Urbanisation drives inter‐ and intraspecific variation in flight‐related morphological traits of aquatic insects at different landscape scales

Urbanisation, as an unstoppable global phenomenon, has led to decreasing connectivity between habitats, which gives strong pressure on organisms. Current research has barely investigated urban effects on aquatic insect species traits. Here, we investigated how inter‐ and intraspecific variations of flight‐related morphological traits change along an urban gradient in three species of Dytiscidae at different landscape scales. We collected specimens in 30 urban wetlands in Helsinki, Finland. We measured flight‐related traits, including body length, pronotum length, elytron length, and the hind wing length and area. With linear models, we modelled how flight‐related traits of the three species responded to the percentage of impermeable surfaces in buffers with nine different radii: 100, 200, 300, 400, 500, 600, 800, 1000 and 1200 m. Our results show that (a) Ilybius ater was not affected by urbanisation in traits, indicating that some species are good dispersers and pre‐adapted to urbanisation. (b) The other two species, Acilius canaliculatus and Hydaticus seminiger, exhibited different patterns along the urban gradient, suggesting species utilise different strategies to cope with movement barriers in urban landscapes. (c) Species were affected by urbanisation at different scales, suggesting species have different ability to adjust morphological plasticity to endure movement barriers caused by urbanisation. This study reveals that urban land‐use change can have complex effects on inter‐ and intraspecific variation of aquatic insects. We highlight that urban planning should consider at which scale target species can endure urbanisation, to create an effective network of urban ponds for aquatic biodiversity conservation.


INTRODUCTION
Urbanisation has been an unstoppable global phenomenon that has transformed adjacent forests, agricultural lands and natural habitats into built-up areas, such as buildings and roads (Antrop, 2004;Liu et al., 2014;Niemelä, 1999;United Nations, 2018).Such land-use change in urban landscapes has led to a multitude of effects on the environment, such as pollution, increasing temperature, and habitat loss and fragmentation, which has decreased the quantity and quality of urban habitats (Grimm et al., 2008;Hamer & McDonnell, 2008;Jacobson, 2011;Oke, 1973Oke, , 1982;;Piano et al., 2020).Many species have undergone local extinction in urban landscapes (Cordier et al., 2021;Dri et al., 2021;Duncan et al., 2011;Fattorini, 2011), and urban flora and fauna show a declining species richness pattern at the global scale (Concepci on et al., 2015;Piano et al., 2020).The decreasing urban biodiversity has hazardous effects on the ecology of cities and the well-being of urban residents.Neglecting the importance of urban biodiversity will hinder sustainable development (Elmqvist et al., 2013;Lambert et al., 2021;Marselle et al., 2021;Naeem et al., 2016;Taylor & Hochuli, 2015).
Species traits refer to any measurable morphological, phenological, physiological, behavioural and reproductive characteristic of an individual (Cadotte et al., 2011;Kissling et al., 2018).These traits are important aspects to investigate for urban biodiversity conservation, because species traits allow quantification of how species respond to environmental changes and can be utilised to predict extinction risk (Bui et al., 2020;Chichorro et al., 2022;Genner et al., 2010;Pearson et al., 2014;Salas-Lopez et al., 2017).In urban contexts, organisms have shown species trait shifts along urban gradients.For example, dispersive and thermophilic species are more likely to occur in urban habitats than other species (Piano et al., 2017;Theodorou et al., 2020).Such shifts in traits may lead to species loss or replacement in urban landscapes (Piano et al., 2017) and have consequential effects on ecosystem functioning, such as food web dynamics (Dahirel et al., 2017;Eggenberger et al., 2019;Gianuca et al., 2017;Merckx et al., 2018).
Morphological traits are highly associated with the performance of organisms and their fitness, and dispersal-related morphological traits are highly relevant to urban contexts (Bellwood et al., 2019;Bertossa, 2011;Bui et al., 2020;Koehl, 1996).In urban landscapes, the dramatic change in land use has increased habitat isolation and decreased landscape connectivity, which hinder the movement of organisms among resource patches (Concepci on et al., 2015;Johnson & Munshi-South, 2017;Liao et al., 2022).Decreased landscape connectivity gives strong pressure on organisms, which may trigger eco-evolutionary feedback at multiple levels (Pelletier et al., 2009;Tüzün et al., 2017).For example, at the community level, some taxa, such as bumblebees and moths, show increased body size along urban gradients because of their better dispersal capacity (Merckx et al., 2018;Theodorou et al., 2020).At the species level, urbanisation can induce intraspecific variation in morphological traits (Thompson et al., 2021), which may drive the evolution of organisms (Koehl, 1996;Kriegman et al., 2018;Lambert et al., 2021;Wund, 2012).As the increasing temperature and habitat fragmentation in cities affect the development and the dispersal of organisms (Banaszak-Cibicka et al., 2018;Berger-Tal & Saltz, 2019;Miles et al., 2019;Sukhodolskaya et al., 2019), it is important to understand how urbanisation drives intraspecific variation of dispersal-related morphological traits in urban organisms.
