Map of 20 field site locations distributed around central Kansas. Fields are color-coded by crop type. Field dots are enlarged and shifted to maintain the anonymity of the participating farms. Map data from www.openstreetmap.org.

Example of an insect event’s signal and clustering. a) An example of an insect event recorded from the sensor. The wing beats are visible as modulations on top of the signal. The dashed red, solid magenta and dash-dotted blue curves show the body, diffuse- and specular wing signals respectively. The BWR is the ratio between the magnitude of the body- and diffuse wing signal. b) Clustered insect events recorded by the sensor in a soybean field (Field R) in July. The grey events are too sparse to form clusters and are therefore discarded.

The number of insects collected using sweep nets (top panel) and Malaise traps (middle panel), and insect flight events recorded with the sensor (bottom panel) per field. Insect numbers are separated by month with insects observed in June visualized with blue bars and in July with red bars.

Scatter plots of measured insect abundance comparing the monitoring methods on a logarithmic scale. a) Scatter plot of the number of insects captured with sweep nets and Malaise traps. b) Scatter plot of the number of insects captured with sweep nets and the number of insect flight events recorded by the sensor. c) Scatter plot of the number of insects captured with Malaise traps and the number of insect flight events recorded by the sensor. No correlations were found on insect abundance for any of these methods.

The number of insects collected with sweep net sampling and Malaise trap monitoring, aggregated by order.

Correlations between the automated biodiversity metrics calculated from sensed insect data, and those obtained from Malaise trap and sweep net collections. Rows in the table denote which data was used to fit the clustering algorithm, whereas columns indicate which parameters the obtained correlations refer to. Correlations with a p-value below 0.05 are significant and marked in bold.

Scatter plots and Spearman correlations for the species richness estimations across all models. The sensor results are from the model fitted to the total richness in both Malaise traps and sweep nets. a) Species richness calculated from Malaise traps vs. sweep net samples, b) species richness calculated from sensor clusters vs. sweep net samples, c) species richness calculated from sensor clusters vs. Malaise trap samples, and d) species richness calculated from sensor clusters vs. total richness across traps and sweeps.

Spearman’s correlation table between richness metrics calculated from sweep nets, Malaise traps, combined conventional methods, and sensor clusters (automated biodiversity metric) compared to ecosystem services of percent waxworm predation, total number of predators, Johnsongrass predation, Pigweed predation, Lambsquarter predation, and all seed predation.

A table describing the crop type and number of insects observed in each field in June and July across all three methods.

Measured insect abundance per crop and monitoring method. Mean and standard deviation.

Co-correlations of all biodiversity metrics.

Model parameters for each fitted metric.

A box plot depicting richness metrics from the Malaise traps, sweep nets, and sensors by field type.

A scatterplot depicting the correlation of the species richness metrics at each field, separated by the June and July timepoints.