Introduction

Expanding and intensifying urbanisation is a significant threat to the condition of ecosystems globally (Crist et al. 2017; Elmhagen et al. 2015). Land-use change from natural to urban and agricultural areas clears and fragments ecosystems and is arguably the most significant anthropogenic stressor for many ecosystems worldwide (Fischer and Lindenmayer 2007; LaPoint et al. 2015). Land-use change can reduce biodiversity, change the structure of animal assemblages, affect the rate and distribution of key ecological functions, and degrade the capacity of ecosystems to support the services upon which people rely (Alberti 2010; Olds et al. 2018b; Reddell et al. 2023). Boundaries between ecosystems have unique linkages whereby impacts in one ecosystem can have implications for adjacent ecosystems (Aguilera and González 2023; Bugnot et al. 2019; Cadenasso et al. 2003; Pittman and Armitage 2016). For example, urbanisation within one ecosystem can affect diversity in nearby ecosystems because animals move across landscapes, or because patterns of nutrients or energy moving from one ecosystem to the other are disrupted (Mosman et al. 2020; Töpfer 2001). Disentangling the effects of urbanisation on one ecosystem that might affect the condition of nearby ecosystems (e.g. through changed subsidies, animal movement patterns or pollutants) and identifying species whose presence or abundance might be indicative of these trends is a challenging, but fundamental, knowledge gap that should be addressed to maximise the effectiveness of biodiversity conservation and landscape management (Savage et al. 2012). More thoroughly understanding such patterns and indicators can assist in modifying coastal planning (Elliott et al. 2022), and conservation and restoration techniques (Carignan and Villard 2002; Lindenmayer et al. 2000) to optimise biodiversity outcomes along impacted coastlines.

Coastal ecosystems support a diversity of terrestrial and marine life and are valued for their cultural, ecological, and recreational benefits (He and Silliman 2019). However, they are increasingly fragmented and replaced by urban land as human populations increase near the seashore (Bishop et al. 2017; Brown et al. 2014; LaPoint et al. 2015). With increasing human populations and expanding urban footprints, impacts caused by habitat alterations, overfishing, and marine pollution are also increasingly evident in many marine systems (Dafforn et al. 2013; Halpern et al. 2019). Coastal urbanisation significantly affects the composition of subtidal animal assemblages by reducing subtidal habitat extent, condition and connectivity (Bishop et al. 2017; Goodridge Gaines et al. 2022), degrading coastal water quality and concentrating the effects of fishing (Gilby et al. 2020). Given these significant changes to ecosystems in both realms of the coastal strip and the capacity for impacts to affect specific species, indicator groups and species are often identified and used for the purposes of tracking both the effects of impacts, and the outcomes of management (Gilby et al. 2017; Hunter et al. 2016; Tulloch et al. 2016). Urbanisation therefore has irrevocable and widespread consequences for coastal biodiversity, fragments and reduces animal populations, changes the structure of food-webs, and reduces the productivity of coastal ecosystem services (Firth et al. 2016; LaPoint et al. 2015; McPhearson et al. 2016; Yan et al. 2022).

Coastal landscapes are comprised of both marine and terrestrial ecosystems that meet at a boundary (Harris et al. 2019; Pittman and Armitage 2016). This close spatial proximity means that human impacts can be manifest in both realms (Bugnot et al. 2019; Savage et al. 2012). For example, increases in urbanised land extent in the coastal strip can reduce water quality, modify sediment budgets, and increase nutrient concentration and eutrophication risk in coastal waters (Brodie et al. 2005; McKinley et al. 2011; Packett et al. 2009; Schroer et al. 2018). Similarly, overfishing reduces food availability for a diversity of animals in both marine and terrestrial realms, including terrestrial predators and scavengers that rely upon the ocean for food sources, and such effects can be centralised closer to larger human populations (Huijbers et al. 2015). It is well established that the extent of urbanised land affects animals at different spatial scales for different species in different ecosystems (Alvarez-Romero et al. 2011; Gilby et al. 2022). However, whether human impacts are expressed equally on either side of ecological boundaries, which species are the most effective indicators of these impacts, and whether any observed environmental change is in the same direction and of the same magnitude remains poorly understood (Bingham et al. 2018).

