Structural complexity of seagrass and environmental variables as a determinant of fish larvae assemblages in tropical coastal waters: Implications for seagrass management and conservation

Anthropogenic activities including climate change affect the development of structural complexity in seagrass and the state of environmental variables. It remains unclear whether these variables, alone or in combination, have an important influence on fish larvae assemblages. This study examined the influence of the structural complexity of seagrass meadows and various environmental variables on fish larvae assemblages in tropical coastal waters of Tanga, Tanzania. The study was conducted in four Thalassia hemprichii dominated seagrass meadows from June 2019 to January 2021 . Multiple regression analysis indicated that the structural complexity of seagrass (canopy height, seagrass cover, and shoot density) and environmental variables (dissolved oxygen, temperature and salinity) were the foremost predictors for fish larvae assemblages; abundance and richness (R 2 = 0.75, p = 0.0185, R 2 = 0.54, p = 0.0396, respectively). Based on these findings, the structural complexity of seagrass and environmental variables are both important determinants of fish larvae assemblages in tropical coastal waters. The findings suggest that reducing anthropogenic activities that affect the development of structural complexity of seagrass and negatively impact environmental variables in seagrass meadows through more effective governance would result in increased production of fish larvae in meadows and, as a result, increased fish recruitment in tropical coastal waters.


