Abstract
Studies have reported widespread declines in terrestrial insect abundances in recent years1,2,3,4, but trends in other biodiversity metrics are less clear-cut5,6,7. Here we examined long-term trends in 923 terrestrial insect assemblages monitored in 106 studies, and found concomitant declines in abundance and species richness. For studies that were resolved to species level (551 sites in 57 studies), we observed a decline in the number of initially abundant species through time, but not in the number of very rare species. At the population level, we found that species that were most abundant at the start of the time series showed the strongest average declines (corrected for regression-to-the-mean effects). Rarer species were, on average, also declining, but these were offset by increases of other species. Our results suggest that the observed decreases in total insect abundance2 can mostly be explained by widespread declines of formerly abundant species. This counters the common narrative that biodiversity loss is mostly characterized by declines of rare species8,9. Although our results suggest that fundamental changes are occurring in insect assemblages, it is important to recognize that they represent only trends from those locations for which sufficient long-term data are available. Nevertheless, given the importance of abundant species in ecosystems10, their general declines are likely to have broad repercussions for food webs and ecosystem functioning.
Similar content being viewed by others
Main
Scientific, public and policy-related interest in the plight of insects has soared in recent years11,12, owing largely to reports of considerable losses of insect abundance and biodiversity1,2,3,4 and fears of concomitant declines in associated ecosystem services11,13. Often confused, however, is exactly which measures of insect biodiversity are being considered. Biodiversity is not a single metric, but rather a generalized concept that encompasses the numbers of individuals and species, species’ relative abundances (for example, evenness and numbers of rare and common species), as well as the identities of species and their interactions. All of these aspects of biodiversity inform the changes that are occurring to insect biodiversity, and might reveal hitherto-overlooked—but crucial—changes. However, although declines in insect abundance and biomass have been shown in large-scale studies and syntheses2,5,6, trends in other biodiversity metrics have been less clear. Species richness, for example, has been found to decline along with abundance in some large studies5,14, whereas in other studies, insect richness was reported to be stable6,7.
A better understanding of declines in insect abundance can emerge through the study of multiple (complementary) measures of biodiversity15, and, in particular, by examining trends in rare and abundant species. For example, even if overall abundances of terrestrial insects are declining2, there are several possible scenarios by which other metrics of biodiversity can concurrently change depending on how species with different relative abundances respond16. Box 1 presents some of these scenarios, particularly focusing on changes in initially abundant and rare species that can strongly influence various biodiversity metrics. We here define ‘abundant’ and ‘rare’ species along the continuum of abundances in a local assemblage (that is, where the species falls within the ‘species abundance distribution’17 (SAD)), rather than according to their regional occupancy or range size. Abundant species contribute the most individuals to any given assemblage, and their trends should therefore have the strongest influence on the changes in total abundance16. In fact, a decline in total abundance is unlikely without declines of the most abundant species. If we assume that the most abundant species decline on average, three simplified scenarios are conceivable for the rarer species. Relative to abundant species, rare species may, on average, (1) decline proportionally, (2) decline less or (3) decline more. Each of these scenarios gives distinctive signatures in various biodiversity metrics and in the shape of the SAD (see Fig. 1 and Box 1). There are other possible scenarios consistent with total abundance declines—for example, the most abundant species showing no declines—but this would require the unlikely scenario of extinction of most of the rarer species.
Here we examine the multidimensional nature of biodiversity trends in assemblages of terrestrial insects, arachnids (spiders and mites) and Entognatha (springtails and allies)—hereafter described collectively as ‘insects’, for brevity—during the past decades using a large compilation of insect surveys through time18. Although previous work has examined some of these multifaceted patterns for subsets of taxa (for example, butterflies14 or hoverflies16), locations6,7 and/or habitat types19, our synthetic approach allows us to assess the prevailing trends in multiple biodiversity metrics from the openly available long-term insect assemblage data across locations, taxa and habitats. Our focus here is on terrestrial insects only, because two recent studies from Europe20 and North America21 have used extensive data compilations to assess changes to the diversity of freshwater invertebrates. We identified 89 terrestrial studies from our previous data compilation16 in which one or more metrics of biodiversity either were provided or could be calculated from the provided data (see Methods). We added newly published years to the existing data, and fixed some minor data issues (updates available in the online repository22). To this, we added 17 new studies from the literature published between 2018 (the year of compilation by ref. 18) and 2021. We also searched the literature for studies in Spanish, Portuguese and Chinese to complement our original search in English and Russian, which allowed us to add two more studies. In all, we were able to analyse data from 106 studies with 923 total sampling sites (Extended Data Fig. 1 and Supplementary Tables 1 and 2), spanning between 9 and 64 years. Of these, 75 studies provided measures of at least two metrics of biodiversity (usually abundance and richness) and 57 studies (from 551 sites) provided full community data (that is, all taxa resolved to a consistent taxonomic level, usually species level) so that we could consider trends in multiple metrics of biodiversity simultaneously. We analysed the trends in total abundance (largely the same analyses as described previously2, but with the newly added studies), several metrics of diversity (for example, species richness, Simpson diversity), a metric of evenness, and trajectories of populations of species within the assemblage. We used hierarchical Bayesian autoregressive models to account for study- and site-level variation. Our models present the yearly change estimates and the 80%, 90% and 95% credible intervals (CI) of these estimates. Following previous studies2,23, we interpreted any 95% CI that did not overlap zero as strong evidence for a directional trend (that is, we are more than 97.5% certain that there is a directional trend in our data), whereas a CI overlapping zero between the 90 and 95% CIs was interpreted as moderate evidence, and an overlap between the 90% and 80% CIs was interpreted as weak evidence (that is, a 10% chance that the actual mean of the data is zero).
a–d, Effects of three conceptual scenarios by which species change over time (all species decline in proportion to each other (scenario 1, red); abundant species decline more than rare species do (scenario 2, turquoise); or rare species decline more than abundant species do (scenario 3; gold)) on population abundances (a), four biodiversity metrics (b), the numbers of species in different SAD intervals (c) and the mean population abundance trends of species in these abundance intervals (d). For biodiversity metrics that depend on species counts (species richness and evenness (b) and species richness per SAD interval (c)), a small amount of error was added. The insets in b represent modelled slope estimates in relation to 0 at the dashed line. Simpson diversity was converted to its effective number of species28. See ‘Extraction and calculation of biodiversity metrics’ in the Methods for an explanation of the biodiversity metrics. Note that because this simple conceptual model is intended to mimic a real dataset to some extent, the trends should be interpreted qualitatively rather than quantitively.
As expected from our previous analysis2, we found strong evidence for an overall decline in the total abundance of insects in this updated and expanded dataset of terrestrial assemblages. The overall change in abundance was −1.49% per year (95% CI: −2.18, −0.79) (Fig. 2), but there was considerable variation at the dataset level (interquartile range of the dataset-level predictions: −2.28% to −0.53%; Extended Data Fig. 1). This general trend was similar but slightly weaker (−1.22% (CI: −2.01, −0.42)) when we restricted the analyses to datasets that provided full community data (n = 57 studies; dotted lines in Fig. 2). We next determined whether these abundance declines translated into changes in other metrics of biodiversity, and, if so, which scenarios from Box 1 were most or least consistent with those changes. We found moderate evidence for an overall decline in species richness across all assemblages (−0.29% annually (95% CI: −0.64, 0.06); −0.20% (−0.57, 0.15) on the full community dataset; Fig. 2). This negative overall trend in species richness, albeit slight, contrasts with several analyses of insects and other taxa that found little evidence for consistent directional trends6,7,24.
Shading represents the posterior probability of the mean slope estimate, corresponding to the 80%, 90% and 95% CIs from the hierarchical Bayesian models. Numbers indicate the numbers of studies and sites available for each metric. Abundance is the total number of individuals observed at each time point, richness is the number of taxa observed at each time point, rarefied richness represents the expected number of species to be seen for the minimum number of individuals observed in any year in the time series, coverage richness is the expected number of species if 80% of the assemblage had been surveyed, Shannon and Simpson diversity were converted to their effective number of species and evenness was calculated as the ratio between the inverse Simpson index and species richness. The dotted lines for abundance and richness represent the results based on only the 57 studies with full community data.
Although concomitant declines in abundance and species richness are expected from a sampling process (that is, fewer species would be expected when fewer individuals are sampled from a species pool15,25), species’ relative abundances might also change. We found no evidence for any trends in diversity on the basis of individual- and coverage-based rarefaction26,27, nor in Shannon’s and Simpson’s diversity metrics (converted to their effective number of species28), which differentially weight abundant species (Fig. 2), suggesting no changes in relative abundances. Sensitivity tests showed that these trends were robust to the exclusion of very long and very short time series (Extended Data Fig. 2). However, when we restricted our analyses to studies that included many sites (10, 20, or 50), we found weaker declines in abundance and richness (Extended Data Fig. 3).
Our finding of a lack of temporal trends for these diversity metrics could imply that there was no change in the relative abundances of species through time15,25. However, we did find strong evidence for an increase in a metric thought to reflect evenness29, calculated as the ratio of the inverse Simpson index to the number of species (Fig. 2). Because this metric is calculated by dividing changes in species richness (which was declining) by inverse Simpson diversity (which was not changing), an increase could be expected. However, because our observations of changes in species richness appear to be driven largely by changes in the total abundance of insects (as indicated by the lack of trends in rarefied richness), changes in this metric need not reflect changes in relative species abundances.
