Morphometric regressions for the endangered Knysna seahorse , Hippocampus capensis , in the Swartvlei Estuary from mass stranding events

The Knysna seahorse, Hippocampus capensis, is listed as endangered on the IUCN Red List of Threatened Species, with the only known, apparently isolated, populations occurring in the Knysna, Swartvlei and Keurbooms estuaries along the south coast of South Africa. The Swartvlei Estuary population is the most genetically distinct (Mkare et al. 2017). Despite all three known populations occurring within protected areas and there being no commercial or subsistence use of the species, population sizes are still thought to be declining because of habitat degradation across much of its range (Pollom 2017). Environmental management authorities have outlined strategies for the conservation of H. capensis, including listing it as a species of special concern in the Estuarine Lower Level Plan of the Garden Route National Park (South African National Parks 2020) and setting management objectives for the Keurbooms population in the Keurbooms–Bitou Estuarine Management Plan (Western Cape Government 2018).


Introduction
The Knysna seahorse, Hippocampus capensis, is listed as endangered on the IUCN Red List of Threatened Species, with the only known, apparently isolated, populations occurring in the Knysna, Swartvlei and Keurbooms estuaries along the south coast of South Africa. The Swartvlei Estuary population is the most genetically distinct (Mkare et al. 2017). Despite all three known populations occurring within protected areas and there being no commercial or subsistence use of the species, population sizes are still thought to be declining because of habitat degradation across much of its range (Pollom 2017). Environmental management authorities have outlined strategies for the conservation of H. capensis, including listing it as a species of special concern in the Estuarine Lower Level Plan of the Garden Route National Park (South African National Parks 2020) and setting management objectives for the Keurbooms population in the Keurbooms-Bitou Estuarine Management Plan (Western Cape Government 2018).
Assessments of population size and habitat use of H. capensis have focussed mainly on the relatively larger and more stable population in Knysna Estuary (Bell et al. 2003;Claassens 2016;Claassens & Hodgson 2018;Teske et al. 2007). The lesser studied Swartvlei Estuary population is subject to large fluctuations in abundance, with past densities ranging from 0.04 individuals per m 2 to 0.26 individuals per m 2 (Lockyear et al. 2006). The Swartvlei Estuary is the only intermittently open host estuary, with elevated water levels and prolonged flooding of land adjacent to the shores common during closed periods. Periodic mass strandings and associated die-offs of H. capensis have been recorded in Swartvlei Estuary when the estuary is breached (Russell 1994;Skelton 1987), mainly by artificial means to prevent inundation of properties on the estuarine floodplain. The breaching protocol and criteria are contained with the Garden Route National Park Management Plan (SANParks 2020).
South African National Parks (SANParks) as the management authority of Swartvlei Estuary has since 1991 intermittently collected dead H. capensis from the banks of the Swartvlei Estuary after breaching events. This now regular sampling has evolved into a SANParks co-ordinated citizen science programme.
The only known morphometric descriptions of H. capensis come from small samples of floodstranded individuals from the Knysna Estuary (n = 57) and Swartvlei Estuary (n = 39) (Toefy 2000), despite the usefulness of such data in allowing comparison between studies and the conversion of data where different measures of length are used (Lourie 2003).
This article aims to use the seahorse morphometric data collected following two mass stranding events (> 500 individuals) in 1991 and 2017 to provide more robust estimates of the relationship between key morphometric characteristics of H. capensis. In addition, data from mass strandings are compared to those collected from seine netting of live individuals in 2019 to determine the feasibility of using these periodic events as a means of assessing population attributes.

Collection protocol
Dead H. capensis were collected after mass stranding events in February 1991 and November 2017. Specimens collected in 1991 were preserved in 10% neutral freshwater-buffered formalin immediately after collection (Russell 1994), whereas in 2017 dead individuals were frozen immediately after collection. In 2017, samples were collected within 1 day of stranded individuals first being noticed, and for 1 week after estuary breaching. Sampling effort was spread throughout the estuary, although repeat sampling, where sites were searched daily over the sampling period, was undertaken mostly in areas with previous high abundance (Lockyear et al. 2006

Morphometrics
Preserved samples were photographed and measured using ImageJ software. Where necessary, individuals were flattened before being photographed to minimise bias in measurements because of body arching (discussed in Sotola et al. 2019).
All individuals were sexed and then measured using the protocol set by Lourie (2003). Head length (HdL), curved trunk length (TrL), curved tail length (TaL) and height (Ht) were measured, with standard length (SL) calculated as the sum of the first three measurements. Straight-line SL (SL straight ) and straight-line Ht (Ht straight ) were also determined for a subsample of individuals whose heads were positioned at a right angle to their trunk. The trunk measurements for the straight-line sizes were measured as straight-line distances from the respective anterior starting points ending at the last trunk ring. Regression equations of size measurements of H. capensis included measurement of Ht straight and Ht because Ht straight is often the size measurement used when sampling live specimens (Foster & Vincent 2004;Lourie 2003), likely to avoid handling stress, and Ht may be useful if the snout of a dead seahorse is damaged. Live individuals collected in 2019 were sexed and Ht straight was measured before the animals were released at the location of capture.
The effect of preservation techniques on morphometric relationships was investigated by comparing the regression equations between SL and HdL, TrL, TaL and Ht straight of formalin preserved (collected 1991) and frozen (collected 2017) specimens. Preservation of samples in a fixative can cause shrinkage, and different preservation techniques may result in different shrinkage rates of seahorse body parts (Nadeau, Curtis & Lourie 2009). Also, further shrinkage is known to occur in fish when multiple preservation techniques, in this case freezing or preservation in formalin after desiccation, are applied (Sotola et al. 2019).