Invertebrates utilise active flight or/and passive transport to disperse (Verberk et al., 2008).Previous research has mainly explored how urbanisation has affected intraspecific variation in flight-related morphological traits of terrestrial fauna (e.g., Bui et al., 2020;Eggenberger et al., 2019;Sukhodolskaya, 2013), whereas only a few studies have investigated aquatic and semi-aquatic organisms (e.g., Brans et al., 2017;Villalobos-Jiménez & Hassall, 2019).Aquatic organisms, however, may have received different impacts than terrestrial organisms.For instance, urban blue space, that is, urban surface waters and its surroundings have lower temperature increases than urban areas without aquatic components (Alikhani et al., 2021;Ampatzidis & Kershaw, 2020;Peng et al., 2020).The different increased temperatures in aquatic and terrestrial habitats may affect the development of aquatic organisms differently from terrestrial ones.Thus, patterns of intraspecific variation along urban gradients found in terrestrial organisms can differ between aquatic/semiaquatic organisms.
Aquatic insects play an important role in nutrient cycling and connecting food webs between aquatic and terrestrial ecosystems (Kraus, 2019;Wesner, 2010).They are sensitive to environmental changes caused by anthropogenic activities and are, thus, often utilised as a biomonitoring tool in aquatic ecosystems (Bonada et al., 2006;Hare, 1992).In this study, we focus on urban effects on inter-and intraspecific variation of aquatic insects.We utilise three diving beetle species (Dytiscidae) as models, because dytiscids are an indicator taxon of wetland biodiversity (Becerra-Jurado et al., 2014;Bilton et al., 2006).We aim to answer the following research questions: (1) How does urbanisation drive intraspecific variation in flightrelated morphological traits of dytiscids?(2) Do different species exhibit the same pattern?and (3) At which landscape scale(s) do dytiscids respond to urbanisation in respect of intraspecific variation in their flight-related morphological traits?

Study species
Dytiscidae is a family of water beetles widely distributed worldwide.
A. canaliculatus (Nicolai, 1822) is widely distributed in Palearctic (Nilsson & Hájek, 2022).The species often occurs in stagnant water with at least some vegetation and occasionally occurs in small creeks and rivers without vegetation (Bergsten & Miller, 2006;Nilsson & Holmen, 1995).The species has dimorphic females, that is, with both a grooved and a smooth morph of elytra, but smooth females have been rarely recorded; thus, the species is not considered as truly dimorphic in females (Bergsten & Miller, 2006).The species hibernates in water as adults (Nilsson & Holmen, 1995).Adults are good fliers and relatively tolerant to urbanisation (Iversen et al., 2017;Liao et al., 2020;Nilsson & Holmen, 1995).The life cycle is normally univoltine, that is, producing one brood per year (Bergsten & Miller, 2006).
It occurs in densely vegetated ponds and ditches and sometimes also in streams and rivers (Nilsson & Holmen, 1995;Temreshev, 2018).Larvae of the species are found in vernal pools (Galewski, 1990;Liao's personal observation).Adults are ready fliers and overwinter on land.The life cycle is univoltine (Nilsson & Holmen, 1995).
I. ater (De Geer, 1774) is widely distributed in the West Palearctic.
It occurs in densely vegetated stagnant waters and is often associated with muddy bottoms with rich detritus (Frelik & Pakulnicka, 2015;Friday, 1988;Nilsson & Holmen, 1995).Females of this species lay their eggs in leaves and stalks of living aquatic plants.The ovipositors are saw-like and used to insert eggs into living plant tissues (Jackson, 1960).Adults are good fliers and overwinter on land.The species is considered semivoltine, that is, passing the first winter as either eggs or larvae and the second winter as adults (Nilsson & Holmen, 1995).

