Dragonflies and damselflies (Odonata) in urban ecosystems: A review

The expansion of urban areas is one of the most signifi cant anthropogenic impacts on the natural landscape. Due to their sensitivity to stressors in both aquatic and terrestrial habitats, dragonfl ies and damselfl ies (the Odonata) may provide insights into the effects of urbanisation on biodiversity. However, while knowledge about the impacts of urbanisation on odonates is growing, there has not been a comprehensive review of this body of literature until now. This is the fi rst systematic literature review conducted to evaluate both the quantity and topics of research conducted on odonates in urban ecosystems. From this research, 79 peer-reviewed papers were identifi ed, the vast majority (89.87%) of which related to studies of changing patterns of biodiversity in urban odonate communities. From the papers regarding biodiversity changes, 31 were performed in an urban-rural gradient and 21 of these reported lower diversity towards built up city cores. Twelve of the cases of biodiversity loss were directly related to the concentrations of pollutants in the water. Other studies found higher concentrations of pollutants in odonates from built-up catchments and suggested that odonates such as Aeshna juncea and Platycnemis pennipes may be candidate indicators for particular contaminants. We conclude by identifying current research needs, which include the need for more studies regarding behavioural ecology and life-history traits in response to urbanisation, and a need to investigate the mechanisms behind diversity trends beyond pollution.

gate current research on the impacts of urbanisation using odonates as a model group.First, we will present a brief overview on the impacts of urbanisation on terrestrial and freshwater ecosystems, then we conduct a systematic literature review regarding odonates in urban environments and outline the plausible effects on odonates, followed by the use of odonates as bioindicators, the conservation value of urban wetlands for odonates, a summary of evolutionary strategies for species to cope with urban stressors, and fi nally we conclude by identifying research needs.

THE EFFECTS OF URBAN STRESSORS ON TERRESTRIAL AND FRESHWATER ECOSYSTEMS
The wide range of impacts associated with urbanisation indicates that it exerts a considerable effect on terrestrial biodiversity.Most studies show that species richness and evenness is reduced in highly urbanised regions, depending on the taxonomic group observed, degree of urbanisation, and spatial scale of analysis (McKinney, 2002(McKinney, , 2008)).For example, butterfl ies (Ruszczyk, 1987;Ruszczyk & De Araujo, 1992;Blair & Launer, 1997) and ground arthropods (McIntyre et al., 2001) show this tendency.The general pattern of biodiversity decrease in cities is probably due to the fact that over 80% of land in city cores is covered by buildings and pavements, with the remain-

INTRODUCTION
Urbanisation is one of the main drivers of ecosystem change.The expansion of urban areas is a main cause of many environmental stressors: (i) removal of native vegetation leading to degradation and fragmentation of the natural landscape, (ii) modifi cation of hydrological systems through water extraction for human use, (iii) increase of local minimum temperature creating a "heat island" effect, and (iv) accumulation of pollutants to a point at which the environment cannot fully process and degrade them, causing eutrophication and affecting biogeochemical cycles (Grimm et al., 2008;McDonald, 2008).Moreover, the estimated percentage of the world population living in urban areas now exceeds 50% (Grimm et al., 2008) and almost 60% will be living in urban areas by the year 2030 (Seto et al., 2012;Güneralp & Seto, 2013).Given the range of impacts and rapid growth of urban development, it is essential that we understand in depth the implications for natural systems both in and around cities. Dragonfl ies and damselfl ies (the Odonata) have been suggested as barometers for environmental change due to their sensitivity to other anthropogenic stressors such as climate change (Hassall, 2015) and variation in habitat quality (Clark & Samways, 1996).However, the impacts of urbanisation on odonates remain poorly understood.In this review, we will investi-sensitivity to a range of urban stressors means that odonate communities have the potential to act as barometers of environmental change in urban areas, as has been demonstrated successfully in South Africa for habitat quality (Clark & Samways, 1996) and as has been suggested for climate change (Hassall, 2015).In this review we outline the stressors associated with urban environments that have the potential to affect odonates considerably across their life cycle.

DRAGONFLIES AND DAMSELFLIES AS A MODEL SYSTEM
Dragonfl ies and damselfl ies (the Odonata) are highly suitable model organisms for the study of urban ecosystems because (a) they are sensitive to different stressors, such as pollutants (Ferreras-Romero et al., 2009) and temperature changes (Hassall & Thompson, 2008), therefore can be a powerful tool to assess the general conditions of the city environment; (b) they are aquatic as larvae and terrestrial as adults, hence can be used as bioindicators in both aquatic and terrestrial habitats (Oertli, 2008); (c) they have an important role as predators, hence have a wide range of interactions with different organisms in both aquatic and terrestrial ecosystems (Knight et al., 2005); (d) they are ideal for studying movement through the landscape, as their adult stage exhibits a high dispersal ability and are very conspicuous (Conrad et al., 1999), and (e) their biodiversity, ecology and evolutionary biology is well-documented, providing a robust foundation for drawing general conclusions (Córdoba-Aguilar, 2008).

