Introduction

Coastal ecosystems are increasingly impacted by an intensifying and diversifying range of anthropogenic impacts that change the distribution, size and connectivity of coastal landscapes (Airoldi and Beck 2007; Halpern et al. 2015; Kennish 2002; Waltham and Connolly 2011). For example, coastal habitats such as mangroves and saltmarsh are regularly replaced by hard, engineered structures (e.g. boat ramps, rock walls) (Dafforn et al. 2015; Todd et al. 2019), and surrounding estuarine channels are dredged (Van Maren et al. 2015), resulting in greater wave energy and suspended sediments and reduced seascape connectivity, biodiversity and key ecosystem services (Bishop et al. 2017; Borland et al. 2022; Gedan et al. 2009; Gittman et al. 2016; Munsch et al. 2017). Impacts affecting coastal catchments, including expanding urbanisation and agriculture, can result in substantial declines in coastal water quality that further reduce coastal conditions (Freeman et al. 2019; Matson et al. 1997). While coastal restoration is increasingly adopted to overcome these challenges (Beck et al. 2001; Bishop et al. 2017; Granek et al. 2010), existing impacts across coastal seascapes that have changed coastal abiotic conditions reduce the efficacy of restoring some coastal habitats (Abelson et al. 2020; Simenstad et al. 2005). Consequently, for restoration planning, it is crucial to quantify the value of habitats across coastal seascapes to better prioritise actions for sites according to their feasibility and potential ecological and ecosystem service benefits (Bayraktarov et al. 2016; Gilby et al. 2021a).

Anthropogenic impacts can significantly reduce the abundance and diversity of fish across coastal seascapes, and this can significantly affect recreational and commercial fisheries (Brown et al. 2018; Kennish 2002) and modify the functioning of coastal habitats (Henderson et al. 2020). Consequently, the objectives of many coastal management and restoration programs are to improve habitat value to enhance fish and fisheries (Creighton et al. 2015; Gilby et al. 2018b). Increasing connectivity across estuarine seascapes by restoring steppingstones both between habitats and from the estuary mouth to upper estuarine reaches ensures safe refuge for fish that make tidal or ontogenetic movements across seascapes (Barbier et al. 2011; Boström et al. 2011; Nagelkerken et al. 2015; Sheaves et al. 2015). Indeed, strategically selecting ecological restoration sites within seascapes to maximise this connectivity has been shown in previous studies to result in order of magnitude differences in benefits for fish biodiversity and abundance (Gilby et al. 2021b) and support ecological functioning (e.g. carrion consumption) (Henderson et al. 2020). However, extensive verge urbanisation which expands across long stretches of some estuaries can restrict this connectivity (Munsch et al. 2017; Waltham and Connolly 2011) and the capacity for the large-scale restoration of some key habitats (Gilby et al. 2021a; Simenstad et al. 2005). Within habitats, the availability of key resources (e.g. food resources) modifies their value for fish and can also be a significant predictor of fish diversity and abundance (Perry et al. 2023). Therefore, understanding the spatial and habitat condition drivers of fish assemblages and their functional groups both within habitats and across seascapes can help optimise strategic restoration actions (Goodridge Gaines et al. 2020; Henderson et al. 2017; Jones et al. 2020).

Log snags support a significant abundance and diversity of fish species in coastal seascapes (Goodridge Gaines et al. 2022; Lyon et al. 2019). Log snags are hard, often structurally complex habitats comprised of submerged or semi-submerged trees and/or branches. They support the growth of food resources such as invertebrates (i.e. barnacles, oysters) and algae and can centralise prey fish species, thereby providing significant benefits for fish that frequent them (Lester and Boulton 2008; Lyon et al. 2019). Log snags can provide structure for fish on vast, low-complexity intertidal areas (Green et al. 2012) as well as form important spawning habitats within estuaries (van der Meulen et al. 2013). Log snags also support a range of other valuable services including wave attenuation (Lester and Boulton 2008). Crucially, their efficacy for fish and other ecosystem services is not as easily impacted by changes in water quality compared to vegetated habitats (Gilby et al. 2018a; Waycott et al. 2009). However, log snags are often actively removed from waterways to improve navigability and boating safety, especially following flood events (Barrett and Ansell 2003; Erskine and Webb 2003; Lester and Boulton 2008). While efforts to re-snag rivers to replenish native fish habitat and secure estuarine banks do occur (Bond and Lake 2005; Brooks et al. 2004; Lester and Boulton 2008; Lyon et al. 2019), these actions predominantly occur in freshwater and not in coastal or estuarine settings (Lyon et al. 2019; Nicol et al. 2004).