On the basis of the patterns of the different biodiversity metrics (Fig. 2), a scenario of steeper declines of rare species is highly unlikely (scenario 3, Box 1). However, these patterns could be consistent with the proportional declines in scenario 1 (for example, a lack of change in the diversity indices), or with the steeper declines of initially abundant species (increasing evenness) in scenario 2.
To distinguish between these scenarios, we compared trends in the number of species in different abundance categories (as in Fig. 1d). For this, we first divided the total SAD of each site into five equally sized intervals of abundance, and then calculated the number of species in each interval each year (for a visual explanation, see Extended Data Fig. 4). We found the strongest declines (−0.80% annually (−0.96, −0.65%)) in the number of species in the highest abundance interval (Fig. 3), but there was also moderate to strong evidence for declines of species in the intermediate abundance intervals (between −0.33% and −0.43% annually). The lowest abundance interval (0–20%) showed little evidence for any trend. This interval contains the highest number of species (Extended Data Fig. 4), and represents several categories of trends, such as stable rare species, species that were previously more abundant, species that went locally extinct and newly colonizing species. We found similar results using alternative ways to bin species (see Extended Data Fig. 5). These patterns are again consistent with either the proportional declines scenario, or the scenario of stronger declines for abundant species, but not with the scenario of greater declines of locally rare species from Box 1.
The five (log10-transformed) SAD intervals were assigned separately for each time series, scaled to the highest abundance of any taxon observed in any year. Shading as in Fig. 2.
Declines in the number of species in the higher abundance intervals might give support to the scenario of disproportionate declines for initially abundant species. However, proportionate declines are still possible, because this would also result in the greatest loss of species from the highest abundance interval, as losses from lower abundance intervals are partially offset by gains of species that were formerly in higher abundance intervals (Box 1). To disentangle these possibilities, we tested whether differences in trends were detectable at the population level by calculating the mean population trends of all species within each initial SAD interval. A common pitfall when relating population trends to their initial values is detecting false trends owing to a regression-to the-mean effect30. That is, if, for example, an extremely high value is found at the start of a time series owing to stochasticity in population dynamics or sampling, the values for the following years are likely to be lower, which would lead to the false detection of a declining slope over time. Hence, the species with highest starting values are likely to show the strongest declines, and species with the lowest starting values the strongest increases, simply by returning to their long-term averages. To avoid this, we corrected the detected trends for each abundance interval by adding an expected regression-to-the-mean effect calculated from the longest and most robust time series (see Methods) and also conducted a series of sensitivity analyses (Extended Data Figs. 6 and 7).
Our population-level analyses indicated that species that were most abundant at the start of the time series showed, on average, the strongest declines (−7.72% annually; Fig. 4), despite considerable variability. Species in all other initial abundance intervals also on average showed strong evidence for declines, but with smaller magnitudes (between −4.63% and −6.14 annually) (Fig. 4; for the results of other ways of classifying species by their initial abundances, see Extended Data Fig. 7). We found that the most abundant species still showed the strongest declines when we took the most conservative approach to defining abundance intervals (based on species’ mean abundances across the whole time series, hence avoiding regression-to-the-mean effects; Extended Data Fig. 7).
Black horizontal bars and error bars represent the mean, 80% and 95% CIs of the mean estimates. The dots represent the population-level random effects, coloured by trend direction and certainty. The most extreme values are not shown, to aid visualization.
Among the populations with the steepest declines were species from across the insect tree of life (for example, beetles, moths and grasshoppers), and they included both agricultural pests, such as the corn aphid Rhopalosiphum maidis, and beneficial insects, such as predatory beetles (Supplementary Data 3 in the code repository31). Indeed, such declines in the populations of formerly abundant insect species have been reported from several locations, and include monarch butterflies (Danaus plexippus)32 and other butterfly species33, the meadow spittlebug (Philaenus spumarius)34 and the now-extinct Rocky Mountain grasshopper (Melanoplus spretus)35. This potentially common phenomenon of strong declines in formerly abundant species could offer an explanation for the observations that there are generally fewer insects in terrestrial systems than there used to be2. To test whether there is indeed a strong association between the mean population trends and the total abundance trends in a dataset, we related the dataset-level random effects of these metrics for all initial abundance groups. Although we found that all abundance groups had strong associations with total abundance loss, the strongest associations were for the more abundant species (Extended Data Fig. 8), supporting our assertion that losses of abundant species underlie the overall declines in insect abundance. Overall, on the basis of our compilation of openly available insect time-series data, we find that changes in local insect biodiversity are dominated by losses in total abundance and weaker losses of species richness, with the strongest declines observed for species that were more abundant at the start of the time series. Of the scenarios described in Box 1, these findings are most in line with scenario 2 (stronger declines for formerly abundant species), through stronger trends of taxa with a higher initial abundance; they also have some characteristics of scenario 1 (proportional declines), with stable diversity metrics and declines in richness. We argue that the most likely explanation for these patterns lies somewhere in between the two: initially abundant species decline strongly, but proportionally to each another, whereas the declining rarer species are often—but not always—replaced by other (now) rare species, leading to declines in richness. This result counters studies showing that biodiversity change is characterized by common species faring better (or less badly) than rarer species8,9. These studies used range size, not local abundance, as a measure of rarity, and did not analyse changes in species abundances. Still, reconciling these contrasting findings will require further work.
Despite these overall trends, we must clearly identify the limitations of data availability for our study. Our data, like those in most syntheses, are not a representative global sample, but have a strong bias towards Europe and North America (Extended Data Fig. 1 and Supplementary Table 2). When analysed without data from Europe and North America, we found no trend in abundance or richness (Extended Data Fig. 9), which probably reflects the limited data available and the high variability of trends in other regions, rather than an absence of change. Hence, our estimated mean trends cannot be geographically extrapolated; instead, they simply represent the most up-to-date state of the available knowledge one can gain from existing (open) data.
Although our work here focuses on trends of terrestrial assemblages, which have so far dominated the discussion on insect declines, trends of freshwater assemblages are similarly important. Our previous analysis indicated a widespread increase in the abundance of freshwater insects2, which was confirmed in a more extensive analysis for European streams that also revealed a mean increase in species richness20. By contrast, a recent synthesis of stream ecosystems in the USA21 indicated strong declines in the abundance but increases in the richness of freshwater insects during the past decades. Much still needs to be done to compare and contrast the trends of different geographical regions and to understand the differences in the changes of terrestrial and freshwater insect assemblages.
Our main finding—the disproportionate declines of initially abundant insect species—could help to explain the frequent observations that there are fewer insects now than in the past. Given the nature of our synthetic analysis across many taxa, systems and locations, we can only speculate on the underlying causes, which are likely to be associated with recent anthropogenic changes. For example, case studies have attributed the declines of some abundant species to climate change34,36, land-use intensification37 and decreases in plant nutritional quality38. Some species might also have been abundant in the past because they benefited from certain types of historic land use (for example, traditional, low-input agriculture), but have declined more recently as land uses have changed37,39. Abundant species are often disproportionately important for ecosystem structure10, functioning40,41 and services42,43, as well as for the diversity and abundance of higher trophic levels10,36,44, so their declines are likely to have already led to a broad-scale rewiring of ecosystems, and will continue to do so.
Methods
Data collection and selection
The basis for the current analysis was the InsectChange database18, which we developed to compile openly available long-term (ten or more years; nine or more years for datasets from under-represented regions) standardized surveys of insect, springtail and arachnid groups (families or assemblages). For the current analysis, we updated the database, which initially focused on assemblage-level abundance and biomass, by extracting species richness and other diversity metrics from these same publications, and including new data that met our criteria published until February 2021. This information included both data added to open-access repositories since our last download, and data from newly published studies. For a new study to be included in our compilation, we required that studies provided two or more calculated biodiversity metrics, over a time span of at least ten years, or the full set of taxa and their abundances for each time point so that we could compute multiple metrics. We searched for new publications fitting these criteria by searching Web of Science (for English language) and https://www.elibrary.ru (for Russian language), using the same search terms that were used to compile the InsectChange database18 (see Supplementary Methods for the exact search strings). For this study, we increased our global coverage by searching papers published in Spanish and Portuguese using https://www.scielo.org/ and in Chinese using https://www.cnki.net/. This yielded two additional studies with appropriate data. In addition, we searched all papers citing references1,2. We also used cross referencing and personal networks to find data. In total, we were able to add 17 new studies from which we could extract data to the InsectChange database, and we updated nine studies that provided more recently published years of data.
For the purposes of our analyses here, we only included studies of terrestrial insects because many of our datasets on aquatic insects within the InsectChange database18 did not have a consistent taxonomic resolution over time to calculate reliable biodiversity metrics, and larger, more comprehensive syntheses of freshwater insects have recently been published20,21. We also excluded datasets that had only biomass measurements, as well as experimental sites at which researchers manipulated the environmental conditions. Within each sample, we excluded invertebrates that did not belong to the classes Insecta, Entognatha (springtails and allies) or Arachnida. For studies in which the provided total abundance or richness values might have included non-insects, we checked that these assemblages consisted of at least 90% insects, arachnids or springtails and excluded them if that proportion was unknown or smaller. Finally, we excluded one study (DatasourceID = 70) that was originally derived from the Global Population Dynamics Database46 and later included in BioTIME47 because its methods and methodological consistency could not be verified (M. Dornelas, personal communication).
