Quantifying rapidly declining abundance of insects in Europe using a paired experimental design

Abstract The abundance of insects has decreased for the last decades in many parts of the world although so far few studies have quantified this reduction because there have only been few baseline studies dating back decades that have allowed comparison of ancient and recent population estimates. Such a paired design is particularly powerful because it reduces or eliminates bias caused by differences in identity and experience of observers, identity of study sites, years, time of season, and time of day, and it ensures identity of sampling procedures. Here, I compiled information on the reduction in abundance of insects in Europe and Algeria by the same persons compiling the abundance of insects from the same 21 study sites during 1951–1997 and again a second time in 1998–2018. There was a reduction by 47% in the abundance of insects. The difference in abundance in old compared to new samples declined with latitude, with a significant variance among taxa. This reduction in abundance of insects was of such a magnitude that it must have consequences for insectivores and the role that insects play in ecosystems.


| INTRODUC TI ON
Extensive surveys of insects have shown dramatic reductions in abundance by as much as 80%, even in nature reserves (Hallmann et al., 2017;Møller, 2019;Morrissey et al., 2015;Sánchez-Bayo & Wyckhuys, 2019). These changes have been attributed to a diversity of drivers including farming practice, land-use, and the associated factors such as use of pesticides, biological interactions, and climate change (Sánchez-Bayo & Wyckhuys, 2019;Vogel, 2017). While some studies have documented such reductions in insect abundance over time (Hallmann et al., 2017;Sánchez-Bayo & Wyckhuys, 2019), others have shown little or no change in abundance of insects (Conrad, Warren, Fox, Parsons, & Woiwod, 2006;Conrad, Woiwod, & Perry, 2002;Shortall et al., 2009). This raises questions about generality, but also the underlying mechanisms accounting for heterogeneity in such effects. In other words, which are the factors that account for these reductions in abundance, and do taxa differ in their reduction in abundance in a predictable way.
Biodiversity and abundance generally show latitudinal gradients although the causes of such differences among taxa are poorly known (Ball-Damerow, M'Gonigle, & Resh, 2014;Deutsch et al., 2008;Jacobson, Tucker, Mathiasson, & Rehan, 2018;Rohde, 1992;Tierno de Figueroa et al., 2010). Although a number of hypotheses have been proposed to account for such gradients, differences in life history such as generation time and differences in the relative importance of interspecific interactions have been hypothesized to be particularly important. Likewise, climate change has caused a strong decline in abundance of many taxa including | 2447 MØLLER insects (Parmesan & Yohe, 2003;Root et al., 2003). Recent dramatic changes in climatic conditions have been shown experimentally to affect normal sperm function in insects (Sales et al., 2018). Therefore, I investigate the extent to which latitude and taxa affected the decline in abundance of insects. Because farming practices and land-use have changed recently (Møller, 1980(Møller, , 1983(Møller, , 2019Sánchez-Bayo & Wyckhuys, 2019), this may greatly have impacted insect abundance since just a small amount of remaining pristine habitats is outside the reach of the decrease in diversity and abundance of insects in farmland. Changes in land-use by plants may drastically reduce the biomass of caterpillars when invasive plants invade hedgerows (Richard, Tallamy, & Mitchell, 2018).
Although there are several ways in which quantitative insect surveys can cause bias and even systematic bias over time (Møller, 2019;Sánchez-Bayo & Wyckhuys, 2019), there have been few attempts to quantify such bias (The Economist, 2019). The ideal procedure for tests of population trends depends on several strict criteria that reduce or minimize biased estimates. These include (

| ME THODS
The main study sites in Denmark were at Kraghede (57°N, 10°00°E), where open farmland habitats were surveyed extensively for insects in 1970-1974 and again in 2015-2018. The main crops in this agricultural landscape were grass, potatoes, wheat, beets and to a smaller extent barley, oats, rye, and other crops. Less intensely cultivated or uncultivated habitats were streams, ditches, ponds, road verges, hedgerows, and plantations. Changes in the extent of these natural habitats and other components of land-use are shown in Møller (1980), The present study of 45 km 2 farmland areas in Northern Denmark was part of an initiative for studying the effects of intensified farm land-use and farming practices on diversity and abundance of insects and birds (Møller, 1980(Møller, , 1983. Old studies that were originally made around 1970 were repeated in 2015-2018 in the same study plots by the same scientist as originally in order to ensure that there was no heterogeneity due to differences in study sites and among observers. These 18 study plots were combined with seven published data sets (reported in Conrad et al., 2006;Conrad et al., 2002;Hallmann et al., 2017;Hofmann, Fleischmann, & Renner, 2018;Møller, 2019;Schuch, Bock, Krause, Wesche, & Schaefer, 2012;Shortall et al., 2009) that were similarly investigated twice by the same scientists during an early study in [1970][1971][1972][1973][1974][1975] (Møller, 2013(Møller, , 2019. These studies have shown a high degree of consistency among sampling methods (windshields, sweep-nets, feeding rates of barn swallow nestlings, sticky traps, and samples of insects derived from windshields and the sound when large insects have an impact with the windshield (Møller, 2013(Møller, , 2019. Finally, it is possible to include car brand as a random factor in a statistical model and thereby adjust for differences in sampling effort among car brands. Such analyses only showed weak effects of car brand on insect abundance (Møller, 2019).  (Møller, 1987). These samples were collected in farmland surrounding 23 farms at Kraghede, Denmark in June-July 1984 and on the same 23 farms in June-July 2017 (Møller, 1987). Insect abundance was estimated as the number of individuals at the end of a transect.
Barn swallows mainly feed on insects living on middens and in farms with domestic animals, mainly cattle. Because barn swallows mainly feed on insects within a distance of 100 from farms with distances up to 500 m (Møller, 1987(Møller, , 2001 (Møller, 1991).
In addition to the insect samples described above. I used pub- I used GLM with normally distributed data and an identity link function to analyze the data. Analyses were weighted by sample size to take differences in abundance among analyses into account. I estimated chance in insect abundance by dividing log 10 -transformed abundance of insects in the early period by log 10 -transformed abundance of insects in the recent study period. I included latitude longitude and insect taxon in the analyses (listed in Table S1). I used SAS (2012) for the statistical analyses.

