Springer-VerlagTokyohttp://www.springer.de101640289-07711439-5444Journal
of EthologyJ
EtholLifeSciences18110.1007/s10164-005-0181-3
J Ethol (2006) 24:205–212 DOI 10.1007/s10164-005-0181-3
© Japan Ethological Society and Springer-Verlag Tokyo 2006
ARTICLE
Jianqiang Xiao • Ding Wang
Respiratory pattern of captive Yangtze finless porpoises (Neophocaena phocaenoides asiaeorientalis)
Received: March 28, 2005 / Accepted: September 21, 2005 / Published online: January 13, 2006 Japan Ethological Society and Springer-Verlag 2006
Abstract Cetacean respiration usually happen in bouts. The most widely applied quantitative method used to analyze the structure of these bouts is the loge-survivorship analysis, based on the assumption that the respiratory intervals are distributed as negative exponentials. However, for the data collected from three captive Yangtze finless porpoises (Neophocaena phocaenoides asiaeorientalis), we failed to obtain a convergent result with the application of logesurvivorship analysis. However, the two-Gaussian model, which was recently proposed to analyze the feeding behavior of cows, was successfully fitted to the data. According to the fitting results, the overall respiratory pattern of the captive Yangtze finless porpoises can be described as a dive with a mean duration of around 30–40 s, followed by two or three ventilations with a mean interval of approximately 9 s. The average intra-bout intervals during both active and inactive periods are constant at 7.7–9.9 s for all individuals. However, when shifting from active to inactive states, the adult male and female decrease their mean numbers of respirations per bout and average length of inter-bout respiratory intervals, while the estimates of both parameters increase for the juvenile female. It was pointed out that the two-Gaussian model might be more adequate for cetacean respiratory-bout structure analyses than the loge-survivorship technique. Key words Neophocaena phocaenoides asiaeorientalis · Yangtze finless porpoise · Respiratory pattern · Behavior · Loge-survivorship analysis · Two-Gaussian mixture model
J. Xiao1 · D. Wang (*) Institute of Hydrobiology, The Chinese Academy of Sciences, 430072 Wuhan, People’s Republic of China Tel. +86-27-68780178; Fax +86-27-68780123 e-mail:
[email protected] Present address: 1 Graduate School, The Chinese Academy of Sciences, 100039 Beijing, People’s Republic of China
Introduction Being exclusively aquatic mammals, cetaceans exhibit most of their behavior under water but periodically surface to breathe. Compared with submerging behavior, the respirations of cetaceans at surface are more evident and easier to observe. However, the submerged behavior can affect the pattern of respiration (e.g. Kopelman and Sadove 1995; Jahoda et al. 2003). Understanding the respiratory pattern therefore may increase understanding of the “invisible” underwater behavior. Besides, accurate measures of respiratory intervals (RIs) and quantitative analysis on the breathing pattern can provide valuable information for estimating cetacean abundance by visual “cue-counting” techniques (Hiby and Hammond 1989; Stern 1992; Kopelman and Sadove 1995; Beasley and Jefferson 2002) and for understanding the animals’ physiological adaptation to the aquatic environment (Sumich 1983; Watson and Gaskin 1983; Chu 1988; Kooyman 1989). Yangtze finless porpoise (Neophocaena phocaenoides asiaeorientalis), the only known freshwater sub-species of finless porpoise, is found in the Yangtze River and the connected Poyang and Dongting Lakes in China (Zhang et al. 1993; Yang et al. 2000; Xiao and Zhang 2002). The size of the population was estimated to be around 2,700 based on survey data collected before 1991 and has been declining remarkably during the past decade (Zhang et al. 1993; Wang et al. 2000). Many (but mainly descriptive) studies on the respiratory pattern of Yangtze finless porpoises have been carried out since 1980 (e.g. Zhou et al. 1980; Hou 1993; Zhang et al. 1996; Wang 1998), but no attempt to quantitatively relate the respiratory pattern with the underwater behavior of this species has been made, because of the difficulty of the field observation on underwater behavior. Although Akamatsu et al. (2002) reported the bimodal distribution of dive depth in a wild Yangtze finless porpoise, with attached velocity–time–depth recorders, no description of the respiratory pattern and the correlation between respirations and underwater behavior was provided,