Dataset description
In total, for the analysis presented here, the database consists of 106 studies, 75 of which provided at least 2 diversity metrics, and 57 provided raw data from which we could calculate all biodiversity metrics (Supplementary Table 1). The data were heavily skewed towards Europe and North America (69% of studies providing abundance data, and 74% of studies providing compositional data; Supplementary Table 2). The median duration of the studies was 20 years, with a median of 11 sampling years. The number of sites per study ranged between 1 and 138, with a median of 3. More details on the dataset characteristics are provided in Supplementary Table 3, with a full account of all studies48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63,64,65,66,67,68,69,70,71,72,73,74,75,76,77,78,79,80,81,82,83,84,85,86,87,88,89,90,91,92,93,94,95,96,97,98,99,100,101,102,103,104,105,106,107,108,109,110,111,112,113,114,115,116,117,118,119,120,121,122,123,124,125,126,127,128,129,130,131,132,133,134,135,136,137,138,139,140,141,142,143 included in Supplementary Table 4. All data used are available as Supplementary Data 1 and 2 in the code repository31.
Extraction and calculation of biodiversity metrics
Metrics of biodiversity change
For datasets that provided derived metrics of species abundances or different metrics of species diversity (for example, species richness), we simply included those values in our analyses after first ensuring that investigators performed appropriate standardization in their sampling methodology, or after the fact, in their calculation of metrics. For datasets in which the numbers of individuals of each species were provided, we calculated several biodiversity metrics after ensuring appropriate standardization of sampling (for example, sampling effort, time of year; further discussed below).
For each study, we calculated as many of the following metrics as possible given the data:
-
Abundance: Total number of individuals observed per year.
-
Species richness: Number of species, morphospecies or higher taxa observed in each year.
-
Rarefied richness: Richness accounting for differences in numbers of individuals; that is, the expected number of species in each year, had the number of individuals caught been equal to the lowest total abundance observed in any year in the time series (if the lowest number was lower than ten individuals, we calculated the rarefied richness for ten individuals).
-
Coverage-based richness: Expected number of species if 80% of the community had been sampled26.
-
Effective numbers of species given Shannon diversity (H′): A diversity metric that quantifies the uncertainty (entropy) in the prediction of species’ identity converted to its effective number of species (eH′) (ref. 28).
-
Effective number of species given Simpson diversity (D): A diversity metric that measures the probability of intraspecific encounter (with replacement) converted to its effective number of species: 1/D (ref. 28).
-
Evenness: The ratio between the effective number of species given Simpson diversity (see above) and species richness (1/D)/S. This metric of evenness is among those recommended previously29 (I1/D).
We calculated coverage-based richness using the iNEXT144 and all other biodiversity metrics using the vegan145 package for R.
Changes in the SAD
To investigate changes in the number of species that were differentially common or rare in the community, we studied changes in the SAD. The SAD describes the distribution of species’ relative abundances in an assemblage17, here quantified through bins of log10-transformed abundances plotted on the x axis, and the number of species in each bin on the y axis. Although the precise shape of this distribution and what this means for communities has been discussed at length146, we were only interested in changes in the number of species in different SAD intervals.
On a per-site basis, we defined intervals of the SAD and counted the number of species falling into them as follows. First, we created five equally sized intervals of the log10 abundances (Extended Data Fig. 4b), which ranged between zero and the maximum abundance observed at that site over the course of the time series. In other words, the baseline for ‘locally abundant’ at a site is the highest abundance of any species recorded from the site over the course of the entire time series. Our second way of slicing up the SAD was to assign four quartiles based on all abundance values in the time series (Extended Data Fig. 4c), each quartile having (approximately) the same number of observations (some variation exists because of rounding). Given the skewed SADs typical for insect assemblages45, the fourth quartile of our second approach is naturally the widest, and can occupy over 50% of the abundance range (Extended Data Fig. 4c). By contrast, with our first approach, the uppermost quartile contained the lowest number of species, composed of the small number of very common species. For each year, we then calculated the number of species whose log10 abundances fell in each interval or quartile. For ties, we took the number of species smaller than the quartile value. We present the results of the first approach in the main text and those of the second approach in the Extended Data Fig. 5, as both lead to similar conclusions.
Mean population changes
To test whether species that were at first more abundant showed more negative temporal trends, we modelled the mean population trend of all taxa in relation to their initial abundances. For this, we used 34,317 populations at 584 sites from 56 studies. We grouped species per site into 6 initial abundance intervals: 5 groups based on the range of log10 abundances of the species found at a site in the classification year(s) (similar to the division for the SAD brackets above, but described in more detail in the next paragraph), and a group for the species that were not detected in the classification years. Because we were interested in local rarity, we allocated species by their initial abundances per site. An alternative way of studying trends in relation to species’ rarity would have been to look at regional abundance or occupancy, but this was not possible for most of our datasets, For an analysis studying occupancy changes including part of the presently used data, see ref. 8.
To allocate species at each site to an initial abundance interval, we created 5 equally sized intervals between 0 and the log10-transformed highest observed abundance value at the start of the time series. We allocated each species to one of the five intervals according to its initial abundance. The species that were absent in the classification year(s) were not analysed because their mean trends would in most cases be positive, leading to a positive mean trend. In the main text, we used the data collected in the first sampling year to define species’ initial abundances, but other ways of allocating species by their initial abundance values are possible; for example, on the basis of their average abundance values over years 1 and 2 or years 1 to 5, or their average abundance across the whole time series (that is, assigning generally abundant and rare species, as has been done previously147). The effect of such choices on the slope estimates is shown in Extended Data Fig. 7.
We discarded sites with no or only one species detected in the classification year(s), or with only a small number of equally abundant (usually singleton) species (that is, when no vector of relative abundance values could be calculated without infinitely large numbers). In all, between 5 and 25 sites in 10 studies were discarded, depending on the number of years used for classification.
When relating trend slopes to their initial values, one is likely to find the steepest slopes for the species with the most extreme (highest and lowest) starting values, owing to a regression-to-the-mean (RtM) effect30. This occurs when, owing to stochastic variation in population dynamics, or random sampling effects, a high value that is drawn in the first year is followed by lower values in subsequent years, closer to the long-term average, leading to an overall apparent negative trend, even when no such decline exists. This can be dealt with by calculating an expected RtM effect from a control group, and adding this to the estimate of a treatment group148. Because an independent control group for real-world monitoring data, without stochasticity or sampling variation, is not available, we used the best sampled sites in our dataset to estimate an expected RtM effect. We first selected only those sites with at least 15 years of sampling data (260 sites from 26 datasets), and modelled the mean population trends of each initial abundance group. To estimate the potential contribution of RtM to estimated long-term trends, we separated the data used to estimate initial abundance from the data used to estimate the long-term trends. Thus, we left-censored the data by one year (and three years as a sensitivity test), and modelled the population trends again, while using the censored year to classify the initial abundance. The difference in year slope estimates between the full time series and the censored time series was, for each initial abundance group, used as a correction factor, and added to the estimates of the full analysis, to shrink the final estimates towards zero. In the main text, we present the corrected data based on one-year censoring; the potential effects of more censoring and the number of years used for classification can be found in Extended Data Figs. 6 and 7 and the calculated correction factors can be found in Supplementary Table 5.
Note that our most conservative method for assigning species to abundance intervals—on the basis of their mean abundances across all years—is not affected by RtM effects. Qualitatively, our results remain the same across these sensitivity analyses.
For datasets that needed sample-based rarefaction to equalize sampling effort over years (see below), we used the rounded median values per species per year as input for these population models.
Extraction of biodiversity metrics from publications
We extracted pre-calculated biodiversity data from studies that did not provide raw data. We could extract the following metrics across the time series: total abundance (n = 89 studies), number of species (n = 28), rarefied richness (n = 3), Shannon diversity (n = 4) and Simpson diversity) (n = 1).
Although we endeavoured to run our models on one value per year (to avoid complications from species’ phenologies), this was not always possible, because some studies provided these metrics per month or per season. In these cases, any potential seasonal patterns were dealt with by including a random intercept on time period of sampling (see ‘Statistical analysis’).
Preparation of raw data
For studying changes in biodiversity metrics, it is essential that sampling effort over the years is equal. Sampling methods were standardized within sites (for example, type of trap used or type of survey), but could vary in sampling effort (for example, duration of sampling period or number of subsamples). To equalize this variation, we processed all studies that provided raw community data in such a way that each year in each site provided one value for each biodiversity metric, on the basis of the same sampling effort (that is, same number of weeks sampled, same number of traps active and/or same number of subsamples taken), summed over the same season(s). In this way, any phenological differences or shifts among years in sampling effort should be accounted for in the best possible way, and stochastic variation should be minimized. Because every study had a different design, we had to decide on the standardization strategy on a case-by-case basis.
The community data available to us can be divided into three types, with different processing requirements:
-
(1)
The study provided one species-by site-by-year matrix. We used this directly to calculate all biodiversity metrics for analysis.
-
(2)
Species-by-site by-year data were provided together with an account of the sampling effort (number of samples underlying each row is reported, but no raw data were provided per individual sample). Here, we subsampled the years with higher trapping effort to provide the expected community composition if the number of samples had been the same as in the year with the lowest sample size. We repeated this subsampling procedure 100 times, and calculated the means of each biodiversity metric, to be used as model input. In some cases, it was clear that subsampling to the lowest sample size would lead to a large loss of data, and in such cases, it was more economical to discard a year than lose a large amount of data. A practical example: if year 1 has 300 individuals in 6 traps and year 10 has 500 individuals in 4 traps, we subsampled 200 individuals (4/6 of 300) from year 1, so that each year has the number of individuals equivalent to 4 traps. We repeated this resampling 100 times to derive 100 hypothetical community compositions, and calculated the biodiversity metrics and population abundances for each iteration.