It is well established that both the spatial (Henderson et al. 2020) and habitat condition attributes (Perry et al. 2023) of log snags affect their value for fish, but their combined influences remain poorly understood in many settings, including in estuaries. Estuarine seascapes that are comprised of highly connected and diverse habitats typically harbour more abundant and diverse fish assemblages and functional groups due to supporting species that use different habitats throughout their lifecycles for varying resources (e.g. nursey and feeding) (Pittman et al. 2004; Sheaves 2009; Whitfield 2017). Therefore, the abundance and diversity of fish species in log snag habitats may be modified by changes in connectivity and/or the context of habitat patches nearby, similar to that of other estuarine habitats (Gilby et al. 2018a; Green et al. 2012). For example, previous research focusing on broad seascape influences on fish in estuaries found that the variance in fish assemblages on log snags was best explained by the proximity to the estuary mouth (Goodridge Gaines et al. 2022; Henderson et al. 2019a). However, these effects were mediated by other spatial influences such as proximity to urban structures (Henderson et al. 2019a) and the area of saltmarsh and an intertidal flat nearby (Goodridge Gaines et al. 2022). Conversely, the variations in habitat conditions and complexity of log snags can affect their habitat value for fish (Newbrey et al. 2005). For example, species richness and harvested fish abundance on log snags can increase with greater oyster cover, while total fish abundance is higher on snags that had a lower percentage of algae cover (Perry et al. 2023). Disentangling how the spatial and condition attributes of log snag habitats affect fish assemblages will allow management to optimise restoration efforts in estuaries where this type of restoration has rarely been implemented.

The use of fish as indicators to assess the functioning, health and management of habitats is increasingly common for coastal seascapes (Gilby et al. 2018b; Whitfield and Elliott 2002). Within estuaries, there are six key functional groups that perform vital roles which maintain ecosystem resilience, health and functioning across seascapes: detritivores, herbivores, omnivores, piscivores, zoobenthivores and zooplanktivores (Elliott et al. 2007b; Whitfield et al. 2022). Here, some functional groups are generally associated with broad functional roles and feeding niches (e.g. zoobenthivores and omnivores), while others have more specialist roles (e.g. herbivores, detritivores, zooplanktivores and piscivores) (Elliott et al. 2007b; Whitfield et al. 2022). The distribution of these functional groups has previously been found to respond differently to the level of urbanisation in estuaries (Henderson et al. 2020) and general seascape patterning (Swadling et al. 2019). Here, identifying the distributions of fish and their functional groups can identify the spatial and/or habitat drivers of fish to help prioritise placement (Teichert et al. 2018; Zellmer et al. 2019) and/or monitor the progress or success/failure of restoration sites (Gilby et al. 2021b). For example, on log snags, establishing patterns of spatial positioning and habitat condition drivers for fish assemblages and different functional groups can help to prioritise the placement of log snags in estuaries. These findings, however, may result in differing or opposing spatial and condition drivers for different fish metrics (Perry et al. 2023) or their functional groups (Henderson et al. 2020; Swadling et al. 2019), and optimising restoration will depend on the desired outcome (e.g. fisheries value, herbivore abundance) (Duarte et al. 2020; Nash et al. 2016).

Coastal habitats are impacted by expanding and diversifying anthropogenic impacts globally (Halpern et al. 2015), resulting in an increased desire for active interventions that improve services such as fisheries (Creighton et al. 2015; Taylor et al. 2017). The resnagging of estuaries represents a possible strategy for restoring fish and fisheries across seascapes that is more resilient to long-term declines in coastal abiotic conditions such as water clarity and sedimentation. However, more information about how and where log snags most enhance fish and fisheries is required so that future actions can be optimised. In this study, we quantify the spatial and habitat attributes which most affect fish abundance and diversity, functional groups and harvested fish abundance of log snags in coastal seascapes. We hypothesise that log snags in closer proximity to the estuary mouth and to other estuarine habitats will support a higher abundance and diversity of fish. Similarly, log snags that have a greater food availability, especially greater invertebrate and/or algal cover, and that are more structurally complex will result in variable trends of the abundance of fish from different functional groups (e.g. higher abundance of herbivorous fish at log snags with greater algal cover).

Methods

Survey Sites

We surveyed fish assemblages and the condition of log snags across 13 estuaries in southeast Queensland, Australia, (Fig. 1) between the months of June and September for three consecutive years (2020–2022). This survey period was selected to align with the periods of maximum water clarity within the region, with these months experiencing lower rainfall on average compared to other times of the year (Bureau of Meteorology 2023). The estuaries stretched along approximately 220 km of coastline from Noosa River in the north to Tallebudgera Creek in the south and varied significantly in the distribution and area of natural habitats (e.g. mangrove, seagrass saltmarsh), extent and intensity of urban land (e.g. residential and commercial development) and proximity to large coastal cities. These differences in context both within and between estuaries allowed us to effectively disentangle how the area, connectivity and distribution of estuarine habitats affect fish assemblages on log snags. The estuaries also varied significantly in their tidal influence extent, with the tidal limit ranging from 7 km (Tallebudgera Creek) to 77 km (Brisbane River) upstream and averaging 23 km across estuaries.

