A New Approach for Detecting Sublethal Effects of Neonicotinoids on Bumblebees Using Optical Sensor Technology

Abstract

Among insects, bees are important pollinators, providing many vital ecosystem services. The recent pollinator decline is threatening both their diversity and abundance. One of the main drivers of this decline is the extensive use of pesticides. Neonicotinoids, one of the most popular groups of pesticides, can be toxic to bees. In fact, numerous studies have found that neonicotinoids can cause sublethal effects, which can impair the biology, physiology, and colony survival of the bees. Yet, there are still knowledge gaps, and more research is needed to better understand the interaction between neonicotinoids and bees, especially in the field. A new optical sensor, which can automatically identify flying insects using machine learning, has been created to continuously monitor insect activity in the field. This study investigated the potential use of this sensor as a tool for monitoring the sublethal effects of pesticides on bumblebees. Bombus terrestris workers were orally exposed to field-realistic doses of imidacloprid. Two types of exposures were tested: acute and chronic. The flight activity of pesticide-exposed and non-exposed bumblebees was recorded, and the events of the insect flights recorded by the sensor were used in two ways: to extract the values of the wingbeat frequency and to train machine learning models. The results showed that the trained model was able to recognize differences between the events created by pesticide-exposed bumblebees and the control bumblebees. This study demonstrates the possibility of the optical sensor for use as a tool to monitor bees that have been exposed to sublethal doses of pesticides. The optical sensor can provide data that could be helpful in managing and, ideally, mitigating the decline of pollinators from one of their most major threats, pesticides.

Keywords: acute exposure; bumblebees; chronic exposure; convolutional neural networks; infrared sensor; insect monitoring; pesticide.

Figure 1

Two bumblebee events recorded from the sensor. In the left bee event, the bee flew across the 4 quadrants of the photodiodes, and due to the 2 wavelengths of the sensor, 8 data points correspond to this bee event. In the right bee event, the bee flew only across the 2 quadrants of the photodiodes, so 4 data points correspond to this bee event.

Figure 2

Wingbeat frequency of the bumblebees (control bumblebees = dash line; treated bumblebees = solid line) flying in the flight cages during the acute exposure experiment. All events from all repetitions were pooled together.

Figure 3

Confusion matrix for imidacloprid-exposed (acute) and control insect events.

Figure 4

(A): t-SNE map representing all insect events recorded. Orange dots = imidacloprid treatment. Green dots = control treatment. (B): t-SNE map with wingbeat frequency of the bumblebees overlaid as a colour gradient on top of the data. Wingbeat frequency scale ranges from 160 Hz to 220 Hz.

Figure 5

Wingbeat frequency of the bumblebees (control bumblebees = dash line; treated bumblebees = solid line) flying in the flight cages during the chronic exposure experiment. All events from all repetitions were pooled together.

Figure 6

Confusion matrix for imidacloprid-exposed (chronic) and control insect.

Figure 7

(A): t-SNE map representing all insect events recorded. Red dots = imidacloprid treatment. Green dots = control treatment. (B): t-SNE map with wingbeat frequency of the bumblebees overlaid as a gradient on top of the data. Wingbeat frequency scale ranges from 160 Hz to 220 Hz.