Proceedings of the Royal Society B: Biological Sciences
You have accessResearch article

Cross-taxa generalities in the relationship between population abundance and ambient temperatures

,
Peter Haase

Peter Haase

Senckenberg Research Institute and Natural History Museum Frankfurt, Gelnhausen, Germany

Faculty of Biology, University of Duisburg-Essen, Essen, Germany

Google Scholar

Find this author on PubMed

,
Christian Hof

Christian Hof

Senckenberg Biodiversity and Climate Research Centre, Frankfurt am Main, Germany

Google Scholar

Find this author on PubMed

,
Ingrid Kröncke

Ingrid Kröncke

Department of Marine Research, Senckenberg am Meer, Wilhelmshaven, Germany

Google Scholar

Find this author on PubMed

,
Léon Baert

Léon Baert

Royal Belgian Institute of Natural Sciences, Brussels, Belgium

Google Scholar

Find this author on PubMed

,
Wouter Dekoninck

Wouter Dekoninck

Royal Belgian Institute of Natural Sciences, Brussels, Belgium

Google Scholar

Find this author on PubMed

,
Sami Domisch

Sami Domisch

Department of Ecology and Evolutionary Biology, Yale University, New Haven, CT, USA

Leibniz-Institute of Freshwater Ecology and Inland Fisheries, Berlin, Germany

Google Scholar

Find this author on PubMed

,
Frederik Hendrickx

Frederik Hendrickx

Royal Belgian Institute of Natural Sciences, Brussels, Belgium

Google Scholar

Find this author on PubMed

,
Thomas Hickler

Thomas Hickler

Senckenberg Biodiversity and Climate Research Centre, Frankfurt am Main, Germany

Institute of Physical Geography, Geosciences, Goethe University Frankfurt, Frankfurt am Main, Germany

Google Scholar

Find this author on PubMed

,
Hermann Neumann

Hermann Neumann

Department of Marine Research, Senckenberg am Meer, Wilhelmshaven, Germany

Google Scholar

Find this author on PubMed

,
Robert B. O'Hara

Robert B. O'Hara

Senckenberg Biodiversity and Climate Research Centre, Frankfurt am Main, Germany

Department of Mathematical Sciences, Norwegian University of Science and Technology, Trondheim, Norway

Google Scholar

Find this author on PubMed

,
Anne F. Sell

Anne F. Sell

Thünen Institute of Sea Fisheries, Hamburg, Germany

Google Scholar

Find this author on PubMed

,
Moritz Sonnewald

Moritz Sonnewald

Senckenberg Research Institute and Natural History Museum Frankfurt, Frankfurt, Germany

Google Scholar

Find this author on PubMed

,
Stefan Stoll

Stefan Stoll

Senckenberg Research Institute and Natural History Museum Frankfurt, Gelnhausen, Germany

Department of Environmental Planning and Technology, University of Applied Sciences Trier, Environmental Campus Birkenfeld, Neubrücke, Germany

Google Scholar

Find this author on PubMed

,
Michael Türkay

Michael Türkay

Senckenberg Research Institute and Natural History Museum Frankfurt, Frankfurt, Germany

Google Scholar

Find this author on PubMed

,
Roel van Klink

Roel van Klink

Division of Conservation Biology, University of Bern, Institute of Ecology and Evolution, Bern, Switzerland

Google Scholar

Find this author on PubMed

,
Oliver Schweiger

Oliver Schweiger

Department of Community Ecology, UFZ—Helmholtz Centre for Environmental Research, Halle, Germany

Google Scholar

Find this author on PubMed

,
Rikjan Vermeulen

Rikjan Vermeulen

Willem Beijerinck Biological Station, Loon, The Netherlands

Google Scholar

Find this author on PubMed

and
Katrin Böhning-Gaese

Katrin Böhning-Gaese

Senckenberg Biodiversity and Climate Research Centre, Frankfurt am Main, Germany

Institute of Ecology, Evolution and Diversity, Goethe University Frankfurt, Frankfurt am Main, Germany