-
(3)
Exact data were provided per sampling event (number of individuals per species per trap/site per date), but the number of traps, dates or periods sampled is variable among years (owing to trap malfunctions or changes in sampling design or sampling frequency). Here, we first divided the years into monthly or bimonthly periods, and then decided whether we could maximize data use by including more months or more years (that is, whether to exclude periods that were not sampled each year, or exclude years that did not have samples in each period). The aim here was to maximize data use, so that we could attain the highest quality of data for a maximum number of years. For each period, we then subsampled the minimum number of samples observed in any year from the available samples, and added these together to form one yearly sample. This was repeated 100 times, and the biodiversity metrics were calculated each time. The mean values of these 100 iterations were used for model input. A practical example: ten pitfall traps per site were usually emptied weekly, but sometimes fortnightly, the trapping period changed from year-round at the start to May–September in later years, and sometimes samples were missing owing to flooding or trampling by livestock. Here, we first discarded all months other than May–September. Then we allocated five monthly periods and calculated the number of sampling days (sum of sampling period × number of traps) for each month. We then drew random samples (without replacement) from the available samples per month, until the total number of sampling days added up to the minimum number sampling days observed. We added the months up to form one yearly sample and calculated our metrics. This was repeated 100 times.
The sampling effort is thus equal among all years in each site, but not necessarily among different sites within a study.
Taxonomic considerations
Insects make up the largest class of eukaryotic organisms, but at least 80% of species remain undescribed. In addition, the expertise required to identify insects to species level is high. No person can identify all species in an area, and there are often life stages in which a species can’t morphologically be distinguished from other species (for example, juveniles or one of the sexes).
It is therefore no surprise that, in our amassed data, there is a large variability in taxonomic precision among studies, and that there is a trade-off between taxonomic breadth (the number of insect groups assessed) and taxonomic precision (the taxonomic level at which all organisms are classified). To avoid biasing our analysis to those taxa in those regions that can reliably be identified to species level, we took a two-pronged approach:
-
(1)
For studies in which almost all individuals were identified to species level (85% of studies), we avoided artificially increasing our biodiversity-change metrics owing to a small number of imprecise identifications (usually genus or family level). To do this, we developed a simple algorithm to clean the taxonomy: We probabilistically allocated individuals at genus level to one of the observed species in the same genus. If no congenerics were present in the time series, these genera were treated as species. This cleaning was always performed just before calculating the biodiversity metrics, also in randomization models.
-
(2)
For studies that were not identified to species level (15% of studies), we accepted this taxonomic imprecision, and assumed that the data collectors identified all organisms to the best of their abilities. We checked that the taxonomic precision had remained stable over time. The use of taxonomic groupings other than species as a proxy for diversity changes is justified, because various studies have shown that identification to higher taxonomic groupings, or morphologically distinct taxa (morphospecies), is generally sufficient to detect differences in richness and composition over environmental gradients149. Of the 34,317 populations, 84.1% were identified to species level, 6.9% to genus, 5.1% to family, 2.7% to suborder and less than 1% to higher taxonomic levels.
Statistical analysis
To test for temporal changes in these biodiversity metrics, we used autoregressive hierarchical Bayesian models. The fixed model structure was simple, with the biodiversity metric of choice as the response variable, and ‘Year’ as the only fixed independent variable.
By focusing only on the temporal slopes, we could account for the different scales of measurement (from one to thousands of individuals, and up to hundreds of species per year). The coefficient of the ‘Year’ variable (the temporal slope) can be back-transformed to the percentage change per year, and we report both in all graphs.
The response data of all biodiversity models with positive exact, estimated or averaged counts (total abundance, species richness, rarefied richness, coverage richness, Shannon diversity, inverse Simpson diversity and the number of species in each of the SAD quantiles) were log10(N + 1)-transformed for analysis, and the error structure was assumed to follow a Gaussian distribution. The evenness values were bounded between 0 and 1, and the error structure followed a beta distribution.
We accounted for the non-independence of the repeated measurements at each site and the expected autocorrelation among sites part of the same study by including a series of nested random intercepts and slopes:
We included random intercepts for study; study area (in cases when sites were clustered in different study areas of the same study); site (the smallest reported sampling unit); and within-year time period (finest resolution: month; when samples were collected repeatedly within year, nested within study). We included random slopes for the effect of year (at the levels of study, study area and site), and for temporal autocorrelation by adding an autoregressive term of order 1 (AR1) (as a continuous Ornstein–Uhlenbeck process), on which we placed a random effect at the site-level to allow site-level variation in the strength of autocorrelation.
We fitted these models using integrated nested Laplace approximation (INLA)150 in R4.2.2 (ref. 151), a Bayesian method that efficiently and accurately approximates Bayesian posterior distributions, without using Markov chain Monte Carlo methods, and which allows for complex layered random effects, including autoregressive terms.
As priors, we used penalized complexity (pc) priors with as hyper-parameters a sigma of 3 × s.d. of the data and α = 0.01, meaning that there is a probability of 0.01 that the mean lies outside 3 standard deviations of the data.
The prior for the AR1 correlation is defined by INLA on the logit lag one correlation scale (where ρ is the correlation coefficient) and was given a pc prior with a sigma of 0.5 and an α value of 0.01 (probability of 0.01 that the mean is larger or smaller than 0.5).
The general model structure in INLA annotation is:
inla(x ~ Year+ f(Period, model='iid', hyper = prior.prec))+ # random intercept season f(Plot_ID, model='iid', hyper = prior.prec))+ # random intercept site f(StudyArea_ID, model='iid', hyper = prior.prec))+ # random intercept area f(Datasource_ID, model='iid', hyper = prior.prec))+ # random intercept study f(Plot_ID_slope, iYear, model='iid', hyper = prior.prec))+ # random slope site f(StudyArea_ID_slope, iYear, model='iid', hyper = prior.prec))+# random slope area f(Datasource_ID_slope, iYear, model='iid', hyper = prior.prec))+# random slope study f(iYear, model='ou', replicate=as.numeric(Plot_ID_4INLA), hyper = list(theta1 = list(prior='pc.prec', param=c(0.5, 0.01))))# AR1 term family=“gaussian” # normal distribution for the log-transformed count data
where x = log10 (metric + 1) for the univariate metrics, and untransformed beta diversity values for the turnover models.
Models of population change
To estimate the mean population trends of all initial abundance intervals, we adjusted model structure for the aggregate biodiversity metrics above, by adding an additional layer of random intercepts and slopes (for unique species × site combinations), allowing random variation among species. The autoregressive term was retained at the site level, because the population-level autoregression required excessive memory use (more than 1 TB) and never converged when included at the species × site level. Here, we used a Poisson error structure (that is, no log transformation) to allow for zero abundances. For the rarefied datasets, we used the median value over the 100 iterations for each species. We rounded any non-integer species abundances, and removed any species that had no non-zero observations left at a site. After modelling, we adjusted for RtM effects by adding the expected RtM effect for each initial abundance class to the slope estimates and to the estimates of the random effects. As expected, the RtM effect was strongest for the highest initial abundance intervals (Supplementary Table 5). We excluded one dataset (Datasource_ID 1396) because the taxonomy was not consistent over time. One model (abundance interval 2 with classification based on year 1) failed to converge, possibly owing to an excess of zeros, so we ran a simpler model excluding one study (Datasource_ID 1560) and removed the random effect on location. In a small test, this same simplified model provided similar results to those of other models, with a difference in mean slope of 0.00096 (4% of the 95% CI) and 0.00032 (10% of the 95% CI) for the abundance intervals 4 and 5.
Sensitivity analyses
We performed sensitivity analyses on the biodiversity metrics to test for the effect of time-series length and the number of sites per dataset, as well as the influence of Europe and North America, from which most data originated.
To see the effect of time-series length on the mean slope estimates, we (1) selected only the sites with at least 20 years between the start and the finish of the time series; and (2) used only the last 10 years of each time series (discarding all sites for which fewer than 10 years between start and end remained). We ran our standard model (above) for the variables abundance, richness and Simpson diversity (see Extended Data Fig. 2).
For the effect of the number of sites per dataset, we retained only the 50, 20 and 10 best sampled (highest number of years data) sites per datasets. All other sites were discarded. We repeated our standard model (above) for the variables abundance, richness and Simpson diversity (see Extended Data Fig. 3).
To test the effect of the well-sampled continents Europe and North America, we ran our standard model for the variables abundance, richness and Simpson diversity for the data from Europe and North America, and for the data excluding these continents. The result of this analysis is presented in Extended Data Fig. 8.
Conceptual model
The aim of this simple model was to show possible changes in biodiversity metrics and population abundances under three scenarios: (1) all species decline proportionally to each other; (2) abundant species decline more than rare species; and (3) rare species decline more than abundant species. To do this, we created an idealized SAD of 43 species as starting values for each scenario. We chose abundances such that there would be realistic numbers of species in each of five starting SAD intervals (12 singletons, 10 doubletons and 21 species with higher abundances, with a highest abundance of 32).