Fig. 1
figure 1

Map of the survey estuaries in southeast Queensland, Australia. Inset of Pumicestone Passage shows the spread of the log snag RUVS deployments, with the colour indicating the year in which the survey was undertaken

Full size image

Fish Surveys

Fish were surveyed on ten log snags in each estuary, with sites spread as evenly as possible from the mouth of the estuary to the most upstream tidal limit in each estuary (Fig. 1). Surveys were replicated at each site in each year of the study where possible. However, log snags can break down over time and/or move due to water velocity changes from tidal or rainfall effects. Therefore, new log snags were selected in each subsequent year at sites with similar seascape positioning if previous sites were no longer able to be surveyed. This ensured that ten replicates were sampled per estuary each year. Fish were surveyed at each site each year using remote underwater video stations (RUVSs), which consist of a GoPro video camera recording in high definition (1080 p) mounted to the top of a 1 kg weight and buoyed at the surface for retrieval and to ensure the rope remains out of the view of the camera. RUVSs were deployed 2 h on either side of a daylight hightide to ensure consistency in water clarity as well as to ensure intertidal habitats had sufficient water depth. Each RUVS was deployed for 30 min and positioned so that the camera field of view was angled along the length of the log snag to capture fish transitioning along or moving in and out of the structure while ensuring the view was not obstructed by the log snag itself.

Fish assemblages were calculated from RUVS using the MaxN metric—the maximum number of individuals of the same species counted in a single frame at each site (Murphy and Jenkins 2010). From here, fish species richness was calculated by counting the number of unique species at each site, total fish abundance was calculated as the sum of MaxN values for each species at each site, and harvested fish abundance was calculated as the sum of MaxN values for each species harvested in local fisheries at each site. Fish functional group abundances were calculated as the sum of MaxN values for each species categorised into each functional group according to information on FishBase (Froese and Pauly 2000) and using the estuarine fish feeding categories and definitions described in Elliott et al. (2007b) (see Table S2).

Environmental Variables

Spatial Metrics

We quantified nine spatial metrics for each site using QGIS (QGIS Development QGIS Development Team 2022), including the area of mangrove, saltmarsh, seagrass, intertidal flats and urban land in 500 m buffers around each site, as well as the proximity of the sites to the estuary mouth, seagrass meadows, saltmarsh habitats and urban structures (Table 1). We chose 500 m buffers as our focal spatial scale as this represents the likely tidal movements of fish species commonly found within these seascapes (Gilby et al. 2018a; Olds et al. 2013). Log snags in these estuaries predominantly occur either upon intertidal flats and/or along the fringes of mangroves. Therefore, we excluded the proximity of our sites to these habitats as these measures had a significant number of both larger outliers and zero values that could not be corrected for in statistical analyses. We also measured water depth (in m) at the time of RUVS deployment using a Garmin depth sounder mounted to the survey vessel.

Table 1 Environmental variables included in the analysis
Full size table

Habitat Condition and Complexity

We quantified four variables indexing the condition, complexity and food availability of each log snag habitat in the field. Condition metrics included visual estimates of algae, barnacle and oyster cover (in percent of the total surface) of the total log snag (Table 1). These metrics were chosen as they represent important food resources for estuarine fish species (Anderson and Connell 1999; Whitfield 2017). Log snag complexity was calculated on a scale of 1 being a single log, up to 5 being a highly complex tree or a combination of trees with multiple branches (Table 1). Habitat condition and complexity metrics were quantified for each site after the retrieval of the RUVS and were repeated within each sampling event to account for changes that may occur within one site over the multiple years of the survey.

Water Quality Metrics

We sourced water quality data from the local waterways monitoring program (Ecosystem Health Monitoring Program (EHMP)) conducted by Healthy Land and Water (Healthy Land and Water 2023). Water quality is quantified for each system in standardised monthly surveys from multiple locations along the estuarine extent of the estuary (Healthy Land and Water 2021). As these sites do not align directly with our RUVS sites, water quality data was interpolated to survey sites using IDW interpolations in QGIS for both 6-month and 3-month periods prior to sampling commencing. Therefore, we included both the 6-month average and 3-month average of sampling in statistical models to account for changes in water quality across seasons prior to fish sampling (6-month average) and short-term retention time during the drier months of the year (3-month average). Three- and 6-month averages were chosen due to the large variations that estuaries experience over short time periods (e.g. tidal variations in nutrients) as well as to account for any discrepancies in water quality drivers between estuaries (Eccles et al. 2020).