Google Scholar

Find this author on PubMed

Published:https://doi.org/10.1098/rspb.2017.0870

    Abstract

    Identifying patterns in the effects of temperature on species' population abundances could help develop a general framework for predicting the consequences of climate change across different communities and realms. We used long-term population time series data from terrestrial, freshwater, and marine species communities within central Europe to compare the effects of temperature on abundance across a broad range of taxonomic groups. We asked whether there was an average relationship between temperatures in different seasons and annual abundances of species in a community, and whether species attributes (temperature range of distribution, range size, habitat breadth, dispersal ability, body size, and lifespan) explained interspecific variation in the relationship between temperature and abundance. We found that, on average, warmer winter temperatures were associated with greater abundances in terrestrial communities (ground beetles, spiders, and birds) but not always in aquatic communities (freshwater and marine invertebrates and fish). The abundances of species with large geographical ranges, larger body sizes, and longer lifespans tended to be less related to temperature. Our results suggest that climate change may have, in general, positive effects on species’ abundances within many terrestrial communities in central Europe while the effects are less predictable in aquatic communities.

    1. Background

    Climate change is now recognized as one of the key drivers of global biodiversity dynamics [13]. Signals of climate change on individuals, populations, and communities are visible in marine [4], freshwater [5], and terrestrial ecosystems [6,7]. Impacts on population abundances are particularly central to how an organism is affected by climate change because abundance is tightly linked to extinction risk and changes in abundance are linked with changes in ranges [8]. Climate change is composed of a broad range of changes—including trends in mean and variance of temperature and precipitation. Here, we focus on the effects of changing average temperatures, which is among the most consistent, directional, and widespread change [9] and has the potential to lead to persistent long-term population change. Species whose annual abundances are more tightly linked to ambient temperatures may be more affected by climate change. Despite the long interest in the role of weather in species' population fluctuations [10], there have been few attempts to explicitly compare the relationships between temperatures and population abundances across major taxonomic groups and environmental realms [11,12]. Nonetheless, such a comparison would help identify any generalities in how species abundances are affected by climate change.

    Ambient temperatures may affect species' abundances along direct pathways, dependent on species’ physiological tolerances, metabolism, and energy requirements. For many ectotherms living within the temperate zone, the direct effect of increasing temperature is likely to be an increase in fitness as current ambient temperatures are usually below their physiological optima [13]. Research on endotherms using physiological measurements of thermal tolerance (i.e. the thermal neutral zone) suggest that ambient temperatures generally lie within those tolerated by temperate species [14]. However, high temperatures, such as during heatwaves, could lead to periods of thermal stress and increased mortality [15]. There are also various indirect pathways through which temperatures can affect population abundances, for example, when temperature affects food availability, or the abundances of competitors and natural enemies [16]. Even if temperature change is in a direction that is favourable for an organism, its population abundance may not increase if the temperature change is also favourable for its competitors or natural enemies. For aquatic organisms, indirect effects can also include oxygen limitation [17], lower nutrient concentrations via thermal stratification of the water column [18], and changes in river discharge [19]. Most likely, both direct and indirect effects of temperature occur simultaneously. Examination of the observed relationship between ambient temperatures and population abundances could provide insight into the relative strengths of these pathways.

    Even when exposed to the same pattern of temperature change, species are likely to be affected in differing ways and to differing degrees. Species' responses to temperature change will vary according to their specific physiology and ecology [2023]. Direct effects of temperature are likely to be mediated by species temperature niche [24]. In particular, species’ cold tolerances have been shown to greatly vary among species, more so than heat tolerances [14,25]. Species' habitat breadth could also determine tolerance to changing environmental conditions [24,26]. Environmental changes may be less important for species with greater niche breadths. By contrast, species with a narrower niche breadth might benefit or suffer depending on the direction of change and the species’ niche position. Dispersal ability may also play a role by affecting a species' ability to track preferred temperature conditions and resources as well as buffering local population fluctuations via metapopulation processes [27]. Body size may affect sensitivity to temperature variability through covariation with other species attributes, such as life history and temperature niche [7] and thermoregulation [28]. Species with a faster life history may respond more quickly to temperature fluctuations, via changes in reproductive success, than species with a slower life history [29].

