The abundance and species richness of woodlice and millipedes were analysed using mixed effects models with location, watering treatment and survey time (May and September) treated as fixed effects, and plot pair as a block (random effect). Abundances were log-transformed prior to analysis to stabilize the variance. The full
models were fitted, and then non-significant interaction terms dropped from the model until the minimal adequate model was obtained as described above. The simpler models had lower AIC values than the more complex models in all cases.