Statistical Analysis

We tested all explanatory variables for collinearity using Pearson’s r correlations in R (R Core Team 2023). Where any two variables correlated with a Pearson’s r value of more than 0.7, or less than − 0.7, one variable was removed based on the strength of underlying hypotheses. This test was undertaken for the three broad environmental groups and then between the simplified environmental groups. In our water quality variables, the 3-month average and 6-month average values were collinear for all three water quality variables, chl-a (r = 0.726), turbidity (r = 0.891) and DO saturation (r = 0.902). Therefore, 3-month averages for all water quality metrics were selected for inclusion in the analysis. We chose the 3-month averages to account for short-term retentions in water quality metrics (e.g. variations in turbidity) that may have driven fish distributions in the months prior to surveys. In our spatial variables, distance to seagrass and distance to estuary mouth were colinear (r = 0.964) resulting in distance to seagrass being excluded from the analysis. Distance to estuary mouth was selected as it is commonly the most important variable in shaping fish assemblages within these estuaries (Goodridge Gaines et al. 2022; Henderson et al. 2019a; Jones et al. 2020).

We used ManyGLMs in the mvabund package (Wang et al. 2012) of R to identify environmental variables that best explained the structure of fish assemblages at sites. We ran a ManyGLM to quantify effects for fish assemblage structure (i.e. a matrix of sites by fish species and their MaxN values) on the full suite of environmental variables (Table 1) along with the main effects of estuary (fixed factor, 13 levels) and year (fixed factor, three levels). The ManyGLM was fit with a negative binomial distribution. The best-fit ManyGLM was identified using reverse-stepwise simplification on Akaike’s information criterion (AIC) values and was that with the lowest AIC. Indicator fish species were identified as the species that best correlated with the suite of variables in the best-fit model. The best-fit ManyGLM was visualised using a non-metric multidimensional scaling ordination (nMDS) plotted using R.

We further interrogated relationships between all environmental variables and the three compound fish assemblage metrics (i.e. fish species richness, total fish abundance and harvested fish abundance) and the abundance of six fish functional groups (omnivore, zooplanktivore, detritivore, zoobenthivore, herbivore and piscivore) using generalised linear mixed models (GLMMs) in the mgcv package (Wood 2018) of R. Model structure for each dependant variable (i.e. three compound fish assemblage metrics and six functional groups) was the same and included all 16 environmental variables (Table 1) with the random effects of estuary and year. All response variables were tested prior to analysis for the most appropriate distribution which resulted in variables being fit with negative binomial error distributions. We minimised GLMM overfitting by running all possible combinations of five variables or fewer using the MuMIn package (Barton 2019). Best-fit models were those with the smallest Akaike’s information criterion corrected for small sample sizes (AICc) (Barton 2019; Wood 2017), and only significant variables from the best-fit GLMM were plotted. Factor importance was calculated for each independent variable as the summed weighted AICc value for each model (Anderson and Burnham 2002). Here, a value closer to 1 indicates a stronger effect of the variable on overall patterns for that dependent variable, while a value of 0 indicates no effect.

Results

Influence of Environmental Metrics on Fish Assemblages

We counted 5803 fish from 59 species on 363 log snag sites across 13 estuaries over the 3 years of surveys in southeast Queensland (Table S2). We identified two species which were the most abundant throughout this study which accounted for 52.4% of the total abundance: yellowfin bream Acanthopagrus australis (18.9%) and estuary glassfish Ambassis marianus (33.5%) (Table S2). Twenty-six species were only identified once or twice across the 363 sites, with 69% being non-fisheries targeted species and predominantly from functional groups with broad feeding niches (i.e. 85% were a zoobenthivore or omnivore) (Table S2).

The structure of fish assemblages was best explained by the combined effects of the proximity of sites to the estuary mouth (χ2 = 250.8, p < 0.01), the proximity to urban structures (χ2 = 110.2, p = 0.011), area of urban land (χ2 = 180.4, p = 0.037) and water depth (χ2 = 124.4, p < 0.01) (Fig. 2). The best-fit ManyGLM also included three non-significant variables: 3-month average turbidity (χ2 = 146.4, p = 0.065), 3-month average chl-a (χ2 = 143.2, p = 0.060) and 3-month average DO sat (χ2 = 124.1, p = 0.123); however, the removal of these variables lowered the overall fit of the model, and therefore, they were included in the final best-fit model. We identified four species which best explained the variation in assemblage structure across sites and variables in the best-fit model: yellowfin bream, fineline surgeonfish Acanthurus grammoptilus, silver biddy Gerres subfasciatus and black rabbitfish Siganus fuscescens (Fig. 2). Here, there was evidence to indicate that silver biddy was more abundant and prevalent at sites with a greater area of urban land that was an intermediate distance from urban structures and the estuary mouth and had a lower than average 3-month turbidity. Yellowfin bream, fineline surgeonfish and black rabbitfish were more abundant and prevalent at sites that were closer to the estuary mouth and urban structures, sites that were deeper and encompassed an intermediate area of urban land nearby, as well as sites with a higher 3-month average dissolved oxygen, and a lower 3-month average chlorophyll-a concentration (Fig. 2).

