How does the Insect Calculator work?

To create the models, we used all the parameters that can be found as buttons in “Expert mode” as fixed effects: mowed (yes/no), mower type, mowing height, mowing width, number of cuts, meadow size, coordinates, day of the year (date) and proportion of sealed area (isolation). We ran a model with abundance as the response variable for each arthropod order. The glmmTMB package in R was used to run the negative binomial models (1). The “walk mode” is similar, but without the variables “mowing width” and “mowing height”. Finally, to simplify the “walk mode” even further, we have defined default settings for the examples of 4 different meadow types. This allows users who are not familiar with the subject to use the calculator initially without any knowledge of the mower type, for example. 

Consequently, we extracted coefficients of the model outputs to calculate the predictions manually in a simple linear function (prediction = a+ b*x + c*x...) with the values entered by the calculator. We used this approach to be able to work together with a company that helped creating this website. As for them it was more difficult to implement R, we found the most parsimonious way to simply provide a prediction function and providing a csv file that contains all model estimates.

Based on the extensive species level data of our project BioDivKultur and of the Biodiversity Exploratories, we additionally predicted the richness per arthropod order based on the abundance. Here, we used a nonlinear model (nls) to take into account a saturation of the number of species.