Ocean beaches are hotspots of coastal urbanisation, being prime locations for recreation and expanding urban residential developments (Nel et al. 2014; Soto et al. 2021). Urban development on land backing ocean beaches destroys and fragments vegetation (e.g. Aguilera et al. 2020; Elliott et al. 2022), alters the population structure, distribution and condition of animal assemblages inhabiting dunes (e.g. Stelling-Wood et al. 2016), and increases the abundance of synanthropic taxa (e.g. Huijbers et al. 2013) both on land and in the sea. Expanding extents of urbanised land along the seaward-located dunes can disrupt natural sediment budgets and dynamics, requiring ecologically harmful beach nourishment to address erosion seen to threaten human infrastructure (de Schipper et al. 2021; Defeo et al. 2009). Dunes and beaches are irreplaceable breeding and feeding habitat for numerous bird and mammal species (Ehmke et al. 2016; Maslo et al. 2019) and also support large and iconic raptors that are functionally important consumers in beach food webs (Huijbers et al. 2015). The surf zones of ocean beaches create a unique habitat for diverse and abundant fish assemblages, including many species that have commercial, recreational or conservation significance (Olds et al. 2018b). The extent of urbanised land in the coastal zone affects the composition of fish assemblages in surf zones through a combination of anthropogenic impacts, including declining water quality and fishing (Mosman et al. 2023; Vargas-Fonseca et al. 2016). Coastal urbanisation can also change (both positively and negatively) the delivery of terrestrially sourced energy and nutrients into nearshore waters, which can change coastal productivity and food availability for fish (Gorman et al. 2019; Olds et al. 2018b). Therefore, the extent of urbanised land in the coastal strip, and the proximity of these sites to animal assemblages have been shown to be significant predictors of animal assemblages on many surf-beach-dune landscapes globally (e.g. Gilby et al. 2022; Huijbers et al. 2016).

Beyond anthropogenic drivers of coastal vertebrate and surf zone fish assemblages, the distribution and abundance of animals in surf-beach-dune landscapes is modified by a suite of natural drivers (Sperandii et al. 2019). For example, the distribution, diversity and abundance of surf zone fish is affected by the proximity of sites to rocky reefs, headlands and estuaries (Mosman et al. 2020, 2023; Vargas-Fonseca et al. 2016), and the distribution and abundance of coastal raptors is affected by the proximity of sites to headlands and the area of remnant vegetation nearby (Gilby et al. 2021, 2022; Huijbers et al. 2013). Similarly, the abundance of larger bird species in coastal dunes might be affected by the size and complexity of coastal dune habitats themselves, including the presence of tall perching trees. Disentangling the impacts of coastal urbanisation from natural drivers of animal abundance and distribution on beaches therefore requires data over spatial scales that covers the breadth of coastal urbanisation extent within a region, and analyses that disentangle the effects of urbanised land extent and proximity from natural population drivers.

Substantial and widespread ecological impacts of urbanisation in the coastal zone have been reported for many types of species, assemblages and habitats (Aguilera and González 2023; Firth et al. 2016; Halpern et al. 2019). Surprisingly, whether the ecological response trajectories of urban impacts are similar in abutting habitat types that form long ecological boundaries, including in the surf zones and coastal dunes of beaches, remains poorly tested. Here we test for cross-boundary effects of coastal urbanisation, contrasting the response of coastal vertebrates (including birds, mammals and reptiles) and fish in the terrestrial and aquatic parts of the surf-beach-dune interface, respectively, and identify species whose abundance is an indicator for these patterns. We test three complementary hypotheses: (1) sites with a larger area of urbanised land nearby will have a lower abundance and diversity of both costal vertebrate and surf zone fish than sites more isolated from urbanisation, (2) sites that are more proximal to urbanised land nearby will have a lower abundance and diversity of both costal vertebrate and surf zone fish than sites more isolated from urbanisation and (3) the presence of extensive natural habitats (especially rocky headlands, subtidal reefs and healthy vegetated coastal dunes, including large, mature trees) adjacent to urbanised beaches can dampen or ameliorate the effects of urbanisation.

