Superfrogs in the city: 150 year impact of urbanization and agriculture on the European Common Frog

Despite growing pressure on biodiversity deriving from increasing anthropogenic disturbances, some species successfully persist in altered ecosystems. However, these species' characteristics and thresholds, as well as the environmental frame behind that process are usually unknown. We collected data on body size, fluctuating asymmetry (FA), as well as nitrogen stable isotopes (δ15N) from museum specimens of the European Common Frog, Rana temporaria, all originating from the Berlin–Brandenburg area, Germany, in order to test: (a) if specimens have changed over the last 150 years (1868–2018); and (b) if changes could be attributed to increasing urbanization and agricultural intensity. We detected that after the Second World War, frogs were larger than in pre‐war Berlin. In rural Brandenburg, we observed no such size change. FA analysis revealed a similar tendency with lower levels in Berlin after the war and higher levels in Brandenburg. Enrichment of δ15N decreased over time in both regions but was generally higher and less variable in sites with agricultural land use. Frogs thus seem to encounter favorable habitat conditions after pollution in postwar Berlin improved, but no such tendencies were observable in the predominantly agricultural landscape of Brandenburg. Urbanization, characterized by the proportion of built‐up area, was not the main associated factor for the observed trait changes. However, we detected a relationship with the amount of urban greenspace. Our study exemplifies that increasing urbanization must not necessarily worsen conditions for species living in urban habitats. The Berlin example demonstrates that public parks and other urban greenspaces have the potential to serve as suitable refuges for some species. These findings underline the urgency of establishing, maintaining, and connecting such habitats, and generally consider their importance for future urban planning.