Fig. 2
figure 2

Non-metric multidimensional scaling (nMDS) ordination with Pearson’s vector overlays illustrating the significant best-fit environmental variables (black vectors), non-significant best-fit environmental variables (grey vectors and italicised text) and indicator species (dashed vectors) from the ManyGLM. Fish images courtesy of efishalbum.com

Full size image

Species Richness, Total Fish Abundance and Harvested Fish Abundance

Variability in fish species richness was best explained by the proximity of sites to the estuary mouth (χ2 = 27.19, p < 0.001) and urban structures (χ2 = 7.26, p = 0.007), the depth of the site (χ2 = 6.39, p = 0.011) and the random effect of estuary (χ2 = 106.95, p < 0.001) and year (χ2 = 38.02, p < 0.001) (Tables 2A, S3, Fig. 3A). Fish species richness was highest at sites within 20,000 m of the estuary mouth, at distances greater than 2500 m from urban structures and at depths greater than 3 m. The random effect of estuary and year were the most important variables affecting fish species richness (I = 1.00), followed closely by distance to estuary mouth (I = 0.99) and then water depth (I = 0.67) and distance to urban structures (I = 0.67) (Tables 2A, S4).

Table 2 Best-fit generalised linear mixed models (GLMMs) showing associations between (A) fish compound metrics and (B) the functional groups with environmental variables
Full size table
Fig. 3
figure 3

Generalised linear mixed model (GLMM) outputs illustrating effects of variables in the best-fit models for the fish assemblage compound metrics A species richness, B total fish abundance and C harvested fish abundance. Fish images courtesy of efishalbum.com

Full size image

Variability in total fish abundance was best explained by the proximity of sites to the estuary mouth (χ2 = 22.94, p < 0.001), algae cover (χ2 = 30.49, p < 0.001), the random effect of estuary (χ2 = 145.60, p < 0.001) and year (χ2 = 12.15, p = 0.005) (Table 2A, S3, Fig. 3B). Total fish abundance was higher at sites within 20,000 m of the estuary mouth and on log snags with less than 25% algae cover. The random effect of estuary and the proximity to estuary mouth were the most important variables affecting total fish abundance (both I = 1.00), followed closely by algae cover (I = 0.99) and lastly year (I = 0.19) (Tables 2A, S4). The 3-month average turbidity was selected in the best-fit GLMM but did not significantly explain total fish abundance (χ2 = 3.37, p = 0.06) and was not as important as most of the other variables (I = 0.30).

Variability in harvested fish abundance was best explained by the proximity of sites to the estuary mouth (χ2 = 11.76, p < 0.001), percent of algae cover (χ2 = 13.66, p < 0.001) and the random effect of estuary (χ2 = 129.20, p < 0.001) (Tables 2A, S3, Fig. 3C). Harvested fish abundance was higher at sites within 20,000 m of the estuary mouth and on log snags with less than 25% algae cover. The random effect of estuary was the most important variable affecting the abundance of harvested fish (I = 1.00), followed by algae cover (I = 0.98) and proximity to the estuary mouth (I = 0.79) (Tables 2A, S4). Log snag complexity and 3-month average DO sat were selected in the best-fit GLMM but did not significantly explain harvested fish abundance (χ2 = 2.62, p = 0.105 and χ2 = 3.03, p = 0.82, respectively), and both variables were not as important as the other variables (I = 0.18, I = 0.29, respectively).

Functional Groups as Indicators

Variability in omnivore abundance was best explained by the proximity of sites to the estuary mouth (χ2 = 4.85, p = 0.027), the area of mangrove (χ2 = 10.22, p = 0.001) and saltmarsh (χ2 = 5.01, p = 0.025) nearby, the 3-month average chl-a concentration (χ2 = 5.23, p = 0.022) and the random effect of estuary (χ2 = 24.83, p < 0.001) (Tables 2B, S3, Fig. 4A). Omnivore abundance was higher at sites within 10,000 m of the estuary mouth, with less than 70,000 m2 of mangrove and more than 175,000 m2 of saltmarsh nearby and less than 5 µg/L 3-month average chl-a concentration (Fig. 4A). The random effect of estuary was the most important variable (I = 0.96) affecting omnivore abundance, followed by mangrove area (I = 0.67), 3-month average chl-a concentration (I = 0.60), distance to the estuary mouth (I = 0.51) and saltmarsh area (I = 0.34) (Tables 2B, S4).