Methods

Study area

Fish of surf zones and coastal vertebrates of coastal dunes (including any species captured by our sampling techniques- we were not specifically targeting any one type of species) were sampled along the Sunshine Coast in eastern Australia (∼ 26.65° S, 153.06° E) (Fig. 1). The region’s beaches encompass a gradient of urban land conversion in the dunes, ranging from highly urbanised locations with no to narrow (< 5 m) and fragmented dune vegetation, to beaches backed by wide (> 200 m) swathes of vegetated dunes. We sampled fish of surf zones and coastal vertebrates of dunes at 100 sites distributed over 20 beaches at five sites per beach (separated by a minimum of 200 m) between May and June in both 2019 and in 2020 for n = 200 baited camera surveys in each ecosystem (Fig. 1). The 200 m distance between sites is greater than the minimum distance recommended between sites for baited camera surveys for fish (Cappo et al. 2006; Unsworth et al. 2014) and reflects the study design of previous studies in the region (e.g. Gilby et al. 2021; Gilby et al. 2022) and beyond (e.g. Huijbers et al. 2016) for quantifying similar patterns in coastal vertebrate assemblages. Baited camera deployments were designed to sample the greatest possible suite of animals occurring within the focal landscapes, and not to census all animals within a location. Of course, all conclusions are therefore limited to the suite of species surveyed by our methods.

Fig. 1
figure 1

Location of study locations sites on the Sunshine Coast, Australia. Inset provides an example of the distribution of survey sites at each location

Full size image

Surveying coastal vertebrates

We used motion detecting trail cameras (ScoutGuard Zero Glow and Browning DarkOps Pro, deployed randomly) to quantify terrestrial vertebrate assemblages at each site; a method used previously in multiple studies for the same assemblage and habitat types (e.g. Bingham et al. 2018; Gilby et al. 2022). Each site was baited with a single sea mullet (Mugil cephalus). Sea mullet are the ideal bait species for this region as they are widespread in the surf zone of ocean beaches and are readily detected and consumed by a wide variety of scavengers, carnivores and omnivores (Borland et al. 2017; Schlacher et al. 2013). Two trail cameras were placed 2–3 m apart and 5–6 m from the bait to provide complementary fields of view and detection. Cameras took 10 s high-definition videos when triggered and were set to the highest sensitivity levels to maximise detection probability. Cameras and baits were positioned in the foredunes above the high tide line. We deployed cameras at sunset on the first day of surveys, and replaced the bait every 12 h, giving a total exposure period of 48 h per camera deployment.

Surveying fish in the surf zone

We surveyed surf zone fish using baited remote underwater video stations (BRUVS) deployed in line with the survey sites used for the coastal vertebrates (i.e. on the same latitude; see inset in Fig. 1). BRUVS are made of a high-definition GoPro camera mounted on a 3 kg weight with a length of PVC pipe holding a bait bag approximately 1 m from the camera. The bait bag contained approximately 500 g of pilchards (Sardinops sagax). While we therefore used different baits and survey durations in surf zones (pilchards for 1 h) and in dunes (sea mullet for 48 h), consistency was not required because: (1) the survey methods for each ecosystem have been optimised, and the different bait types and survey times used, in multiple previous studies in each ecosystem globally (see Olds et al. 2017), and (2) we were not directly comparing assemblages between realms, but rather quantifying assemblage-level responses to urbanised land extent and proximity, and so it was more important to optimise surveys for individual ecosystems than to match methods between ecosystems. Five BRUVS were deployed at each beach seaward of the breaking waves and along the same latitude as trail cameras (Fig. 1). BRUVS was deployed for one hour, yielding 100 h of video footage across all sites per year.

Data analysis

The structure of coastal vertebrate and fish assemblages was quantified from both BRUVS and trail camera videos using the MaxN metric: the maximum number of individuals of each species observed in a single video frame over the recording period. MaxN is the standard approach for extracting abundance and diversity data from camera deployments globally and ensures that the same individual is not counted more than once at each site during each sampling event (Dorman et al. 2012). These data were then used to calculate the compound metrics of species richness (being the number of unique species identified at each site) and total abundance (being the sum of MaxN values for all species identified at each site).

