February 10, 2021

The future of agricultural decision making

The transition to data-driven farming

A lot has changed in the last 100 years when it comes to agriculture. When it comes to making decisions, many additional sources of information have been added for the farmer but in many cases he makes the decision himself. Farm21 has a vision in which we also help the farmer in this decision-making process. How we see this and what the advantages are we will explain three scenarios.

We would like to share our vision about the future of decision-making in agriculture. In this blog we explain how it evolves and the needed steps to get there. The evolution of crop advice is explained through three scenarios.

For each scenario, we look at information that precedes cultivation decisions and how this can determine subsequent actions. It shows how the intricacy of the information used to make farming decisions affects when you can undertake those decisions.

To properly compare scenarios with each other, we score both the information input and the resulting cultivation actions.

Scoring information

Ideally, you would monitor every individual plant always real-time. You then would know exactly when a plant needs what. Having fine-grained information also allows you to decide how local you want to act.

Resolution: How fine-grained is the available information.

Region (1) - Acreage (2) - Field (3) - Part of field (4) - Plant (5)

Frequency: The frequency of information being updated, giving insight into the correct current status of that information.

Season (1) - Month (2) - Week (3) - Day (4) - Real time (5)

Scoring of cultivation actions

All incoming information is translated into insights and ultimately actions. Actions are scored on 3 points.

The resolution indicates to which scale the actions (theoretically) are performed. Acting more locally means using fewer resources, realizing a gain in efficiency. The fineness of the grower's final actions then depend largely on the resources/machinery available to the grower. The grower can also determine whether it is cost-effective to purchase new equipment that would allow for finer-grained actions.

The timing of the actions determine whether the action is in time or not; never too late, never too early. This again could result in a more efficient use of resources.

The priority of the action indicates whether the actions are performed in the right order. For instance, which field to irrigate first. In occasions where many actions must be performed at the same time, the system would indicate which action to perform first to receive the best possible result.

Resolution: How fine-grained you (theoretically) can implement the action.

Region (1) - Acreage (2) - Field (3) - Part of field (4) - Plant (5)

Timing: Are you too early/too late or just at the right time.

Too late/too early (1) - Exactly right time (5)

Priority: The certainty that the action you are taking is the most relevant.

No right priority (1) - Exactly right priority (5)

Three scenarios

Scenario 1 - Traditional

Without data support, decisions come primarily from visual inspection, weather information, and (agricultural) knowledge. With this approach, there is room for improvement. The Wageningen University & Research states that Dutch growers are on average at 80% of their maximum potential yield. Next to that there might even be an even bigger gap between the average efficiency and the potential efficiency if you take all possible reduction of costs into account.

With the traditional approach, a lot of time is spent on visual inspection, the priority of actions is not optimal, and you cannot take targeted action. Detection of potential problems comes late and the window to react is short. A lot of work is done pre-emptively or too late (e.g. protecting all your crops from a certain disease right at the start of the season, or acting on a disease only once you've visually detected it).

Scenario 2 - Generic Decision Supporting System

Knowledge and insights are supplemented by measurement data from weather stations. Generic DSS models are used to draw conclusions. Generic DSS models are algorithms which describe conditions in which events occur that need the farmer's attention. As an example: if you have 90% or higher relative air humidity and temperature between 12 and 18oC for a period of 5 days, you have an increased chance of phytophthora infections in potatoes.

This approach provides additional supportive insights. It improves the timing and prioritization of cultivation actions. The downsides are that the model is static, Basic algorithms like this use a combination of 2 or 3 factors as basis for their outcome, but have no room to take additional data into account. Also, there is no option for feedback, which would allow the user to control/improve the outcome.

Comparing this scenario with scenario 1 you see that opportunity to detect potential problems is earlier, resulting in a longer period to act. The resolution of the data is still on acreage level, however, meaning that the scale at which the farmer could potentially preform an action also still on acreage level.

Scenario 3 - Personalized crop advice

This scenario describes the ultimate vision; fine-grained and frequent data collection from multiple sources, all linked together in one platform. The large amounts of data are backed up by DSS models which are further enhanced with machine learning.

Fine-grained measuring allows cultivation actions to be carried out at a very precise level. Due to all the information that is fed into the system and analyses by machine learning algorithms, the level of insights is vastly increased. This results in hyper localized, personalized farming advice.

Automated tailor-made cultivation advice is extremely detailed, works preventive but is targeted. It gives the chance to perform each action at the right time, right place and right quantity to optimize efficiency in yield and use of resources.

In conclusion

The comparison between the 3 scenarios shows the need for fine data gathering but also the need for combining multiple data sources. Only when agricultural data is shared it becomes truly relevant. From there the data needs to be properly translated and at Farm21 we see this as our job. We provide the platform and together with our partners and users we work to analyse and translate this data into proper farming advice. Only through that level of collaboration added benefit can be achieved.

Building our future

The quality of actions increases greatly due to the capabilities of machine learning. To this end, Farm21 is developing a self-learning system in which data sources are combined, provided with user feedback and thereby continuously improved.

To learn how we're building our future vision, please read this blog.

Continue reading

Soil moisture graphs in mm

Florence joins the team

Grow more

Smart precision farming has never been so accessible. Buy our starter pack to start right away.
“Farm21 helps us give our farmers better advice, save on costs and improve yields. It makes our services more scalable and growers love the insights.”
Sander Dekker
Crop consultant precision farming