Insect pests and diseases cause annual losses of 20-40% in crop production. A wide range of pesticides has been developed to combat these invaders. However, regulations are increasingly imposing constraints on the use of these chemicals, which raises the question: how can agriculture manage pests and provide enough food while also complying with increasingly strict regulations?
Keeping Pests in Check
Agriculture could try to shift towards a more organic, pesticide-free, and sustainable approach to growing crops. This has huge benefits for consumers and biodiversity, but current methods also come with some drawbacks. Organic agriculture often relies on natural predators for pest control, such as ladybugs that prey on aphids. However, natural predation generally works more slowly than chemical pesticides and consequently requires very early release in order to be sufficiently effective.
“Species-specific pesticides could reduce the overall need for chemicals and the rise in pesticide resistance.” – Broes Laekeman
“Species-specific pesticides could reduce the overall need for chemicals and the rise in pesticide resistance,” says Broes Laekeman, predoctoral fellow in AI and Vertical Farming working on the ‘Ornamental cultivation moves to a higher level’ project at Flanders Research Institute for Agriculture, Fisheries and Food (ILVO) and Ghent University. “The issue is that species identification needs to be done quickly to enable targeted applications of pesticides, before the infestation spreads, but this is an expensive and very labor-intensive process. AI is our best bet to solve this problem.”
Keeping Up with Whiteflies and Thrips
Some of the most common insect pests in greenhouse agriculture are whiteflies and thrips. It is interesting to note that these common names do not align taxonomically. While whiteflies refer to the taxonomic family of Aleyrodidae, the name thrips refers to the order of Thysanoptera, which contains multiple families of thrips. “There are about 1,500 species of whiteflies and 7,700 species of thrips,” says Laekeman. “Luckily, most of these species are not harmful. Some thrips species can even be beneficial, serving as pollinators.”
The problem lies with a minority of species, which damage crops by feeding on the plants, and can even carry viruses that lead to further harm or plant death, resulting in significant losses for growers.
There are two reasons for these insects’ success as agricultural pests. Firstly, they have short generation times: the span between an egg being laid, hatched, and then that individual laying their own eggs is just two to four weeks. Secondly, whiteflies and thrips can produce hundreds of eggs in their lifetimes.
“Short generation times allow these species to evolve quickly, which means they can rapidly develop resistance to widely used chemicals,” says Laekeman. “Biological control with predatory insects or species-specific pesticides would reduce the development of acquired resistance by relieving pressure on non-target organisms. This would also diversify control methods, making them more targeted and thereby reducing the total amount of pesticides used in horticulture.”
A Sticky Solution
Fluorescent insect sticky traps are widely used in horticulture, similar to the coils of sticky tape used to trap flies in kitchens. A common misconception is that these traps are solely used to suppress pest populations, while the primary goal of these agricultural traps is to detect problematic pests in a timely manner. Early pest detection makes it easier to control infestations, either by early release of predatory insects or by allowing growers to apply small, localized amounts of pesticides to the affected areas before the insects spread.
These traps are usually checked manually, with growers scanning the sticky yellow cards to see what insects are present and in what abundance. However, identifying pests in these traps at the species level is challenging and requires expert knowledge. “Not only are the insects very small, ranging from 0.5-3mm, but you often only get a partial or angled view of the insect, depending on how the insect sticks to the trap. It makes this method of pest detection difficult, time-consuming, and costly for growers,” says Laekeman.
Monitoring with insect sticky traps is often most intense in greenhouses or systems like vertical farming, where growers aim to optimize space by positioning plants in close proximity to one another. Unfortunately, this also creates ideal conditions for pests to thrive, spread, and go unnoticed. “These are the growers most in need of a cost-effective solution for frequent and intensive insect monitoring with accurate identification of individual pest species,” he says.
The AI Revolution
The use of AI in agriculture has been increasing and evolving in recent years, with applications ranging from irrigation management to detecting the need for fertilizer. As the technology’s complexity has grown over time, so has the diversity and number of its applications.
Read this article to learn more about Belgium’s ‘Positive for the Planet’ evolution through agritech!
In the early 2000s, pest-identification software was limited to thresholding color-scale images and counting the number of insects on a sticky trap in a controlled research setting. Later machine learning algorithms allowed for distinguishing general insect categories, but they couldn’t provide species-level identification. The software might have been able to tell a whitefly and a thrips apart; however, it was unable to distinguish whether that thrips was harmless or a pest species.
“Since about 2020, advances such as deep learning algorithms using neural networks have allowed new AI models to identify increasingly complex visual patterns,” says Laekeman. “This has finally sparked the possibility of a system for automated, species-level pest identification in agriculture using simple camera images.”
The Future of Pest Detection
At ILVO, researchers like Laekeman are now using cameras aimed at fluorescent sticky traps and deep learning algorithms in an attempt to automate insect identification. For growers, this system reduces manual labor and time spent checking traps, while increasing the accuracy of pest identification. “Importantly, the system also enables continuous monitoring, allowing for more timely detection of infestations and early, targeted intervention with species-specific pesticides or biological controls like ladybugs,” says Laekeman.
Read this article to find out how microalgae could help us move towards a more sustainable tomorrow!
However, not everything is said and done. The technology is still in development, the algorithms have so far only been tested in specific case studies, and the initial cost of the system is still considerable. But as the field progresses and the amount of training data increases, systems like this will be optimized and likely become commercially available.
Then, AI pest detection can help reduce the large-scale use of pesticides in agriculture, “to the benefit of growers, consumers, and the environment.”