In scenario 2, the annual decline of each species was proportional to its starting abundance \(\left({N}_{t}={N}_{t=0}^{1\,-\,0.018t}\right.\), where N0 = the number of individuals at time 0, and t is the time of measurement). Here, the most abundant species declined by around 6% per year, and the rarest species did not decline. In scenario 3, the decline of the rarest species was around 6% per year and that of the most abundant species 0.76% per year: \({N}_{t}={N}_{t=0}\times {({10}^{-0.027}+0.035\times {\log }_{10}({N}_{t=0}))}^{t}\).
We extrapolated these species abundance changes for 20 years, added a random error of (μ = 0, σ = 0.5) to the yearly values, and rounded them to integers. The error was not intended to be a realistic simulation of population dynamics, but rather to add some variation for statistical model fitting. For visualization, we declared all species with an abundance below 0.6 to be extinct (abundance = 0).
We analysed these scenarios using the same approach as for the real data. All code is available at https://doi.org/10.5281/zenodo.8369189 (ref. 31).
Reporting summary
Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.
Data availability
All analytical code and data (Supplementary Data 1 and 2) used for the analysis are available at Zenodo: https://doi.org/10.5281/zenodo.10115304 (ref. 31), and the metadata for these files is available in Supplementary Methods 1. The raw data are available at https://doi.org/10.5063/F1ZC817H (ref. 22) in so far as they are not openly available elsewhere.
References
Hallmann, C. A. et al. More than 75 percent decline over 27 years in total flying insect biomass in protected areas. PLoS One 12, e0185809 (2017).
Van Klink, R. et al. Meta-analysis reveals declines in terrestrial but increases in freshwater insect abundances. Science 368, 417–420 (2020).
Forister, M. L. et al. Increasing neonicotinoid use and the declining butterfly fauna of lowland California. Biol. Lett. 12, 20160475 (2016).
Dirzo, R. et al. Defaunation in the Anthropocene. Science 345, 401–406 (2014).
Seibold, S. et al. Arthropod decline in grasslands and forests is associated with landscape-level drivers. Nature 574, 671–674 (2019).
Pilotto, F. et al. Meta-analysis of multidecadal biodiversity trends in Europe. Nat. Commun. 11, 3486 (2020).
Crossley, M. S. et al. No net insect abundance and diversity declines across US Long Term Ecological Research sites. Nat. Ecol. Evol. 4, 1368–1376 (2020).
Xu, W.-B. et al. Regional occupancy increases for widespread species but decreases for narrowly distributed species in metacommunity time series. Nat. Commun. 14, 1463 (2023).
Newbold, T. et al. Widespread winners and narrow-ranged losers: land use homogenizes biodiversity in local assemblages worldwide. PLoS Biol. 16, e2006841 (2018).
Gaston, K. J. Common ecology. BioScience 61, 354–362 (2011).
Wagner, D. L. Insect declines in the Anthropocene. Annu. Rev. Entomol. 65, 457–480 (2020).
Saunders, M. E., Janes, J. K. & O’Hanlon, J. C. Moving on from the insect apocalypse narrative: engaging with evidence-based insect conservation. BioScience 70, 80–89 (2020).
Schowalter, T. D., Noriega, J. A. & Tscharntke, T. Insect effects on ecosystem services—introduction. Basic Appl. Ecol. 26, 1–7 (2018).
Kucherov, N. B., Minor, E. S., Johnson, P. P., Taron, D. & Matteson, K. C. Butterfly declines in protected areas of Illinois: assessing the influence of two decades of climate and landscape change. PLoS One 16, e0257889 (2021).
Blowes, S. A. et al. Local biodiversity change reflects interactions among changing abundance, evenness, and richness. Ecology 103, e3820 (2022).
Hallmann, C. A., Ssymank, A., Sorg, M., de Kroon, H. & Jongejans, E. Insect biomass decline scaled to species diversity: general patterns derived from a hoverfly community. Proc. Natl Acad. Sci. USA 118, e2002554117 (2021).
Preston, F. W. The commonness, and rarity, of species. Ecology 29, 254–283 (1948).
Van Klink, R. et al. InsectChange: a global database of temporal changes in insect and arachnid assemblages. Ecology 102, e03354 (2021).
Schuch, S., van Klink, R. & Wesche, K. Is less simply less? A comparison of abundance and biomass losses in auchenorrhynchan grassland communities and their different impacts on trait composition and taxonomical diversity. Ecol. Indic. 146, 109743 (2023).
Haase, P. et al. The recovery of European freshwater biodiversity has come to a halt. Nature 620, 582–588 (2023).
Rumschlag, S. L. et al. Density declines, richness increases, and composition shifts in stream macroinvertebrates. Sci. Adv. 9, eadf4896 (2023).
Van Klink, R. et al. A global database of long-term changes in insect assemblages. Knowledge Network for Biocomplexity https://doi.org/10.5063/F1ZC817H (2020).
Muff, S., Nilsen, E. B., O’Hara, R. B. & Nater, C. R. Rewriting results sections in the language of evidence. Trends Ecol. Evol. 37, 203–210 (2022).
Dornelas, M. et al. Assemblage time series reveal biodiversity change but not systematic loss. Science 344, 296–299 (2014).
Gotelli, N. J. & Colwell, R. K. Quantifying biodiversity: procedures and pitfalls in the measurement and comparison of species richness. Ecol. Lett. 4, 379–391 (2001).
Roswell, M., Dushoff, J. & Winfree, R. A conceptual guide to measuring species diversity. Oikos 130, 321–338 (2021).
Chao, A. et al. Rarefaction and extrapolation with Hill numbers: a framework for sampling and estimation in species diversity studies. Ecol. Monogr. 84, 45–67 (2014).
Jost, L. Entropy and diversity. Oikos 113, 363–375 (2006).
Smith, B. & Wilson, J. B. A consumer’s guide to evenness indices. Oikos 76, 70 (1996).
Galton, F. Regression towards mediocrity in hereditary stature. J. Anthropol. Inst. G. B. Irel. 15, 246–263 (1886).
Van Klink, R. & Bowler, D. E. Code for: ‘Disproportionate declines of formerly abundant species underlie insect loss’. Zenodo https://doi.org/10.5281/zenodo.10115304 (2023).
Brower, L. P. et al. Decline of monarch butterflies overwintering in Mexico: is the migratory phenomenon at risk? Insect Conserv. Divers. 5, 95–100 (2012).
Van Dyck, H., Van Strien, A. J., Maes, D. & Van Swaay, C. A. M. Declines in common, widespread butterflies in a landscape under intense human use. Conserv. Biol. 23, 957–965 (2009).
Karban, R. & Huntzinger, M. Decline of meadow spittlebugs, a previously abundant insect, along the California coast. Ecology 99, 2614–2616 (2018).
Lockwood, J. A. & Debrey, L. D. A solution for the sudden and unexplained extinction of the Rocky Mountain grasshopper (Orthoptera: Acrididae). Environ. Entomol. 19, 1194–1205 (1990).
Salcido, D. M., Forister, M. L., Garcia Lopez, H. & Dyer, L. A. Loss of dominant caterpillar genera in a protected tropical forest. Sci. Rep. 10, 422 (2020).
Habel, J. C., Trusch, R., Schmitt, T., Ochse, M. & Ulrich, W. Long-term large-scale decline in relative abundances of butterfly and burnet moth species across south-western Germany. Sci. Rep. 9, 14921 (2019).
Welti, E. A. R., Roeder, K. A., de Beurs, K. M., Joern, A. & Kaspari, M. Nutrient dilution and climate cycles underlie declines in a dominant insect herbivore. Proc. Natl Acad. Sci. USA 117, 7271–7275 (2020).
Roubik, D. W. et al. Long-term (1979–2019) dynamics of protected orchid bees in Panama. Conserv. Sci. Pract. 3, e543 (2021).
Genung, M. A., Fox, J. & Winfree, R. Species loss drives ecosystem function in experiments, but in nature the importance of species loss depends on dominance. Glob. Ecol. Biogeogr. 29, 1531–1541 (2020).
Smith, M. D. & Knapp, A. K. Dominant species maintain ecosystem function with non-random species loss. Ecol. Lett. 6, 509–517 (2003).
Kleijn, D. et al. Delivery of crop pollination services is an insufficient argument for wild pollinator conservation. Nat. Commun. 6, 7414 (2015).
Winfree, R., Fox, J. W., Williams, N. M., Reilly, J. R. & Cariveau, D. P. Abundance of common species, not species richness, drives delivery of a real-world ecosystem service. Ecol. Lett. 18, 626–635 (2015).
Kamiya, T., O’Dwyer, K., Nakagawa, S. & Poulin, R. What determines species richness of parasitic organisms? A meta-analysis across animal, plant and fungal hosts. Biol. Rev. 89, 123–134 (2014).
Fisher, A. R. A., Corbet, A. S. & Williams, C. B. The number of animals in a random sample of an animal population. J. Anim. Ecol. 12, 42–58 (1943).
Prendergast, J. et al. The global population dynamics database. Knowledge Network for Biocomplexity https://doi.org/10.5063/F1BZ63Z8 (2010).
Dornelas, M. et al. BioTIME: a database of biodiversity time series for the Anthropocene. Glob. Ecol. Biogeogr. 27, 760–786 (2018).
Pollard, E., Hall, M. L. & Bibby, T. J. Monitoring the Abundance of Butterflies 1976–1985 (Joint Nature Conservation Committee, 1986).
