivation of food in the long term. Yet we must not forget that technical development including mass data and AI only alows the optimization of agricultural yields to a certain efficiency. Thereafter, the gene-modified varieties will have to help out.
CB Insights therefore identifies 5 key use cases for AI in AgTech:
- Analysis of satellite images
- Monitoring of cultivation areas (drones, computer vision)
- Analysis of the condition / health of sowing and cultivation area
- Analytical prediction (e.g. seasonal fluctuations): Using big data and predictive analytics to address farm-related issues and make better farm-related decisions in order to save energy, increase efficiency, optimize herbicide and pesticide application
- Agricultural robots
In the last 5 years, startups have raised more than 800 billion dollars in funding (source: CB Insights). However, this number contains, inter alia, Orbital Insights with a total funding of USD 78 million. However, the company does not exclusively deal with the agricultural sector, but uses proprietary algorithms for satellite-based evaluation of all possible objects, also from agricultural land to predict crop yield.
The Munich start-up Acrai, on the other hand, has the goal to replace chemical weed killers, also called herbicides.
- The three main problems of herbicide use in agriculture are:
- Development of a new selective herbicide (combats only the weeds, well tolerated for cultivated plants) now costs up to EUR 300 million with an approval period in the EU of up to 12 years
- Social acceptance of chemistry on crops falls. We all know this slightly bitter smell that is carried into the landscape after the application of herbicides with the wind. And we do not want imagine this to be on and in our food.
- Resistance to selective herbicides increases
- The use of herbicides reduces the yield per individual plant by approximately 12.5% because growth is inhibited by herbicides
The startup uses intelligent software to distinguish between crops and weeds. The team built a first prototype under laboratory conditions and tested and further developed the prototype in cooperation with now 40 conventional and biologically managed farming companies. The idea itself was born when one of the founders had attempted to persuade his parents of renouncing herbicides in their agricultural company.
A key role is played by the fact that the use of labor in Agriculture is becoming increasingly unprofitable. In Germany, the minimum wage rises to EUR 9.10 per hour by the end of 2017. At the same time it will be harder to hire seasonal workers to remove weeds, and the work results will be very inaccurate after a few hours of hard work on the field. The costs for the plucking of weeds are now between about 1,200 and 2,000 euros per hectare and season.
Acrai aims at 4 central use cases:
- Differentiation of crops and weeds with computer vision and then selective hoeing. The plan is to use a trailer behind a non-autonomous tractor
- Early detection of diseases
- Phenotyping (recognition of the characteristics of a plant at a very early stage), particularly interesting for developers of seeds
- Selective fertilization and irrigation
Other teams also focus on maximizing yield per hectare. BlueRiver Technologies from the US, after detecting the plant type, sprays fertilizer individually on every single plant. Eco Robotics from Switzerland sends an autonomous robot for the spreading of herbicides over the field and promises a reduction of the herbicide quantity by a factor of 20. Dr. Julio Pastrana and Tim Wigbels from the Institute for Geodesy and Geoinformation at the University of Bonn are combating weed by means of a laser pulse.
The basis of all the procedures mentioned is the use of computer vision, deep learning and robotics. Acrai uses Python and Tensorflow. The actual IP is, however, the data set (image material) and the procedure for enriching the data with feature vectors for exact location of the plant and the robot, but also for integration of weather data which allow conclusions to be drawn about the size of the plants.
Acrai starts to use its technology on sugar beet fields, since contribution margin on sugar beet is over 700 euros per hectare and thus exceeds every other field plant by lengths.
Still believe AgTech is not hot?
- Artificial Intelligence