Summaries of GLMs testing the effects of landscape factors (distance to city centre or percentage of sealed area in the surroundings), garden size (area with vegetation) and local garden characteristics, as well as of the interaction between the landscape factors and garden size on species richness and abundance of different taxonomical groups.

As the two landscape factors were intercorrelated, separate models were used to assess their effects. To test whether the assignment to classes with different distance to city centre, different percentages of sealed area in the surroundings, or different garden sizes affected the outcomes, analyses were repeated using models that treated all factors as continuous variables. This resulted in four different model types (models 1–4). Full models were stepwise reduced by omitting variables explaining little variation (F < 1.0) staring with the variable with the lowest F-value. However, the main landscape factors and garden size were always retained in the model. Species richness was log-transformed. All models used a quasipoission error distribution and log-link function.
(XLSX)