Abstract
The sugarbeet root maggot (SBRM), Tetanops myopaeformis (Röder) is a significant economic pest of sugarbeet in the Red River Valley growing area. It is also an important pest of the crop in several other states, and in the Canadian province of Alberta. Producers and pest management advisors have relied on manual sticky-stake traps or similar technology to monitor SBRM fly activity in commercial sugarbeet fields for several decades. However, artificial intelligence (AI) technology has the potential of improving the practice of insect pest monitoring. This technology involves the creation of a teachable machine (e.g., computer program and associated hardware) that is trained to work or react in response to information like a human being. The machine “learns” to react to specific information patterns and then performs a response function controlled by an algorithm. The machine continues to learn based on feedback into the algorithm, which can substantially improve its performance. This experiment involved field testing of two AI-based insect traps for monitoring sugarbeet root maggot fly activity in the Red River Valley during the 2018, 2019, and 2020 growing seasons. The following traps were compared for accuracy and efficiency with conventional sticky-stake traps: 1) the DTN SMART Trap, a modified Delta insect trap equipped with a high-definition camera mounted in its roof that collected images of insects captured on a sticky card placed inside the floor of the trap housing; and 2) the DTN Z-trap, which functioned as a low-voltage “bug zapper”, that counted SBRM flies based on electrical impedance produced by their bodies making contact with its electrodes. Both traps were equipped with cellular-enabled mobile reporting technology integrated with a data storage website and mobile device app, which collectively allowed for real time data access. Preliminary results from processed samples indicate that the camera-equipped SMART Trap exceeded 80% accuracy (before providing the algorithm with feedback), whereas the Z-trap tended to significantly overestimate counts as a result of flies in catch containers recovering from shock, repeatedly landing back on electrodes, and being recounted. That issue was resolved by adding a killing agent to the catch container in subsequent testing. Although sample processing and data compilation are ongoing, findings thus far suggest that this technology has strong potential to revolutionize the process of SBRM monitoring in sugarbeet, and it may also have applications in monitoring other important crop insect pests.