Introduction
Anthropogenic activities, including climate change, are increasingly affecting the health and function of seagrass meadows (Dunic et al., 2021), with significant impacts on the recruitment of fish stocks (Waycott et al., 2009;Brodie and De Ramon N'Yeurt, 2018;Hedberg et al., 2019). Threats directly affecting the health and functions of seagrass meadows include destructive fishing methods such as drag-net fishing, the use of beach seines, ring nets, gleaning, trampling, pulling or pushing boats towards deeper waters, surface rain runoff, and excessive nutrient and sediment fluxes from activities related to agriculture Erzad et al., 2020). Also, increased populations of grazers, such as sea urchins, can cause local overgrazing of seagrasses. Increases in grazers are associated with overfishing of predatory fish that feed on sea urchins (Eklöf et al., 2008).
These threats underscore the need for effective conservation and governance to address the pressures that impact the ecosystem function of seagrass meadows in coastal waters.
They are distinguished by an abundance and diversity of fish larvae, which play an important role in recruitment of fish stocks in coastal waters (Cullen-Unsworth and Unsworth, 2016;Unsworth et al., 2019). While relatively high abundance of fish larvae in seagrass meadows is often attributed to the availability of prey (Vonk et al., 2010), the structural components of seagrass meadows could diminish predator foraging efficiency (Lugendo et al., 2007;Muhando and Rumisha 2008;Hedberg et al., 2019) and be important in attracting fish larvae seeking refuge (Gillanders, 2006;Lugendo et al., 2007;Gullström et al., 2008;Jones et al., 2021;Tarimo et al., 2022). The degree of structural complexity in seagrass meadows is influenced by the local environment (Huwer et al., 2016), which also has an impact on the fish larvae assemblages. Furthermore, seagrass plays a crucial role in combatting climate change (Uku et al., 2022), ensuring food security, protecting coastlines, and biodiversity enhancement (Nordlund et al., 2016;Brodie and De Ramon N'Yeurt, 2018).
Seagrass cover, shoot density, canopy height, length and width of leaves, and number of leaves per shoot are used as a structural complexity measure or indicators, and have been shown to decrease with disturbance in previous studies (Hedberg et al., 2019;Jones et al., 2021;Mwaluma et al., 2021). Research on how the complexity of seagrass structures affect fish larvae assemblages is lacking. A few studies in the Western Indian Ocean (WIO) region have examined the impact of seagrass complexity on juvenile, sub-adult, and adult fishes but not on fish larvae (Gullström et al., 2006;Palmqvist et al., 2013;Hedberg et al., 2019;Jones et al., 2021). Other studies focused on seasonal patterns of fish larvae in mangrove creeks, and inshore seagrass meadows (located adjacent to mangroves) (Lugendo et al., 2007;Tarimo et al., 2022). In other geographical areas, studies focused on the complexity of vegetated areas on fish larvae distribution and variability (Rappe et al., 2013;Erzad et al., 2020). Despite these studies, there is limited information on the impact of tropical seagrass structural complexity alone or in conjunction with environmental variables on fish larvae assemblages, making it difficult to determine which characteristics are crucial for setting management priorities (Molina et al., 2020). The present study was designed to examine the relative importance of seagrass structural complexity and environmental variables on fish larvae assemblages (abundance and family richness) in tropical coastal waters. The explicit hypotheses was tested that abundance and family richness of fish larvae are determined by (1) seagrass structural complexity (seagrass percentage cover, shoot density, and canopy height), and (2) environmental variables (temperature, dissolved oxygen, pH, salinity, and water depth).
The selection of sites was based on the presence of seagrass meadows influenced by varying degrees of anthropogenic disturbance, affecting the development of seagrass structural complexity. In general, seven seagrass species were present in the surveyed areas of which Thalassia hemprichii was dominant.
The data collected were for the single species Thalassia hemprichii, based on the finding of Jones et al., (2021) that seagrass diversity (both functional and species) had minimal effect on fish assemblages. Therefore, in this study it was decided to concentrate on the single dominant species.
These sites experience varying degrees of anthropogenic disturbances that impact on the development of seagrass structural complexity. Fungu ya Kaangoni (ST2) and Mwamba Karange (ST4) were characterized by reduced intensity and frequency of fishing, and anthropogenic activities that impact on seagrass beds, as well as natural factors like the influence of seasonal streams inflow, which brings sediments from land sources, as these sites are comparatively far from the coastline (about 10 km away from the coast). Kitanga (ST1) and Nyonza (ST3) are located nearshore, where the majority of damaging fishing practices (e.g., drag nets) are carried out and streams flow directly into these sites, bringing sediments and wastes from agricultural and industrial activities and contribute to impacts on these sites. While Nyonza (ST3) is influenced by the Kisare stream, Kitanga (ST1) is influenced by the Koreni stream. These streams transport domestic waste, sediment from land-based operations, nutrients, or fertilizers from sisal estates during the rainy season. Furthermore, these sites are impacted by fishing activities (the use of ring nets, gleaning, beach seines and other fishing methods), trampling, and pulling or pushing boats towards deeper waters.
The study sites are influenced by southeast and northeast monsoon winds (Peter et al., 2021), which affects water temperature, wind, rainfall, water circulation, wave action, and biological processes. The southeast monsoon season (SEM), from May to September, is characterized by strong winds (blowing relatively strongly from the southeast towards the northwest, at a speed of about 9 ms -1 ), heavy rains, and low air temperatures. The northeast monsoon season (NEM), from November to March (Peter et al., 2018), is characterized by steady winds (blowing from the northeast towards the southwest at about 5 ms -1 ), short rainy periods, and high air temperature (Peter et al., 2021). Field surveys in seagrass meadows were conducted during spring low tides while fish larvae sampling was conducted over the seagrass meadows during the corresponding high tides.
During the SEM, sampling was conducted for four months; June and August (2019) as well as July and September (2020). During the NEM season, sampling was also conducted for four months; December (2019), February and November (2020) as well as January (2021).

Field sampling and laboratory procedures
Environmental variables, including temperature, salinity, pH, dissolved oxygen, and water depth were measured directly in the field. Temperature and dissolved oxygen (DO) were measured using a thermometer with a temperature sensor and a DO meter (ECOSENSE DO 200A), respectively. Salinity was measured using a refractometer (RS 20). The pH was recorded using a pH meter (HANNA S8128) and water depth was recorded using an echo sounder (speed tech instrument 4308055). All equipment used were handheld.
At each seagrass meadow site, two transects were established perpendicular to the shoreline covering upper, middle, and lower zones. These transects were set 100 m apart to capture site representation. On each transect, three plots in each zone were randomly selected using a 0.25 x 0.25 m quadrat, for nine plots in total. In each quadrat, data for seagrass cover, canopy height, and shoot density were recorded. Seagrass species were identified in situ using field manuals appropriate for the region (Richmond, 2002). Shoots of the dominant seagrass species, Thalassia hemprichii, were counted and then used to estimate shoot density. Seagrass shoot density was determined as the number of individual seagrasses in a quadrat, which was expressed in a square meter area (m 2 ) (Erzad et al., 2020). Seagrass percentage cover of T. hemprichii was determined by visual estimate using 0.25 x 0.25 m quadrats (Saito and Atobe, 1970). Within the quadrat, canopy height of T. hemprichii was measured using a ruler (30 cm).