LITERATURE REVIEW
We conducted a literature review of the prior work regarding urbanisation and odonates, in order to evaluate the potential of odonates as a model system for studying urbanisation.During January-April 2014, we used standardised search terms with three online databases: Web of Science, Scopus and ScienceDirect.The words "urban" and "Odonata OR dragonfl * OR damselfl *" were searched in title, keywords and abstract (except in Web of Science, where the search was in Topic).Only peer-reviewed papers including odonates in an urban environment were selected, reviews were excluded.According to the nature of the study, each paper was classifi ed into one, two or three of the following categories: diversity (variables considered here were taxa richness, abundance, probability of occurrence for creating habitat suitability models, diversity indices such as Shannon-Wiener, etc.), toxicology (such as physicochemical features of the water bodies and heavy metal concentration), behavioural ecology (dispersal, trophic dynamics), and life-history traits (reproduction, life growth).We went on to examine the references from the papers found in the standardised search and also added those to the analysis where appropriate.
The search returned a total of 79 papers, most of which (65.82%) were published after 2004 (see Fig. 1, and Table S1 for a complete list of papers with categories).We were also interested in analysing geographical patterns in re-ing area used for lawns, trees, and shrubs (Blair & Launer, 1997), thus fragmenting the landscape and homogenising vegetation composition (Blair & Launer, 1997;Faeth et al., 2011;Faeth et al., 2012;McKinney, 2002).However, plant biodiversity peaks at intermediate levels of urbanisation, perhaps due to the high numbers of species associated with suburban gardens (McKinney, 2008).
Urban freshwaters have not been explored as much as terrestrial urban ecosystems, despite the high degree of vulnerability that freshwaters exhibit to the pressures of urbanisation (Paul & Meyer, 2001;Dudgeon et al., 2006).Urban streams, for example, are usually modifi ed to carry out storm water into natural streams, washing away industrial and human wastes, as well as road runoff (Forman, 2014).Streams are subjected to toxins, temperature change, siltation, organic pollutants, and the replacement of riparian vegetation and substrate for rocks or concrete and, consequently, insects, molluscs, crustaceans, and annelids tend to have decreased species richness and abundance (Paul & Meyer, 2001;Forman, 2008Forman, , 2014;;Vaughan & Ormerod, 2012).The presence of organic pollution has been associated with increased abundance of chironomids and oligochaetes, making them the dominant members of urban stream communities (Campbell, 1978;Seager & Abrahams, 1990;Paul & Meyer, 2001).
Ponds are very frequent in parks and gardens for aesthetic purposes and to improve human well-being (Forman, 2014;Hassall, 2014), and are greatly infl uenced by a variety of factors such as the land use of the surroundings, the runoff to the pond, subsurface groundwater fl ow, and shoreline conditions (Forman, 2014).Parks and gardens are heavily managed and fertilisers are commonly used, and due to their limited capacity for the processing of pollutants and nutrients, ponds often suffer from eutrophication with an accompanying decline in biodiversity (Forman, 2014).However, the relative isolation of urban ponds generates a substantial degree of habitat heterogeneity.Each pond possesses its own physical, chemical, and hydrological features, facilitating the presence of a range of species of specialised habitat requirements, hence increasing the differentiation among these ecosystems and the landscape diversity (β and γ diversity, respectively) (De Meester et al., 2005).They also work as "stepping stones", thus facilitating landscape connectivity and potentially acting as metapopulations (De Meester et al., 2005;Hassall, 2014).
The biodiversity of odonates in urban areas is expected to be different from their rural counterparts, as is the case for other taxa.However, as already discussed, urbanisation is a complex process that encompasses a wide range of selective pressures, and each of them is expected to have a different impact on odonates according to the species and the life stages.Sensitive odonate species may be severely affected by stressors such as contaminants from sewage input, but generalist species are expected to tolerate these conditions, leading to decreased species richness.Moreover, generalist species may exploit the resources available due to decreased interspecifi c competition, resulting in high abundance of tolerant species.This variation in search regarding odonates in urban ecosystems and found most studies were conducted in United States and Brazil, though plenty were from Austria, South Africa, Germany, Japan, and others (see Fig. 2).
Of the 79 papers found, 71 (89.87%) investigated patterns in biodiversity, of which 31 were executed in an urban-rural gradient or compared diversity in streams before and after the water fl ow reached an urbanised area.Of the studies regarding biodiversity in an urban-rural gradient, 21 reported less diversity in cities, and 12 cases of diversity loss were related to pollutant concentration in the water (see Table 1).Below, we discuss the general trend within these papers and give example studies to illustrate those trends; a comprehensive list of studies and summaries of fi ndings can be found in the Supplementary Information (Table S1).