Data collection
From 2018 to 2021, using activity traps and handnets, we collected specimens of the three dytiscid species, A. canaliculatus, H. seminiger and I. ater, in 30 ponds in the Helsinki Metropolitan Area (60.17 N, 24.94 E), Finland.Fifteen ponds were fishless, whereas the other 15 had fish (Liao, 2018a(Liao, , 2018b(Liao, , 2019;;Liao et al., 2022).Pond size varied from 0.01 to 7.45 ha (0.57 ± 1.38 ha), with a perimeter of 50-1110 m (250 ± 236 m).The ponds occurred in sites with a gradient of urbanisation (Supplement 1).We set 5-15 activity traps in most of the study ponds, the number on each pond's perimeter length (Liao et al., 2020).Traps were operated for 48 h.We determined the sex of the specimens in the field, males having broadened pro-and mesotarsomeres with suckers beneath while females do not have such broadened tarsomeres (Miller & Bergsten, 2014;Nilsson & Holmen, 1995).
In ponds with very biased sex in the beetles caught by the traps, we utilised a D-shape handnet with 500 μm mesh to sweep water to catch the three study species, to get equal or almost equal numbers of male and female specimens of a study species.We collected comparable numbers of specimens for both sexes, so that the sex could be included in the statistical models and to avoid the bias in results caused by the dimorphism of the species and the sex ratio being sometimes unequal in urban habitats (Emets, 1983).We recorded the presence or absence of fish by operating a fish trap in each pond during the sampling years, as fish predation can induce dispersal of aquatic organisms (Dahl & Greenberg, 1999) and intraspecific variation in species traits (Relyea & Werner, 2000).
For the study of A. canaliculatus, we collected 103 specimens from all the habitats (N = 21 males, N = 28 females in ponds with fish; N = 27 males and N = 27 females in ponds without fish).For the study of H. seminiger, we collected 98 specimens from all the habitats (N = 23 males, N = 23 females in ponds with fish; N = 25 males, N = 27 females in ponds without fish).For the study of I. ater, we collected 76 specimens only from ponds without fish (N = 41 males, N = 35 females), as the species appears to avoid habitats occupied by fish (Liao et al., 2020) and we were unable to collect adequate specimens from ponds with fish for this study.All specimens of the three species had normal flight muscles and no external parasites.
We measured seven morphological traits that are possibly related to the flight performance: the total length without head (TL-H), elytron length, pronotum length, total wing length, wing length I (the wing length from the base to the bending zone), wing length II (the wing length from the bending zone to the wing tip; Figure 1) and wing area.We excluded head length from the total length, that is, measuring the TL-H, because dytiscids' heads had different positions and easily caused bias in measurement.We measured the right elytron and the right hind wing of each specimen for the wing traits.The traits were measured in the laboratory with LEICA ® microscope M205C.We photographed the right hind wing of each specimen and measured the wing area with the software ImageJ (Schneider et al., 2012).
We utilised the percentage of unnatural impermeable surfaces, such as buildings and roads, as an indicator of the extent to which the surroundings of each pond might be considered 'urbanised' (Liao et al., 2020).Although Helsinki is a coastal city, the seawater does not hinder the dispersion of dytiscids in the same way as impermeable surfaces: First, water beetles detect water with polarised light (Schwind, 1991(Schwind, , 1995)).Impermeable surfaces in urban landscapes consist of deceptive features, such as glass, that polarise a much higher percentage of light than natural waterbodies and are ecological traps for aquatic insects (Horváth et al., 2009).The sea is not an ecological trap for dispersing dytiscids.Second, impermeable surfaces, such as buildings, are the main movement barriers in urban landscapes (New, 2015).Therefore, we included the sea as urban green-blue space and did not calculate the percentage of the sea separately.

