The orchard of the future will be data-driven

Whoosh. Whoosh. Whoosh. If you drive past an orchard in the Netherlands, you will pass fruit trees with mechanical regularity. Those neat rows seem to have been designed by computer, but the management of Dutch orchards is set to become even more digitised in the future. Fruit growers will soon be taking decisions about each individual tree on their land based on smart data. Which tree has rather a lot of blossom? Which one isn’t growing very well? And how big will the harvest be?

While more and more machines are already being used in horticulture, for example in harvesting and pruning, the next step is innovation in managing the orchard. To achieve that goal, scientists are joining forces with growers and tech companies to implement new practices embedded in data. This latest development in horticulture is called ‘precision fruit cultivation’.

‘We see precision fruit cultivation as having the potential to make orchards not only more efficient but also more sustainable,’ says researcher Pieter van Dalfsen. Together with Peter Frans de Jong of the Fruit Trees research team at  Wageningen Plant Research, Van Dalfsen studies the developments in data-driven fruit growing. Which solutions work well and what are their benefits?

Precision fruit cultivation is part of a broader trend in agriculture and horticulture in which food production is underpinned by data and technology. The principles of the precision approach are the same in all forms of agriculture: measure, analyse, decide and implement. Growers gather data using sensors — for example on a drone or tractor — and then analyse it to determine targeted strategies. Dashboards give growers an overview and help them take tailored decisions. At present, they take those decisions themselves, but given current developments in robotics, that work is likely to become more automated in future.

Wageningen University & Research is a pioneer leading the way internationally in precision agriculture, the umbrella term for this approach. Wageningen scientists do practice-based research at the Farm of the Future, the National Precision Agriculture Living Lab, and the Fruit Research Centre in Randwijk. For example, in Randwijk, Van Dalfsen and his fellow researchers are testing sensors that collect data on individual fruit trees, and specially designed robot arms that can harvest pears. The orchard in Randwijk is monitored all year round, from flowering in the spring to harvesting in autumn and pruning in the winter months.

‘When the trees flower, that marks the start of a new growing season,’ says Van Dalfsen. ‘So let’s begin there.’ In the seasonal cycle of an orchard, the blossom is the preliminary announcement of the harvest. After all, the number of flowers is more or less directly related to the number of fruits hanging from the tree later in the year. ‘You don’t want too few flowers, but you don’t want too many either,’ explains Van Dalfsen. ‘If the tree is covered in blossom, not all the fruits will develop properly and you’ll end up with a lot of variation in ripening and size.’

To get a picture of the flowering, the grower can drive a tractor around the orchard that is fitted with a camera system. The camera takes photos of all the trees and an image recognition model then counts the number of flowers in each photo. This data gives an estimate of the number of flower clusters on each individual tree. Combined with GPS positioning, the result appears on the grower’s dashboard as a ‘blossom map’. The trees with the most blossom are shown as green and the ones with the least are displayed as red.

The grower then draws up a plan for the deepest green and deepest red squares on the dashboard. Trees with too many flowers are ‘thinned out’ by spraying them with a product that dries out the stigmata and styles of some of the flowers. That process is almost fully automated because it uses the blossom-map data: a machine attached to the back of the tractor driven by the grower sprays only those trees that were selected on the blossom map. The grower uses the same approach to spray the ‘red’ trees (with very little blossom) with growth products. ‘Growers spray less in total using this approach because they only spray the trees that need it.’

Ideally, over the course of the growing season the remaining flowers will develop into fine, uniform apples, pears and so on. The sooner growers knows how much fruit they can expect, the better they can prepare for the harvest. Van Dalfsen, De Jong and their colleagues are also working on a solution for this. As with the blossom mapping, the starting point is a camera system. It takes photos of the orchard and artificial intelligence (AI) is then used to count the fruits. ‘That will always be an estimate because a few fruits will be hidden behind the tree’s leaves.’

The method still needs a lot of testing, but the benefits are already clear. ‘This will give growers a good prediction of the yield and let them prepare effectively for the harvest. They will know roughly how many crates they need, how much space in the cold rooms and — perhaps most important of all — how much labour.’ In short, the data puts growers in control of the harvest and lets them make arrangements for only the cooling, space and resources they actually need.

