330 



Fishery Bulletin 105(3) 



averaged within and among tagged males observed on 

 the grid in 2001 and 2002. 



For comparison, the annual frequency of clutches 

 prepared by tagged females was also estimated by moni- 

 toring the time required to prepare eggs, t^, and the 

 interval between inferred clutches, t^^ (Fig. 1). Female 

 trunk girth was used as an indicator of reproductive 

 status (Vincent and Sadler, 1995; Perante et al., 2002). 

 Females whose trunks were bulging between trunk 

 rings (occasionally with visible eggs) were assumed to 

 be preparing a clutch of eggs for mating. Within mo- 

 nogamous pairs of//, comes (Perante et al., 2002) and 

 H. whitei (Vincent and Sadler, 1995), changes in female 

 trunk girth were significantly correlated with changes 

 in the reproductive state of their male partner. There- 

 fore trunk girth was assumed to be a reliable proxy for 

 clutch preparation in H. guttulatus. Because female sea 

 pony {Hippocampus fiiscus), prepared and released their 

 eggs within approximately three days (Vincent, 1990), 

 i,^ was estimated only during periods when female H. 

 guttulatus were observed on average at least once every 

 two to three days. 



Empty 



2 May 



2 Jun 



2 Jul 



2 Aug 



2 Sep 



Swollen n 



Thin 



2 May 



2 Jun 



2 Jul 



2 Aug 



2 Sep 



Figure 1 



Examples of the changes in reproductive state from May to 

 September 2002 in one marked male (top) and one marked 

 female (bottom) Hippocampus guttulatus (long-snouted sea- 

 horse) during the census period in 2002. Males have full pouches 

 when brooding eggs and embryos. Females whose trunks were 

 bulging between trunk rings (i.e., swollen) were assumed to be 

 preparing a clutch of eggs. Brackets marked with an A indicate 

 the maximum duration of three brooding periods, and brackets 

 marked with a B indicate minimum durations. One brief and 

 one lengthy interbrood interval are evident. The interclutch 

 intervals, inferred from changes in female trunk girth are also 

 denoted with brackets marked with a C. 



Intermittent and seasonal reproduction (ISR) model 



This model is more realistic than the two previous 

 models because information about seasonal trends in 

 reproductive activity is incorporated, rather than the 

 assumption that there is constant reproduction through- 

 out the reproductive season. This model was adapted 

 from the spawning fraction method used to estimate the 

 average number of spawnings per year among female 

 northern anchovy (Hunter and Leong, 1981). A Gaussian 

 curve was fitted to the fraction of mature males with 

 full pouches in the catch data plotted against sampling 

 month by using nonlinear regression, as in Curtis and 

 Vincent (2006). The area beneath this curve corresponds 

 to an estimate of the total number of days the average 

 adult male brooded embryos per year, tf^^^. The mean 

 number of broods produced per male each year (i.e., 

 annual spawning frequency), S/^ ^,^, was estimated as 

 t^, ^.,-ltf^. Annual fecundity was then calculated as /^^s^ ^,,,. 

 Similarly, the ISR model was applied to the proportion 

 of adult females preparing eggs, where the number of 

 clutches produced by females each year, s^ , was esti- 

 mated as the total time spent preparing clutches 

 of eggs, /j ,,,, divided by the time required to pre- 

 pare a single clutch, t^. In order to characterize 

 among-population differences in H. guttulatus 

 spawning frequency, this method was also applied 

 to male catch data reported from the Arcachon 

 Basin, France (Boisseau, 1967). 



Model validation 



An independent estimate of the brooding period 

 (21 days; Boisseau, 1967) was used to evalu- 

 ate the predictive accuracy of the CR and ISR 

 models. The CR model was used to predict the 

 expected number of broods produced by males, 

 Sj, ^,p„, during the census periods from 18 July 

 to 26 October (2001, 112 days) and 23 May to 

 12 September (2002, 108 days). Similarly, the 

 curve fitted to the male catch data using the 

 ISR model was used to calculate the predicted 

 number of broods produced by males during these 

 two census periods. Because the only estimate of 

 the time females take to prepare eggs, t^, (used 

 to estimate the number of clutches produced 

 by females in the ISR model) was based on the 

 underwater visual census data, the ISR model 

 could not be used to predict the total expected 

 number of clutches produced by females during 

 the census periods. Therefore, the ISR model 

 was used to predict the total number of days 

 females spent preparing eggs during the census 

 periods, /,.„,„. Expected values of s^,,,„ and ^^ ^.^^ 

 based on the CR and ISR models were compared 

 to the mean values of s^.^^^ and <^ „„ directly 

 observed on the grid during the corresponding 

 census periods. The IR model could not be vali- 

 dated because there was no independent esti- 

 mate of interbrood interval, i^^, for H. guttulatus. 