Thomsen, P. F. et al. Data from: Resource specialists lead local insect community turnover associated with temperature—analysis of an 18-year full-seasonal record of moths and beetles. Dryad Digital Repository https://doi.org/10.5061/dryad.s4945 (2016).
Rennie, S. et al. UK Environmental Change Network (ECN) moth data: 1992–2015. NERC Environmental Information Data Centre https://doi.org/10.5285/a2a49f47-49b3-46da-a434-bb22e524c5d2 (2018).
Ellison, A. M. Ant assemblages in hemlock removal experiment at Harvard Forest since 2003. HF118. Harvard Forest Data Archive http://pasta.lternet.edu/package/doi/eml/knb-lter-hfr/118/30 (2017).
Rennie, S. et al. Environmental Change Network (ECN) butterfly data: 1993–2015. NERC Environmental Information Data Centre https://doi.org/10.5285/5aeda581-b4f2-4e51-b1a6-890b6b3403a3 (2018).
Rennie, S. et al. UK Environmental Change Network (ECN) spittle bug data: 1993–2015. NERC Environmental Information Data Centre https://doi.org/10.5285/aff433be-0869-4393-b765-9e6faad2a12b (2018).
Rennie, S. et al. UK Environmental Change Network (ECN) carabid beetle data: 1992–2015. NERC Environmental Information Data Centre https://doi.org/10.5285/8385f864-dd41-410f-b248-028f923cb281 (2018).
Wolda, H. Trends in abundance of tropical forest insects. Oecologia 89, 47–52 (1992).
Wolda, H., Marek, J., Spitzer, K. & Novak, I. Diversity and variability of Lepidoptera populations in urban Brno, Czech Republic. Eur. J. Entomol. 91, 213–226 (1994).
Lightfoot, D. Long-term core site grasshopper dynamics for the Sevilleta National Wildlife Refuge, New Mexico (1992-2013). Environmental Data Initiative https://doi.org/10.6073/pasta/c1d40e9d0ec610bb74d02741e9d22576 (2010).
Meijer, J. & Barendregt, A. Forty years of undisturbed change in the ground dwelling fauna in the Lauwersmeer, a reclaimed tidal estuary of the Dutch Waddensea. Entomol. Ber. 78, 122–151 (2018).
Honek, A., Martinkova, Z., Kindlmann, P., Ameixa, O. M. C. C. & Dixon, A. F. G. Long-term trends in the composition of aphidophagous coccinellid communities in Central Europe. Insect Conserv. Divers. 7, 55–63 (2014).
Belovsky, G. Grasshopper density. National Bison Range LTREB Database https://belovskylab.nd.edu/national-bison-range-ltreb-database/survey-data/grasshopper-data/ (2018).
Valtonen, A. et al. Long-term species loss and homogenization of moth communities in Central Europe. J. Anim. Ecol. 86, 730–738 (2017).
Lightfoot, D. Small Mammal Exclosure Study (SMES) ant data from Chihuahuan desert grassland and shrubland at the Sevilleta National Wildlife Refuge, New Mexico (1995–2005). Environmental Data Initiative https://doi.org/10.6073/pasta/c4963aa3363d18ac99bd40307db2249d (2010).
Grimm, N. & Childers, D. Long-term monitoring of ground-dwelling arthropods in central Arizona–Phoenix, ongoing since 1998. Environmental Data Initiative https://doi.org/10.6073/pasta/74d30fdbb17e0f76b54548ce74bf27e4 (2018).
Pennings, S. C. Long-term mid-marsh grasshopper abundance and species diversity at eight GCE-LTER sampling sites. Environmental Data Initiative https://doi.org/10.6073/pasta/ec9d524b8f13c000e8e8a225d5a23c7b (2016).
Gandhi, K. J. K., Epstein, M. E., Koehle, J. J. & Purrington, F. F. A quarter of a century succession of epigaeic beetle assemblages in remnant habitats in an urbanized matrix (Coleoptera, Carabidae). ZooKeys 147, 667–689 (2011).
Pizzolotto, R., Gobbi, M. & Brandmayr, P. Changes in ground beetle assemblages above and below the treeline of the Dolomites after almost 30 years (1980/2009). Ecol. Evol. 4, 1284–1294 (2014).
Roubik, D. W. Ups and downs in pollinator populations: when is there a decline? Ecol. Soc. 5, art2 (2001).
Grøtan, V., Lande, R., Chacon, I. A. & Devries, P. J. Seasonal cycles of diversity and similarity in a Central American rainforest butterfly community. Ecography 37, 509–516 (2014).
Grøtan, V., Lande, R., Engen, S., Sæther, B. E. & Devries, P. J. Seasonal cycles of species diversity and similarity in a tropical butterfly community. J. Anim. Ecol. 81, 714–723 (2012).
Meserve, P. L., Vásquez, H., Kelt, D. A., Gutiérrez, J. R. & Milstead, W. B. Patterns in arthropod abundance and biomass in the semiarid thorn scrub of Bosque Fray Jorge National Park, north-central Chile: a preliminary assessment. J. Arid Environ. 126, 68–75 (2016).
White, E. The changing abundance of moths in a tussock grassland, 1962–1989, and 50- to 70-year trends. N. Z. J. Ecol. 15, 5–22 (1991).
Quintero, I. & Roslin, T. Rapid recovery of dung beetle communities following habitat fragmentation in central Amazonia. Ecology 86, 3303–3311 (2005).
Rybalov, L. & Kamayev, I. Comparative analysis and long-term dynamics of soil macrofauna in forest-tundra ecotone of the Khibiny mountains. Russ. Entomol. J. 21, 179–183 (2012).
Kočíková, L., Čanády, A. & Panigaj, L. Change in a butterfly community on a gradually overgrowing site. Russ. J. Ecol. 45, 391–398 (2014).
Shafigullina, S. M. The role of floods in the long-term dynamics of geobiont and chortobiont communities on islands of the Kuibyshev Reservoir. Russ. J. Ecol. 40, 218–226 (2009).
Valtonen, A. et al. Tropical phenology: bi-annual rhythms and interannual variation in an Afrotropical butterfly assemblage. Ecosphere 4, 1–28 (2013).
Daghighi, E., Koehler, H., Kesel, R. & Filser, J. Long-term succession of Collembola communities in relation to climate change and vegetation. Pedobiologia 64, 25–38 (2017).
Gallé, L. Climate change impoverishes and homogenizes ants’ community structure: a long term study. Community Ecol. 18, 128–136 (2017).
Hodecek, J., Kuras, T., Sipos, J. & Dolny, A. Post-industrial areas as successional habitats: long-term changes of functional diversity in beetle communities. Basic Appl. Ecol. 16, 629–640 (2015).
Ananin, A. A. & Ananina, T. L. Long-term dynamics of birds and ground beetles population density in catena of Barguzinskiy Ridge (Northern Pribaikalye) [in Russian]. Izv. Samar. Nauchnogo Cent. Ross. Akad. Nauk Proc. Samara Res. Branch Russ. Acad. Sci. 1, 1041–1044 (2011).
Tsurikov, M. N. Long-term dynamics of the species composition of herpetobiont and hortobiont beetles (Coleoptera) in the Galichya Gora Nature Reserve. Entomol. Rev. 96, 191–198 (2016).
Fedyunin, V. A. On population dynamics of Ichneumon flies in the Visim Reserve. Russ. J. Ecol. 39, 225–228 (2008).
Aarhus University. Greenland Ecosystem Monitoring Database http://data.g-e-m.dk/ (2018).
Martikainen, P. & Kaila, L. Sampling saproxylic beetles: lessons from a 10-year monitoring study. Biol. Conserv. 120, 171–181 (2004).
Nemkov, V. A. & Sapiga, E. V. Impact of fires on the fauna of terrestrial arthropods in protected steppe ecosystems. Russ. J. Ecol. 41, 173–179 (2010).
Korobov, E. D. The influence of windfall disturbances on the structure and dynamics of ground beetle populations (Coleoptera, Carabidae) in the spruce forests of the Central Forest Nature Reserve. Russ. J. Ecol. 46, 595–599 (2015).
Steinwandter, M., Schlick-Steiner, B. C., Seeber, G. U. H., Steiner, F. M. & Seeber, J. Effects of Alpine land-use changes: soil macrofauna community revisited. Ecol. Evol. 7, 5389–5399 (2017).
Kwon, T.-S., Kim, Y. S., Lee, S. W. & Park, Y.-S. Changes of soil arthropod communities in temperate forests over 10 years (1998–2007). J. Asia-Pac. Entomol. 19, 181–189 (2016).
Holmes, R. Long-term trends in abundance of Lepidoptera larvae at Hubbard Brook Experimental Forest and three additional northern hardwood forest sites, 1986–1997. Hubbard Brook Data Archive http://data.hubbardbrook.org/data/dataset.php?id=82 (2018).
Bisevac, L. & Majer, J. D. Comparative study of ant communities of rehabilitated mineral sand mines and heathland, Western Australia. Restor. Ecol. 7, 117–126 (1999).
Blandenier, G., Bruggisser, O. T. & Bersier, L.-F. Do spiders respond to global change? A study on the phenology of ballooning spiders in Switzerland. Ecoscience 21, 79–95 (2014).
Ernest, S. K. M. Portal ant data. GitHub https://github.com/weecology/PortalData/tree/master/Ants (2018).