Data analysis
Before statistical analyses, the assumption of homogeneity of variance was checked by using Shapiro-Wilk's test at the significance level of p < 0.05. Fish larvae abundance and environmental variable data were log 10 (x+1) transformed when necessary based on the values of skewness. This was carried out using R statistical programming version 4.1.2 software. When the data remained heteroscedastic despite transformations, hypotheses were rejected at alpha levels lower than the p-values of the Shapiro-Wilk's test. Data of seagrass structural complexity measurements (i.e., percentage cover, shoot density, and canopy height) and fish larvae abundance were analyzed using Analysis of Variance (ANOVA) to compare the means and state significant differences, followed by Tukey's post hoc test in the four sites. A two sample t-test was used to test the seasonal difference between the SEM and the NEM seasons. Multiple linear regression analysis was used to explore the relative importance of various continuous variables: seagrass structural complexity (percentage cover, shoot density, and canopy height); and environmental variables (temperature, dissolved oxygen, pH, salinity, and water depth) on fish larvae assemblages. Moreover, before the analysis, all predictor variables were checked for collinearity. The data for testing the response of fish larvae abundance and fish family richness were grouped into two distinct groups: (1) seagrass structural complexity; and (2)  Linear mixed-effects were used to explore the relative importance of three seagrass structural complexity variables and five environmental variables on two fish response variables; fish larvae abundance and fish family richness. Multivariate analysis of the fish larvae assemblage was performed using PRIMER ver. 6.1.2 software (Plymouth Routines in Multivariate Ecological Research) (Clarke and Warwick, 2001).
Two-way crossed analysis of similarities (ANOSIM) was used to test for differences in fish larvae assemblages among sites. Patterns of similarities were visualized using non-parametric multidimensional scaling (nMDS) based on the Bray-Curtis similarities measure (a well-suited similarities index since it does not require exclusion of rare species or family), calculated by means of square root-transformed data.
The similarity of percentages (SIMPER) procedure was carried out to determine which fish larvae family contributed most to dissimilarities among the different study sites. To determine the degree of correlation between 2 independent distance (dissimilarity or similarity) matrices, the Mantel test was applied whereby a randomization technique to test whether dissimilarity matrices of fish assemblages and habitat variables (i.e., seagrass structural complexity, environmental variables) showed association among sites (Mantel, 1967

General description of environmental variables, fish larvae assemblages and seagrass structures
Variations in environmental variables in the different seagrass meadow sites and seasons are presented in Table 1. There were no statistically significant differences in environmental variables among sites (p > 0.05). However, a two-sample t-test revealed a significant seasonal difference in environmental variables (p < 0.05), except for the depth, as presented in Table 1. During the SEM season, dissolved oxygen, and salinity levels were higher than during the NEM season. Temperature and pH, on the other hand, were significantly higher in the NEM season than in the SEM season. In the present study, there were no significant seasonal differences in fish larvae assemblage and seagrass structures between SEM and the NEM (p > 0.05) ( Table 2).
For structural complexity variables (mean seagrass percentage cover, and canopy height), there was a significant difference among seagrass meadow surveyed sites (p = 0.000, p = 0.022 respectively). In contrast, estimates of the mean shoot density were comparable with no significant differences among sites (p = 0.16).
There was significantly higher seagrass cover at sites ST2 and ST4 than ST1 and ST3, while canopy height was significantly higher at sites ST1 and ST3 than at ST2 and ST4.
A total of thirty-eight (38) fish larvae families were identified (Fig. 3). One-way ANOVA showed a significant difference (p = 0.013) in fish larvae abundance (number of individual families per m 3 ) among study sites (Fig. 4).      Table 5. Fish families contributing (by > 5%) to dissimilarities (cumulative limit of 68%) among sampling sites (legend described in Fig. 1  showed a positive prediction of fish larvae family richness, but they were not statistically significant (p > 0.05).
On the other hand, environmental variables significantly predicted fish larvae family (R 2 = 0.73, p = 0.013). Temperature and salinity were found to be negatively correlated with fish larvae family richness (p = 0.03 and p = 0.04, respectively). Dissolved oxygen positively correlated with fish larvae family richness while pH and depth gave a negative correlation, but all were not statistically significant (p > 0.05; Table 4). are all seagrass residents. These were also the families that contributed most to dissimilarities in the fish larvae assemblage structure among study meadow sites (Table 5). The study was further confirmed by nMDS analysis that reflected the analogous pattern of grouping among the sites as observed in the cluster analysis ( Figs. 6a and b). The group average similarity between sampling sites ST1 and ST3 showed a similar pattern, comprising 80% similarity (Fig. 6a). Furthermore, sites ST2 and ST4 formed a separate group of less than 50% as shown in Figure 6 (a and b). The stress value was less than 0.1, which is a good ordinance pattern and a perfect description of the observed data for distances between sample sites. Both plots are based on the Bray-Curtis similarities index using square-roottransformed fish larvae abundance data.