206
because of the difficulty to identify the exact moment of respirations precisely from the data and the very limited information on other behavior. Fortunately, captive animals provide us a unique opportunity to record RIs precisely, by visual observation, and to make a simultaneous observation of their behavior. The respirations of cetaceans, including those of Yangtze finless porpoise (Zhou et al. 1980; Zhang et al. 1996; Wang 1998), usually occur in bouts. Therefore, in most respiratory pattern analyses, the ventilation–dive cycles are firstly divided into surfacing periods, which are composed of consecutive intra-bout RIs, and diving periods, which correspond to the inter-bout RIs with relatively longer durations than the intra-bout ones, and then analyzed respectively to obtain estimates of parameters, such as the average length of inter- and intra-bout RIs, and the average number of respirations per bout (e.g. Watson and Gaskin 1983; Chu 1988; Kopelman and Sadove 1995). However, the result of the analysis depends strongly on the manner by which the bout criterion (the time criterion for respiratory bout definition) is decided. To make the results from different studies comparable, the bout definition should be objective and repeatable. However, in the majority of cetacean studies, researchers arbitrarily define the respiratory bout criteria as a value ranging from 15 s to 3 min (Watson and Gaskin 1983; Würsig et al. 1984; Dolphin 1987a; Chu 1988; Silber et al. 1988; Dorsey et al. 1989; Watkins et al. 2002). Although behavioral characteristics like the depth of dives or an indication of initiation of deep foraging dives were considered in addition to the arbitrary criteria in some studies (Würsig et al. 1984; Dolphin 1987a; Dorsey et al. 1989), which reduced the level of subjectivity in the definition, the method is highly individual and experience-dependant. While a bout definition based on clear biological significance is unavailable, a statistical manner should be a better choice than the arbitrary one for the definition. Loge-survivorship analysis has been applied to discriminate intra-bout RIs from inter-bout ones (e.g. Kopelman and Sadove 1995; Jahoda et al. 2003). In the loge-survivorship analysis, the loge-transformed frequency of RIs that are longer than t is plotted against RI length t, and the bout criterion can be estimated as the break point of the loge-survivorship curve (Fagen and Young 1978). The method is based on the assumption that the intra- and inter-bout intervals are produced by Poisson processes with different coefficients of rate, hence the survivorship can be described as the mixture of two negative exponential distributions, and the logesurvivorship curve is in the form of a broken-stick (Slater and Lester 1982). However, the assumption of the underlying Poisson processes and the corresponding negative exponential distributions for this technique are not always met, where the method could not be successfully applied (Tolkamp et al. 1998; Tolkamp and Kyriazakis 1999; Frische et al. 2000). Alternatively, in the analysis of cow feeding behavior, Tolkamp et al. (1998) loge-transformed the feeding intervals and then fitted the frequency distribution with a two-Gaussian mixture model, resulting in a good fit and a biologically significant bout criterion. Frische et al. (2000)
also applied this technique in the analysis of the respiration of snapping turtles (Chelydra serpentina). In this study, RIs and underwater behavior of three captive Yangtze finless porpoises were recorded simultaneously. The objectives of the study are: (1) to provide respiratory statistics of Yangtze finless porpoise based on precise visual observation, (2) to compare and examine the suitability of the loge-survivorship method and the twoGaussian model for describing the respiratory bout structure of the species, and (3) to investigate the difference in the respiratory bout structure under different behavioral states.