Data analysis was undertaken using two key steps. Firstly, we identified patterns in the assemblage structure of the animals sampled using multivariate generalised linear models (ManyGLMs) to identify overall patterns with key environmental variables and species whose abundance correlated best with overall patterns (i.e. indicator species). Secondly, we further analysed patterns in the abundance of these indicator species and the compound metrics of species richness and total abundance using generalised additive models (GAMs). This allowed us to explore potentially contrasting patterns of abundance between key species, metrics and ecosystems.

We assessed the effects of 12 environmental variables on the assemblage structure of coastal vertebrates and surf zone fish separately using multivariate generalised linear models (ManyGLMs) in the mvabund package (Wang et al. 2012) of R (R Core Team 2024) calculated using the negative binomial distribution. ManyGLMs seek correlations between a suite of explanatory variables and a matrix of species abundance by replicate, providing results in the form of a standard GLM table. ManyGLM also identifies species whose abundance and prevalence correlates significantly with the best fit model; these species are considered indicator species in this study. Statistical significance of variables in the best fit models was tested using analysis of variance on the chi squared test statistic (i.e. χ2). Best-fit ManyGLMs (that incorporate information regarding best-fit variables and the indicator species that best correlate with these effects) were identified using reverse stepwise simplification on Akaike’s Information Criterion (AIC) and were visualised using non-metric multidimensional scaling (nMDS) ordinations calculated on Bray-Curtis dissimilarity.

The 12 environmental variables included in the models included the distance (in meters) of each site to the nearest headland, reef, estuary and urbanised land as well as the area (in hectares) of urbanised land within 1, 2 and 5 km radii. We included the maximum tree height (in meters) of dune vegetation at each site by running a 100 m transect directly west through the foredune. The height of each tree whose canopy intersected the transect was measured using a Suunto clinometer (Elliott et al. 2022). We included the width of both the beach and surf zone as proxies for beach exposure to wind and waves (Mosman et al. 2020; Ortodossi et al. 2019). We also included the survey year in ManyGLMs to identify effects between data collected in 2019 and 2020 and beach name to account for site clustering. The same environmental variables and model structure was used for ManyGLMs assessing both coastal vertebrate and fish assemblages; these assemblages were analysed separately as data from the fish and coastal vertebrate surveys cannot be easily combined as they were sampled over different time frames (i.e. 48 h for terrestrial vertebrates, 1 h for fish). No environmental variables included in the models covaried with a Pearson’s correlation coefficient of greater than 0.7 or less than − 0.7. Urban land maps were sourced from the Queensland Government (Queensland Government 2021), and all spatial variables were calculated from these layers and aerial imagery from NearMap (NearMap 2020) in QGIS (QGIS Development Team 2022).

Relationships between environmental variables and indicator species identified from the best-fit ManyGLMs, along with species richness and total animal abundance were further assessed using GAMs calculated using thin plate splines in the mgcv package (Wood 2019) in R using a negative binomial distribution. This error distribution was chosen for consistency with the multivariate analyses, and resulted in good model fits. GAMs were preferred over linear models in this study as Pearson’s vector correlations from nMDS ordinations indicated the likely presence of non-linear effects in the dataset. Best-fit GAMs were again identified using reverse stepwise simplification on AIC and model overfitting was reduced by limiting the number of model knots to 4 or fewer (i.e. k = 4).

Results

Animal assemblages

Variation in the composition of coastal vertebrate assemblages (see Table S1) was best explained by the distance of sites to the nearest headland (χ2 = 111.86, p = 0.001), the year of survey (χ2 = 94.89, p = 0.001) and the area of urbanised land within 5 km (χ2 = 93.34, p = 0.001) (Fig. 2A). All other variables included in the initial model were removed during reverse stepwise simplification, and so do not significantly affect coastal vertebrate assemblages in this study. Variability in the abundance of four species accounted for most of these patterns: Torresian crows (Corvus orru), domestic dogs (Canis lupus familiaris), brahminy kites (Haliastur indus), and black rats (Rattus rattus; Fig. 2A), with an apparent divide in multivariate space between species found at sites with less extensive urban areas nearby (Torresian crows and brahminy kites) and those found at sites with intermediate levels of urbanised land nearby (Fig. 2A).

