Generalized mixed-model analysis

Generalized mixed models were used to assess the association of the main two factors investigated in the study (grass height, canopy cover) with insects’ total count as well as the number of species found. Time was also included as a covariate categorical factor. Sites was used as a random effect (random intercepts) to account for insects’ diversity in the various sites. The mixed model analysis was ideal as repeated measurements were made in each site. The fit of the model was assessed using Cox-Snell’s R² as well as residuals plot. P- values less than 0.05 were considered statistically significant. Statistical analysis was performed in R version 3.3.1. Statistical modelling was performed using the gamlss package (Rigby & Stasinopoulos, 2005), while data visualization was performed using ggpubr package (Kassambara, 2017).

Ground beetles diversity tests

Shannon’s diversity index was used to assess the diversity across the eighteen sites as well as across various times and in a correlation test against percentage canopy cover and sward height. The same sampling scheme was used across all sites. This test is essential for comparing diversity among different habitats (Clark & Warwick, 2001), in this case, 18 grassland sites. The Spearman’s correlation was also used to assess the strengths of the monotonic relationships between paired data. To measure how the distribution of individuals over the species was, the Pielou’s evenness index was used. Evenness was ideal in this study because as it gave a detailed perspective into how equal species where in the community (Heip et al. 1998). This analysis was achieved using the vegan package in R (Oksanen 2017).

Ordination analysis

Ordination was used to show how sites and species relate with each other by taking into account the measured external environmental variables (Ramette 2007). In this study, a correspondence analysis (CA) was used to project species association with sites and then using envfit function to map the environmental variables in the biplot. The fitted vectors being canopy cover and sward height, the arrows pointed to the direction of most rapid change. The analysis also shows how much each species is affected by the environmental factors. For plotting, the axes were scaled by the square root of R² (squared correlation coefficient). The p-values were based on random permutations of the data. To run the CA, the package vegan was used (Oksanen 2017).