Long-term variation in the effect of lizards on spider density is linked to rainfall

The objective of this study was to ascertain whether the impact of lizards on spiders varies temporally, and if so, whether this variability is related to rainfall. We compared annual censuses of orb-spider populations on 24 islands with diurnal lizards present, and on 20 islands with diurnal lizards absent, to rainfall over a 10-year period. A strong positive correlation (Pearson r=0.877) was found between mean spider density on nolizard islands and the number of days of rain that occurred 2 months prior to spider censuses; correlation coefficients declined with longer and shorter time periods prior to censuses. Correlation coefficients between mean spider density on lizard islands and rainfall showed a similar pattern but were generally lower than those for no-lizard islands. The strength of the impact of lizards on spiders, measured as the ratio of mean spider density on no-lizard islands to mean spider density on lizard islands, varied considerably and was positively correlated with rainfall; the correlation was highest with number of rain days 6 months prior to spider censuses (r=0.741). Repeated-measures analysis of variance on the time series of spider densities showed that the lizard factor (present versus absent) varied significantly among years. Split-plot ANOVA with rainfall as a covariate indicated that spider density and the impact of lizards on spider density were both significantly correlated with rainfall. We discuss several hypotheses on the causal mechanisms that integrate abiotic and biotic factors.