Fig. 2
figure 2

Non-metric multidimensional scaling ordination of best-fit ManyGLM models for (A) coastal vertebrates and (B) surf fish driving this pattern. Solid black Pearson’s vectors denote significant indicator species, while dashed grey Pearson’s vectors indicate environmental variables. Panel A MDS stress = 0.09 and panel B MDS stress = 0.12

Full size image

Variation in the composition of fish assemblages (see Table S2) was best explained by the distance of sites to reefs (χ2 = 421.9, p = 0.001) and urbanised land (χ2 = 114.9, p = 0.012), the year of survey (χ2 = 256.3, p = 0.001), and the width of the surf zone (χ2 = 148.8, p = 0.002) (Fig. 2B). All other variables included in the initial model were removed during reverse stepwise simplification, and so do not significantly affect fish assemblages in this study. These patterns mostly result from variation in the abundance and distribution of brown-banded bamboo shark (Chiloscyllium punctatum), gold-spotted sweetlip (Plectorhinchus flavomaculatus), Gunther’s wrasse (Pseudolabrus guentheri), moon wrasse (Thalassoma lunare), sieve-patterned moray eel (Gymnothorax cribroris), Sydney cardinalfish (Ostorhinchus limenus), weeping toadfish (Torquigener pleurogramma) and yellowfin bream (Acanthopagrus australis (Fig. 2B), with an apparent pattern in multivariate space of most species (except for weeping toadfish) being found in greater abundance at sites further from urbanised land (Fig. 2B).

Patterns in animal abundance and diversity

Coastal vertebrate species richness was lower at sites with greater area of urbanised land within 5 km (estimated degrees freedom (edf) = 1, χ2 = 5.3, p = 0.04) (Fig. 3A) and increased further from headlands, plateauing at sites farther than 3000 m from headlands (edf = 1.48, χ2 = 6.1, p = 0.03) (Fig. 3B)(deviance explained = 25.1%). All other variables were removed during reverse stepwise simplification, and so do not significantly affect coastal vertebrate richness in this study. Coastal vertebrate abundance was highest at sites within 5 km of 1,700 ha (i.e. 21.6% of the 5 km radius attributed to urban land) of urbanised land and lowest when urbanised land extent exceeded 3,000 ha (i.e. 38.2% of the 5 km radius attributed to urban land) (edf = 2.74, χ2 = 36.58, p < 0.001) (Fig. 3C) (deviance explained = 23.2%). Coastal vertebrate abundance was also highest at sites 3,500 m from headlands, with abundance being lowest at sites nearest to headlands and at approximately 6000 m from headlands (edf = 2.8, χ2 = 67.66, p < 0.001) (Fig. 3D). All other variables were removed during reverse stepwise simplification, and so do not significantly affect coastal vertebrate abundance in this study.

Fig. 3
figure 3

Generalised additive model (GAMs) outputs illustrating best-fit models for the species richness (A-B, E-G) and abundance (C-D and H-J) for coastal vertebrates (A-D) and surf fish (E-J). Shaded regions indicate 95% confidence interval. Areas are provided in hectares. Points on the graphs are the raw data values

Full size image

Surf zone fish species richness declined with greater distances from urbanised land (edf = 1, χ2 = 7.62, p = 0.005) (Fig. 3E), declined with greater distances to reefs, plateauing at sites beyond 700 m from reefs (edf = 2.7, χ2 = 19.42, p < 0.001) (Fig. 3F), and was lowest at intermediate surf zone widths (edf = 2.46, χ2 = 19.87, p < 0.001) (Fig. 3G) (deviance explained = 19.3%). All other variables were removed during reverse stepwise simplification, and so do not significantly affect fish species richness in this study. Models for fish abundance against best-fit environmental variables showed similarly shaped relationships for both 2019 and 2020 but higher average abundance in 2019 (Figures H-J). Surf zone fish abundance peaked in each year (edf = 1, χ2 = 386, p < 0.001) at sites approximately 150 m from urbanised land (edf = 2.96, χ2 = 248.8, p < 0.001), with abundance declining sharply as distance to urbanised land increased (Fig. 3H), increased to a plateau at approximately 1000 m from reefs (edf = 2.31, χ2 = 47.06, p < 0.001) (Fig. 3I), and was highest at sites with 100 m surf zone width (edf = 2.99, χ2 = 371.71, p < 0.001) (Fig. 3J) (deviance explained = 38.52%). All other variables were removed during reverse stepwise simplification, and so do not significantly affect fish abundance in this study.