    In Europe, there has been consistent warming across all environmental realms [3032]. Few studies have tested for commonalities in the effects of temperatures on annual population abundances across a broad range of taxa, and particularly not across different environmental realms [11,33]. Present studies show that temperatures influences population dynamics of terrestrial and freshwater species and that time series of insects were less predictable by weather variables than those of vertebrates [11,12]. However, we lack a systematic comparison of the effects of weather on the population dynamics of across different taxa and realms, as well as an understanding of how species attributes modify species’ responses to climatic fluctuations. Here, we studied the relationship between temperatures in different seasons and the annual population abundances of different freshwater (invertebrates and fish), marine (benthic invertebrates and benthic/demersal fish), and terrestrial communities (spiders, beetles, and birds) using long-term local population data collected in western/central Europe (Belgium, Denmark, Germany, The Netherlands, and the southern part of the North Sea). Our aim was to identify any generalities, according to species attributes, major taxonomic group, or environmental realm, in how species are affected by temperature (electronic supplementary material, figure S1).

    First, we asked whether there was an average relationship between temperature and annual abundance across all species in a community. We hypothesized that temperatures, especially winter temperatures, would be most commonly positively related to abundances [5,12,3436]. Next, we studied the variation among species in the relationship between temperature and abundance within each community. Because species vary more in their cold than in their heat tolerances [14,25], we hypothesized that interspecific variation in the effects of temperature on species' abundances would be greater in winter than in other seasons. We then tested whether species attributes explained interspecific variation in the temperature–abundance relationships. We predicted that the population abundances of species with a broad temperature range, large geographical range size (as a proxy of niche breadth [37]), broad habitat niche, high dispersal ability, and large body size and long lifespans would be less affected by temperature than the respective counterparts.

    2. Material and methods

    (a) Population datasets

    We used population abundance data from nine datasets that have been collected annually in recent decades, with a time span of at least 15 years, within central Europe or the North Sea (table 1). These data comprise terrestrial (2 × ground beetles, spiders and birds), freshwater (insects and fish), and marine (2 × invertebrates and fish) communities.

    Table 1.Characteristics of the nine population datasets used in our analysis, including the taxonomic group sampled (labelled as ‘dataset’), the general location of sampling, the timespan over which population data were collected, the number of species used in our analysis (species seen in at least 25% of the census years), the season when sampling was undertaken (and frequency with which each sampling site was visited), and the methods used to sample the community.

    dataset location timespan no. sites no. species season methods
    beetles(1) Dwingelder Veld, The Netherlands 1991–2013 2a 59 all year (fortnightly) pitfall traps [38]
    beetles(2) Ijzermonding, Belgium 1990–2006 3b 50 all year (fortnightly) pitfall traps
    spiders Ijzermonding, Belgium 1990–2013 3c 96 all year (fortnightly) pitfall traps
    birds Danish common bird survey 1976–2012 370 79d May/June (once) point count survey [39]e
    freshwater insects Breitenbach, Germany 1987–2005 1 39 all year (every 1 or 2 days) capture of emerging adults in a tent over the stream [40]f
    freshwater fish River Lippe, Germany 1993–2007 2 25 Sept/Oct (once) data collected using electrofishing by the German federal state agencies
    marine invertebrates(1) Dogger Bank, North Sea 1991–2012 37g 36 July–Aug (once) beam trawl catches in July/August [41]
    marine invertebrates(2) southern North Sea 1998–2012 24 32 July–Aug (once) beam trawl catches in July/August [42]
    marine fish southern North Sea 1991–2014 24 32 July–Aug (once) North Sea International Bottom Trawl Survey [43]h

    aWe focused on unmanaged sites. Such site-level knowledge was not available/applicable (i.e. no management) for the other datasets.

    bTwo sites were not surveyed in two survey years.

    cTwo of the sites were not surveyed in four and six of the survey years.

    dTrans-Saharan migrants were excluded.

    eData available at: http://www.dof.dk/.

    gTwenty-three sites were not surveyed in one or two survey years.

    hData available at: http://datras.ices.dk/.

