LAUSR

Abstract Sampling flower-visiting insects in agricultural fields at large spatial and temporal scales is significant for understanding local insect pollinator communities. The most commonly used method, pan trap, has been criticized due to its attractant bias. A window trap (also referred to as the flight-intercept trap) is a non-attractant sampling method, which has been applied in forests and grasslands, but rarely in agricultural fields. We aim to test whether we can replace pan traps with window traps in agricultural fields by comparing species richness and species composition between the two methods, and to show whether flower-visiting insects collected in both traps can reflect flower-visiting activity recorded by camera observation. We conducted a 2-year study to compare the performance of these sampling methods in an oilseed rape field. Results showed that the relative abundance of dominant flower-visiting species was highly correlated between the window trap and the pan trap samples, while window traps caught more individuals and higher (rarefied) species richness than pan traps. The species composition of window traps was more similar to each other than that of pan traps. The proportion of honey bees (Apis spp.) collected in both traps underestimated their flower-visiting activity recorded by camera observations, while sweat bees (Halictidae) and butterflies (Lepidoptera) were overestimated. Our study suggests that the window trap has the potential to serve as an alternative sampling method of flower-visiting insects to the pan trap. However, we need to be cautious when using specimens caught in both traps as a proxy of their flower-visiting activity.