Data analysis
We applied linear models (LMs) to analyse how each study trait was affected by urbanisation for the three species.The response variables are TL-H, elytron length, pronotum length, total wing length, wing length I, wing length II and wing area, which were continuous positive values.In data exploration, we noticed the sex of specimens and the presence or absence of fish affected some of the traits significantly (Tables 2 and 3); therefore, we included the sex of specimens for all three species, and the presence or absence of fish for A. canaliculatus and H. seminiger, as covariates in the models.In data exploration of A. canaliculatus and H. seminiger, we detected interactions between urbanisation and the presence or absence of fish; thus, we also included the interaction as a covariate in the LMs.In the full LMs of A. canaliculatus and H. seminiger traits, the covariates include urbanisation (i.e., the percentage of impermeable surfaces of a pond's surroundings), the sex of the specimens, the presence or absence of fish, and the interaction between urbanisation and the presence or absence of fish (Appendix 1).In the data analysis of I. ater, the covariates include urbanisation and the sex of specimens; we detected no interaction between urbanisation and sex (Appendix 2).To investigate (1) if and how each trait responds to urbanisation and (2) at which scales (some of) the traits respond to urbanisation, we modelled the responses of each trait to urbanisation.To avoid Type II error caused by collinearity between covariates (Zuur et al., 2010), the collinearity between urbanisation at different radii in this study, we modelled the response variables against urbanisation at different scales in separate models (Booth et al., 1994;Neter et al., 1996;Zuur et al., 2009; Appendices 1 and 2).We detected little effect of the sampling year on the measured species traits, and the sampling year was excluded from further models.
To validate a model, we simulated fitted values, to ensure that the model does not produce non-positive values, obviating the need to run generalised LMs with Gamma distributions; we plotted residuals against each covariate to check whether there was any pattern between the residuals and the covariates, and to ensure there was no need to run more complex models (Zuur & Ieno, 2016) robustness of the results, we rescaled the seven traits with Z-score normalisation and applied principal component (PC) analysis, computing the PCs with the 'vegan' package (Oksanen et al., 2022).We used the resulting PCs as response variables and modelled if PCs responded to urbanisation at the study scales, to avoid type I errors caused by multicollinearity.To validate the results, we also calculated Pearson's r correlation in the meta-analysis, which measures the degree of linear relationship between two quantitative variables (Sullivan & Feinn, 2012) We conducted all data analysis with statistical software R version 4.0.3(R Core Team, 2020).

Traits of Acilius canaliculatus changed along an urban gradient in fishless habitats
The pronotum length, elytron length, total wing length and wing length I of male A. canaliculatus differed significantly from females (Table 2, Supplements 2 and 3).Individuals from ponds without fish had significantly larger values in all traits except pronotum length than individuals from ponds with fish (Table 3).
In ponds without fish, two traits responded to urbanisation: The TL-H of A. canaliculatus responded negatively to urbanisation with radii from 400 to 1200 m ( p < 0.05), with the strongest effect at the scale of 1000 m (Figure 2a,b, Supplement 3).The size effects were small but not negligible (À0.45 ≤ Pearson's r correlation ≤ À0.26; Table S8.1).Pronotum length was positively affected by urbanisation with buffer radii from 100 to 300 m ( p < 0.05), but the size effects were trivial (Pearson's r correlation ≤ 0.14; Table S8.1).The Wing Length II was positively affected by urbanisation with buffer radii from 100 to 1200 m ( p < 0.05), with the strongest effect at the scale of 300 m (Figure 2a,b, Supplement 3).The size effects were between small and moderate (0.25 ≤ Pearson's r correlation ≤ 0.44; Table S8.1).In ponds with fish, no traits responded to urbanisation (Figure 2c,d, Supplement 4); the size effects were mostly trivial (Pearson's r correlation < 0.2; Table S8.2).
The responses of the computed PCs to urbanisation confirm the robustness of the results (Tables S11.1 and S11.2):In ponds without fish, PC2 was positively correlated with urbanisation from the scale of 100 to 800 m, whereas PC3 was negatively correlated with urbanisation from the scale of 200 to 1200 m (Figure 3a,b).In ponds with fish, the PCs were not correlated with urbanisation (Figure 3c,d).
T A B L E 3 Mean ± SD of the studied morphological traits of the three species in ponds with and without fish.

Aclius canaliculatus Hydaticus seminiger Ilybius ater
Fish Fishless Fish Fishless Fish Fishless T A B L E 2 Mean ± SD of the studied morphological traits of the three species between the two sexes.
The estimated parameters and p-values of urbanisation in the linear models regarding how dytiscid morphological traits respond to urbanisation at nine different scales; a trait is modelled at each scale in a separate model: (a, b) Acilius canaliculatus in ponds without fish, (c, d) Acilius canaliculatus in ponds with fish, (e, f ) Hydaticus seminiger in ponds without fish, (g, h) Hydaticus seminiger in ponds with fish, (i, j) Ilybius ater in ponds without fish.'TL-H' stands for 'total length without head'.The estimated parameters mean the slope of linear models.The error bars in (a), (c), (e), (g) and (i) represent standard error.The red lines in (b), (d), (f), (h) and (j) represent p-value equal to 0.05.When an estimated parameter is negative, the corresponding trait responds negatively to urbanisation, that is, the value of the trait decreases with the increasing percentage of impermeable surfaces within the buffer; and vice versa.For example, in (a) the body length of A. canaliculatus in ponds without fish decreases with increasing urbanisation, that is, the smaller specimens are found in highly urbanised ponds while bigger specimens are found in less urbanised ponds.