    (b) Species attributes

    We took an a priori approach to selecting the attributes for analysis, testing those related to concepts that have broad relevance across organisms. However, this approach meant that data on the desired species attribute were not always available for all taxonomic groups. We aimed to collect information for each species on its distribution (used to calculate its temperature range and geographical range size), habitat breadth, dispersal ability, body size, and lifespan (or related measures of life-history speed, e.g. age at maturity or life cycle duration). Species attribute data were collected from databases and the literature (see electronic supplementary material, table S1 for details on sources). The type of information available for each taxonomic group varied—for instance, the habitat breadth of beetles was coded as either stenotopic or eurytopic while the habitat breadth of birds could be calculated as a continuous variable using species affinities to different habitat classes. For freshwater insects, only genus level data were available for many attributes. The few species that were not listed in the main attribute data sources were excluded from the analysis. We additionally obtained taxonomic information for each species and used this information to create a taxonomic tree (electronic supplementary material, table S1 for sources). For birds, we used the phylogenetic tree from Jetz et al. [3].

    In a small number of cases, some attribute data were missing for certain taxa (15% of habitat breadth data for spiders; 3% of age at maturity for marine fish; 8% of longevity data for marine benthic invertebrates). These data were imputed using a random forest model, including all the species attribute variables used in the subsequent regression models and the first eigenvector of the decomposed phylogenic/taxonomic tree as predictors [44]. Some attributes were not tested for whole datasets: habitat breadth (marine benthic invertebrates—no appropriate data); dispersal ability (birds—Kipp's distance has been previously used [45] but such data are not found within databases; freshwater insects—all were flying insects; marine fish—no data within databases independent of size); lifespan (ground beetles/spiders—no data but little species variation expected).

    Distribution data (electronic supplementary material, table S1) were used to calculate temperature range and geographical range size for each species. Using distribution data delineated to Europe, we overlaid each species' distribution with a temperature data layer. We used a European extent because the best distribution data were usually limited to this extent. We used a mean daily temperature map from the E-Obs gridded dataset on a 25 km grid [46] for the terrestrial and freshwater datasets and a bottom surface temperature map from Aquamaps on a 50 km grid [47] for the marine datasets. Since we only intended to create a variable that placed species on a gradient from small to large ranges, this difference in resolution is expected to be unimportant. Temperature range was calculated for each species as the difference between the maximum and minimum temperature over its range (mean of the five warmest occupied grid cells minus the mean of the five coolest occupied grid cells). Range size was estimated as the number of climatic grid cells intersecting with each species’ distribution. For the bird dataset, we used range maps delineated to breeding and/or resident areas and temperature data for April–June.

    (c) Temperature data

    We retrieved temperature data at the specific sites of population data sampling. For the terrestrial datasets, air temperatures came from national weather service agencies (Deutsche Wetterdienst in Germany (DWD, www.dwd.se); Royal Meteorological Institute of Belgium for Belgium (www.meteo.be); the European Climate Assessment and Dataset (http://www.ecad.eu) and local weather stations (http://www.weerstation-eelde.nl) for The Netherlands). The data provided were either daily or monthly mean daily temperatures. For the marine realm, we obtained mean weekly sea surface temperature data compiled by the Federal Maritime and Hydrographic Agency, which had been collected from drifting buoys at fixed stations (Helgoland in the southern North Sea and the Dogger Bank) in the North Sea. For the freshwater datasets, we used water temperature data when it had been collected as part of the population monitoring (insects [40]) and air temperature (from the DWD) when not (fish). When possible (all except for marine), we also obtained data on maximum and minimum temperatures but we focused on average temperatures because mean summer and winter temperature were usually correlated with mean maximum and minimum temperature during summer and winter (i.e. the summers/winter that were warmer/cooler on average also had warmer/cooler extremes). Temperature data missing for some dates were imputed using a generalized additive model with a spline for day of year and an additive site effect.

    These data were used to calculate average daily temperatures (°C) per season per dataset. Seasons were winter (Dec–Feb), spring (Mar–May), summer (June–Aug), and autumn (Sept–Nov) (electronic supplementary material, figure S2). For the terrestrial and freshwater datasets, we also obtained precipitation data from the same data sources and calculated average monthly (millimetres) precipitation amounts. For the marine datasets, the seasons were shifted ahead by one month (i.e. winter was Jan–Mar) to correspond to the seasonal peaks and troughs of temperatures in the North Sea.

