Long-term changes in insect abundance: a method of their detection and its application to nocturnal Lepidoptera

The decreasing abundance of insect species is a common phenomenon of the present era. To detect the change in species abundance, it is essential to have a sufficiently long time series over which the change can be detected. We examined the time series of the abundance of 110 moth species using annual captures in a light trap operated nightly from April to November of 1967-1995. For each species, the time series was fitted by a state-space model allowing for a time-varying slope (ie for a locally linear trend with the slope allowed to change over time). We then determined the position in time and the length of the period when the estimate of the slope became significant. No significant change in abundance was detected for 65 species. Significant negative trends were established in 29 species; at first after 17 yr from the beginning of the study, significant positive trends were established in 16 species, with the earliest trend being detected after 15 yr. The trends were either transient or remained significant for the duration of the study (and possibly longer), demonstrating that the state-space modelling approach is suitable for detecting long-term changes in species abundance. The time series needed to be long (≥15 yr) to detect significant trends in abundance.