Traits of Hydaticus seminiger changed along an urban gradient mainly in fishless habitats
Male H.seminiger were slightly larger than females in several flightrelated measures (TL-H, pronotum length, and elytron length; Table 2, Supplement 4).There were no significant differences between individuals from ponds with fish or without fish (Table 3, Supplement 5).
In ponds without fish, urbanisation had negative effects on the study traits mainly from the scale of 100 to 500 m (Figure 2e,f):   S9.2).Other traits of H. seminiger did not respond to urbanisation ( p > 0.05; Figure 2g,h; Supplement 6).
The responses of the computed PCs to urbanisation confirm the robustness of the results (Tables S12.1 and S12.2):In ponds without fish, PC1 had a significantly positive correlation with urbanisation from the scale of 100 to 500 m (Figure 3e,f).In ponds with fish, the PCs were not correlated with urbanisation (Figure 3g,h).

Traits of Ilybius ater had little change along an urban gradient
Female I. ater were generally longer than males, but mostly without significance (Table 2, Supplement 2).There was little evidence of variation in the measured traits along the study urban gradient (Figure 2i,j), except that elytron length had a positive correlation with the urban gradient at the scale of 100 m (estimated parameter = 0.06 ± 0.03, p = 0.0335; Figure 2i,j; Supplement 7).
The size effects of elytron length against urbanisation were small (À0.21 ≤ Pearson's r correlation ≤ À0.18; Table S10).The responses of the computed PCs to urbanisation confirm the robustness of the results (Tables S13.1 and S13.2):Only PC5 was significantly correlated with urbanisation at the scale of 100-300 m (Figure 3i,j).

DISCUSSION
In

Not all species respond to urbanisation
Urbanisation can drive intraspecific variation in flight-related morphological traits of aquatic insects.In our study, we found many traits of A. canaliculatus and H. seminiger have significant changes along the urban gradient (Figure 2a,b,e-h).We think this is because these dytiscids need to disperse for breeding and/or overwintering.
A. canaliculatus and H. seminiger may breed in temporary waterbodies (Carr, 1988;Galewski, 1990;Liao, 2017).The emerged adults, especially those that emerge from temporary waters, need to disperse and seek habitats for overwintering.As urbanisation creates movement barriers, individuals with better flight capacity are favoured in urban landscapes (Piano et al., 2017).For better fitness, organisms adjust their phenotypic plasticity to adapt to local conditions (Beldade et al., 2011;Sinclair et al., 2012).Urban habitats are suffering from the urban heat island (UHI) effect, that is, significantly higher temperature in urban areas than in rural surroundings, to different extents, which depends on many spatio-temporal factors, such as building density and season (Deilami et al., 2018;Grimmond, 2007;Li et al., 2020).
Inhomogeneity of urban surfaces creates microscale and local scale advection, and highly urbanised areas have serious aerodynamic isolation of urban street canyons, unlike in rural areas (Ampatzidis & Kershaw, 2020;Oke et al., 2017).The aerodynamic isolation may negatively affect the dispersal of insects.Thus, urbanisation may have driven the intraspecific variation of flight-related traits in dytiscids to select for better flight performance in urban landscapes.
Although urbanisation drives intraspecific variation in flightrelated morphological traits, not all species are affected.In our study, I. ater exhibited no obvious responses to urbanisation (Figure 2i,j).Our result is in line with other studies showing that urbanisation affects the intraspecific variations in some but not all species within a family (e.g., Papp et al., 2020;Tommasi et al., 2022;Tüzün et al., 2017;Villalobos-Jiménez & Hassall, 2019).For example, in ground beetles (Carabidae), urban individuals of Carabus violaceus have shorter elytron length than their non-urban counterparts, whereas urbanisation has no effects on the elytron length of Abax parallelepipedus (Papp et al., 2020).In damselflies (Zygoptera), a significant difference was found in dispersal-related traits of Coenagrion puella between urban and rural males (Tüzün et al., 2017) but not in Ischnura elegans in the same family Coenagrionidae (Villalobos-Jiménez & Hassall, 2019).In our case, we think no responses in the flight-related traits of I. ater may have resulted from the fact that it has long hind wings with a large area, compared with the other two species with similar body length (Table 2).The large hind wing helps the species to be preadapted to urbanisation.Such interspecific variation may have resulted in different responses to urbanisation, and urbanisation may not affect those species with traits good for dispersal.