    (d) Statistical analysis

    We took a standardized approach to analyse each dataset (electronic supplementary material, figure S1). Because temperatures across sampling sites within datasets were highly correlated, we fitted the models at the dataset level, which meant first aggregating data from different sampling sites within each dataset to create total annual abundances. Missing site-level values (e.g. when an annual survey was undertaken but a particular site was not included in that year) were first imputed in a simple AR(1) model with year and site as fixed factors. After this aggregation step, each dataset comprised data of a single population count for each species in each census year.

    The basic model fitted to each dataset was a Gompertz population model with an autoregression term of order 1, i.e. AR(1):

    Display Formula
    where x is logged abundance in the year of census (t) and for species (s); temp is the mean daily temperature in a specific season; other represents one or more terms of additional temperature (or precipitation for the terrestrial or freshwater datasets) variables like temp; year is a continuous variable; and sp and obs are random effects for species (overall intercept and random slopes for each covariate) and observation. Finally, a and b are the intercept and slope coefficients of the fixed effects being estimated. Models were fitted using a Bayesian approach—integrated nested Laplace approximations (INLA), which uses an integrated nested Laplace approximation [48].

    Using this model, we tested whether there was an average relationship between temperature and abundance across all species in a community (i.e. dataset). Thus, we call this analysis ‘community level’ because species differences were only modelled as random effects around the community average effect. Different seasonal temperature parameters were tested: average daily temperature during the winter immediately preceding the survey, and autumn, summer, spring, and winter in the year before the survey (electronic supplementary material, figure S3). We focused on the effects of temperature preceding the survey, and hence lagged effects of temperature mediated through survival and/or reproduction, to avoid complicating effects of temperature affecting detection probabilities during the time of survey. However, some organisms may respond quickly to current temperatures with no or at least a shorter time lag than we assume here.

    We present the analysis of multiple regression models in which all five variables were tested simultaneously in the same model, and insignificant variables were removed sequentially. The effect sizes of insignificant variables were then obtained by adding them to the reduced model. For comparison, we also tested the effect of each temperature variable in a simple regression model (electronic supplementary material, figure S4). The effects of average monthly precipitation were also tested in the terrestrial and aquatic datasets, and any significant effects were controlled for when testing the effect of temperature. However, in the main paper, we focus discussion on the effects of temperature because trends in precipitation are much more variable than those in temperature [9], which complicates the link to climate change. Because other environmental variables are likely to have affected abundances, we also included a continuous ‘year’ predictor in each model. This aimed to separate the temperature effects from any long-term population trend driven by an omitted variable, which would be estimated by the year effect (the results were similar without the year effect; electronic supplementary material, figure S5). However, this would not account for the effects of more complex and fluctuating environmental change.

    Most of the datasets contained data on observed abundances (integers) and the model was fitted as a Poisson model (with a log-link) with abundance as the response. In the bird and marine fish dataset, the data were instead abundance indices; in these cases, the response was logged and a Normal model was fitted. Since the Poisson model used a log-link, the coefficients from both models represent change in log abundance per degree Celsius change in temperature and were, therefore, comparable. Random intercepts and random slopes for each species were included to allow the effects of temperature to vary among species, and the AR(1) correlation. We also included an observation-level random effect to account for overdispersion. Correlations between the random slopes and intercepts were considered by either they had little effect on the results (i.e. whether or not the 95% CIs overlapped zero) or their inclusion was not supported by the deviance information criteria (DIC). We calculated R2 following published methods [49]. To test whether we could detect any generalities in the effects of temperature across datasets, we also used meta-analysis to combine the coefficients (effect sizes) of the models. We calculated the average effect as a weighted mean of the effect sizes across all datasets (or terrestrial/aquatic), weighting each by the precision (i.e. inverse variance) of the effect size, see [50] for further details.

    We next estimated the effect of temperature on abundances at the species level. We fitted a similar AR(1) population model as above, but separately for each species. To avoid overfitting time series, we tested each temperature variable, controlling for any effect long-term year effect, in separate models.