Indicator species

We identified species who exhibited positive, negative and non-linear responses with urbanised land in both coastal vertebrate and fish assemblages. We selected three species from coastal vertebrates (Fig. 4A-C) and surf zone fish (Fig. 4D-F) assemblages to represent these three key relationships (positive linear, negative linear, non-linear) between animal abundance and urbanised land in each ecosystem. For example, the area of urbanised land best explained the abundance of Torresian crows (edf = 2.65, χ2 = 50.6, p < 0.001, deviance explained = 28.3%), black rats (edf = 1, χ2 = 9.73, p = 0.001, deviance explained = 44.4%) and brahminy kites (edf = 1, χ2 = 9.44, p = 0.002, deviance explained = 21%) within 5 km of sites, but each species responded in different trajectories (i.e. some positive linear, some negative linear, some non-linear). Torresian crow abundance was highest where the area of urbanised land was approximately 1700 ha (i.e. 21.6% of the 5 km radius attributed to urban land) within 5 km, and then declined with increasing urbanised land, and were more abundant in 2020 (Fig. 4A). The abundance of black rats increased where urban areas were more extensive and was higher in 2020 (Fig. 4B). Finally, brahminy kite abundance decreased as the area of urbanised land increased (Fig. 4C).

Fig. 4
figure 4

Generalised additive models (GAMs) outputs illustrating best-fit models for the abundance of key indicator species for (A-C) coastal vertebrates and (D-F) surf fish. Shaded regions indicate 95% confidence interval. Areas are provided in hectares. Points on the graphs are the raw data values

Full size image

The abundance of yellowfin bream (edf = 1, χ2 = 44.33, p = 0.01, deviance explained = 37.6%), weeping toadfish (edf = 2.94, χ2 = 103.2, p < 0.001, deviance explained = 47.3%) and Gunther’s wrasse (edf = 2.93, χ2 = 28.31, p < 0.001, deviance explained = 54.8%) was best explained by the distance of sites to urbanised land, but once again these three species showed fundamentally different relationships with this variable (i.e. some positive linear, some negative linear, some non-linear). Yellowfin bream abundance decreased with increasing distance to urbanised land and was higher in 2019 (Fig. 4D). Weeping toadfish abundance increased with increasing distance from urbanised land and was again higher in 2019 (Fig. 4E). Finally, Gunther’s wrasse abundance spiked at sites approximately 170 m from urbanised land and flatlined as the distance to urbanised land increased (Fig. 4F).

Discussion

Coastal urbanisation is a significant threat to the condition of ecosystems and can change the animal assemblages they support (Defeo et al. 2009; He and Silliman 2019). Given different species assemblages and impacts, there is strong potential for differential effects of urbanisation on the terrestrial and aquatic realms of coastal landscapes. This study supports the notion that the conversion of natural dunes to urbanised land heavily affects the distribution and changes the assemblage structure of both terrestrial vertebrates living on beaches and marine fish in the sea. We show that the abundance and diversity of coastal vertebrates sampled using our techniques was significantly lower at sites abutting larger areas of urbanised land, which aligned with our first hypothesis. Conversely, the abundance and diversity of fish sampled using our techniques were typically higher in surf zones located closer to urban areas, which ran counter to our hypothesis. This finding suggests that coastal vertebrates were most influenced by the area of urbanised land in this study, that fish assemblages were most influenced by the proximity of urbanised land in this study, and that the shapes of these relationships differ. Subtle species-specific effects can often be masked by the compound metrics of total abundance and richness, as was the case in this study, where highly abundant and prevalent synanthropic species might have a disproportionately larger effect on results. Crucially, urban effects at the assemblage level and for key compound metrics were also significantly affected by the effects of natural drivers of animals assemblages (e.g. connectivity with natural features), meaning that disentangling the effects of urbanisation requires detailed knowledge of the spatial ecology of target assemblages (Mosman et al. 2020; Ortodossi et al. 2019). While our results are limited by the scope of our survey methods (as with all studies, and in this case using baited camera arrays), results highlight the capacity for urbanisation to have implications for the composition and condition of ecosystems across ecological boundaries and ecological realms. Quantifying such cross-boundary effects on animals more broadly might assist ecologists and managers in more thoroughly understanding the implications of human impacts on ecosystems (Kark 2013).