Grechanichenko, T. Linear and cyclic long-term trends in the dynamics of ground beetles activity (Carabidae, Coleoptera). Aktual. Probl. Gumanit. Estestv. Nauk 4–1, 44–49 (2014).
Kuznetsova, N. Organization of Communities of Soil-Dwelling Collembola [in Russian] (Prometei, 2005).
Gryuntal, S. Y. Organization of Communities of Ground Beetles (Coleoptera; Carabidae) in Forest Biocoenoses of East-European (Russian) Plain. [in Russian] (Gallea-Print, 2008).
Guseva, O. G. Rove beetles (Coleoptera, Staphylindae) in agricultural landscape of Leningrad region. Plant Prot. News 94, 39–42 (2017).
Mutin, V. A. in A.I. Kurentsov’s Annual Memorial Meetings: Vladivostok (ed. Storozhenko, S. Yu.) 325–337 (Danaulka, 2015).
Sasova, L. E. Population of Day Lepidoptera (Lepidoptera, Diurna) of State Nature Reserve «Ussuriysky» Named After V.L. Komarov (Far-Eastern State University, 2008).
Shlyakhtenok, A. S. Studying the dynamics of the complex of the digging wasps (Hymenoptera, Sphecidae) in the abandoned zone of the Chernobyl power station. Ecologiya 5, 391–394 (2007).
Chen, I.-C. et al. Asymmetric boundary shifts of tropical montane Lepidoptera over four decades of climate warming. Glob. Ecol. Biogeogr. 20, 34–45 (2011).
Shlyakhtenok, A. S. Hymenoptera Aculeates of raised bogs in Belarus. Zool. Zhurnal 86, 295–306 (2007).
Ploquin, E. F., Herrera, J. M. & Obeso, J. R. Bumblebee community homogenization after uphill shifts in montane areas of northern Spain. Oecologia 173, 1649–1660 (2013).
Shlyakhtenok, A. S. Aculeate Hymenoptera of the family Chrysididae of Byelorussia. Vestn. Zool. 41, 433–438 (2007).
Nitochko, M. I. Structure and dynamic of population of ground beetles and tenebrionid beetles (Coleoptera: Carabidae, Tenebrionidae) of sand steppe of Black Sea Biosphere Reserve NAS of Ukraine. Optim. Prot. Ecosyst. 7, 62–73 (2012).
Szabó, S., Árnyas, E., Tóthmérész, B. & Varga, Z. Long-term light trap study on the macro-moth (Lepidoptera: Macroheterocera) fauna of the Aggtelek National Park. Acta Zool. Acad. Sci. Hung. 53, 257–269 (2007).
Schuch, S., Bock, J., Krause, B., Wesche, K. & Schaefer, M. Long-term population trends in three grassland insect groups: a comparative analysis of 1951 and 2009. J. Appl. Entomol. 136, 321–331 (2012).
Schuch, S., Wesche, K. & Schaefer, M. Long-term decline in the abundance of leafhoppers and planthoppers (Auchenorrhyncha) in Central European protected dry grasslands. Biol. Conserv. 149, 75–83 (2012).
Cormay, O. et al. Analysis of monitoring data where butterflies fly year round. Ecol. Appl. 30, e02196 (2020).
Schowalter, T. Canopy invertebrate responses to Hurricane Hugo. Environmental Data Initiative https://doi.org/10.6073/pasta/82d334e74866175c791e557d8c303a62 (2018).
Hu, G. et al. Mass seasonal bioflows of high-flying insect migrants. Science 354, 1584–1587 (2016).
Aebischer, N. J. Assessing pesticide effects on non-target invertebrates using long-term monitoring and time-series modelling. Funct. Ecol. 4, 369–373 (1990).
Benton, T. G., Bryant, D. M., Cole, L. & Crick, H. Q. P. Linking agricultural practice to insect and bird populations: a historical study over three decades. J. Appl. Ecol. 39, 673–687 (2002).
Brown, P. M. J. & Roy, H. E. Native ladybird decline caused by the invasive harlequin ladybird Harmonia axyridis: evidence from a long-term field study. Insect Conserv. Divers. 11, 230–239 (2018).
Irish National Biodiversity Data Centre. Irish Butterfly Monitoring Scheme. Occurrence dataset. GBIF https://doi.org/10.15468/4wwpyc (2018).
Zhang, X. et al. Changes in assemblages and diversity patterns of Carabidae (Coleoptera) from 1997 to 2014 in a desalinized, intensively cultivated agricultural landscape in northern China. Coleopt. Bull. 72, 597–611 (2018).
Blanchet, F. G. et al. Related herbivore species show similar temporal dynamics. J. Anim. Ecol. 87, 801–812 (2018).
Blanchet, F. et al. Data from: Related herbivore species show similar temporal dynamics. Dryad Digital Repository https://doi.org/10.5061/dryad.sh02b (2018).
Homburg, K. et al. Where have all the beetles gone? Long‐term study reveals carabid species decline in a nature reserve in Northern Germany. Insect Conserv. Divers. 12, 268–277 (2019).
Cuesta, E. & Lobo, J. M. A comparison of dung beetle assemblages (Coleoptera, Scarabaeoidea) collected 34 years apart in an Iberian mountain locality. J. Insect Conserv. 23, 101–110 (2019).
Gran, O. & Götmark, F. Long-term experimental management in Swedish mixed oak-rich forests has a positive effect on saproxylic beetles after 10 years. Biodivers. Conserv. 28, 1451–1472 (2019).
Wepprich, T., Adrion, J. R., Ries, L., Wiedmann, J. & Haddad, N. M. Butterfly abundance declines over 20 years of systematic monitoring in Ohio, USA. PLoS One 14, e0216270 (2019).
Petersen, H., Jucevica, E. & Gjelstrup, P. Long-term changes in collembolan communities in grazed and non-grazed abandoned arable fields in Denmark. Pedobiologia 48, 559–573 (2004).
Hallmann, C. A. et al. Declining abundance of beetles, moths and caddisflies in the Netherlands. Insect Conserv. Divers. 13, 127–139 (2020).
Antão, L. H., Pöyry, J., Leinonen, R. & Roslin, T. Contrasting latitudinal patterns in diversity and stability in a high-latitude species-rich moth community. Glob. Ecol. Biogeogr. 29, 896–907 (2020).
Antão, L. H., Pöyry, J., Leinonen, R. & Roslin, T. Contrasting latitudinal patterns in diversity and stability in a high-latitude species-rich moth community. Dryad Digital Repository https://doi.org/10.5061/dryad.905qfttgj (2020).
Costa, F. R. C. et al. Effects of climate change on central amazonian forests: a two-decades synthesis of monitoring tropical biodiversity. Oecologia Aust. 24, 317–335 (2020).
Filho, W. M., Flechtmann, C. A. H., Godoy, Wesley, A. C. & Bjornstad, O. N. The impact of the introduced Digitonthophagus gazella on a native dung beetle community in Brazil during 26 years. Biol. Invasions 20, 963–979 (2018).
Pereira, F. W., Carneiro, L. & Gonçalves, R. B. More losses than gains in ground-nesting bees over 60 years of urbanization. Urban Ecosyst. 24, 233–242 (2021).
da Rocha-Filho, L. C., Montagnana, P. C., Boscolo, D. & Garófalo, C. A. Species turnover and low stability in a community of euglossine bees (Hymenoptera: Apidae) sampled within 28 years in an urban forest fragment. Apidologie 51, 921–934 (2020).
Harris, J. E., Rodenhouse, N. L. & Holmes, R. T. Decline in beetle abundance and diversity in an intact temperate forest linked to climate warming. Biol. Conserv. 240, 108219 (2019).
Luk, C.-L., Basset, Y., Kongnoo, P., Hau, B. C. H. & Bonebrake, T. C. Data from: Inter-annual monitoring improves diversity estimation of tropical butterfly assemblages. Dryad Digital Repository https://doi.org/10.5061/dryad.j35fq7r (2019).
Luk, C.-L., Basset, Y., Kongnoo, P., Hau, B. C. H. & Bonebrake, T. C. Inter-annual monitoring improves diversity estimation of tropical butterfly assemblages. Biotropica 51, 519–528 (2019).
Guo, J. et al. Long-term shifts in abundance of (migratory) crop-feeding and beneficial insect species in northeastern Asia. J. Pest Sci. 93, 583–594 (2020).
Gibb, H., Grossman, B. F., Dickman, C. R., Decker, O. & Wardle, G. M. Long-term responses of desert ant assemblages to climate. J. Anim. Ecol. 88, 1549–1563 (2019).
Gibb, H., Grossman, B. F., Dickman, C. R., Decker, O. & Wardle, G. M. Data from: Long-term responses of desert ant assemblages to climate. Dryad Digital Repository https://doi.org/10.5061/dryad.vc80r13 (2019).
Choi, S.-W. Long-term (2005–2017) Macromoth Community Monitoring at Mt. Jirisan National Park, South Korea https://db.cger.nies.go.jp/JaLTER/metacat/metacat?action=read&qformat=jalter-en&sessionid=&docid=ERDP-2019-02.1 (ERDP, 2019).
Choi, S.-W., An, J.-S., Kim, N.-H., Lee, S. & Ahn, N. Long-term (2005–2017) macromoth community monitoring at Mt. Jirisan National Park, South Korea. Ecol. Res. 34, 443–443 (2019).
Koivula, M. J., Venn, S., Hakola, P. & Niemelä, J. Responses of boreal ground beetles (Coleoptera, Carabidae) to different logging regimes ten years post harvest. For. Ecol. Manag. 436, 27–38 (2019).