    Display Formula

    From these models, we extracted the species-specific temperature effects (i.e. b1, the slope between temperature and abundance). First, we calculated the variances of these effects within each dataset for each season as a weighted variance (weighting each effect size by its inverse standard error). A Levene's test was used to examine whether the variances significantly differed among seasons. Next, we used these species-specific temperature effects to test our hypotheses about how they are influenced by species attributes. Because our hypotheses regarding the effects of species attributes were based on how much species responded to temperature rather than in what direction, we were interested in the absolute size of the effects of temperature. Thus, we took a two-step modelling approach. The absolute values of the coefficients (|b1|) estimated by these species-level population models were used as the response in further regression models that tested whether variation in |b1| could be explained by species attributes. We analysed species response to whichever seasonal temperature resulted in a model with the lowest DIC (repeating it also without the year term in the model as well as for whichever seasonal temperature resulted in the largest interspecific variation; see electronic supplementary material, figure S6). Species were weighted by the inverse standard error of b1 in this analysis so that species' estimates with lower confidence were downweighted. We also used robust regression [51] to ensure that no outlier or single influential species drove the results. Model residuals were checked graphically. When species attribute data were ordinal but represented a continuous variable, data were treated as continuous if there were at least five levels and graphical exploration suggested a linear relationship was reasonable. For categorical variables, we present the largest pairwise difference. We present the results of simple regressions of the effects of species attributes on abundance responses to temperature but further analysis found that building multiple regressions did not affect the conclusions. Finally, we tested whether there was a phylogenetic signal in the residuals of our models using an Abouheif's [52] test but it was insignificant in all cases.

    To create a comparable effect size metric for the effect of each species attribute, we converted the t-statistic of each coefficient from the robust regression into a correlation coefficient, r:

    Display Formula

    To calculate its 95% CIs, r was Fisher's z-transformed and its standard error calculated as:

    Display Formula
    Display Formula

    Finally, the 95% CIs were back-transformed from Inline Formula to r. The correlation coefficients from each dataset were then combined to produce an average effect size for each attribute by a standard random-effects meta-analysis, see [50] for further details. All analysis was conducted within R v. 3.2.3 [53].

    3. Results

    (a) Community level

    Winter (or previous winter) temperature had the most common positive effect, seen in all taxa except freshwater organisms (figure 1). This means that, in these cases, population abundances tended to be greater in the year following warmer winters. Previous spring temperatures were also positively related to abundances for birds, and previous spring and summer for freshwater insects. However, there were also some negative associations with temperature but these were almost only in the aquatic datasets (figure 1). Negative effects were seen for winter and previous autumn temperatures in the marine benthic invertebrates, for summer temperatures in the freshwater fish, and for previous winter temperatures in the freshwater insects. The only negative effect of temperature in the terrestrial datasets was found for the effect of summer temperatures on beetle abundances. When averaging the model coefficients across datasets (all, or for terrestrial and aquatic data separately), we only found non-zero average effects in the terrestrial datasets for the effects of winter and previous winter (electronic supplementary material, figure S7).

    Figure 1.

    Figure 1. Community-level effects of seasonal temperatures on species' abundances of different taxonomic groups. Species differences were modelled as random effects around the average community-level effect in a multiple regression model (see equation in main text). Cross-bars are means (log change in abundance per °C change of temperature) ± 95% CIs. Bars not overlapping zero are regarded as significant—shaded in grey.

    Despite this evidence for significant effects, the amount of variation explained by the fixed effects of these models was very low (marginal R2 of the best variables: 0.01–9%), which can be partly explained by species-level variation around these community-level effects. Mostly similar patterns were seen when testing the effect of each temperature variable in simple regression models (electronic supplementary material, figure S4), or without the continuous year predictor in the model (electronic supplementary material, figure S5). The terrestrial and freshwater analysis shown in figure 1 also controlled for any precipitation effects on abundance. Precipitation effects, both positive and negative, were found in the beetle community and in both the freshwater datasets (electronic supplementary material, figure S8).