Environmental variables
Distribution of fish larvae within seagrass nursery areas differ between families and species, and depend on environmental variables (Palmqvist et al., 2013).
Environmental variables play an important role in fish larvae assemblage structure (Molina et al., 2020).
Fish larvae are distributed across a wide range of environmental conditions, yet the presence or abundance of some families or species is limited by factors such as dissolved oxygen, pH, temperature, salinity, and water depth . The variation in environmental variables is influenced by a range of factors including climatic, hydrological, geological, and anthropogenic stress (Hedberg et al., 2019).
From the present study there was no differences in the water temperature, dissolved oxygen, pH, and salinity among the sites; this lack of variation may have been caused by constant water mixing, and the patterns of the current within the relatively shallow tropical seagrass habitat (Perez-dominguez et al., 2006). The seasonal difference in environmental variables are commonly related to seasonal monsoonal weather (most pronounced in the upper layer of the water column) and oceanographic conditions (McClanahan, 1988).
The average water temperature in this study was higher during the NEM season due to longer exposure to sunlight radiation (McClanahan, 1988).
Additionally, in the SEM, lower temperatures are caused by strong winds that cause deep mixing, thereby bringing colder waters to the surface (Peter et al., 2021). There was a significant seasonal variation in dissolved oxygen, and pH was higher during the SEM than the NEM season. Similarly, salinity was slightly higher during the SEM than during the NEM season, probably due to surface runoff caused by the rains during the NEM, which is supported by the work of Giering et al., (2019) and Peter et al., (2021). During the NEM season, however, pH was slightly higher than during the SEM season, presumably due to runoff from nearby agricultural areas carrying organic wastes (Levinton, 2001;Dhanam et al., 2016).