The effects of urban stressors on odonates
Our results suggest that increasing urbanisation seems to affect odonate diversity negatively, on the whole, as has been shown in other groups (McKinney, 2008).However, urbanisation is a multi-faceted problem and the studies reviewed herein offer some perspectives on which aspects of urbanisation are having the greatest impacts on Odonata across their life cycle and how this may refl ect the biodiversity loss in cities (see Fig. 3).It is worth mentioning that the stressors listed may not only have a considerable impact in a given stage of their life cycle, but the effects can also be transferred to other stages via carry-over effects (e.g.maternal effects from adult to larva, or persistence of impacts between life history stages from larva to adult), except when a decoupling mechanism interferes (for a detailed review see Stoks & Córdoba-Aguilar, 2012).

Fragmentation
Fragmented landscapes with little vegetation cover and few water bodies may not provide suffi cient corridors to facilitate dispersal, thus limiting odonate connectivity within cities (Chovanec et al., 2000;Watts et al., 2004;Sato et al., 2008).Urban landscapes were found to function as barriers for Paracercion calamorum, Ischnura senegalensis, and I. asiatica in Japan when the genetic differentiation among populations was analysed (Sato et al., 2008), showing that the effect of fragmentation in urban areas is consistent within at least three zygopteran species.These results are also consistent between methods of dispersal analysis: Watts et al. ( 2004) provide evidence from genetic techniques and a mark-release-recapture study that urban areas cause a strong negative effect in the dispersal of Coenagrion mercuriale in UK.Although we did not fi nd any studies using odonates linking habitat selection and urban landscapes, fragmentation may also affect habitat selection negatively by constraining access to optimal sites, as has been shown in a study using two model species representing territorial animals (van Langevelde, 2015).

Vegetation removal and/or modifi cation
Increased vegetation cover and biodiversity of plants was associated with increased odonate richness and overall evenness in Austria (Chovanec et al., 2002), Germany   ( Goertzen & Suhling, 2013), France (Jeanmougin et al., 2014), and South Africa (Samways & Steytler, 1996;Pryke & Samways, 2009), among others.For example, Goertzen & Suhling (2013) studied odonate diversity patterns in ponds across an urban-rural gradient using a multivariate approach and found that vegetation was not only one of the major drivers of alpha diversity, but also trampling vegetation had a signifi cant negative effect.Other studies have reported similar results, but suggested percentage cover of submerged macrophytes was the main component of vegetation affecting odonate diversity (Jeanmougin et al., 2014).Additionally, indigenous plant species have also been highlighted as a key factor shaping odonate communities in urban wetlands, and leaving a strip of indigenous riparian vegetation of at least 20 m between the stream edge and commercial plantations seems to facilitate the presence of sensitive odonate species such as Chlorolestes tessellatus in South Africa (Samways & Steytler, 1996).
The strong management of vegetation in terrestrial ecosystems in cities has a considerable impact on the adult phase of odonates (Silva et al., 2010) because vegetation is one of the main cues used for habitat selection and it infl uences a vast range of behaviours such as basking, foraging, roosting, sheltering, among others (Buchwald, 1992).Removal or modifi cation of aquatic vegetation from urban freshwater habitats also might have a considerable effect on the behaviour of odonates not only on the adult phase, but throughout their life cycle.Adults use aquatic vegetation as an oviposition substrate (Buchwald, 1992), often by inserting eggs into submerged plant stems (Corbet, 1999).Once the eggs hatch, the larvae use the submerged plant stems for perching, hiding from predators, and for sit-andwait ambush hunting (Buchwald, 1992).During emergence, the larva climbs upwards out of the water along vertical plant stems and proceeds to full metamorphosis once it leaves the water.The composition of plant communities may also exclude some specialist species that rely on particular plants, such as Aeshna viridis that only oviposits on Stratiotes aloides (Dijkstra, 2006).However, it is worth mentioning that sensitivity to aquatic vegetation loss may vary across odonates, since not all species require aquatic vegetation for oviposition.