of roads and railways are fragmenting the landscape, and cities become heat islands characterized by higher temperatures compared to their surrounding (Liu et al., 2016;Ward et al., 2016).
The world's human urban population has grown rapidly and is associated with an expansion of urban areas but also with an increase of global food demand, and accordingly, with an agricultural expansion.
Agriculture is a major cause for deforestation, increase in greenhouse gas emissions, and the consumption of a substantial amount of freshwater (Ramankutty et al., 2018). The invention of the Haber Bosch procedure in the early 20th century permitted the synthesis of nitrogen fertilizer from atmospheric nitrogen. The resulting increased use of fertilizer, in combination with pesticides, led to a boost in agricultural productivity but with the result in decrease of freshwater quality with detrimental effects to freshwater ecosystems and their terrestrial surroundings (Ramankutty et al., 2018).
However, some species, such as wild boars, red foxes, house sparrows, American toads, or whistling tree frogs, persist in habitats altered by urbanization or intense agriculture (Bateman & Fleming, 2012;Hamer & McDonnell, 2010;Isaksson, 2018;Koumaris & Fahrig, 2016;Stillfried et al., 2017). They are traditionally referred to as urban-adapters (McKinney, 2002). Species in both types of these modified ecosystems have usually been considered generalists with a broad environmental tolerance. Although scientific attention has rather focused on identifying reasons for the exclusion of certain species from human-modified landscapes (Ducatez et al., 2018;Hassall, 2014;Howard et al., 2020), the characteristics behind a species' ability to persist, as well as the limits in environmental changes that still allow persistence, have rarely been investigated (Hamer & McDonnell, 2008;Jung & Threlfall, 2016;Marques et al., 2019).
However, it is essential to understand the organisms' responses to anthropogenic impacts and environmental features facilitating their persistence in order to develop sustainable management strategies, which might promote and maintain biodiversity and associated ecosystem services in human-modified landscapes (Alberti, et al., 2017;Donihue & Lambert, 2015;McDonnell & Hahs, 2015). General responses of organisms to rapid environmental changes, natural as well as those deriving from human actions, are either moving into more suitable areas, or stay and adjust physiology, behavior, life-history, and/or morphology (Alberti, 2015;Alberti et al., 2017;Lowry et al., 2013;Sparkman et al., 2018).
These adjustments are facilitated via phenotypic plasticity and contemporary evolution, that is, evolution occurring over less than a few hundred years (Palkovacs et al., 2012), with humans acting as major driving forces (Hendry et al., 2008;Hendry et al., 2017).
Phenotypic trait changes, in turn, can alter ecosystem function and, in case of heritability, create the potential for eco-evolutionary feedback, with consequences for human well-being (Alberti, 2015;Alberti, et al., 2017;Rivkin et al., 2019;Rudman et al., 2017).
The study of potential adaptive trait changes in persisting species requires long-term data, which are not easily accessible for physiological, behavioral, or life-history traits. However, changes in morphological and biochemical traits of preserved specimens may be used as a proxy for species' responses to rapid environmental changes (Holmes et al., 2016;Kern & Langerhans, 2018;Meineke et al., 2018;Pergams & Lawler, 2009;Schmitt et al., 2018;Stumpp et al., 2016). Several morphological characters have been shown to be affected by urbanization and agricultural landuse, for example, body sizes of birds, amphibians, and arthropods as an indicator of habitat quality and life history (Jennette et al., 2019;Meillère et al., 2015;Merckx et al., 2018). Fluctuating asymmetry (FA), defined as small, random deviations from perfect bilateral symmetry, has been used as a measure of developmental instability caused by environmental stress in reptiles and fish (Lazić et al., 2013;Lutterschmidt et al., 2016). Additionally, stable isotopes can serve as an indicator of nitrogen enrichment in amphibians, fish, invertebrates, and plants (Donázar-Aramendía et al., 2019;Jefferson & Russell, 2008). Notably, so far only few studies investigating human-induced morphological changes in mammals and fish included the temporal component (Kern & Langerhans, 2018;Pease et al., 2018;Pergams & Lawler, 2009;Tomassini et al., 2014).
We herein aim at identifying specific trait changes reflecting the changing living conditions of a species that have persisted in rapidly changing environments. To this end, we tested the hypotheses that in the widespread European Common Frog, Rana temporaria Linnaeus, 1758, morphological traits, that is, body size and FA, and nitrogen stable isotopes (δ 15 N) as a reflection of the environment the frogs lived in, (a) have changed over a period spanning the last 150 years (1868-2018); and (b) that changes in these traits can be attributed to urbanization and agricultural intensity. More precisely, we expected decreasing body sizes, increasing levels of FA and increasing δ 15 N values in response to decreasing habitat quality, increasing environmental stress and enrichment of nitrogen through artificial fertilizers and air pollution, respectively. If we could detect such changes, we assumed that they were associated with the change of specific land use features.
We selected R. temporaria for three reasons. First, amphibians have highly specific ecosystem demands, low dispersal abilities either due to physical constraints, a high breeding site fidelity, or to anthropogenic barriers, and are therefore particularly affected by landscape modifications (Arntzen et al., 2017;Hamer & McDonnell, 2008;Stuart et al., 2008). They rely on terrestrial and aquatic environments due to their biphasic life cycle, making them very vulnerable to changes in both ecosystem types (Becker et al., 2010). Consequently, they have experienced a global decline since the second half of the 20th century (Beebee & Griffiths, 2005). Second, the European Common Frog has persisted in a large range of habitats (Sillero et al., 2014), including cities and agricultural landscapes, during the entire Anthropocene, despite fundamental environmental changes (Carrier & Beebee, 2003;Schlüpmann et al., 2004), making the species suitable for tracking the effects of environmental changes (Vander Wal et al., 2013).
Third, historical series of museum-preserved specimens were available, making long-term analyses possible. We chose specimens originating from the Berlin-Brandenburg region, Germany, which comprises both urban areas and areas with intense agricultural activities, and thus potentially huge environmental change during the last 150 years.