Effect of seagrass structural complexity on fish larvae assemblages
Previous studies have identified that individual factors, such as the characteristics of seagrass meadows (Palmqvist et al., 2013;Zerrato and Giraldo, 2018;Jones et al., 2021;Mwaluma et al., 2021) and environmental variables (Reynalte-tataje et al., 2012;Molina et al., 2020), affect the spatial patterns and variability in seagrass fish larvae assemblages. The present study showed that, when looking at abundance and richness, a number of predictor variables affect fish larvae assemblages in tropical coastal waters. In terms of seagrass structural complexity, it was discovered that seagrass cover, shoot density and canopy height all have a significant impact on fish larvae abundance, while canopy height has a significant impact on family richness. These findings concur with those of Gullstrom et al., (2008) and Jones et al., (2021), who observed that the seagrass cover and canopy height, which served as a measure of the complexity of the seagrass, had an impact on fish abundance and richness in coastal waters. The abundance of fish larvae and family richness were found to be strongly related to the canopy height of the seagrass meadows. One explanation for the strong positive relationship between canopy height and fish larvae abundance and family richness is that a higher seagrass canopy provides shelter, which leads to a higher survival rate by providing protection from predators (Unsworth et al., 2019;Tarimo et al., 2022). Furthermore, a higher seagrass canopy height supports a variety of fish larvae species because of reduced currents which favor organic matter deposition that support high primary productivity and enhance food availability (Arshad et al., 2012). Hedberg et al., (2019) found that fish larvae assemblages increased with seagrass canopy height. Similar results were also reported by Palmqvist et al. (2013) and Jones et al. (2021), attributing the increased fish abundance to their ecological function as nurseries and shelter. Similarly, Erzad et al. (2020) noted the abundance of fish larvae in seagrass ecosystems is influenced by shelter availability. This supports the current findings that high canopy height provides shelter and food availability (Gullstrom et al., 2008). It has been reported that greater fish abundance was observed in seagrass species with lower shoot density ( Jones et al., 2021), which contrasts with the current findings that an increase in seagrass shoot density could result in an increased fish abundance; however, the Jones et al.
(2021) study was based on juveniles and adult fishes, whereas the current findings are based on fish larvae.
Seagrass cover had a positive relationship with family richness but a negative relationship with fish larvae abundance. Such a negative relationship might be due to other factors such as reproduction patterns and fish species preferences (Tarimo et al., 2022), which were not investigated in the current study. This is in contrast to previous studies, which discovered that seagrass cover is an important factor in determining fish assemblages regardless of fish larvae stage (Arshad et al., 2012;Erzad et al., 2020). However, Rappe et al. (2013) reported that the validation of such a relationship is only possible in areas with high seagrass species richness and fish assemblages. Additionally, in contrast to earlier studies by Gullstrom et al., (2008), Rappe et al., (2013) and Jones et al., (2021), the relative significance of seagrass habitat structure that was dominated by T. hemprichii was apparent in in the present study.
This implies that high seagrass percentage cover, shoot density, and canopy height attract more fish larvae families to occupy an area. These findings are similar to that of Jones et al. (2021) who reported that the complexity of seagrass with extensive coverage, and high leaf canopy provide strong shelter capacity and a variety of food resources. Moreover, high seagrass cover attracts various fish species because of the avoidance of predators and wide space for forage.
These results are supported by Gullström et al. (2006) and Jones et al. (2021), who also found that high cover and canopy height is a harbour for a variety of faunal assemblages and support greater fish diversity and richness. Therefore, seagrass cover, shoot density and canopy height influence fish larvae richness. Overall, a complex canopy structure (high canopy height, long and more numerous leaves, but moderate shoot density) had greater fish richness as observed, and shoot density predicted fish larvae families richness, similar to what was reported previously (Rappe et al., 2013;Erzad et al., 2020;Jones et al., 2021). Generally, seagrass structural complexity provides a favorable environment for fish larvae survival and recruitment Ramli et al., 2013).

Effects of environmental variables on fish larvae assemblages
Regression analysis revealed that environmental variables influence fish larvae abundance and family richness. When combined and using PCA values however, there was no significant influence on fish larvae assemblages, but when treated separately there was an influence. This means that the fish larvae abundance and family richness can be determined by the environmental variables. However, other factors need to be taken into account. Average water temperature was negatively correlated with the abundance of fish larvae and family richness. This could imply that an increase in temperature affects the fish larvae assemblage, abundance and family richness (Zerrato and Giraldo, 2018). Temperature influences the physiological processes in seagrass and fish larvae growth (Nordlund et al., 2016;Mwaluma et al., 2021). The average water temperature in the seagrass ecosystem in the study sites was around 27.15 °C, which is deemed ideal for fish larvae growth and survival and for the photosynthesis process of seagrass (Erzad et al., 2020). The DO was found to positively predict fish larvae assemblages. According to Perez-dominguez et al., (2006), pH had a negative correlation with family richness, but a positive correlation with fish larvae abundance.
According to Molina et al. (2020) this could be due to differences in sensitivity and responses among fish families. Moreover, a small shift in pH can have significant impacts on fish larvae assemblages. The negative relationship between fish larvae assemblages and salinity could imply that fish larvae cannot tolerate a wide range of salinity (Arshad et al., 2012). Similar results were reported by Zerrato and Giraldo (2018). Another predictor variable, depth, was negatively correlated with fish larvae abundance and family richness. The most likely explanation for this is that the majority of the fish larvae reside in nearshore habitats in shallow waters. This is in contrast with previous findings which show juvenile, subadult, and adult individual fish to be positively correlating with water depth ( Jones et al., 2021). This is due to fact that the occurrence of post larvae fish primarily depends on the tidal regime and species-specific mobility (Tarimo et al., 2022). Large fish have a preference for slightly deeper subtidal seagrass habitats which provide a suitable environment for foraging and increased space for protection against predators Jones et al., 2021).
This disparity could be explained by the fact that the current investigation was based on fish larvae, which are small and with limited mobility. Additionally, it is necessary to carry out an extensive comparable study that would include all fish life histories in the seagrass ecosystem in order to assess and contrast their diversity and abundance under various tidal regimes.