Fig. 4
figure 4

Generalised linear mixed model (GLMM) outputs illustrating effects of variables in the best-fit models for the abundance of fish in different functional groups: A omnivore, B zooplanktivore, C detritivore, D zoobenthivore, E herbivore and F piscivore. Pictures A–F show well-known species within that functional group that were abundant throughout the study. Fish images courtesy of efishalbum.com

Full size image

Variability in zooplanktivore abundance was best explained by the proximity of sites to the estuary mouth (χ2 = 3.85, p = 0.049), algae cover (χ2 = 30.07, p < 0.001), the 3-month average DO sat (χ2 = 12.93, p < 0.001) and the random effect of estuary (χ2 = 32.86, p < 0.001) (Tables 2B, S3, Fig. 4B). Zooplanktivore abundance was higher at sites within 20,000 m of the estuary mouth, on the log snag with less than 25% algae cover and at sites with a greater than 100% 3-month average DO sat percent (Fig. 4B). Algae cover was the most important variable (I = 0.97) affecting zooplanktivore abundance, followed by the 3-month average DO sat (I = 0.78), the random effect of estuary (I = 0.64) and lastly distance to the estuary mouth (I = 0.48) (Tables 2B, S4). The oyster cover was included in the best-fit model but did not significantly explain zooplanktivore abundance (χ2 = 3.08, p = 0.079) and was the least important variable (I = 0.27).

Variability in detritivore abundance was best explained by the proximity of sites to urban structures (χ2 = 4.50, p = 0.034), depth (χ2 = 9.61, p = 0.002) and the random effect of estuary (χ2 = 65.77, p < 0.001) (Table 2B, Fig. 4C, Table S3). Detritivore abundance was higher at sites more than 3000 m from urban structures and greater than 4 m depths (Fig. 4C). The random effect of estuary was the most important variable (I = 1.00) affecting detritivore abundance, followed by distance to urban structures (I = 0.75) and depth (I = 0.62) (Tables 2B, S4).

Variability in zoobenthivore abundance was best explained by the proximity of sites to the estuary mouth (χ2 = 25.26, p < 0.001), urban structures (χ2 = 4.56, p = 0.032), the 3-month average DO sat (χ2 = 6.70, p = 0.009), the random effect of estuary (χ2 = 152.49, p < 0.001) and year (χ2 = 17.72, p < 0.001) (Tables 2B, S3, Fig. 4D). Zoobenthivore abundance was higher at sites within 15,000 m of the estuary mouth, more than 3000 m from urban structures and at sites with a greater than 80% 3-month average DO sat percent (Fig. 4D). The distance to estuary mouth and the random effect of estuary were the most important variables (I = 1.00) affecting zoobenthivore abundance, followed by the random effect of year (I = 0.97), 3-month average DO sat (I = 0.39) and lastly distance to urban (I = 0.27) (Tables 2B, S4).

Variability in herbivore abundance was best explained by the area of intertidal flat nearby (χ2 = 14.63, p < 0.001), depth (χ2 = 11.12, p < 0.001), 3-month average DO sat (χ2 = 4.84, p = 0.028), 3-month average chl-a concentration (χ2 = 9.03, p = 0.002) and the random effect of estuary (χ2 = 52.38, p < 0.001) (Tables 2B, S3, Fig. 4E). Herbivore abundance was higher at sites with greater than 500,000 m2 area of intertidal flat nearby that were greater than 4 m of depth and sites that had a greater than 30 µg/3-month average chl-a concentration and greater than 100% 3-month average DO sat percent (Fig. 4E). The random effect of estuary was the most important variable (I = 1.00) affecting herbivore abundance, followed by intertidal area (I = 0.81), 3-month average chl-a concentration (I = 70), depth (I = 0.64) and lastly 3-month average DO sat (I = 0.52) (Tables 2B, S4).

Variability in piscivore abundance was best explained by depth (χ2 = 13.30, p < 0.001) and the random effect of estuary (χ2 = 47.11, p < 0.001) (Tables 2B, S3, Fig. 4F). Piscivore abundance was higher at sites with greater than 4 m depth. The random effect of estuary was the most important variable (I = 1.00) affecting piscivore abundance, followed closely by depth (I = 0.98) (Tables 2B, S4).

Discussion

Coastal ecosystems are facing diverse and intensifying impacts (Halpern et al. 2008; Kennish 2002), causing declines in the extent, condition and connectedness of natural habitats and reductions in the supply of key ecosystem services (Bishop et al. 2017). Consequently, there is an increasing desire globally for active interventions such as ecological restoration that work to reverse these impacts (Abelson et al. 2020; Duarte et al. 2020; Waltham et al. 2020). In this study, connectivity to the estuary mouth was the main driver in structuring fish species richness, total fish abundance, harvested fish abundance and the abundance of three functional groups: omnivores, zooplanktivores and zoobenthivores. Similarly, several key functional groups (e.g. omnivores, zoobenthivores, zooplanktivores and herbivores) were each modified by different connectivity (e.g. distance to estuary mouth and urban structures) and some context variables (e.g. area of intertidal flat and urban land), but additionally, a series of water quality variables including chlorophyll-a concentration and dissolved oxygen saturation. We highlight one habitat condition attribute that had significant influences on fish assemblages, with a lower algae cover resulting in a higher abundance of total and harvested fish, and the abundance of zooplanktivores. This study sought more specific recommendations for resnagging projects while complementing the findings of previous research in this region which highlighted the importance of seascape connectivity between multiple estuarine habitats (Gilby et al. 2018a; Henderson et al. 2019a). Strategically placing log snag habitats throughout seascapes increases their effectiveness for enhancing fish diversity, abundance and fisheries values across estuaries.