Sewage discharge and stormwater runoff
Sewage is formed primarily of domestic, commercial, and industrial waste.Wastewater is collected through the sewage system and treated before releasing into freshwater.However, sewer overfl ows are not uncommon, and some countries lack these facilities or treat wastewater only partially (Rosa & Clasen, 2010) and in many occasions, treatment is not enough to remove all the contaminants (Paul & Meyer, 2001).
Several studies found in our literature search examined the effects of sewage input or "urban pollutants" in general rather than particular contaminants.In these cases, diversity was negatively related to sewage input (Solimini et al., 1997;Henriques-de-Oliveira et al., 2007), although most of the papers from the search were focused on aquatic macroinvertebrates and the overall group of odonates as bioindicators rather than particular species.Urban pollution was associated with increased abundance and dominance of Libellulidae and other taxa (see Table 2), but decreased abundance of Gomphidae (Ferreras-Romero et al., 2009) and overall evenness (Henriques- de-Oliveira et al., 2007).However, it is worth mentioning that even though there are many tolerant species within Libellulidae and sensitive species in Gomphidae (Ferreras-Romero et al., 2009), the sensitivity or tolerance of a stressor depends on the particular species rather than a larger taxon, therefore caution must be taken when considering libellulids or gomphids as bioindicators.
Wastewater contains a wide variety of contaminants ranging from metals, organic and inorganic fertilisers and pesticides to polychlorinated biphenyls (PCBs) and polycyclic aromatic hydrocarbons (PAHs) (Paul & Meyer, 2001), forming a cocktail of pollutants which may affect odonate larvae -and other aquatic macroinvertebrates -in a number of ways.First, high contents of organic matter and fertilisers cause eutrophication, leading to algal and bacterial blooms and decreasing levels of dissolved oxygen in water (Forman, 2008).Second, organic and inorganic contaminants from both sewage discharge and stormwater runoff may be toxic for some species.For example, pesticide has been shown to increase mortality in odonate larvae and in sublethal concentrations, it also increases fl uctuating asymmetry (e.g.Chang et al., 2007).A study looking at the life history of Coenagrion puella under pesticide stress suggested that the sensitivity of life-history traits depends on the type of pesticide and the exposure time, being increased development time the only life-history trait which showed consistency across pesticide treatments in the long-term exposure experiment (atrazine, carbaryl and endosulfan in this case) (Campero et al., 2007).On the other hand, exposure of Ischnura elegans larvae to a pesticide (chlorpyrifos) decreased immune function signifi cantly, and adult heat stress caused a stronger decrease in immune function, indicating that temperature and pesticide stress interact across metamorphosis (Janssens et al., 2014).Moreover, urban populations of C. puella larvae showed increased activity levels when exposed to chlorpyrifos at 20° and 24°C while showing decreased food intake under pesticide stress at 24°C, compared to rural populations which showed decreased activity and food intake in all treatments (Tüzün et al., 2015).These results suggest that urban larvae populations are locally adapted to higher contaminant levels (Tüzün et al., 2015).These studies are particularly useful when analysing the impacts of urbanisation due to their multiple-stressor approach, which highlight the effect of interactions between urban stressors and contribute to a more integral understanding of urban ecosystems.Certain contaminants such as PCBs and heavy metals may bioaccumulate in tissues.There are 209 different types of PCBs, and the mid-chlorinated congeners (e.g.PCB-153 and PCB-138) have been found in high concentrations among chironomids and dragonfl ies in an urban riparian ecosystem in Beijing (Yu et al., 2013).Similar results have been found with other organic persistent pollutants such as polybrominated diphenyl ethers (PBDEs) and hexachlorobenzene (HCB) in I. elegans larvae in several ponds across Flanders, Belgium (Van Praet et al., 2012).Heavy metals may come from both sewage input and road runoff, and they also tend to accumulate in the exoskeleton of odonate larvae (Meyer et al., 1986).Although lead and copper can cause deformities in other insects, e.g.Chironomus mentum (de Bisthoven et al., 1998), Aeshna juncea showed tolerance of high concentrations of manganese and nickel, while Platycnemis pennipes showed a sensitivity to cadmium, boron, and iron (Girgin et al., 2010), providing evidence of their use as suitable candidates for biomonitoring programmes.

ODONATES AS BIOINDICATORS IN URBAN ENVIRONMENTS
Aquatic biomonitoring programmes are aimed at evaluating water quality and identify water pollution in an early stage before it leads to decreased environmental health and ultimately affects human health and the security of public water supplies, which is particularly important in cities (Jones et al., 2010).Odonates have been used extensively as a bioindicator group of habitat quality (Clark & Samways, 1996;Sahlén & Ekestubbe, 2001;Foote & Rice Hornung, 2005;Subramanian et al., 2008;Clausnitzer et al., 2009;Dolný et al., 2011) and are one of the taxa included in the Biological Monitoring Working Party (BMWP) score system (Biological Monitoring Working Party, 1978).The BMWP score system has been used as a biological classification system for river pollution surveys in UK since 1980 (Paisley et al., 2014).The revised BMWP system includes Calopterygidae, Platycnemididae, Coenagrionidae, Cordulegastridae, Aeshnidae and Libellulidae (see Paisley et al., 2014 for the scores of each group).However, some species from these groups seem to be tolerant to urban stressors, such as Ischnura elegans, I. graellsii, Erythrodiplax fusca, among others (see Table 2).The presence of tolerant taxa in this system, as well as the absence of sensitive taxa such as Gomphus vulgatissimus (Brooks, 2004), might result in a misleading BMWP score.For instance, if in a given freshwater habitat there are abundant odonates, but 50% of the abundance includes tolerant species such as I. elegans, the result would be a high BMWP score which might not indicate water quality appropriately.Conversely, if there were abundant G. vulgatissimus, the BMWP score would be low, although the water quality in fact might be high.In order to reduce error using the current BMWP score system, we suggest that this system include sensitive species such as G. vulgatissimus and reduce the value of tolerant taxa.It has also been suggested that BMWP scores relying on Odonata might be infl uenced by shifting distributions under climate change (Hassall et al., 2010).

