Performing Department
(N/A)
Non Technical Summary
Entomologists have used microphones since 1901, but these tools are challenging for farmers and scientists due to complexity in identifying diverse insect species. Applying machine learning to audio sensors, though rare in entomology, shows promise. Specifically, using contact microphones with tailored algorithms offers a cost-effective way to monitor economically important insect stages, as demonstrated by this team (Mehrotra et al., 2023). Practitioners at the farmer, regional government, commodity board, and industry levels independently identified the barrier to the use of the Insect Eavesdropper in decision-making as the link to economic thresholds. Therefore, we propose to (1) develop an algorithm correlating Insect Eavesdropper signals to pest density and (2) develop a decision tree framework to interpret the algorithm output as an economic threshold. We will evaluate the algorithm and framework for leaf defoliating, sap-feeding, root feeding, and seed feeding insect pests on three major commodities. If successful, this project will result in a new type of economic threshold based on autonomous sensing of insect damage to crops. By correlating the sensor signals with pest density and crop damage, we aim to empower agricultural stakeholders to make informed decisions.This research involves using a device called the Insect Eavesdropper to listen to the sounds insects make when they feed on crops. By analyzing these sounds, the researchers hope to estimate the number of insects and the damage they're causing. The study covers different types of insects and crops. For example, they're looking at beetles on soybeans and potatoes, and corn earworms on corn. They're also studying how insects feed at different times of day. The researchers are also trying to understand the relationship between the number of insects and the damage they cause. For instance, they're studying how much leaf damage beetles cause on soybeans, and how many aphids are on soybean and sorghum plants. In addition, they're looking at how corn rootworms affect corn roots and how corn earworms damage corn kernels. They're using the Insect Eavesdropper to continuously monitor these insects and record their feeding sounds. The researchers are developing algorithms to interpret the sounds recorded by the Insect Eavesdropper. These algorithms will help them understand the relationship between the number of sounds (or 'events') and the amount of damage caused by the insects. Finally, they're creating a decision tree, a type of model that can help make decisions based on the data. This decision tree will classify the damage as 'below', 'at', or 'above' the economic threshold - the point at which it becomes cost-effective to take action against the insects. This could help farmers make better decisions about managing pests on their crops. This project has the potential to revolutionize pest monitoring in agriculture, significantly reducing costs and labor while enhancing the efficacy and sustainability of pest management strategies, ultimately bolstering global food security, and mitigating the impact of herbivorous insects on agricultural crops.
Animal Health Component
50%
Research Effort Categories
Basic
(N/A)
Applied
50%
Developmental
50%
Goals / Objectives
This project aims to facilitate insect sensors incorporation in agricultural decision-making.Economic thresholds used for agricultural decision-making are based on costly labor-intensivemonitoring of pests. The 'Insect Eavesdropper' sensor enables autonomous detection andidentification of insect herbivores through recordings of their feeding sounds with contactmicrophones interpreted with machine learning algorithms. This $15 sensor has great potential toreduce the cost and labor of pest monitoring drastically while increasing the precision and efficacyof pest interventions. To bridge the gap between practitioner and their adoption of this technology,we propose the following objectives to develop economic thresholds to inform agriculturaldecision-making:Obj. 1 Develop algorithms to interpret Insect Eavesdropper signals as density and damageObj. 2 Develop a decision tree framework to interpret algorithm output as an economicthreshold
Project Methods
The methods described involve the use of an Insect Eavesdropper to estimate insect numbers and their impact on crops. The methods are divided into several sections:Analysis of Existing Data: The Insect Eavesdropper's recordings are analyzed to extract individual insect bites. The correlation between the number of bites and the number of insects is evaluated using a Pearson correlation.Leaf Defoliation Injury: The Insect Eavesdropper records feeding sounds from Japanese beetles on soybean plants. The aim is to correlate each insect bite sound with an amount of leaf defoliation at different insect densities.Sap-Feeding Injury: The Insect Eavesdropper records feeding sounds from aphids on soybean and sorghum plants. The aim is to correlate aphid density with feeding sounds in these different production crops.Root Feeding Injury: The Insect Eavesdropper monitors plants continuously at sites with high corn rootworm presence. The correlations between total feeding pressure and peaks in sound files are analyzed.Seed-Feeding Injury: The Insect Eavesdropper monitors sweetcorn genotypes during multiple weeks of corn tasseling. Corn ears are destructively sampled to quantify the corn earworm densities and will be correlated with sound recordings.Density and Defoliation Analysis: Algorithms are developed to interpret Insect Eavesdropper signals as density and damage. An algorithm parameterizes the relationship between plant defoliation and the frequency of Insect Eavesdropper events.Economic Threshold Decision Tree: A non-parametric supervised learning decision tree framework is implemented to interpret the density and defoliation algorithm outputs as economic thresholds capable of guiding agricultural decisions.