Search

Highlights • The butterfly diversity is high in farmland of Romania and Spain, and less in Sweden. • The coefficient of variance of transect-specific species richness declined beyond 800 m transect length. • The required sample size to record a 10% loss of species was less than 30 transects per region. • The time to be invested in field work differed between countries. • The final effort to document a change of a certain magnitude was similar between countries. Abstract In agro-ecosystems, environmental monitoring is fundamental to detect and survey changes related to land use change and management practices. Butterflies and moths have often been suggested as suitable indicators for monitoring environmental effects on biodiversity in farmlands. Here, we estimated the required sample size and monitoring effort necessary to run a Lepidoptera survey in European farmland, assessing in particular if monitoring investment would differ between representative bio-geographical regions. We operated linear 1-km long transect routes in farmland of Romania, Spain and Sweden from 2013 to 2015, and recorded butterflies and burnet moths (Papilionoidea, Zygaenidae). The transects were walked back and forth four times a season, and replicated yearly. The lepidopteran diversity was high in farmlands of Romania and Spain, but comparatively low in Sweden. The coefficient of variation (CV) of recorded species number differed between countries being lowest in Sweden and highest in Spain. In general, the CV dropped above a transect length of 400–800 m, thus indicating an increase in statistical power. Assuming a non-parametric test for matched samples, power calculations were conducted with the raw count data and with log-transformed count data for comparison. When using log-transformed data, the required sample size to detect an effect was less than 10 transects per country or region (in order to detect a 10% loss of species or a decrease of 30% in total abundance). Specific subgroups of species, e.g. protected species or specific indicator groups, showed a higher variance, thus requiring a higher sample size to detect effects ranging from 12 to 16 transects (equivalent to 21–29 working days per country and year). When using original, untransformed count data a considerably larger sample size would be needed. Actual time to be invested in field work differed between countries due to contrasting regional constraints and conditions. Nevertheless, the final monitoring effort in working days was similar between countries as the factors involved balanced out each other, in particular due to the differing year-to-year variations. Our study demonstrated the feasibility of an environmental monitoring programme in arable land using farmland butterflies across Europe. We present a suitable approach and guidelines as well as the necessary effort to be invested in future Europe-wide monitoring programmes of butterflies in agro-ecosystems, based on predictions of statistical power. [ABSTRACT FROM AUTHOR]