| Study area
The study area comprises Germany's capital, the city of Berlin The Berlin sites used in this study are predominantly characterized by a certain amount of built-area and developed urban green spaces. Very few Berlin sites are located within forested areas or close to agricultural fields at the edge of the city. The Brandenburg sites are predominantly, but not exclusively, characterized by agriculture. Some are also located close to forested patches, within grasslands (not to be confused with developed urban greenspace) or urbanized areas (e.g., the city of Potsdam).

| Specimen selection
We used all ethanol-preserved adult voucher specimens of R. temporaria, unambiguously originating from the Berlin-Brandenburg area in the collection of the Museum für Naturkunde Berlin (ZMB, https://doi.org/10.7479/5tm4-9r29). Snout-vent length (SVL) of at least 5 cm was used as a criterion to define adults (Dittrich et al., 2018;Miaud et al., 1999). This provided us with an initial sample size of n = 124 specimens, divided into n = 56 specimens collected at 15 different locations in Brandenburg, and n = 68 specimens collected at 17 different locations in Berlin, covering a time span from 1868 to 2017.

| Morphological traits and stable isotopes
Body size was measured as SVL (Watters et al., 2016) with a digital caliper, always by the same observer. In addition to the measurements of preserved specimens, we also took SVL of 34 live adults from Berlin in 2018. Total sample size for SVL analysis over time was We used VG Studio Max 3.0 with the distance measurement tool for the measurements. Broken bones were excluded from analyses.
For determining nitrogen stable isotope values, thigh muscle tissue was extracted from preserved frogs (n = 104). Samples were dried at 60°C in a drying chamber for 72 hr. Stable isotope F I G U R E 1 Limb characters measured for fluctuating asymmetry assessment in Rana temporaria (male, ZMB 87968). (a) Micro-3Dcomputed tomography scan of entire body with, (b) right humerus (purple), (c) right radio-ulna (yellow), (d) right femur (green), and (e) right tibio-fibula (orange) analysis of 1 mg dried tissue per frog was performed with a THERMO/Finnigan MAT V isotope ratio mass spectrometer (Thermo Finnigan), coupled to a THERMO Flash EA 1112 elemental analyzer via a THERMO/Finnigan Conflo IV-interface in the stable isotope laboratory of the Museum für Naturkunde, Berlin.
Stable isotope ratios are expressed in the conventional delta notation (δ 15 N) relative to atmospheric nitrogen (Mariotti, 1983).
Standard deviation for repeated measurements of laboratory standard material (peptone) was generally better than 0.15 per mille for nitrogen.

| Land use features
We assessed the proportion of built-up area (impervious surfaces e.g., buildings, roads, industrial areas), greenspace (public parks, cemeteries, private gardens, sports grounds), and agricultural fields