| Total sample sizes
The total abundance of insects for the old surveys was on aver-   Table S1. The ratio of log-transformed abundance of insects in old minus log-transformed abundance of insects in recent samples was 0.537, SE = 0.084, t = 6.43, df = 20, p < .0001). Thus, there was a significant reduction by a factor 3.443 between the two sets of samples.

Abundance in old compared to new samples declined with latitude
showing a particularly strong decline in abundance at low latitude with a significant variance among taxa (Table 1). The mean reduction in abundance was 47%, SE = 8%, t = 5.69, df = 20, p < .0001. There was a significant difference in decrease in abundance among taxa (Figure 2), but also a significant decrease with increasing latitude (Figure 2). The reduction in abundance was similar for all taxa as shown by a negative slope that did not differ among taxa (Figure 2). The difference in intercept is revealed by the lines in Figure 2

| D ISCUSS I ON
The abundance of insects has decreased considerably in many areas across the world (review in Sánchez-Bayo & Wyckhuys, 2019). For example, Hallmann et al. (2017) have documented declines in insect abundance in nature reserves in Germany by 80%. Here, I have compiled 21 sets of paired samples of insects that were collected by the same persons in the same study sites and with the same sampling procedures on average in 1976 and again on average in 2011. This paired study method is particularly powerful for quantifying whether there is a systematic decline in insect abundance while controlling for potentially confounding variables. I am unaware of any other study using a similar rigorous approach (e.g., the review by Sánchez-Bayo and Wyckhuys (2019) did not provide other similar studies).
Land-use in the study site at Kraghede, Denmark, that accounted for most of the datasets reported here, has been relatively modest with 80%-90% of crops being grass, potatoes, wheat, and beets.
Detailed information on land-use including ditches, ponds, and plantations are provided in detail in Møller (1983).
Insect abundance has been shown to vary with latitude in a number of studies although differences in sampling method and heterogeneity in sampling procedures among observers were never or rarely considered (Rohde, 1992 (Nebel, Mills, McCracken, & Taylor, 2010). This is consistent with insect abundance driving changes in the abundance of insectivores and not the reverse.
Pesticides have been suggested to play a crucial role in controlling the abundance or resulting in a decline in insect abundance F I G U R E 1 Mean insect abundance in recent samples in relation to mean insect abundance in old and recent samples of insects from the same locations. See the main text for definitions of abundance. The blue line is Y = X. Data are located in Table S1 0 2 4 6 8 10 0 2 4 6 8 10 Insect abundance from old samples Insect abundance from new samples F I G U R E 2 Abundance of insects in old samples relative to the abundance in recent insect samples at the same 21 sites collected by the same persons. Lines of different color show relationships for different taxa at different latitudes based on the statistical model in Table 1 in recent years (Fox et al., 2014;Hallmann et al., 2014;Nocera et al., 2012;Poulin, Lefebvre, & Paz, 2010;Sánchez-Bayo & Wyckhuys, 2019). This decline varied among insect taxa. I am still unaware of any large-scale studies experimentally testing for such effects.
Insects are consumed by insectivores, and reductions in population size of insects result in matching reductions in the number of insectivores. Indeed, there is a tight positive association between population size of insectivorous birds and population size of insects (Møller, 2019).
The findings reported here must have severe implications for insectivores such as many birds (Hallmann et al., 2014;Møller, 2019), but also for pollination and other biological interactions (Sánchez-Bayo & Wyckhuys, 2019). Indeed, avian insectivores have been shown to closely track the abundance of insects in Danish farmland during a long-term study of 23 years (Møller, 2019). It is no surprise that many plants with a main distribution in farmland have shown dramatic reductions in diversity and abundance.

ACK N OWLED G M ENTS
William Carøe Årestrup and Nouara Benslimane helped collect data on insect abundance.

CO N FLI C T O F I NTE R E S T
The author declares no competing interests.

AUTH O R S CO NTR I B UTI O N S
The author was responsible for the entire study.

E TH I C A L A PPROVA L
Collection of insect samples did not require ethics approval.

DATA AVA I L A B I L I T Y S TAT E M E N T
All data are available in Table S1 in this paper.

Anders Pape Møller
https://orcid.org/0000-0003-3739-4675 TA B L E 1 Abundance of insects in old samples relative to the abundance of insects in recent insect samples at the same sites collected at 21 site years by the same persons Note: The model had the likelihood ratio (LR) statistics χ 2 = 33.12, df = 13, p = .0016.