LAUSR

Abstract Coccinellid beetles are important functional components of many terrestrial ecosystems. Coccinella septempunctata L. and Harmonia axyridis (Pallas) (Coleoptera: Coccinellidae) are dominant, predacious beneficial beetles that both contribute to the suppression of many different pests, especially in temperate regions. Here, we analyse temporal trends in the populations of these two species over a 28-year period in wheat agroecosystem of northern China. To this, we used three machine learning algorithms (support vector machine [SVM], random forest [RF] and multilayer perceptron [MLP]) to relate occurrences of C. septempunctata and H. axyridis to selected environmental variables. SVM was the algorithm that most accurately forecast species abundance of these two lady beetles, followed by RF and MLP, using the coefficient of determination (R2), root mean square error (RMSE) and mean absolute error (MAE). Average temperature, average wind velocity, and average relative humidity were the most explanatory factors associated with the absolute abundance of C. septempunctata and H. axyridis in agroecosystems.