It is conceivable that harvesting will become even more automated in the future with the use of fruit-picking robots. ‘In 2023, we carried out tests with a robot arm that we’d designed specifically to pick pears,’ says Van Dalfsen. ‘Robot designers aren’t as keen on pears as they are on apples because pears have a more complex shape and need to be harvested using a special tilting movement. Our robot was fitted with a silicone suction cup and a vacuum pump to grab the fruit at the broadest point and lift it up.’

A camera was used to tell the robot which pears it should pick. The system was driven by an image recognition module that gave a precise calculation of the best orientation for gripping the pear. ‘The robot essentially created a 3D model of each individual pear,’ says Van Dalfsen. The robot is not yet being used in orchards because the system is currently not good enough at recognizing the branches that can get in the way when picking fruit. ‘But it probably won’t be long now.’

The growers of the future will also collect data during the harvest. ‘One of our partners developed a camera system that collects information on every apple or pear that is carried along the conveyor belt into a storage crate,’ says Van Dalfsen. ‘The system is integrated into the picking platform. It measures the size of the fruit and the percentage blush coverage, while tracking which fruits end up in which crate. That data is used to calculate a quality label per crate, which is shown on the grower’s dashboard. If they link that to the crate’s location in the cold room, they are ready to sell their produce.’

And because all the data is also linked to the GPS location where the fruit was harvested, it is also possible to create a map of the yields. That shows precisely how much fruit was produced in each part of the orchard, plus the quality of that fruit. ‘The map lets growers check whether their decisions had the expected effects. Perhaps they will conclude they didn’t thin out the flowers enough — the tree was too full of fruit, meaning the apples stayed on the small side.’

Once the harvested fruit is in the cold room — or the supermarket — it is time to prepare the orchard for the next growing season. In other words, it is time for pruning. ‘A good shape for fruit trees is a triangle with lots of branches at the bottom and not many at the top,’ says fruit-tree expert Peter Frans de Jong. ‘Growers prune old wood to create space for new branches and make sure all parts of the tree get enough light.’

Is it possible to fully automate the pruning process? ‘In theory it is,’ says De Jong. ‘We have actually carried out tests with a robot fitted with pruning shears. It was perfectly good at cutting the branches, but the challenge with pruning by robots lies in the decision-making process. What algorithm will let us make sure the robot knows which branches to prune and which to leave intact? That’s a complex conundrum that we have yet to solve.’

Trees that are growing much faster than the ones around them need to be restrained in their entirety. Growers do that by pruning the roots. That’s not a very subtle process. In fact, it essentially involves a deep stab with a long knife to cut off part of the root system. ‘Machines are often already used for root pruning and you could use data to guide those machines,’ explains De Jong. ‘You would then be pruning efficiently, only targeting the trees that really need their growth inhibited.’ 

‘We have been experimenting with LiDAR for collecting data that can guide the root-pruning machines,’ he continues. ‘That technology gives you a picture of an area based on light signals, comparable to how radar works using radio waves. Self-driving cars use LiDAR to get a picture of the road.’ A light pulse is emitted and the longer it takes to be reflected to the light scanner, the greater the measured distance. The researchers were able to use LiDAR to create a detailed 3D map of the orchard, complete with the height and width of each tree.

Incidentally, LiDAR can also help make crop protection more efficient. ‘The light signals also show the spacing between the trees,’ says De Jong. If you enter that data into the spraying system, the grower can make sure the spray nozzles close automatically when they reach those spaces.

To summarise the potential benefits of precision fruit cultivation for orchards, Van Dalfsen cites the ‘biennial bearing effect’ as an example. ‘That is a real problem for fruit trees,’ he explains. ‘If an apple tree produces too much fruit one year, it becomes exhausted. As a survival strategy, the tree produces far less fruit in the following year. The tree then overcompensates in the year after that, and so it needs another year of rest after that.’ One year, the tree is laden with small, green apples; the next year, it has just a few large, soft fruits hanging from the branches.

Precision techniques can help combat that problem. ‘The trees that are out of sync are precisely the ones that are given what they need to get back in line with the rhythm of the orchard. The result is a more homogenous orchard with more and more trees bearing fruit of the same quality,’ says De Jong. What is more, growers use smaller amounts of crop protection products and fertiliser and they require less labour because they only take action where needed. Costs come down and the financial returns increase, and it is good for the environment and the climate.