Greater connectivity with the estuary mouth has been found consistently to enhance fish species richness, total abundances and fisheries value in estuarine habitats and therefore the outcomes of coastal management efforts such as restoration (Nagelkerken et al. 2015; Sheaves 2009; Teichert et al. 2018). Proximity to the estuary mouth is often associated with better and more consistent water quality (Healthy Land and Water 2023), which may be attributed to increased tidal flow (Costa et al. 2018). This has been associated with more consistent fish habitat use in lower estuarine sections (Harrison and Whitfield 2021; Whitfield and Elliott 2002) as well as increased fish distributions that track variations in salinity (Whitfield et al. 2023). In this study, distance to estuary mouth correlated with distance to seagrass meadows, which is likely to be contributing towards variations in fish assemblages given the importance of seagrass in structuring fish assemblages (Gilby et al. 2018a; Swadling et al. 2019) and fisheries production (Unsworth et al. 2019). Furthermore, we chose to use distance to estuary mouth instead of salinity as it provides clearer management targets across spatial scales (Gilby et al. 2021a) that are less variable over daily or seasonal changes (e.g. rainfall) (Healthy Land and Water 2023). Managers often seek habitat restoration opportunities that have the capacity to enhance fish assemblages in more upstream or urbanised locations (Elliott et al. 2007a). Despite this finding, log snags still seem an appropriate selection due to the structures not being restricted by water quality parameters unlike other estuarine habitats (e.g. seagrass). Previous studies have found that these habitats to support more abundant and diverse fish assemblages in upper reaches of estuaries compared to other habitats, in particular unvegetated muds and sands (Goodridge Gaines et al. 2022). Therefore, installing log snags throughout these unvegetated sections will likely increase the fish metrics we are interested in (e.g. harvested fish) relative to the minimum baseline. While spatial variables (specifically distance to estuary mouth) were identified in this study as the key drivers of fish assemblages on log snag habitats, there were some consistent water quality variables identified that helped to disentangle these trends that can assist with prioritising management.

Coastal water quality has been well established as a key driver of estuarine fish assemblages both globally (Teichert et al. 2016) and within Australia (O’Mara et al. 2016). Although some water quality variables were identified in this study as significant drivers of the abundance of individuals in some fish functional groups, this result was not consistent across the broader fish assemblage or compound fish metrics such as abundance and richness. Chlorophyll-a concentration, turbidity and dissolved oxygen saturation were found in the best-fit model for structuring fish assemblages on log snags, and while these variables did not have a significant effect, the influence of chlorophyll-a concentration and turbidity were only marginally insignificant in structuring fish assemblages. Conversely, the abundance of some of the functional groups was significantly influenced by chlorophyll-a (omnivores), dissolved oxygen saturation (zooplanktivore and zoobenthivore) and a combination of both (herbivore). While most of these trends were as expected (e.g. lower chlorophyll-a may indicate better water quality from reduced nutrient input in the estuary resulting in more consistent conditions for fish), herbivore abundance was higher at sites with a greater concentration of chlorophyll-a. Here, the increased nutrients may be providing enhanced feeding opportunities for herbivores, but this pattern requires further investigation. Consequently, these specific drivers for functional groups may highlight the need to incorporate water quality variables into monitoring programs to assist in tracking the distributions of certain fish species (e.g. herbivorous fish). Incorporating potential thresholds of water quality parameters into coastal management planning may also assist in achieving better outcomes for resnagging placement.