Materials and methods Animals and housing The studied colony was composed of one adult male aged 7 years (M), one adult female aged 7 years (F1), and one juvenile female aged 3 years (F2) Yangtze finless porpoises, living in an approximately 20 × 7 × 3 m (length × width × depth) kidney-shaped indoor pool at the Institute of Hydrobiology (the Chinese Academy of Sciences, Wuhan, China). The two adult animals (M and F1) were captured from different groups in the Yangtze River and translocated into the dolphinarium in December 1996; the juvenile female (F2) was introduced from the Tian-e-zhou Reserve into the dolphinarium in December 1999. Each of the animals was fed with about 1.0 kg thawed freshwater fish (mainly Carassius sp. and Cyprinus sp.) by hand four times per day, at around 0700–0800, 1100–1130, 1500–1530, and 1800–1830 h. Each feeding lasted no longer than 10 min and no training was performed throughout the study. The animals have no access to food during the rest of day. The pool was under a weak artificial illumination (t, became convex after about 50 s, which differed greatly from λintra and λinter are the mean values of all intra- and inter-bout the broken-stick model. As a result, fitting of the brokenRIs, respectively, N is the total number of RIs, and Nintra is stick model to the plots failed to converge and no bout the number of all intra-bout RIs. The value of the bout criterion could be obtained with this method. criterion can be calculated as: In contrast, good fitting was obtained by applying the two-Gaussian model for the overall loge-transformed RIs of T = loge ( N intra λ intra ) − loge ( N − N intra )λ inter (λ intra − λ inter ), each individual (R2 = 0.981, 0.987, and 0.963, for M, F1, and which was suggested by Slater and Lester (1982). F2, respectively; all P < 0.01, F test). Similar estimates of the
() (
(
[
(
)) () (
))
(
)
(
)
]
])
[
Table 1. Respiratory data for each studied animal, by state of activity Animal and behavioral state
Total number
Mean ± SD (s)
Median/lower/upper quartile (s)
Range (s)
M Active Inactive F1 Active Inactive F2 Active Inactive
9841 6287 588 10,430 7477 870 8505 3381 1295
18.8 ± 15.66 18.1 ± 15.40 20.6 ± 13.30 18.3 ± 15.25 17.9 ± 14.77 20.2 ± 15.93 20.8 ± 22.35 17.8 ± 15.87 25.8 ± 30.38
12.6/ 8.0/ 24.4 12.0/ 7.7/ 23.5 16.6/ 10.4/ 25.1 12.5/ 8.3/ 22.5 12.4/ 8.1/ 22.0 13.6/ 9.2/ 24.6 10.6/ 7.2/ 24.4 11.2/ 7.7/ 22.0 9.0/ 6.7/ 38.6
0.6–133.2 0.7–133.2 3.0–68.7 0.7–124.8 0.7–124.8 4.0–96.7 0.8–150.0 0.8–108.3 2.1–127.6
Shown are the total number, the mean ± standard deviation (SD), the median, the lower and upper quartiles, and the range of the analyzed respiration intervals
208
Fig. 1. Frequency distributions of respiratory intervals for each studied animal (bin width: 1 s)
proportion of RIs in the intra-bout distribution (~56%), back-transformed mean intra-bout RIs (~9 s) and backtransformed bout criterion (~15 s) were obtained for each individual by fitting the model to the overall RIs (Fig. 3). Therefore, the overall respiratory pattern for the animals can be described as two or three consecutive respirations interspersed with a mean interval of around 9 s followed by a dive with a mean duration of 30–40 s. A good fit was also obtained by applying the twoGaussian model for the loge-transformed RIs of each individual in active or inactive status (R2 = 0.981 (M, active), 0.849 (M, inactive), 0.982 (F1, active), 0.904 (F1, inactive), 0.981 (F2, active), 0.916 (F2, inactive); all P < 0.01, F test). As illustrated by Fig. 4, the respiratory bout structure during the active periods was quite consistent across the individuals. However, the estimates of respiratory pattern during inactive periods differed from those during active periods and differed among the individuals, except that the mean intra-bout RIs were still relatively stable. The mean
Fig. 2. Loge-survivorship curves of respiratory intervals for each studied animal. N is the number of respiratory intervals longer than t. BC potential bout criteria, A intra-bout intervals, B inter-bout intervals
inter-bout RIs decreased by 32.8% (31.7 s vs 21.3 s) for M, decreased by 17.2% (30.9 s vs 25.6 s) for F1, but increased by 149.4% (32.0 s vs 79.8 s) for F2 when shifting from active into inactive states. Similarly, when shifting from active into inactive states, both the mean number of respirations per bout and the bout criteria decreased for M (2.23 vs 1.14; 15.0 s vs 9.1 s) and F1 (2.48 vs 1.44; 16.5 s vs 11.6 s), but increased for F2 (2.29 vs 3.52; 14.6 s vs 25.1 s). Although the two-Gaussian model fitted the data well and both distributions were clearly recognized, a high degree of overlapping between the two distributions made the accurate separation of intra-bout RIs with inter-bout RIs difficult, or impossible (Figs. 3, 4). When classifying the overall RIs into intra- and inter-bout distributions using the calculated bout criteria where the two distributions crossed, 11.7, 14.5, and 10.7% of all RIs would be misclassified for M, F1, and F2, respectively. In the intra-bout distribution, 7.3, 7.9, and 3.8% would be misclassified into the inter-bout one, and 17.1, 23.0, and 19.3% of inter-bout RIs would be
209
Fig. 3. Relative frequency distributions of loge-transformed respiratory intervals with bin width of 0.05 loge units (dots) for each studied animal, and the fit of the probability density function of the twoGaussian mixture model (solid line),
(
) + (1 − p) 1 ( s
(
)
2 y = p 1 s1 2 π exp −2 (t − m1 ) s1
2
)
(
)
2 2 π exp −2 (t − m2 ) s2 ,
where p represents the proportion of respiratory intervals in the first or intra-bout distribution, and m1, s1, m2, and s2 represent the means and the standard deviations of the intra- and inter-bout distributions (dotted lines). BC potential bout criteria, A intra-bout intervals, B inter-bout intervals
misclassified into the first distribution, for M, F1, and F2, respectively (Fig. 3).
