Modelling pollinator trends using multi‐scale environmental data from the UK Pollinator Monitoring Scheme

Declines in pollinating insects are a cause for concern, yet modelling abundance trends remains challenging and most species are not covered by long‐term standardised monitoring
We produce new metrics of change for insect pollinators which incorporate multi‐scale environmental variables into the modelling process to account for various sources of noise and potential bias. Data were generated from three surveys under the UK Pollinator Monitoring Scheme (PoMS) over 5‐ and 6‐year periods, measuring (i) insect abundance and species richness of bees and hoverflies from pan traps across a network of 1 km squares; (ii) insect abundance from Flower‐Insect Timed Counts (FIT Counts) in the same 1 km squares and (iii) FIT Counts by citizen scientists in self‐selected locations, especially gardens (‘public’ FIT Counts).
We compare five models containing different sets of environmental variables across insect groups from the three surveys. Our ‘full’ model, including both large‐scale spatial and temporal environmental variables combined with local‐scale survey variables, had the best fit, highest R 2 and lowest residual variance among sites in 31 of the 35 modelled datasets.
The full model estimated declines in ‘all insect’ abundance between 2017 and 2022 of 38% (average annual decline of 8.5%) from pan traps, and 33% (average annual decline of 7.4%) from 1 km FIT Counts. This reflected declines in hoverflies, ‘other flies’ and wasps on both the 1 km surveys, and bumblebee declines on 1 km FIT Counts, while solitary bee abundance showed no change. There was no change in overall insect abundance between 2018 and 2022 from public FIT Counts but hoverflies declined and several other groups increased. Given the short time series, these trends should be interpreted with caution.
Solution . PoMS represents a step change in evidence on UK pollinating insects. Our study demonstrates the value of including environmental variables to improve models of temporal trends in pollinator abundance and richness from standardised annual monitoring. Inclusion of these variables does not result in dramatic changes in estimated trends, which tend to be robust across model variants. The proposed metrics could help target conservation actions for different species groups or contribute to improved national biodiversity indicators.