Relative importance of seagrass structural complexity and environmental variables
Seagrass structural complexity and environmental variables influence the fish larvae assemblage abundance and richness. Multiple regression analysis indicated that seagrass structural complexity (canopy height, seagrass cover, and shoot density) was the foremost predictor of fish larvae assemblages in tropical coastal waters.
This was followed by variables related to the environment (temperature, dissolved oxygen, and salinity).
The current study found that both seagrass structural complexity and environmental variables are important for fish larvae assemblages in coastal waters, and that conservation efforts should take both into account.
Cluster analysis and general patterns of fish larvae assemblage among sites The hierarchical cluster analysis showed variation in fish larvae assemblages among study sites. The distribution of fish larvae families was closely associated with the environmental variables and seagrass structure (Mwaluma et al., 2021). A similar pattern of grouping among the sites in hierarchical cluster analysis and nMDS in the present study is in line with Arumugum et al., (2016). In their observations, these authors stated that the nMDS plot revealed the same groups as a cluster, which was again demonstrating the variations in different sampling sites. In contrast to ST3 and ST4, where there was unequal dispersion, the group average similarity across sampling sites ST1 and ST3 showed a comparable pattern, suggesting that most fish larvae families are the same and were distributed equally in the two sites. This might be due to tidal and water current fluctuation differences for distributing fish larvae in different areas as reported by Erzad et al., (2020).
This study showed that the structure of fish larvae assemblages varied spatially among seagrass meadows dominated by T. hemprichii, displaying high fish larvae abundance and family richness at ST2 and ST4 sites.

Implications for management and governance
The findings presented here will be of broad interest to fisheries managers, researchers, and other relevant stakeholders, including responsible authorities to ensure effective management and conservation of seagrass and adjacent coastal ecosystems. It has been observed and reported that one of the most direct adverse effects on seagrass beds is the damage caused by fishing or recreational boat activities (e.g., the use of beach seines, cutting by propellers, propeller wash, anchor and mooring damage, and boat groundings), which could result in significant localized impacts on the physical integrity of seagrasses (Turner and Schwarz, 2006;Jones et al., 2021). Propeller scarring, for example, can result in a continuous line of seagrass damage, fragmenting the seagrass bed and increasing the vulnerable bed edge to erosion, thereby leading to more scouring and deepening of the scoured area.
As a consequence of increased seagrass bed fragmen- and governance (Mwaluma et al., 2021). Nevertheless, in the WIO region, climate change is another threat to habitats and it is necessary to improve understanding of the present coastal habitats how climate change will impact fisheries productivity alongside efforts to prepare adaptation for those future changes Sekadende et al., 2020;Mwaluma et al., 2021).

Conclusions and recommendations
The structural complexity of seagrass beds and environmental variables are determinants of fish larvae assemblages in these coastal habitats. The abundance and diversity of fish larvae are determined by seagrass structural complexity, (canopy height, shoot density, and seagrass cover), and environmental variables (temperature, dissolved oxygen, pH, and salinity) which influence fish larvae assemblage in tropical coastal waters. The study recommends that shallow coastal habitats, including seagrass meadows, should be prioritized for conservation and governance efforts in order to protect critical habitats for fish larvae, which help to maintain robust coastal fish stocks and viable coastal fisheries, which is the main occupation of the coastal communities.