    (b) Species level

    Interspecific variation in the effect of temperature on species' abundances tended to be lower for winter (or previous winter) temperatures compared with those for other seasons in the terrestrial and freshwater datasets (figure 2). In the analysis of each dataset, we found negative effects of range size (marine benthic invertebrates), habitat breadth (birds), dispersal ability (freshwater fish), body size (spiders, freshwater fish, marine benthic invertebrates, and fish), and lifespan (marine benthic invertebrates and fish) (figure 3). Thus, in these cases, the relationship between temperature and abundance was weaker for species with a greater range size, habitat breadth or dispersal ability, or a larger body size or a longer lifespan. When combining the results from each dataset by meta-analysis, we found significant average effects of range size, as well as body size and lifespan. The negative average effect of range size was robust to modelling decisions but the effects of body size and lifespan were partially sensitive (electronic supplementary material, figure S6).

    Figure 2.

    Figure 2. Standard deviations of the species-level effects of average temperature on population abundances in each dataset. In all datasets except freshwater fish, marine benthic invertebrates(1), and marine fish, the differences in variance were significant in a Levene's test.

    Figure 3.

    Figure 3. Relationships between species attributes and absolute response (i.e. absolute slope of relationship between species abundance and temperature) to whichever seasonal temperature resulted in a model with the lowest DIC. Cross-bars are means ± 95% CIs of the correlation coefficient. Bars not overlapping zero are regarded as significant—shaded in grey. A negative effect means that the abundances of species with a higher value of that attribute was less related to temperature. We focused on the season with the most significant community effect—beetles(1): previous winter; beetles(2): previous winter; spiders: winter; birds: winter; freshwater insects: previous spring; freshwater fish: previous summer; marine benthic inverts(1): winter; marine benthic inverts(2): previous autumn, and marine fish: previous spring.

    4. Discussion

    We asked whether there are any generalities in the effect of ambient temperature on the population abundances of different taxonomic communities within central Europe. The effect of winter temperature on abundances was consistently positive for terrestrial communities but not for aquatic communities. Within some communities, species with large range sizes, larger body sizes, or longer lifespans tended to show weaker relationships between ambient temperatures and their abundance.

    (a) Community-level effects of temperature on population abundances

    Effects of temperature, especially during winter, on population abundances were positive for the terrestrial communities, consistent with previous studies [34,54]. In the temperate zone, increases in temperature have been predicted to have positive effects on the individual fitness of ectotherms based on physiological relationships [13,55], which could potentially lead to increased abundance, depending on other factors affecting population growth. Thus, our results are consistent with the importance of direct effects of temperature on species’ performance and abundance, perhaps mediated by improved overwinter survival [56]. While other environmental changes (e.g. land use change) are likely to have affected the dynamics of these communities, ambient temperatures still influenced species' abundances. Such effects may contribute to species’ range expansion in response to climate change [8]. Warmer temperatures could also affect the activity of organisms, making them more detectable/catchable during a population survey [57]; however, because we focused our analysis on the effects of temperatures during periods preceding the population survey, this effect should be minimized. A meta-analysis of published studies [12] also found generally positive effects of temperature on demographic responses, particularly for birds and plants. Additionally, at least over medium to high latitudes, the results from this previous study also suggested that the importance of temperature was geographically widespread [12].

    In the aquatic datasets, we found evidence of positive, neutral, and negative effects of temperatures on abundance at the community level. Only marine fish were found to have only neutral or positive relationships (with winter temperature). Temperature effects are likely to be more complex in aquatic systems [36]. Negative correlations between temperature and recruitment have been reported in several North Sea fish [58]. Increasing water temperature decreases the dissolved oxygen concentration, which may lead to hypoxia for aquatic organisms [17]. Warmer water temperatures can also lead to decreased nutrient concentrations via thermal stratification of the water column [59] as well as increased toxicity of pollutants [36]. Thus, any possible direct benefits of warming for temperate species may be counteracted by these indirect pathways. Interestingly, in the freshwater datasets, we also found some evidence that precipitation may be as important as temperature—precipitation can have multifaceted effects including changes in discharge, flow rates, and pollutant concentrations [60].