Understanding the attributes and spatial scales of urbanised land that most affect animal assemblages is important to design and implement effective land-use planning, management interventions and conservation strategies, especially for mobile species which can be exposed to anthropogenic threats over large areas (Oppel et al. 2018). While we found effects of urbanised land proximity and extent on both terrestrial and marine animal assemblages in this study, we found differences in the attributes of urbanisation affecting different ecosystems and the ways in which urbanisation influence assemblages. The assemblage of coastal vertebrates surveyed in this study were affected by the extent of urbanised land, as opposed to the proximity of urbanised land. Here, the coastal vertebrates we predominantly surveyed move throughout (mammals and reptiles) or over (birds) the urban environment, interacting directly with it, meaning that the area and intensity of urbanised land at sites nearby to beaches likely affects different species in different ways. For example, scavenging raptors actively avoid populated areas and human disturbance in some systems (Huijbers et al. 2013, 2015). The lack of diversity and abundance of scavenging raptors over both survey years could be explained by reduced breeding success in urban areas through the demolition of nesting sites, thereby resulting in individuals being driven out of home ranges (O’Donnell and Debus 2012; Rourke and Debus 2016). Alternatively, raptors like those surveyed in this study have large flight initiation distances (Shuai et al. 2024), which might serve to drive these sorts of species further from urbanised areas, or affect the likelihood of them being detected by our methods. Conversely, rats increased in abundance at sites with greater urban extent nearby as these synanthropic species are associated closely with human habitation and infrastructure as opposed to more natural areas (Gilby et al. 2021; Huijbers et al. 2013).

Landscape transformation resulting from urbanisation can alter assemblages by favouring generalist species more capable of adapting to highly modified environments (Gilby et al. 2021; Sorace and Gustin 2009). In this study, coastal vertebrate abundance was dominated by Torresian crows. Generalist species with highly varied diets, such as crows, are more equipped to tolerate habitat alterations and anthropogenic disturbances (Henderson et al. 2020; Kimber et al. 2020), which may explain the peak in coastal vertebrate abundance with a moderate extent of urban area.

Species richness and fish abundance in the surf zone showed relationships with urbanised land that were counter to our hypotheses in this study. The surf zone fish assemblages recorded in this study correlated with the proximity of survey sites to urbanised land, as opposed to the extent of urbanised land. Here, surf zone fish do not interact directly with urbanised areas but are indirectly impacted by its consequences (Olds et al. 2018b). Proximity to urbanised land can correlate with variables like fishing intensity (Lewin et al. 2019), the width of vegetation buffers between beaches and urban areas (Pessoa and Lidon 2013), and therefore the volume of subsidies provided from terrestrial vegetation to surf zones and beach nourishment activities (Schlacher et al. 2016). This means that the effect of urban proximity that we found in this study is likely a combination of these variables (Defeo et al. 2009; Henderson et al. 2020). Coastal urbanisation including human traffic, pollution, beach grooming and nourishment predominantly impact coastal food webs from the bottom up; predators are affected by reduction in the abundance and diversity of their prey (Chapman et al. 2009; Vargas-Fonseca et al. 2016). These effects reduce the prevalence and abundance of sensitive species, especially larger predators that are highly prized fisheries species.

