Speaker
Description
Insects are one of the most ecologically important animals in the world, pollinating, cycling nutrients and decomposing in the ecosystem, but also the backbone of life on Earth with birds, reptiles, amphibians, and mammals providing the backbone of almost all terrestrial environments. Despite this ecological significance, insect populations are in a rapid decline worldwide with over 40% of insect species being threatened with extinction, due to habitat destruction, intensive agriculture, pesticide use, light pollution, invasive species, and climate change. This situation threatens to decimate the ecosystems and services that support life on Earth and insect conservation is a global priority. Traditional methods such as manual trapping and morphological identification are not sufficient to solve this problem, as they are costly, labor-intensive and restricted to a few areas. The current study is an IoT-based passive acoustic monitoring framework (ESP32 microcontroller) that combines an INMP441 MEMS digital microphone and acoustic features and machine learning classifiers of three levels Low, Medium and High for nocturnal insects in a database of 3,000 recordings. The LSTM model produced the best performance, achieving 88.0% with ROC-AUC of 0.971 and the accuracy of all models was within a narrow accuracy range of 9.2 percentage points of the highest level. These results of the study suggest that IoT-based eco-acoustic monitoring is an effective, scalable, and cheap framework for quantifying biodiversity of insects.