Not all traits respond to urbanisation
Urban organisms have complex responses to cope with movement barriers caused by urbanisation, which reflects in different changes in traits in different species.In the specimens that we collected from ponds without fish, A. canaliculatus has decreasing TL-H but increasing wing length II along the urban gradient (Figure 2a,b).The body weight of beetles can affect flight performance (Chen et al., 2011;Song et al., 2021).Light body weight can decrease the load on hind wings (Fedorenko, 2015), and smaller insects, including beetles, have better flight endurance than large members of the same species, possibly a trade-off with fecundity (Brown et al., 2017;Dillon & Dudley, 2004;Lehmann, 2002).In our study, individuals of A. canaliculatus in ponds with fish are on average smaller than individuals in ponds without fish (Table 3), which may be trade-offs with not only fecundity but also the production of anti-predator chemicals (Dettner, 2014(Dettner, , 2019)).Previous research shows that A. canaliculatus is more tolerant to urbanisation than most dytiscid species (Liao et al., 2020).Having smaller individuals with little change in hind wing traits is an advantage for the species to have good flight performance during dispersal in urban landscapes.
As different dytiscid species have different shapes for swimming and diving hydrodynamics, which is a common phenomenon among aquatic insects (Nachtigall, 1974;Nachtigall & Bilo, 1975), species need to adjust different traits to deal with aerodynamic isolation in densely built urban areas.In our study, H. seminiger shows changes in different traits along the urban gradient different from A. canaliculatus: it has significant changes in traits of both elytra and hind wings along the same urban gradient (Figure 2e,f).Hind wings provide beetles with aerodynamic force in flight (Song et al., 2021;Sun et al., 2016), whereas elytra play an important role in stabilising flight (Souza & Alexander, 1997).The interactions between elytra and hind wings help beetles support their weight by improving vertical force (Burton & Sandeman, 1961;Johansson et al., 2012;Le et al., 2013;Dahmani & Sohn, 2018).Such complex interactions of aerodynamics between the two pairs of wings are also found in other aquatic insects, such as dragonflies (Hu & Deng, 2014;Usherwood & Lehmann, 2008).The changes in the measured traits of H. seminiger are complicated and may have resulted in improved flight performance of urban individuals.How exactly the changes in an array of wing traits have affected the flight performance of urban aquatic fauna could be investigated with simulation studies and/or with the assistance of bioinspired robots.
Although sexual dimorphism is common in insects, including diving beetles (Bergsten et al., 2001;Bilton et al., 2008Bilton et al., , 2016;;Kiyokawa & Ikeda, 2022), we did not find different responses to urbanisation between males and females in the study species.Our result is in accordance with the findings of a damselfly study on I. elegans that exhibit strong sexual dimorphic differences but no different response to urbanisation in flight-related traits between sex (Villalobos-Jiménez & Hassall, 2019).However, other studies have found different responses to urbanisation occurring in the same traits between sex in some species of odonates (Tüzün et al., 2017) and carabids (Papp et al., 2020;Sukhodolskaya, 2013;Sukhodolskaya & Saveliev, 2017).Such different responses to urbanisation between species or higher taxon levels may have resulted from the biology and ecology of different taxa.Invertebrates exhibit different patterns in the sexual size dimorphism across taxa during development (Blanckenhorn et al., 2007), and such inconsistent patterns of invertebrates are found in their responses to other geographic factors, such as latitude and altitude (Chown & Gaston, 2010;Horne et al., 2019;Moran et al., 2016;Shelomi, 2012).We assume urbanisation creates as complex environmental changes as non-anthropogenic geographic gradients that can affect morphological plasticity of urban organisms.

