As anticipated when designing the study, the moth populations crashed during the next two years (a–c). Most of the E. autumnata populations had already crashed in 2006. The O. brumata crash, starting from somewhat higher densities, showed more variation and continued until 2007 at most sites. The crash rate was strongly spatially density-dependent in both species (f,g).
Moran's I correlograms, based on variation in crash rate along the transect, depicted monotonically declining synchrony with increasing inter-site distance for both moth species, with a steeper cline for O. brumata than E. autumnata (b,c). Thus, the local populations crashed with variable rates that were distinctly spatially structured. The monotonically decaying autocorrelation implies that the crash rates became gradually more dissimilar towards the opposite ends of the transect. Specifically, the inner sites of the fiord crashed more steeply than the outer sites (a). Significant positive autocorrelation was evident only at inter-site distances up 10 km (b,c).
Both the mean and the variance of the site-specific parasitism showed similar, high levels for the two moth species (O. brumata: mean Pt = 34.0%, s.d. = 10.2%; E. autumnata: mean Pt =38.8%, s.d. = 18.9%). Overdispersion from random binomial variance in the site-specific parasitism rates (residual deviance/d.f.-ratios for E. autumnata: 2.69, p = 0.007; O. brumata: 4.11, p < 0.001) implied significant aggregation of parasitoids along the transect. However, both predictions regarding a regulatory impact of larval parasitoids on moth dynamics were rejected. Firstly, the Moran I correlograms of parasitism (d,e) did not visually match the correlograms of moth dynamics (b,c). Parasitism showed no consistent spatial structuring and autocorrelation was not statistically significant at any spatial scale (d,e). Secondly, parasitism did not predict the population crash rate at the site level in either of the moth species (E. autumnata: R2 = 0.00044, p = 0.94; O. brumata: R2=0.21; p = 0.10; d,e).