Urban effects may proliferate through surf zone fish assemblages in non-linear ways, as was found here for both total fish abundance and the abundance of key indicator species; non-linear effects are indeed common in surf zone fish assemblages (e.g. Mosman et al. 2020; Vargas-Fonseca et al. 2016), and the effects of urbanisation are not always negative. This non-linear effect may be caused by, for example, fishers preferring areas with some natural surrounding vegetation and away from highly populated recreation beaches, but not so isolated such that sites are difficult to access (e.g. Vargas‐Fonseca et al. 2016). In this study, the abundance of smaller, zoobenthivorous species like weeping toadfish and yellowfin bream was higher at intermediate and high levels of urbanisation, thereby potentially explaining our effects for total fish abundance. Such highly abundant zoobenthivorous species are not necessarily desirable and may be indicative of mesopredator release at these sites where fishing pressure can be higher (Mosman et al. 2020; Vargas‐Fonseca et al. 2016). This heightened fish abundance might therefore be an indicator of a more highly stressed ecosystem. Overall, however, effects of urbanisation cannot be discounted for surf zone fish as we identified species-specific negative responses that were supported by the results of our assemblage-level analyses. Disentangling the relative influence of these various consequences of urbanisation requires further experimental study. In any case, appreciating the subtlety in effects and attributes of urbanisation on different ecosystems can help to optimise management in multiple settings. For example, in this study, minimising the total area of urbanised land close to the coast by creating smaller, more dispersed developments that incorporate significant vegetation buffers which increase distance between surf zones and urbanised land might create co-benefits for animals in both realms.

Environmental variables such as the topography and connectivity of landscapes affects the structure, function and dynamics of ecosystems and therefore mediates the effects of urbanisation on animal assemblages (McKinney 2008). In this study, the significance of proximity to headlands reflects the diversity of coastal vertebrates and their different methods of scavenging. For example, raptors use headlands as vantage points and to take advantage of air currents to scan over their environment to optimise scavenging rates and consumption (Beasley et al. 2015; Huijbers et al. 2013). Headlands are popular sites for picnic and recreation areas where human food resources are abundant and therefore attract opportunistic scavengers such as black rats who feed on food scraps at night (Gilby et al. 2022). For example, Torresian crows have been shown in previous studies to have greater survivorship at sites more proximal to urbanised land (Armendariz et al. 2011). Moreover, anthropogenic food subsidies are the predominant incentive for crows to colonise urbanised land (Clifton and Jones 2017; Marzluff and Neatherlin 2006). This coincides with the significance of intermediate proximities to urbanised land and abundance of coastal vertebrates found in this study, as crows dominated the coastal vertebrate assemblage at many sites.

The proximity of sites to reefs was an important variable which influenced the abundance and diversity of surf zone fish. Reefs provide shelter, food, and nursery grounds for an array of species in surf zones and are therefore vital to marine life (Borland et al. 2017; Olds et al. 2018b). Many fish use multiple habitats throughout their lifecycle (Olds et al. 2018a; Vargas-Fonseca et al. 2016); species that are found in the surf zone are often juvenile fishes foraging, seeking shelter in shallow waters or transitioning to other habitats such as estuaries or reefs (Borland et al. 2017; Mosman et al. 2020). Reefs have also been found to affect the level of marine subsidies to terrestrial environments (Ehmke et al. 2016). Understanding the effects of these mediating variables on animal assemblages is crucial, and without incorporating them into statistical models, the effects of urbanisation could potentially be masked.

Many studies report human impacts on animals along beaches, but they do so separately for terrestrial vertebrates inhabiting the dunes and beaches and for fish inhabiting the surf zones (Bingham et al. 2018; Borland et al. 2017; Huijbers et al. 2013). In comparison, we know very little about how human impacts act across boundaries to impact multiple abutting systems in the same seascape (Bugnot et al. 2019; Pittman and Armitage 2016; Stoms et al. 2005). While all conclusions are limited by the type and extent of sampling, our findings support the notion that urbanisation is a pivotal driver of change in coastal animal assemblages. Crucially, urbanisation influences the structure of terrestrial vertebrate and surf fish assemblages in different ways. Whilst ecological effects attributable to urbanisation are observed across the land-sea boundary, the type of urbanisation metric that best explains wildlife responses and the trajectory in which urbanisation changes compound metrics of assemblages is distinct for each of the two realms.