Species have limited morphological plasticity to cope with urbanisation
Species can adjust morphological plasticity for better flight performance to cope with urbanisation, but their capacity is limited.In our study, among the specimens collected in ponds without fish, the body length (TL-H) of A. canaliculatus responds to urbanisation from the buffer zone of 400 to 1200 m and wing length II from 100 to 1200 m (Figure 2a,b), whereas H. seminiger responds to urbanisation in most traits at the scales from 100 to 500 m (Figure 2e,f).We think the phenomena may have resulted from the interactions between different flight capacity of species and their living environments.Although A. canaliculatus and H. seminiger are both considered good fliers (Nilsson & Holmen, 1995), the former species appears to be a better flier than the latter (Liao et al., 2020), as A. canaliculatus has a high flight frequency from spring to autumn (Iversen et al., 2017) and has been recorded by flight traps more frequently than H. seminiger (Lundkvist et al., 2002).The phenomena indicate that even good flier species have limited ability to endure the movement barriers caused by urbanisation.As the landscape scale can affect the inference of species-landscape relationships (Jackson & Fahrig, 2015), urban planning should consider scales when making conservation plans for aquatic biodiversity conservation, especially when focussing on protecting certain species.It is important to retain or even create ponds within the optimal scales to promote the dispersal and establishment success of target species in conservation.
One limitation of this study is that we do not have genetic data of the specimens to explain the phenomena found in this paper in an eco-evolutionary context.The founder effects may have played a role in the morphological variation of flight-related traits in this study.
Some individuals disperse to urban ponds, which have the capacity to support aquatic invertebrates (Liao et al., 2020) but are more isolated habitats compared with wetlands in forested landscapes.The offspring of such individuals having dispersed to and established in urban habitats may experience isolation due to the movement barriers in pond surroundings, which can result in loss of genetic variability (Barton & Charlesworth, 1984;Mayr, 1999;Templeton, 1980).Longterm data of morphological traits and genetic data are needed to explain the morphological variation we found in an eco-evolutionary context in future studies.
Another limitation of this study is that we did not consider whether the area of urban ponds affects the landscape scale(s) at which dytiscids can cope with urbanisation.Unlike terrestrial organisms, dytiscids and other aquatic/semi-aquatic insects live in urban blue space, that is, urban surface waters, which has stronger cooling effects on their vicinity than urban green space (Kleerekoper et al., 2012;Tan et al., 2021).Urban ponds with large water surfaces have stronger cooling effects than their small counterparts (Syafii et al., 2017).Temperature is a determinant of the reproduction and development of aquatic insects, including dytiscids (Aiken, 1986;Gillooly & Dodson, 2000;Inoda et al., 2007;Kiyokawa & Ikeda, 2022).
The cooling effects of urban ponds may mitigate the aerodynamic isolation of urban street canyons (Syafii et al., 2017).Thus, the area of ponds may have effects on the physical environment and the fitness of aquatic insects, both of which determine the stability of populations of urban organisms.

The presence of predators may have complex effects with urbanisation on intraspecific variation
Although this study focuses on urban effects on intraspecific variation of flight-related traits in aquatic insects, the impact of predation on intraspecific variation cannot be ignored in urban organisms.
In our study, we found the presence or absence of fish has led to different responses to urbanisation within the same species.
A. canaliculatus has decreasing body length and increasing wing length II along the urban gradient in ponds without fish (Figure 2a,b), but it has no response to urbanisation in ponds with fish (Figure 2c,d).Similarly, H. seminiger in ponds without fish shows negative responses to urbanisation, but only the wing area responds to urbanisation in ponds with fish (Figure 2e-h).The presence of predators is known to affect the morphological plasticity of many aquatic organisms, such as cladocerans (e.g., Black, 1993), snails (e.g., Brönmark et al., 2011) and frog tadpoles (e.g., Relyea & Werner, 2000), which is an anti-predator strategy.Although aquatic fauna can select habitats with low predation risk, such habitats may involve other environmental stressors.For example, in our case, ponds without fish are sometimes associated with smaller waterbodies (Table 1) and are prone to drought due to the UHI effect and recent climate change, which potentially expose aquatic species to unfavourable conditions.Thus, urban aquatic This study utilises the percentage of impermeable surfaces as an indicator of urbanisation to investigate urban effects on intraspecific variation, instead of categories, such as 'urban', 'suburban' and 'rural'.
Yet, we did not have equal observations of each species along the study urban gradient.An advantage of this study is that we have investigated urban effects on species traits at different landscape scales and found that different species respond to urbanisation in the studied flight-related traits at different scales, which is similar to urban effects on aquatic organisms at the community level (Gledhill et al., 2008).Although we have studied how urbanisation affects flight-related traits of diving beetles in this study, we did not investigate the 3D structure of their elytra and hind wings, nor the aerodynamics when dytiscids are at flight.We call for cooperation among entomologists, ecologists and biomechanists in future studies for biodiversity conservation in the changing world.
where the trait n of specimen m collected in the pond l, Y lmn , follows a normal distribution.The modelled traits were total length without head (TL-H), elytron length, pronotum length, total wing length, wing length I, wing length II and wing area.The term Y lmn is modelled in terms of urbanisation (Urban l ), and the sex of the specimen (Sex lm ).