Discussion For all captive Yangtze finless porpoises studied in this study, similar overall average RIs (18–20 s) were recorded, which were comparable to those recorded three years before for M and F1 by Wang (1998; 15–19 s, calculated
Fig. 4. Estimated probability density functions of the two-Gaussian model during active (solid line) and inactive (dashed line) states for each studied animal
from the numbers of respirations per 15 min in Table 7.1). The values were slightly longer than, but still comparable with, that of another adult male in captivity alone (mean RI = 16 s, calculated by averaging the hourly mean RI in Table 2; Hou 1993). Comparable mean RIs of 17.5 s for wild Yangtze finless porpoises was reported by Zhou et al. (1980, p. 366). However, their small sample size (62 dives) made the value less representative. In addition, according to the results of Akamatsu et al. (2002, personal communication), which were obtained from eight adult free-ranging Yangtze finless porpoises with attached dataloggers, the maximum dive duration ranged from 83s to 248 s (170.0 ± 53.77 s). This result is also comparable to that of this study (133.2, 124.8, 150.0 s for M, F1, F2, respectively). Based on the above reasons, together with the similarity between the respiratory bout structure described in this study and that has been reported qualitatively for free ranging animals (Zhou et al. 1980; Zhang et al. 1996), we may assume that the captive environment did not affect the diving and respira-
210
tion greatly; and the analysis technique used for the animals in captivity can also be applied to the respiratory data obtained from wild animals. In the present study, we failed to fit the traditional broken-stick model to the loge-survivorship curves of RIs. By visual inspection, Winn et al. (1995) also failed to find a clear break-point in the loge-survivorship curve for the respiratory data of right whales (the 1989 dataset). As argued before by Tolkamp et al. (1998) and Tolkamp and Kyriazakis (1999), this was possibly because that the underlying assumption of negative exponential distribution was not met. According to the negative exponential distribution, highest frequency of occurrence should be observed at the shortest RIs (Sibly et al. 1990). However, this contrasts to the respiratory physiology of cetaceans. As reported by Sumich (1983, 1986, 1994), the efficiency of oxygen uptake of gray whales (Eschrichtius robustus) is lower at very short RIs. Therefore, the relative density of RIs is unlikely to be the highest at the shortest length. Similar to Fig. 1 of this study, positively skewed frequency distributions of RIs, with a “shortage” of very short RIs, have been widely observed in cetacean species such as harbor porpoises (Reed et al. 2000, Fig. 1), gray whales E. robustus (Lyamin et al. 2000, Fig. 5; Sumich 1983, Fig. 2), bowhead whales Balaena mysticetus (Würsig et al. 1984, Fig. 2a; Dorsey et al. 1989, Fig. 2a), and humpback whales Megaptera novaeangliae (Dolphin 1987a, Fig. 3b). Corresponding to the positively skewed frequency distribution, flat beginning should be found in the logesurvivorship plots, as presented in this study (Fig. 2). Even though loge-survivorship analysis was seemingly successfully applied by Kopelman and Sadove (1995) to determine the criterion between intra- and inter-bout blow intervals of fin whales (Balaenoptera physalus), after making a close inspection, flat beginnings can also be found in the loge-survivorship curves in their Figs. 2, 4 (
