Estimation of required sampling effort for monitoring the possible effects of transgenic crops on butterflies: Lessons from long-term monitoring schemes in Switzerland

摘要

Butterflies (Lepidoptera) have been suggested for environmental monitoring of genetically modified organisms (GMO) due to their suitability as ecological indicators, and because of the possible adverse impact of the cultivation of current transgenic crops. A critical point is the sampling effort to be invested in such a monitoring. Here, we estimated the required sample size necessary to monitor potential effects of genetically modified crops on butterflies (Lepidoptera). We used data from two Swiss long-term butterfly monitoring surveys applying the common transect count method. The two monitoring surveys differed in several basic aspects such as geographical area covered, landscape context and sampling intensity. We carried out prospective power analyses in order to estimate the required sample size to detect effects of differing magnitude on mean species number, total individual abundance, mobility classes of butterflies and selected individual species. The required sample size decreased substantially when effect sizes above 10% were estimated. For example, a sample size of 79 transects would be sufficient to detect changes of 30% in total individual abundance for both survey types. Detecting effects on mean species number would need much less transects. Considerably more samples would be needed to analyze the abundance of single species. Several options are presented to increase statistical power or reduce required sample size, respectively. Also, we recommend to pool species to different mobility classes, and/or analyze patch occupancy of species instead of their individual abundance. The transect count approach is a suitable method for butterfly monitoring, both on a local as well as on a landscape scale. Consequently, both types of Swiss butterfly monitoring schemes are basically suitable for GMO monitoring. If transects are short and restricted to intensely used landscape, even non-professional field workers may yield data sufficient for effective monitoring, which might be relevant with respect to involved costs.