ECOLOGICAL TRAPS IN CITIES
Several studies have demonstrated that ecological traps can arise for dragonfl ies within urban areas, reducing the fi tness of urban populations (e.g.Horváth et al., 2007).The term "ecological trap" refers to situations in which unsuitable sites unable to sustain a population are preferred over the suitable sites or the unsuitable habitats mimic the cues that species use for selecting ideal habitats, leading species to choose unsuitable habitats over the optimal sites for roosting, feeding, and mostly reproducing (Donovan & Thompson, 2001;Schlaepfer et al., 2002;Robertson & Hutto, 2006).Ecological traps are often induced by anthropogenic modifi cations of the environment (Schlaepfer et al., 2002), hence urban locations could be used as model sites for studying ecological traps (Hale et al., 2015).Alongside anthropogenically-modifi ed water bodies that might be sub-optimal for odonates (Hale et al., 2015), human activity also produces a range of surfaces that refl ect polarised light to an equal or greater extent than water (Horváth et al., 1998).Since odonates (and other semi-aquatic insects) use polarised light as a cue for oviposition site selection, such surfaces represent important ecological traps.For example, it was found in Brazil that the refl ectance of cars imitates the refl ectance pattern of ponds, encouraging the presence of Pantala fl avescens in parking areas and even causing them to oviposit on the cars' surfaces, representing an important energy loss (Van de Koken et al., 2007).This behaviour has not only been observed in response to cars (Watson, 1992;Gunther, 2003;Wildermuth & Horváth, 2005;Blaho et al., 2014), but also to black plastic foil (Wildermuth, 1998), crude oil ponds (Horváth et al., 1998) and grave stones (Horváth et al., 2007).According to Kokko & Sutherland (2001), this behaviour could produce an Allee effect in low population densities.In high densities where there is increased competition, most individuals would compete for the preferred but unsuitable habitats, whereas the losing rivals would settle in the less preferred, high-quality habitat.However, in low population densities, the lack of competition will allow most individuals to make poor decisions on choosing proper habitats, therefore decreasing population density even more and eventually leading to extinction (Kokko & Sutherland, 2001;Schlaepfer et al., 2002).