Statistical analysis
Sex ratio bias within the 1991, 2017 and 2019 datasets was tested using an exact binomial test (Crawley 2007;Zar 2010). Sexes were analysed separately because of known sexual dimorphism (Bell et al. 2003).
Normality of SL (between preservation techniques) and Ht straight (between alive and preserved specimens) measurements for years and sexes were tested using a Shapiro-Wilk test. Homogeneity of variances was tested using a Fligner-Killeen test before differences in means were tested using either a one-tailed Welch's two-sample t-test or a one-tailed Wilcoxon rank-sum test. Differences in mean Ht straight between years were tested using a Kruskal-Wallis rank-sum test. Regression equations of SL and Ht straight and their component measurements for years and sexes were produced separately using ordinary least squares regressions and compared to determine if sex and preservation techniques influenced correlation between variables. Regression equations of SL and Ht straight , separated by year and sex, were tested using an analysis of covariance (ANCOVA) to determine if regression equations obtained differed significantly (Crawley 2007;Zar 2010).
Data analyses were undertaken using the programming language R (R Core Team 2019).

Size class distribution
Mean length of H. capensis collected in 2017 (62 mm -100 mm SL; n = 274) was significantly larger than the mean length of those collected in 1991 (33 mm -100 mm SL; n = 161) when sexes were grouped (W = 7800; p < 0.05) as well as for females (W = 1650; p < 0.05) and males (W = 2227; p < 0.05) separately. The majority of individuals collected in 2017 were above the size of sexual maturity (≥ 65 mm SL [Whitfield 1995 (Figure 1b).

Morphometric regressions
The fit of regression models between SL and HdL, TrL and TaL was better for the 1991 samples compared to those collected in 2017, as evidenced by the high R 2 values in the former (Tables 1 and 2). Test statistics for the regressions between SL and HdL, TrL, TaL and Ht straight are presented in Table 3. HdL, TrL and TaL were significantly correlated to SL for both sexes when years were analysed separately, whilst sex effected all regressions except HdL in 1991. Interaction did not significantly effect the fit of the models in either year (p > 0.05) and the slope of the models did not differ significantly between sexes (p > 0.05), with the exception of HdL in 2017. When differences between years were calculated separately for sexes, interaction effects were evident for the female datasets for TrL and TaL    evident in either year (p > 0.05). When sexes were analysed separately, Ht straight and year significantly influenced the regressions for both sexes, with no interaction effects evident for females or males (p > 0.05). The linear regression for males collected in 1991 (Table 1) was used to calculate Ht straight at sexual maturity (56 mm) (Figure 1b).

Discussion
The 1991 and 2017 datasets provided reliable correlations between the morphometric variables measured for the different preservation techniques. Both preservation techniques used in this study are known to result in significant shrinkage of fish bodies Buchheister & Wilson 2005;Nadeau et al. 2009) with the preservation of samples in formalin and ethanol resulting in ongoing shrinkage of body parts Sotola et al. 2019). To compound the complexity of using preserved specimens in morphometric investigations, shrinkage in fish may not be uniform across body parts and differences can even be observed at an individual level (Sotola et al. 2019). This effect was noted in this study with the regressions of SL with HdL, TrL and TaL of H. capensis collected in 2017, showing greater variation than those collected in 1991. Although the individuals collected in 2017 were frozen before processing, they showed differing levels of desiccation when collected. This, coupled with an extended, staggered collection period for the stranded seahorses, and the relatively shorter period between preservation and processing of samples compared to the 1991 dataset, could have contributed to the larger variation in the body length ratios. The greater correlation between morphometric measurements of SL and HdL, TrL and TaL in 1991 was therefore likely because of more even shrinkage resulting from prolonged exposure of the samples to the fixative. This result, together with the differences in slopes between the 2 years, suggests that different regression equations are needed for each preservation type and that comparisons between samples preserved in different fixatives may require correction factors to be calculated by measuring samples before and at several periods after preservation. In addition, using regression equations obtained from preserved specimens to describe morphometric relationships of live specimens is not recommended.
The mean length (Ht straight ) of H. capensis differed significantly between 1991, 2017 and 2019 for both sexes. Differences could relate to the timing of the breaching events, with both the 1991 and 2019 samples collected towards the end of the breeding period for H. capensis (austral summer -Whitfield [1995]), whereas the 2017 samples were collected at the start of the breeding period, potentially explaining the lower proportion of smaller individuals in that year. The duration of preservation of specimens could also have contributed to differences in mean length, with specimens collected in 1991 being subjected to longer exposure to a fixative likely having shrunk to a greater degree.
Although differences in the size data existed, the slopes of the SL versus Ht straight regressions for 1991 and 2017 did not differ significantly from one another, indicating that the relationship between SL and Ht straight remains relatively constant regardless of preservation technique used. This was, however, not the case for the relationship between SL and HdL, TrL and TaL, where preservation technique, as well as sex, influenced the slopes of the regressions. As various length measurements are often used for measuring seahorses, standardisation of seahorse measurements is difficult (Foster & Vincent 2004). The morphometric regressions presented in Tables 1 and 2 can assist in comparing the results of previous studies where different measurements and preservation methods were used.
Equal sex ratios observed in 1991 and 2017 were similar to those found in in situ studies of natural populations (Lockyear et al. 2006) and in 2019, suggesting that stranding affected both sexes equally. The use of stranding data for H. capensis in the Swartvlei Estuary could be a novel way of post hoc determination of population attributes, as data collected from these events are comparable to those collected from the study of in situ populations. The effects of preservation on morphometric variables also need additional consideration, and shrinkage correction equations should be developed for different preservation methods and durations.