Experimental landscapes
The 3D printed experimental arenas were made from PLA plastic and consisted of four habitats patches (circular: c . 7cm in diameter and 1.5cm in height) connected by four corridors (c . 28cm in length and 1cm in width) in a square (2x2) configuration, with each habitat patch connected by two corridors. Both habitat patches and corridors were filled with a mixture of plaster of Paris and black dye (c. 0.5cm in thickness). Plaster of Paris acts as a substrate for the Collembola to live on which keeps moisture, whilst the black dye makes Collembola more visible for counting (Li et al. 2021). To avoid moisture wicking between the different components of the landscapes (patches and corridors), plastic baffles were printed allowing the moisture of the corridors to be manipulated separately to the moisture of the patches (Fig. 1a).
As Collembola are delicate and not easy to handle, they were inoculated into the four patches of each landscape by tapping individuals (mean ± SD = 423 ± 317) into each mesocosm. During the experimental set up a potential confounding effect was accidentally introduced whereby microcosms were not randomly assigned to treatment after introducing the Collembola. This may have caused systemic differences in population level responses to treatments (Fig. 2), which we have attempted to account for by including microcosm as a random effect in the statistical models. Thus, food (dry dead yeast) was added to each patch each week (7±2mg mean±SD, n=30) throughout the experiment, and thus food was not a limiting factor to growth. Collembola were reared in the microcosms for three weeks prior the start of the treatments to allow populations to stabilise, with 2ml water and 7mg of food added each week to each patch. Clear lids were used to cover the microcosms and stop them drying out too quickly, and all microcosms were placed in dark rearing room to minimise light exposure.