THE CONSERVATION VALUE OF URBAN WATER BODIES FOR ODONATES
Although most studies showed a decrease in biodiversity as a result of urbanisation, there is evidence suggesting that urbanisation per se is not necessarily negative, but rather the negative impacts on biodiversity arise from highly intensive management associated with urban environments.Urban sites with a variety of vegetation composition, along with a proper management to minimise water pollution, were found to host more diverse communities of dragonfl ies and damselfl ies than those with limited vegetation diversity and increased water pollution (Colding et al., 2009;Goertzen & Suhling, 2013).For example, urban drainage systems in The Netherlands with low nutrient content and rich vegetation have been shown to have comparable macroinvertebrate diversity to rural drainage systems (Vermonden et al., 2009).In the Austrian Danube River fl oodplain system located within the city limits of Vienna, a water enhancement programme was implemented (which consisted of restructuring the embankments of an artifi cial island by creating shallow water areas, gravel banks, small permanent backwaters and temporary waters) and showed increased vegetation, odonate diversity and connectivity in the landscape in a long-term monitoring programme (e.g.Chovanec et al., 2000Chovanec et al., , 2002)).A study conducted in South Africa compared odonate diversity (and other invertebrate taxa) in natural and recovering forests and fynbos to alien pine plantations and botanical gardens rich in indigenous plants, and surprisingly the botanical gardens presented the highest species richness and abundance, especially compared with alien pine plantations (Pryke & Samways, 2009), providing evidence that botanical gardens represent a major refuge for invertebrate species.Parks in South Africa have shown high odonate diversity, whereas alien plantation forests where Eucalyptus sp. was the most abundant had the lowest odonate richness (Samways & Steytler, 1996).From a social perspective, increased biodiversity of odonates in urban green areas such as botanical gardens and parks is also important because they help attract tourists and increase awareness of the role of wetlands (Lemelin, 2007).
Additionally, urban ecosystems offer a vast diversity of pond types, ranging from ornamental ponds to drainage systems, each of these being subjected to different management plans (Hassall, 2014).This heterogeneity of pond types provides different hydrological and ecological conditions which may promote a higher β and γ-diversity of odonates, not only generalist species (Goertzen & Suhling, 2013).Likewise, ponds are abundant mostly in suburban areas in Central Europe (Willigalla & Fartmann, 2012), perhaps due to garden ponds.Garden ponds are also common in UK, between 2.5 and 3.5 million garden ponds are estimated in UK (Davies et al., 2009), providing abundant aquatic habitats which facilitate connectivity and promote metapopulations (Hassall, 2014).
In spite of generalist species being abundant in cities, there are some specialist species that can also fi nd refuge in urban habitats.For example, the threatened damselfl y Coenagrion ornatum has been reported to inhabit drainage systems (Harabiš & Dolný, 2015), and golf courses in Sweden have demonstrated to serve as a refuge for endangered species e.g.Leucorrhinia pectoralis, a dragonfl y considered "near threatened" in the Appendix II of the Bern Convention (Colding et al., 2009).
An interesting case is presented by Ischnura gemina.This species is endemic to the San Francisco Bay area, USA, which is highly urbanised.Even though I. gemina has been able to survive despite the stressors in the region (Garrison & Hafernik, 1981a, b), populations are decreasing and now it is threatened by urban development, in spite of repatriation efforts (Hannon & Hafernik, 2007).This species is currently under protection, which is predicted to benefi t the species under climate change scenarios as well (Sánchez-Guillén et al., 2014).This situation represents an important case study in urban invertebrate conservation not only for conservation biologists, but also for the regional authorities and urban planners with whom conservationist must work, since I. gemina and other odonates, as previously discussed, are helpful indicators of habitat quality and can work as "umbrella species" (Bried et al., 2007;Clausnitzer et al., 2009).This is particularly important in cities because by meeting the conditions required for odonates to survive, other species' requirements will be met as well (Bried et al., 2007), thus saving time and effort in developing biodiverse, sustainable cities.They also work as "fl agship species" for wetlands due to their attractiveness (Lemelin, 2007), which may help attract visitors to urban green areas.

ADAPTING TO URBANISATION
As we have mentioned throughout this review, some odonate species may be more tolerant of urban stressors.In order to unravel the specifi c traits driving the success of species in urban ecosystems, some studies have examined the role of the life history of odonates in light of the stressors caused by urbanisation.For example, a study in Italy observed the life history patterns of Erythromma lindenii and Ischnura elegans along an urban tract of a river and suggested the key factors to allow these species to cope with organic pollution is a longer reproductive period, absence of diapause, and tolerance of low oxygen concentration (Solimini et al., 1997).Studies performed in Japan demonstrated that the life cycle of Sympetrum striolatum imitoides was synchronised with the use of swimming pools; this species laid eggs in autumn, the larvae hatched in mid-winter and most adults emerged before the pools were drained and cleaned (Matsura et al., 1995;Matsura et al., 1998).However, the mechanisms underlying the life history of these species in both studies are uncertain, i.e. whether the species possessed pre-adaptations to urban habitats or exhibit plastic responses that allow them to persist despite urban stressors.
However, the Odonata represent one group within which common garden rearing experiments using multiple stressors have yielded considerable insights (Stoks et al., 2015).These studies allow the partitioning of genetic and environmental factors to investigate the relative contributions of plasticity and adaptation, and have suggested that local adaptation to temperature (Dinh Van et al., 2014;Arambourou & Stoks, 2015) and pesticide stress (Tüzün et al., 2015) is common in odonates.Hence, studying phenotypic variation in urban environments represents a great opportunity for researchers in ecology and evolution to observe the complex, underlying processes of how species adapt to new circumstances through the use of urban habitat patch networks as "natural experiments".

RESEARCH NEEDED
While general patterns of diversity are well described and some of the mechanisms underlying those patterns have been quantifi ed, a number of important areas of research are under-represented in the current literature.Below, we outline some of these key areas:

Trophic interactions
We found very few studies regarding trophic interactions of odonates in cities, in spite of being considered key predators in both aquatic and terrestrial ecosystems (Knight et al., 2005).Some studies using turtles (Martins et al., 2010) and bats (Srinivasulu & Srinivasulu, 2005;Kalcounis-Rueppell et al., 2007) encountered odonates as prey, in both urban and non-urban habitats.However, there is little information regarding how their role as predators changes in urban ecosystems.Possible intraguild predation has been suggested, but not observed in detail (Agüero-Pelegrín & Ferreras-Romero, 1994;Sacha, 2011).Other than that, it is commonly known that they feed on chironomids (Matsura et al., 1998).Thus, there is no comprehensive knowledge regarding the diet of dragonfl ies and damselfl ies in cities.Given the fact that aquatic macroinvertebrate communities (i.e.prey) differ in cities from their rural counterparts (Azrina et al., 2006), there might be a signifi cant impact on the energy fl ow in both aquatic and terrestrial habitats from urban environments (Knight et al., 2005) and the potential role for control of disease vectors in densely populated areas (Saha et al., 2012).Moreover, given the fact that odonate larvae tend to bioaccumulate contaminants such as heavy metals and PCBs that come from human activities, odonate larvae can transfer the contaminants to other organisms via trophic interactions, either in aquatic or terrestrial ecosystems, that is, in the absence of a decoupling mechanism preventing the transfer of contaminants from larvae to adults (Yu et al., 2013).

Behaviour
Research on behaviour in relation to urban odonate populations has focused on dispersal and the selection of oviposition sites, but a much wider suite of behavioural traits would be expected to vary across urban-rural gradients.Considering, for example, the different stressors from urban areas (reviewed above), it is common to fi nd tolerant species in cities (i.e.urban adapters and exploiters; McKinney, 2002) such as I. elegans (Solimini et al., 1997).Urban adapters and exploiters among invertebrates would be predicted to exhibit syndromes involving multiple behavioural responses such as increased boldness, that is, being willing to take more risks (Lowry et al., 2013) like those exhibited by vertebrates (e.g.song sparrows, Evans et al., 2010).Only one study to our knowledge has investigated the behavioural response of odonates to urban stressors, and found that exploration behaviour, activity, and food intake of C. puella larvae were affected by pesticide stress except for boldness, although urban populations in general showed increased activity rate compared to rural populations (Tüzün et al., 2015).Therefore, there is a current lack of studies regarding behavioural syndromes in odonates and other invertebrates in response to other urban stressors along urban-rural gradients and the ecological and evolutionary consequences of those syndromes.Further studies regarding odonate behavioural ecology could also provide important insights into how odonates and other animals interact with urban ecosystems and test for a generalisation of urban evolutionary pressures across taxa.

Urban thermal ecology
Furthermore, we must consider that temperature increases in urban environments, creating a regional climate change effect called "Urban Heat Island" (UHI, Grimm et al., 2008).This temperature increase is most noticeable during the night (Karl et al., 1988), and operates both in terrestrial areas where pavements and rooftops absorb and radiate heat to increase the ambient air temperature, and in aquatic systems where heated runoff that passes over hot concrete and asphalt enters water bodies (Jones et al., 2012).Temperature is a major driver of odonate phenology (Hassall et al., 2007), polymorph frequency (Gosden et al., 2011), and body size (Hassall, 2013;Hassall et al., 2014).Additionally, temperature has a great impact on many physiological processes, such as metabolism, respiration, muscular activity, immunology, development, reproduction, and -naturally -thermoregulation (i.e.pigmentation, basking, wing-whirring) (Hassall & Thompson, 2008;Neven, 2000).Therefore, there may be phenological, morphological and physiological shifts in odonates caused by the urban heat island effect, as well as interactions between these shifts.However, the effect of the UHI might not necessarily be negative.Most odonates are likely to tolerate such warm conditions due to their tropical evolutionary origin (Pritchard & Leggott, 1987), although thermal stress can be lethal if the temperatures exceed upper tolerance thresholds.In fact, the UHI may facilitate dispersal of odonates, as has been found in scale insects (Meineke et al., 2013) and mosquitoes (Araujo et al., 2015).It has been suggested that cities in temperate regions may benefi t from the increased temperature, particularly during winter, although the UHI can also increase thermal and drought stress in tropical, subtropical and desert cities (Shochat et al., 2006).The UHI also has a great impact on freshwater habitats.However, there is no research regarding these topics, hence we can only speculate at this point.

Modifi ed hydrogeomorphology of freshwaters
Rivers and ponds are frequently subjected to human alterations mostly by increasing imperviousness, which increases temperature, decreases infi ltration of water and increases stormwater runoff drastically (for a detailed review see Paul & Meyer, 2001).Sediment is also coarser in urban water bodies as a result of alteration of sediment supply and velocities, and also tends to accumulate high metal contents such as lead, zinc, and cadmium (Paul & Meyer, 2001).The lack of natural substrate may interfere with microhabitat use, habitat selection, and a variety of behaviours (e.g.sheltering, hunting).However, to our knowledge no studies have yet been carried out on the importance of fl ow regimes and sediment change to odonates within urban rivers and ponds.