| Data analysis
Prior to FA analyses, measurement error (ME) was quantified by repeating measuring bones on both sides (right and left) in a subset of n = 20-22 individuals per limb character (see Niemeier et al., 2019). Significant ME outliers were identified using the Grubb's test (Grubbs & Beck, 1972) leading to the exclusion of two tibio-fibula measurements. We then applied a mixed-model ANOVA (R package "lme4"; Bates et al., 2015) with Side as a fixed factor, Individual as a random factor, and the Side by Individual interaction as a mixed effect. Significance in the fixed factor Side would indicate directional symmetry, which has to be excluded.
To verify that FA exceeded ME, variance components (σ 2 ) were extracted from the random effects and signal (FA)-to-noise (ME) ratios calculated (Graham et al., 2010;Knierim et al., 2007). The variance component for the interaction ( 2 S×I ) is an estimate for FA.
The residual random variance ( 2 ME ) is an estimate for ME (Table 1).
To check if ME was of similar magnitude for each character, variations in the degree of ME were tested with a mixed-model ANOVA with Character as a fixed factor and Individual as a random factor ( Figure S1). Absence of antisymmetry was validated for the whole dataset by examining the frequency distributions of the signed FA values that is, right side minus left side (R − L), visually for symmetry and kurtosis ( Figure S2) and by using the Anscombe-Glynn kurtosis test (R package "moments"; Komsta & Novomestky, 2015; Table S1). Character-size dependency was tested by Spearman's rank correlation between absolute values of FA (|R − L|) and character-size (averaged (R + L)/2; Table S1).
Directional asymmetry was further excluded for the whole dataset by comparing deviations of the mean FA (R − L) from zero for each character with a one-sample t test (Table S1). Characters suitable for FA analysis were finally selected according to the following criteria: (a) a significant level of FA; (b) a signal-to-noise ratio > 1 to avoid that FA was masked by ME; (c) no significant variation in the degree of ME; (d) absence of directional asymmetry and antisymmetry; and (e) no character-size dependency in signed FA values. Only the humerus and radio-ulna met all the criteria (Table 1). In a recent study, selection pressure on the symmetry of functionally important characters, such as the hind limbs of anurans, has been proposed as an explanation for that observation (Didde & Rivera, 2019). Therefore, we excluded the femur and tibio-fibula from FA analyses. To investigate the effect of agriculture, all sites with a proportion of arable land, that is, intensively/extensively cultivated fields and fallow arable land excluding meadows and orchards, within the surrounding 1 km buffer zone greater than 2% were classified as agricultural sites (n = 33), regardless of the region (Berlin or Brandenburg). These sites covered a range from 2% to 39% of arable land. δ 15 N values of these sites were compared to the sites without agricultural land use, that is, 0%-1.3% (n = 71) using a linear mixed-effects model. Location of the sample was included as a random effect. We decided to use the 2% threshold because it has been shown that agricultural land cover along with the contamination of soils and waterbodies with chemicals from fertilizers can affect amphibian populations over large spatial scales up to at least 1 km due to their migration between aquatic and terrestrial habitats and the circulation of contaminants in surface water systems (Babini et al., 2018;Boissinot et al., 2019;Koumaris & Fahrig, 2016;Marsh et al., 2017). We thus speculated a priori that a proportion of 2% arable land, which was usually located at the edge of the 1 km buffer zone adjacent to agricultural fields continuing outside the buffer zone, might be sufficient to detectably affect our specimens.
Additionally, it was tested if there was a difference in the variability of the δ 15 N values between these two land use types by comparing the coefficient of variation (R package "cvequality"; Marwick & Krishnamoorthy, 2019). TA B L E 1 Summary of Rana temporaria limb characters' suitability examination for fluctuating asymmetry analysis. A two-way mixed model ANOVA (side = fixed factor, individual = random factor, side × individual = mixed interaction) on subset of untransformed repeated measurements was applied All statistical analyses were performed using R (R Core Team, 2018). The significance level was set to p = .05. Visualizations of model outputs were done using ggplot2 (R package "ggplot2"; Wickham, 2016).

| RE SULTS
Body sizes decreased between 1868 and the beginning of the Second World War in 1939 but increased afterward (GAM adjusted R 2 = .55, p < .01; Figure 3a). Comparison of the two time periods including the interaction with the factor Region revealed that body size differed between the two time periods (t = −3.98, p < .001), and that there was a significant interaction with the regions Berlin and Brandenburg (t = 2.693, p < .05; Table 2). Specifically, postwar individuals became larger in Berlin (t = 3.932, p < .01), but not in the rural Brandenburg (t = 0.337, p = .98) as revealed by post hoc pairwise comparisons (Figure 3b).
There was no general trend detectable for the development of FA over time (GAM adjusted R 2 = 0.03; p = .25; Figure 3c).
Comparison of pre-and postwar time periods including the interaction with the factor Region revealed a tendency toward lower CFA levels in Berlin after war, and higher postwar levels in

