But how do you combine plant and data science? As I also mentioned in Chapter 1, just about any data concerning the plant, such as growth data and root zone information, as well as climatic conditions, can be collected as useful data. This can be done manually, but that can be time-consuming and prone to error, or through high-tech systems, such as sensors and imaging systems. There is so much data available, but what to do with it? And how can the best of both worlds be combined, combining valuable greenhouse plant science insights with data science?
Plant science vs. data science
Plant science dates back to ancient times, but data science is relatively new. Plant science is the classical science dealing with the physiology, genetics and growth of crops. It is structured and systematic. On the other hand, data science is the science of extracting useful knowledge from raw data, in response to which action can be taken. Data science can be considered more like an art form, with no set rules.
The sciences combine so well because both use the latest technologies and both contain a strong research element. Data science can be combined with just about any other science or field, such as medical science, financial fields, astrophysics, etc. When data science activities are combined with dynamic plant science, wonderful things happen. The grower has more security and achieves more with less (less water, less fertilisers, less chemicals). Through this combination, better crop quality, higher yields and greater greenhouse efficiency are achieved.
How do you combine plant and data science?
As this is still quite new, I find that there are many misconceptions and assumptions about combining plant science with data science and using new technologies in the greenhouse. Below, I share some of the main misconceptions I often encounter (and related facts).
- Misconception: "My gut feeling is always right. I don't need new technologies and have no interest in applying data science in my business."
Fact: It cannot be denied that data and AI can be very useful for a business, as smart algorithms can effectively demonstrate connections and data technology can be used to record historical data. However, the right combination of plant science and data science needs to be sought. The grower's expertise (knowledge regarding the crop or 'green thumb') must be taken into account. The predictions, suggestions and recommendations only act as data-driven support. I always use the metaphor of the pilot for this. The vast majority of the pilot's work on the plane is done by a computer. However, the pilot still has to make the decisions, and they have to be informed decisions. Even the most successful growers can benefit from data-driven insights, data and forecasts from time to time. We are increasingly seeing different types of greenhouse growers applying data science to improve their growing strategies.
- Misconception: "Generic plant and data science models are suitable for all growers."
Fact: Everyone cultivates differently and therefore specific information and growing conditions must always be considered. Each greenhouse is unique, for example in terms of location, type of greenhouse, lighting, growing medium, etc. These factors all affect growth and yield. It is therefore essential to work with personalised data models and the right program. Grodan has developed a data platform called 'e-Gro', which allows real-time analysis of specific data from your greenhouse. Read more about e-Gro at www.grodan.com/e-Gro.
- Misconception: "If I start doing data science and AI, I'm going to see results right away."
Fact: It is sometimes thought that a solution is generated right away after all the data is entered. However, it is not that simple. Descriptive analytics can indeed be done immediately, but predictive analytics (predictions realised from historical data) and prescriptive analytics (using data to make better decisions) take more time. Data must be collected from at least one complete growth cycle and requires a lot of data science knowledge, experience and skills. The data needs extensive preparation and processing, and existing features need to be selected and new ones added (feature engineering). It is not the raw data that is so valuable, but the insights that can be derived from the processed data. Here, big data and the people with whom it is collaborated are also very important.
- Misconception: "Plant models work directly as data science models and scientific models concerning generic data work directly on plants."
Fact: The actual 'behaviour' of the crop may differ from the theoretical physiological models, as there are often external and unexpected factors affecting growth, such as extreme climatic conditions, price fluctuations, pests and diseases, working conditions, etc. What is observed in a laboratory does not always work the same in practice (not one solution for all situations). This works the same the other way round. You cannot assume that known scientific models concerning generic data are going to work directly in the greenhouse. These predictive models have to be adapted based on the domain knowledge regarding crop and horticulture-specific activities.
Misconception: "It is enough to just hire a data scientist and a developer."
Fact: As mentioned earlier, it is very important to work with the right people, external and/or internal. To develop a valuable data science product, the team should consist of appropriate data scientists, developers, (multiple) domain experts, product managers and user experience experts so that good algorithms can be developed. You achieve this by using an existing platform with an extensive team to work with. In addition, a lot of good data from the greenhouse is needed to develop impactful algorithms. In the next chapter, read more about how to make the most of your data in your greenhouse.
Not sure where to start? Then start by finding out more about data science from the right people. For this, you can read books and articles, have discussions and try things out. It is advisable to follow as many groups on data science, machine learning, AI, plant science, and horticulture as possible.