Seymour, M. et al. Ecological community dynamics: 20 years of moth sampling reveals the importance of generalists for community stability. Basic Appl. Ecol. 49, 34–44 (2020).
Zhou, Z. et al. Diversity and population dynamics of flies in four habitat in Wuhan area. J. Xinjang Norm. Univ. Nat. Sci. Ed. 33, 70–75 (2020).
Guo, H. et al. The structure, spaciotemporal dynamics, and diversity of mosquito communities in Wuhan. Chin. J. Appl. Entomol. 57, 955–962 (2020).
Cardoso, M. C. & Gonçalves, R. B. Reduction by half: the impact on bees of 34 years of urbanization. Urban Ecosyst. 21, 943–949 (2018).
Martins, A. C., Gonçalves, R. B. & Melo, G. A. R. Changes in wild bee fauna of a grassland in Brazil reveal negative effects associated with growing urbanization during the last 40 years. Zool. Curitiba 30, 157–176 (2013).
Hsieh, T. C., Ma, K. H. & Chao, A. iNEXT: an R package for rarefaction and extrapolation of species diversity (Hill numbers). Methods Ecol. Evol. 7, 1451–1456 (2016).
Oksanen, J. et al. vegan: Community Ecology Package. R version 2.6-4 http://CRAN.R-project.org/package=vegan (2020).
McGill, B. J. in Biological Diversity: Frontiers in Measurement and Assessment (eds. Magurran, A. E. & McGill, B. J.) 105–122 (Oxford Univ. Press, 2011).
Daskalova, G. N., Myers-Smith, I. H. & Godlee, J. L. Rare and common vertebrates span a wide spectrum of population trends. Nat. Commun. 11, 4394 (2020).
Barnett, A. G., van der Pols, J. C. & Dobson, A. J. Regression to the mean: what it is and how to deal with it. Int. J. Epidemiol. 34, 215–220 (2005).
Driessen, M. M. & Kirkpatrick, J. B. Higher taxa can be effective surrogates for species-level data in detecting changes in invertebrate assemblage structure due to disturbance: a case study using a broad range of orders. Austral. Entomol. 58, 361–369 (2019).
Rue, H., Martino, S. & Chopin, N. Approximate Bayesian inference for latent Gaussian models by using integrated nested Laplace approximations. J. R. Stat. Soc.B 71, 319–392 (2009).
R Core Team. R: A Language And Environment For Statistical Computing v.4.2.2 (R Foundation for Statistical Computing, 2022).
Wessel, P. & Smith, W. H. F. A global, self-consistent, hierarchical, high-resolution shoreline database. J. Geophys. Res. Solid Earth 101, 8741–8743 (1996).
Sandvik, B. thematicmapping.org. https://www.thematicmapping.org/downloads/world_borders.php (2018).
Acknowledgements
We thank R. Didham for helpful comments; N. Naderi and A. Al-Hemiary for help with the digitization; M. Desquilbet and L. Gaume for thoroughly checking our dataset for instances of errors; A. Sagouis and T. Engel for help with the R code; A. Alzate for help with the Spanish translations; H. Schenk for help with the conceptual model; and S. Blowes for advice on the biodiversity metrics used. Funding: German Research Foundation grant FZT 118 (R.v.K., D.E.B. and J.M.C.) and China Scholarship Council (CSC) 202104910063 (M.S.). The following funding sources were used to collect the original data we used in our analysis: National Science Foundation NSF06‐20443, 8811906, 9411976, 0080529, 0217774, DEB-0423704, DEB-1633026, DEB-1637685, DEB-1256696, DEB-0832652, DEB-0936498, DEB-1832016, DEB-0620652 DEB-1234162, OCE-9982133, OCE-0620959, OCE-1237140 and OCE-1832178, and the German Research Foundation DFG Priority Program 1374 ‘Infrastructure Biodiversity Exploratories’.
Author information
Authors and Affiliations
Contributions
Conceptualization: R.v.K., D.E.B. and J.M.C. Methodology: R.v.K., J.M.C., D.E.B., M.S. and S.R.S. Investigation: R.v.K., J.M.C., K.B.G., M.S. and S.R.S. Visualization: R.v.K. and D.E.B. Writing (original draft): R.v.K. Writing (review and editing): J.M.C., D.E.B., K.B.G., M.S. and S.R.S.
Corresponding author
Ethics declarations
Competing interests
The authors declare that they have no competing interests.
Peer review
Peer review information
Nature thanks Raphael Didham and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. Peer reviewer reports are available.
Additional information
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Extended data figures and tables
Extended Data Fig. 1 Bivariate map of changes in richness and abundance at the study level.
The trend estimates of the individual studies were derived from the random effects of the hierarchical Bayesian model. The colours are coded according to the strength of evidence, in which the middle colour in both axes indicates that the 80% CI includes zero, hence, all other colours indicate weak to strong evidence of a temporal trend on either abundance, richness or both. Map derived from refs. 152 and 153.
Extended Data Fig. 2 Influence of time-series length on the estimated temporal slopes of abundance, richness and Simpson diversity.
Long time series were selected by restricting the data to sites with at least 20 years from the first to the last year. Short time series were created by only retaining the last 10 years of each site. To aid the comparison among rows, the mean estimates of each realm are provided as dotted lines. This shows some shifts in mean estimates, but no differences in the qualitative interpretation of the results.
Extended Data Fig. 3 Influence of studies with a large number of sites on the estimated temporal slopes of abundance, richness and Simpson diversity.
The studies underlying the analysis were subsetted to only the 50, 20 and 10 best sampled sites, and the models were rerun. To aid comparison among rows, the mean estimates of each realm are provided as dotted lines. This shows a progressive shift to more positive slopes for all metrics.
Extended Data Fig. 4 Explanation of changes in the number of species per SAD section.
a, The SAD (the number of species per bin), in years 1 and 10 of a hypothetical time series. Scaled in relation to the highest observed log10-transformed value in the whole time series, we have binned the log10-transformed species abundance values in two ways. b,c, We assigned five equally spaced sections (b), and four quartiles (c), where the quartiles had approximately equal species numbers (variation because of rounding of bins). After the bins were assigned, we calculated the number of species falling in each bin in each year.
Extended Data Fig. 5 Changes in the number of species in each of four quartiles on the basis of the distribution of values of all species in each whole time series.
In comparison to Fig. 4, there are more species and more individuals in the higher quartiles, given the naturally low number of very abundant species (see Extended Data Fig. 4a,c). Here, there isan equal number of observations in each quartile, whereas in Fig. 4, the spacing (in log space) between bins is equal.
Extended Data Fig. 6 Effects of censoring the first year or first three years on the slope estimates for correcting RtM effects.
We calculated the slope estimate (±80, 90 and 95% CI) on the highest quality datasets (260 plots in 26 datasets with at least 15 years of data) for each of the initial abundance groups with no censoring (all data included), excluding the first year, and excluding the first three years. The shrinkage towards zero of the estimates with increasing censoring is assumed to be caused by RtM effects, but we cannot exclude that it’s partially due to true greater declines of populations during the early years of monitoring. We have taken the difference between the mean slope estimate without censoring and the estimate with one-year censoring from the start as the correction factor for RtM effects in the main text (see Supplementary Table 5). Three-year censoring would provide a larger correction factor (Supplementary Table 5).
Extended Data Fig. 7 Effects of different ways of classifying locally rare and abundant species on the estimated mean population trends.
Classification of species starting interval was done based on (i) the abundances observed in year 1, (ii) the abundance of each species averaged over years 1 and 2 (iii) the abundance of each species averaged over years 1-5, and (iv) the abundance of each species averaged across the whole time series. Species that were absent at the start of the time series were not analysed, because their abundance trends will be, on average, positive by definition. The weakest results when based on the whole time series was expected because this is the most conservative approach to assessing rarity. An initially abundant species with a strong decline might not be abundant across the whole time series. Hence, this analysis shows that even when looking at abundance across the whole time series, the most abundant species decline most strongly.
Extended Data Fig. 8 Relation between abundance slopes and the mean population slopes at the dataset level.
Abundance slopes and mean population slopes at the dataset level were both converted to the percentage change per year per initial abundance interval. The initial abundance intervals can be understood as the abundance interval of a species in relation to the log-transformed abundance of the most abundant species in year 1. Dotted lines represent the 1:1 relation, orange lines and slope estimates (β) represent the best fit according to least squares regression models. We removed one data point with an extreme slope from panels 1 and 2 to aid visual interpretation.
Extended Data Fig. 9 Temporal trend slopes for biodiversity metrics excluding Europe and North America.
The probability densities are shown for the slope estimates of abundance, species richness and Simpson diversity (ENS) for the data excluding Europe and North America. Numbers indicate the number of studies and the number of sites underlying each estimate respectively.
Supplementary information
Supplementary Information
This file contains Supplementary Methods (Search terms used in all languages) and the metadata for Supplementary Data 1–3 (available at https://doi.org/10.5281/zenodo.10115304).
Supplementary Tables
This file contains Supplementary Tables 1–5.
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
About this article
Cite this article
van Klink, R., Bowler, D.E., Gongalsky, K.B. et al. Disproportionate declines of formerly abundant species underlie insect loss. Nature 628, 359–364 (2024). https://doi.org/10.1038/s41586-023-06861-4
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1038/s41586-023-06861-4
Comments
By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.