| D ISCUSS I ON
In our study, we aimed at examining potential trait changes across a long-term temporal scale in a species, which has persisted in habitats altered by urbanization and intense agriculture. We herein demonstrated trait change in the European Common Frog, R. temporaria, from the Berlin-Brandenburg area collected over the last 150 years.
Our initial prediction assumed that environmental conditions have Explanations in these studies vary from reduced predation pressure, more stable water temperatures to lower population densities, and thus lower competition and lower risk of infectious diseases in cities (Saenz et al., 2014;Scheffers & Paszkowski, 2015).
It is well known that amphibians grow continuously and body size thus is dependent of the age of an individual. To estimate amphibian ages skeletochronology techniques can be used (Leclair & Castanet, 1987). This method turned out to be inapplicable in the case of preserved museum specimens because the bones splinter when trying to cut them and destruction of the irreplaceable vouchers should be minimized. Furthermore, even in cases of successful identification of Lines of Arrested Growth (LAGs), results indicated to provide unreliable ages for example, large adult individuals with only one or two LAGs (unpublished data). However, there is less evidence for variation in the ages of breeding adults among different populations (Jennette et al., 2019). In our study we only used adult that is, breeding individuals with an SVL of at least 5 cm. To this end, we expected that age classes over time and between regions would be comparable and their role for changes in overall body sizes would be negligible. Even if the observed increase of body sizes was supposed to be driven by older ages of the specimens, this would imply that the frogs on average survived longer, which in turn would be the result of a change of the conditions they lived in for example, habitat quality and predation pressure. Furthermore, female Common Frogs tend to be larger than males at the same age (Miaud et al., 1999). However, we did not detect significant size differences between the sexes within the preand postwar groups neither in Berlin nor in Brandenburg.
Notably, the study sites with intermediate urbanization degrees  (Koumaris & Fahrig, 2016;Sievers et al., 2018;Trochet et al., 2016 non-agricultural areas was higher. We did not analyze the carbon stable isotopes because we did not expect major differences in the origin of the carbon or in the general diet (i.e., trophic relationships) neither over time nor between land use types ( Figure S3).
However, the greater δ 15 N variation in non-agricultural areas indicates that the urbanized areas with their developed greenspaces offer more variable habitats with different nitrogen baselines and respective food resources than the agricultural areas. The loss of habitat heterogeneity as a consequence of agricultural intensification has been identified as a major cause driving biodiversity decline in agriculture-dominated landscapes (Benton et al., 2003).
Opposed to that, urban landscapes including urban green infrastructure increasingly gain importance as they provide a variety of food resources, nesting sites, or hiding places and thus serve as refuges for several species that are expelled from their natural habitats, such as lizards, falcons, bees, and bumblebees (Becker & Buchholz, 2015;Kettel et al., 2018;Samuelson et al., 2018).
In conclusion, the study of frogs in the city of Berlin exemplifies that urbanized areas have the potential to serve as suitable refuges for environmentally sensitive species, such as amphibians.
As indicated by the trait change over time and their association with environmental factors, habitats within cities can be restored whilst at the same time urbanization proceeds continuously. In this, and presumably many other examples, particular attention must be paid to the role of urban greenspaces in facilitating the function of urban ecosystems by providing high-quality habitat for wildlife. These findings underline the need of maintaining, establishing and connecting such green infrastructures by sufficiently large corridors and stepping stones within the urban matrix and considering their importance for future urban planning. However, areas with intense agriculture, such as predominantly but not exclusively found in Brandenburg, still seem to provide habitats of reduced quality. Management actions should further strive on minimizing and compensating for anthropogenic interventions in order to preserve and promote present biodiversity and associated ecosystem services in agricultural landscapes.

ACK N OWLED G EM ENTS
We thank the Technical University of Berlin for providing land use maps; especially A. Hiller, who helped us with the QGIS calculations. Open access funding enabled and organized by Projekt DEAL.

CO N FLI C T O F I NTE R E S T
All authors confirm that they have no conflict of interest to declare.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are openly available in MfN Data Repository at https://doi.natur kunde museum.berli n/ data/10.7479/5tm4-9r29.