    (b) Species-level effects of temperature on population abundances

    While analyses at the community level inform on the average responses of species in the community, it hides the differences at the species level. Understanding interspecific variation in response to changing temperatures is relevant because of the interest in trait-based assessments of species vulnerability to climate change [20]. Despite consistent evidence that species tend to vary more in their cold tolerances than in their heat tolerances [14,25], we found no evidence of greater interspecific variation in the importance of winter temperature for abundances. Together with the community-level effects of winter temperature on abundance, these results suggest that the terrestrial species in our analysis more consistently benefited from warmer winters than from warmer temperatures in other seasons. This could indicate that even species that are able to physiologically tolerate cold weather benefit from warmer temperatures during winter [61].

    Some of the species-level variation in our datasets could be explained by species attributes. We predicted that the population abundances of species with broad habitat and temperature niches, large geographical ranges, high dispersal ability, large body size, and long lifespan would be less affected by temperature than the respective counterparts, i.e. that these attributes would have negative effects on the strength of the relationship between temperature and abundance. We found some evidence to support our hypotheses but the effects were only statistically significant in a few datasets. In general, we found that species with larger bodies and longer lifespans were less affected by temperature change within communities, which would be consistent with slower responses for organisms with slower/longer life histories [62]. There was also an indication that the abundances of species with larger range sizes tend to be less affected by variability in temperature. Species with large ranges are often those with broad niche breadths—whether that is temperature, diet, and/or habitat breadth [37,63]—thus our results may suggest that the abundances of ecological generalists are less affected by temperatures. Ecological generalism [64,65], as well as more stable population trends [8], have been linked with range expansion in response to climate change. If generalist species are less affected by temperature fluctuations within their ranges, they might be more able to maintain stable population trends than specialists, helping range expansions at the cold range margin and limiting range contraction at the warm range margin.

    5. Conclusion

    By focusing on the relationship between abundances and temperatures in the preceding year, our analysis estimated the short-term effects of changes in average temperatures. We did not consider the effect of climatic variability per se and any extreme events—this would be interesting to further examine, especially for precipitation, but would require a different analytical framework. Thus, our approach could point towards the impacts of long-term climate change assuming they are an accumulation of average temperature effects, rather than the consequences of extreme events [66]. Our findings lead to the prediction that warmer winters due to climate change are likely to increase the abundances of many terrestrial species within central Europe; by contrast, the outcome is more complex for aquatic species. By comparing effects across taxa, we could also investigate why species differ in their response to changing temperatures. Our results suggest that populations of species with large geographical ranges, larger bodies, and longer lifespans might be less sensitive to changing temperatures. These factors that affect changes in abundance in response to climate change within species' geographical ranges may also have consequences for changing abundances beyond their range.

    Data accessibility

    Data on species-specific slopes of the relationship between their abundances and ambient temperatures are available in dryad: http://dx.doi.org/10.5061/dryad.23f21 [67].

    Authors' contributions

    D.E.B. and K.B.-G. designed the study. D.E.B. performed the analysis, with input from R.B.O.H. P.H., I.K., L.B., W.D., S.D., F.H., H.N., M.S., S.S., M.T., R.v.K., R.V. supported the analysis by contributing data. D.E.B. drafted the manuscript with K.B.-G., with edits and revisions provided by all authors.

    Competing interests

    We have no competing interests.

    Funding

    This study was funded by a LOEWE excellence initiative of the Hessian Ministry for Science and the Arts and the German Research Foundation (DFG: grant no. BO 1221/23-1).

    Acknowledgements

    We thank the Umwelt und Verbraucherschutz Nordrhein-Westfalen and Hessisches Landesamt für Umwelt sharing their data. Additionally, we appreciate the open access marine data provided by the International Council for the Exploration of the Sea. We thank the following scientists for taxonomic or technical advice: Theo Blick, Christoph Brendel, Rudy Claus, Konjev Desender, Susanne Fritz, Eva-Maria Gerstner, Jean-Pierre Maelfait, Eike-Lena Neuschulz, Steffen Pauls, Hans Turin, and Rüdiger Wagner. We thank Ingolf Kühn and Callum Lawson for commenting on a previous version of the manuscript.

    Footnotes

    Deceased.

    Electronic supplementary material is available online at https://dx.doi.org/10.6084/m9.figshare.c.3869188.

    Published by the Royal Society. All rights reserved.

    References