Aim: Recent studies suggest insect declines in parts of Europe; however, the generality of these trends across different taxa and regions remains unclear. Standardized data are not available to assess large-scale, long-term changes for most insect groups but opportunistic citizen science data is widespread for some. Here, we took advantage of citizen science data to investigate distributional changes of Odonata.

Location: Germany

Methods: We compiled over 1 million occurrence records from different regional databases. We used occupancy-detection models to account for imperfect detection and estimate annual distributions for each species during 1980–2016 within 5 x 5 km quadrants. We also compiled data on species attributes that were hypothesized to affect species’ sensitivity to different drivers and related them to the changes in species’ distributions. We further developed a novel approach to cluster groups of species with similar patterns of distributional change to represent multi-species indicators.

Results: More species increased (45%) than decreased (29%) or remained stable (26%) in their distribution (i.e., number of occupied quadrants). Species showing increases were generally warm-adapted species and/or running water species, while species showing decreases were cold-adapted species using standing water habitats such as bogs. Time-series clustering defined five main patterns of change – each associated with a specific combination of species attributes, and confirming the key roles of species’ temperature and habitat preferences. Overall, our analysis predicted that mean quadrant-level species richness has increased over most of the time-period.

Main conclusions: Trends in Odonata provide mixed news – improved water quality, coupled with positive impacts of climate change, could explain the positive trends of many species. At the same time, declining species point to conservation challenges associated with habitat loss and degradation. Our study demonstrates the great value of citizen science and the work of natural history societies for assessing large-scale distributional change.

Datafile 1 - Traits and trends.csv

(1) Odonata trait data are from various sources as follows:

Distribution: We estimated species’ European geographic range size as the number of occupied grids (50 x 50 km) in a published atlas (Boudot & Kalkman 2015).

Species temperature preference: Species’ temperature preferences were calculated by overlaying each species’ European distribution with an average temperature map from E-OBS v. 19e (Cornes, van der Schrier, van den Besselaar, & Jones, 2018) following other studies (Jiguet, Gadot, Julliard, Newson, & Couvet, 2007). For each species, we calculated the mean of the mean daily temperatures of occupied grid cells. While we call this variable “temperature preference”, its calculation did not aim to estimate species’ optimal temperatures but rather to place species on a gradient from those preferring cooler temperature to those preferring warmer temperatures.

Life-history: Data on voltinism, i.e., number of generations per year, was compiled from Corbet et al. (2006), complemented by expert knowledge within the co-author team. We applied a weighted mean of fuzzed-coded species affinities (values assigned to multiple categories reflecting the relative commonness of that category for the species, summing to 10 across all categories) to voltinism categories: multivoltine (coded as 5), bivoltinie (4), univoltine (3), semivoltine (2), and partivoltine (1). This weighted mean ranged between 1 (for a fully partivoltine species) and 5 (for a fully multivoltine species).

Phenology: Mean start dates of the main flight period were taken from Boudot & Kalkman (2015), which provided start date at a resolution of monthly tertiles. Species’ phenologies vary geographically but the data presented was usually for Bavaria, southern Germany. However, like for temperature preference, the aim was to create a variable that placed species on a gradient from those appearing early in the year to those appearing later.

Habitat: Main habitat preferences were classified according to descriptions in Dijkstra (2006) and Boudot & Kalkman (2015). Each species was coded to whether they use the following habitats: streams, rivers, ponds, lakes, ditches, canals, fenland, bogs, forest and quarries/pits.

Morphology: Hind wing length (median of lower and upper values) was taken from Dijkstra (2006).

Threat-level: We compiled data on the 2015 red list classification for each species in Germany (Ott et al., 2015). We aligned the German categories with the international IUCN categories following Jansen et al. (2020).

(2) Trend data

Trends were estimated using occupancy-detection models (see the main paper for full details). In brief the trends estimates represent the slope of a regression line through the predicted annual occupancy proportions for each species.

Datafile 2   - AnnualOccupancies

Annual occupancies (between 0 and 1) reflect the predicted proportion of survey quadrants (c. 5 x 5 km) occupied by each species each year between 1980 and 2016. Occupancies are derived from an occupancy-detection model, including site as a random effect and year as a fixed effect, fit using JAGS. The predicted occupancies (z), which account for variation in detection, are then summed across all survey qudrants in the analysis to reflect a mean nationwide occupancy for each species each year. The data frame contains summary statistics (mean, sd and quantiles) of the posterior distributions for these occupancy estimates.

Please read the full methods section of the paper in Diversity and Distributions before use. The paper is open-access.