The term Urban l is the value of urbanisation, that is, the percentage of impermeable surfaces, of the pond l; we modelled in terms of urbanisation at nine different buffers, that is 100, 200, 300, 400, 500, 600, 800, 1000 and 1200 m, separately to explore urban effects on the flight-related morphological traits of Ilybius ater at different landscape scales.The term α 2n is the intercept for the LM modelling the trait n, whereas β 5n and β 6n are the parameters of covariates.The term ε lmn is unexplained information.
The estimated parameters and p-values of urbanisation in the linear models regarding how PCs respond to urbanisation at nine different scales: (a, b) Acilius canaliculatus in ponds without fish, (c, d) Acilius canaliculatus in ponds with fish, (e, f) Hydaticus seminiger in ponds without fish, (g, h) Hydaticus seminiger in ponds with fish, (i, j) Ilybius ater in ponds without fish.The TL-H, pronotum length, total wing length, wing length I and wing length II responded to urbanisation from the scale of 100 to 500 m; the size effects were small to moderate (À0.54 ≤ Pearson's r correlation ≤ À0.22; TableS9.1).Wing length I had the strongest response to urbanisation at the scale of 300 m.Other traits have the strongest responses at multiple scales from 100 to 500 m (Supplement 4).Wing area responded to urbanisation from the scales of 100 to 1200 m, with the strongest effect at the scale of 500 m (Supplement 4).In ponds with fish, only wing area of H. seminiger responded to urbanisation, from the scale of 100 to 1200 m ( p < 0.05, Figure2g,h; Supplement 5); the size effects were small (0.26 ≤ Pearson's r correlation ≤0.37;Table this study, we investigated how and at what scale urbanisation affects intraspecific variation in flight-related morphological traits in three diving beetle species.A. canaliculatus, H. seminiger, and I. ater exhibit different patterns: TL-H of A. canaliculatus decreases and wing length II increases along the urban gradient in ponds without fish (Figure 2a,b) but not in ponds with fish (Figure 2c,d).The studied traits of H. seminiger are negatively affected by urbanisation in ponds without fish, mostly from the scales of 100 to 500 m (Figure 2e,f); only wing area increases along the urban gradient in ponds with fish (Figure 2g,h).I. ater from ponds without fish is not affected by urbanisation in the studied traits (Figure 2i,j).The studied traits were affected by urbanisation at different landscape scales.
insects may face complex trade-offs among fecundity, flight performance and anti-predator defences in their habitat selection.It would be interesting to investigate whether such trade-offs have led to the evolution of urban fauna in future studies.CONCLUSIONThis research is one of the first studies that have investigated how urbanisation affects inter-and intraspecific variation of flight-related morphological traits in aquatic invertebrates.Our results show that dytiscids exhibit responses to urbanisation, but not in all species, not in all traits and not at all scales.The phenomena have resulted from the biology and ecology of species and their interactions with complex environmental changes caused by urbanisation, such as the UHI effect.Our study highlights the importance of landscape scales in aquatic biodiversity conservation.Urban planning should consider scales and retain or even create ponds within the optimal scales when making conservation plans for especially targeted species.Future research should pay attention to whether habitat features, such as the area of ponds, affect morphological plasticity of aquatic organisms, to generate research-based recommendations for urban blue space design, to facilitate the dispersal of native species at larger scales.
urban effects on the flight-related morphological traits of dytiscids at different landscape scales.The term α 1k is the intercept for the LM modelling the trait k, while β 1k , β 2k , β 3k and β 4k are the parameters of covariates.The term ε ijk is unexplained information.The traits of Hydaticus seminiger were modelled with the same equation described above at the nine different landscape scales.APP E NDIX 2 : Full model description of LM of studied traits of Ilybius ater: . To test the Measured traits of a dytiscid specimen.'PL' stands for 'pronotum length'.The hind wing area was estimated by tracking the edge of a hind wing.
T A B L E 1 Basic information of habitats where specimens were collected.Note: Urbanisation is based on the percentage of impermeable surfaces in a buffer zone of 500 m.Abbreviation: SD, standard deviation.