Heterogeneous seascapes with diverse coastal habitats are typically found to support higher fish species richness and have greater fisheries value (Meynecke et al. 2008; Sheaves 2009; Whitfield 2017). In this study, while it was found that seascape connectivity between habitats was important for some functional groups (i.e. log snags near a lower area of mangrove and greater area of saltmarsh patches supported a higher abundance of omnivores), the proximity of the log snag to urban structures was important in explaining the variation in species richness and the abundance of two functional groups, detritivores and zoobenthivores. Here, log snags that were further from urban structures supported more diverse fish assemblages and a higher abundance of detritivores and zoobenthivores, indicating that urbanisation had a negative effect on fish on log snags. Conversely, previous research looking at the broader urban land and urban shoreline effects on fish in estuaries found a positive influence of urban shorelines on species richness and the abundance of some functional groups with urban shorelines (Brook et al. 2018; Yabsley et al. 2020). These findings were mediated by key estuarine water quality parameters (e.g. chlorophyll-a concentration) (Yabsley et al. 2020) and the extent of some coastal vegetation (e.g. mangrove area) (Brook et al. 2018) within the estuary. Similarly, we show that the effect of urban structures on fish was mediated by a suite of spatial and water quality variables, particularly the depth of the site. Log snag sites that were deeper supported a greater species richness and greater abundance of detritivores, herbivores and piscivores. These findings are consistent with previous research highlighting the importance of deeper and readily accessible estuarine habitats for fish (Bradley et al. 2017). Having subtidal refuges that are connected to intertidal areas is crucial for fish that utilise habitats that become exposed on a low tide (Teichert et al. 2018). Consequently, prioritising the placement of log snag in deeper sections throughout the estuary and further from urban structures will support increased outcomes for fish species richness and several fish functional groups; however, placement may be restricted by the prioritisation of navigational safety for vessels (Miller and Craig Kochel 2010).

Quantifying how the structure and complexity of estuarine habitats affect fish assemblages can help guide the type, design and placement of estuarine restoration projects (Gilby et al. 2018b; Goodridge Gaines et al. 2022; Perry et al. 2023). For example, the complexity of habitat can increase the rates of important functions performed by fish (Mosman et al. 2023), and a greater complexity of log snags can support important food resources for fish (O’Connor 1991). Understanding the subtleties in different conditions and complexity drivers of fish and functional groups on log snags in estuaries will assist in the design of resnagging efforts. In this study, we show that total and harvested fish abundance, and the abundance of zooplanktivores, were higher at sites with a lower cover of algae which is a prominent food source in estuaries. Here, log snags with a lower cover of algae contain more fish, so placing resnagging locations in areas of less nutrients and high abundance of fish that eat algae (e.g. herbivores) may result in the benefits of log snags being greater. This finding is closely related to the spatial effects regarding the distance to estuary mouth, because upper estuarine locations tend to have poorer water quality (Healthy Land and Water 2023). Therefore, these two variables are likely to be working in concert, and jointly implementing land-based restoration efforts to improve upstream water quality can assist in the success of estuarine restoration efforts; however, this requires further research. Aside from algae cover, we found no significant effects of habitat condition and complexity structuring the broader fish community or fish functional groups in this study, which resembles the findings of previous studies where seascape effects were more important in structuring fish assemblages compared to habitat condition (Goodridge Gaines et al. 2022). Throughout estuaries, prioritising the placement of log snags will likely have the most significant influence on enhancing fish abundance, diversity and the abundance of different fish functional groups.

Previous efforts throughout Australia to restore fish habitat through resnagging have predominantly been undertaken in freshwater systems (Lester et al. 2020; Nicol et al. 2004). Few studies have looked specifically at fish assemblages within log snag habitats (Goodridge Gaines et al. 2022; Henderson et al. 2019a); however, this research was inclusive of all estuarine habitats. Comparatively, other estuarine habitats are often the sole focus of research, subsequently leading to more enhanced management priorities due to the values of these habitats being well understood (Henderson et al. 2019b; Romañach et al. 2018). Consequently, further research into other benefits or ecosystem services that log snags provide both estuarine fish (e.g. nursery) and humans (e.g. fisheries biomass) will enhance the desire to prioritise resnagging efforts.

Quantitatively understanding how attributes of log snags modify their value for fish can help to optimise the design and placement of resnagging actions throughout coastal seascapes (Beck et al. 2001; Goodridge Gaines et al. 2022; Teichert et al. 2018). Log snags are naturally occurring habitats; however, they are poorly studied and often overlooked in coastal restoration planning, even though they show significant potential as important fish habitats (Goodridge Gaines et al. 2022). Log snags offer an opportunity for the financially efficient enhancement of coastal fish assemblage in areas where other estuarine habitats might not be able to be restored. Maximising the benefits of resnagging efforts for fish and fisheries in estuaries requires a significant focus on site selection. Connectivity with the estuary mouth is crucial for maximising benefits, but resnagging might also be highly beneficial in upstream areas where the distribution of other habitats is restricted (i.e. seagrass restricted due to poor water quality parameters). Here, site selection will need to consider water quality parameters to ensure algal growth can be limited or plan to incorporate land-based management to help benefit the outcomes of restoration efforts. Site selection within estuaries will also experience further restrictions due to navigational passageways for boats (Miller and Craig Kochel 2010) and having to be placed away from fast-moving water areas to avoid potential dislodgement or increased amounts of breakdown (Gonor et al. 1988; Hinwood and Mclean 2017). Identifying, understanding and implementing restoration efforts, such as resnagging, in coastal seascapes are crucial to overcome habitat loss and degradation in estuaries. Our findings offer new insights into resnagging efforts in estuaries and show that efforts might be optimised with strategic design and placement across seascapes.