Urban genomics
As mentioned previously, the intraspecifi c variation throughout an urban gradient may be driven via phenotypic plasticity or genetic adaptation.Other than studies looking at dispersal (Watts et al., 2004;Sato et al., 2008), there are no investigations to our knowledge that link genetics to the phenotypic variation in urban odonate populations.Thus, the underlying mechanisms that drive phenotypic variation of odonates in cities are not well-documented.While one study has looked at population differentiation across an urban landscape (Sato et al., 2008), there could be far greater insights through a combination of gene-fl ow models, cost surfaces, and corridor analyses to use odonate movement as an indicator of large-scale habitat connectivity in cities. Modern advances in transcriptomics and metabolomics also offer opportunities to explore the stressors experienced by urban populations through the differential expression of genes (e.g.Chapman et al., 2009).

CONCLUSION
To summarise, this review illustrates that odonate diversity is generally lower in built-up city cores than in surrounding areas.However, with suitable management and design, urban areas could increase diversity and landscape connectivity, which has various promising implications for ecology, evolutionary biology, conservation biology and urban planning.While patterns in diversity are welldocumented, there has been a lack of research into the behavioural, genetic, and life history processes that might act as mechanisms to drive those diversity patterns.The literature reviewed here provides a strong case for the use of odonates as a model taxon in both the lab and fi eld for the study of a wide range of phenomena related to urbanisation.In order to achieve an integrated understanding of urban ecosystems, we encourage further research, specifically using odonates due to their versatility as bioindicators in both aquatic and terrestrial habitats.

Diversity
The BIO-SAFE biodiversity assessment method was developed based on several species of rivers and their fl oodplains, including 14 species of odonates.This method was tested in different ecotopes from urban fl oodplains, and showed that ecotope saturation (TES) and the actual ecotope importance (ATEI) scores must be taken into account to gain insight in the value of ecotopes for endangered species.

Lubertazzi & Ginsberg (2010)
Odonates USA Ponds Community assemblage, taxa richness, and diversity indices along an urban-to-rural gradient Diversity Richness and evenness did not differ signifi cantly between urban and rural sites; many species were more commonly found in urban sites with fi sh.

Fig. 3 .
Fig. 3. Summary of drivers of odonate biodiversity in cities due to heavy management.Dashed lines represent hypothetical effects, since no studies were found to investigate the association between the specifi ed stressor and the corresponding trait(s) on odonates.

Table 1 .
Urban odonate peer-reviewed papers found related to diversity.

Table S1 .
Summary of publications considered in the analysis and the categories in which they were included.
DiversityThe highest richness of adults was recorded during the warmer months; immatures were out of phase.Most ponds accounted less than 25% of the recorded taxa.Only two odonate taxa recorded.

Table S1 (
continued).Odonate richness was higher where fl oating vegetation was found.Taxa richness in temporary pools was not cantly different from permanent ponds, but overall diversity was lower.Only four odonate taxa recorded.Micrathyria spp.and Miathyria sp. were found exclusively in the site with less sewage input.

Table S1 (
continued). the second lowest taxa richness, but had the highest abundance in anisopterans.Parks, on the other hand, were the highest in zygopteran richness and abundance.sediment,waterandinvertebrateswas higher in the newest (urban) wetlands.Odonates were signifi cantly greater in wetlands equal to or less than 2 years old.Odonate MeHg concentrations from new and old wetland were not signifi cantly different.Taxa richness and diversity were lower in urban sites.However, urban sites had high similarity in both headwater and main-stem communities, unlike rural sites.The unique headwater taxa in urban headwater streams belonged only to Odonata (Ischnura spp.andCalopteryx maculata) and Diptera (Aedes and Odontomyia).Damselfl y assemblage was mostly formed by generalist species found in lentic habitats.Ischnura elegans and Erythromma (= Cercion) lindenii were the most abundant in highly polluted wetlands, presenting a longer reproductive period, absence of diapause, and tolerance of low oxygen concentration.All studied samples contained mostly insects and spiders, including odonates.Forest bats presented opportunistic feeding behaviour, while urban ones were selective with their prey.Stenotopic species were highly sensitive to changes in sunlight, water fl ow and vegetation structure.Yet, taxa richness increased over 100% after the creation of the pond.Sympetrum pedemontanum, Coenagrion pulchellum, Ischnura pumilio and Sympetrum danae had declined, although species like Sympecma fusca, Gomphus vulgatissimus, Ophiogomphus cecilia, Orthetrum brunneum and Orthetrum coerulescens had increased in a period of almost 30 years.Cordulegaster heros was found in an urban stream, where it co-occurs with Cordulegaster bidentata and Calopteryx virgo.fl avescens occurrence was positively related to the amount of cars found in the parking area.There was also a preference towards light coloured cars.Oviposition was recorded on the hood surfaces, which represents an energy loss for the females.
ToxicologyZooplankton presented the highest PCB levels, followed by soil-dwelling invertebrates.Chironomids and odonates had high concentrations of mid-chlorinated congeners (PCB-153 and PCB-138).