Spatial synchrony in population dynamics of West African fishes: a demonstration of an intraspecific and interspecific Moran effect

Blackwell Publishing, Ltd.

PABLO A. TEDESCO*, BERNARD HUGUENY*, DIDIER PAUGY† and YVES FERMON‡ *Institut de Recherche pour le Développement (IRD), UMR CNRS 5023, Université de Lyon 1. 43, Bd du 11 Novembre 1918, F-69622 Villeurbanne cedex, France; †Antenne IRD, Laboratory Ichthyologie, MNHN, 43 Rue Cuvier 75231, Paris cedex 05, France; and ‡Laboratory Ichthyologie, MNHN, 43 Rue Cuvier 75231, Paris cedex 05, France

Summary 1. Synchronous fluctuations of the abundance of several populations at a regional scale can be a sign of a climatic effect on their dynamics (Moran effect). However, the interpretation of spatial synchrony is often complicated by the interaction between climatic disturbances and migrations between the studied populations. 2. We addressed this question by studying 24-year time-series of abundance estimates for four populations of four fish species in three different catchment basins in Côte d’Ivoire (West Africa). As localities from two different basins cannot exchange individuals, dispersal cannot be considered a synchronizing factor in this system. 3. A sign test performed on the four site × site correlation matrices rejected the null hypothesis of no spatial synchrony. No relation was detected in relating the level of synchrony in species dynamics (r = 0·58 on average) to the geographical distances between the four sites (176–367 km). A more detailed analysis taking account of sampling error was carried out between two sites, and confirmed the above results. When corrected for sampling noise, high values of spatial synchrony were obtained (r = 0·88 on average). 4. Interspecific spatial synchronies (r = 0·57 on average) were almost as high as intraspecific values, suggesting that the four species are reacting in the same way to the same Moran effect. 5. A good correlation (r = 0·54) between a regional river discharge index and the total catch of the four species 1 year later was observed, suggesting that hydrological variability is one of the main synchronizing factors. Moreover, as expected on theoretical grounds, spatial synchrony in river discharge (r = 0·84 and 0·88 on a log scale) is similar to spatial synchrony in population dynamics (r = 0·88 on average and corrected for sampling error). Key-words: disconnected populations, hydrology, Moran effect, sampling error, within and between species synchrony. Journal of Animal Ecology (2004) 73, 693–705

Introduction The relative importance of intrinsic factors and extrinsic environmental variations in determining population size fluctuations is a major debate in ecology (Royama 1992). While studying the spatiotemporal dynamics of

© 2004 British Ecological Society

Correspondence: Pablo A. Tedesco, Institut de Recherche pour le Développement (IRD), UMR CNRS 5023, Université de Lyon 1. 43 Bd du 11 Novembre 1918, F-69622 Villeurbanne cedex, France. E-mail: [email protected]

lynx populations in Canada, Moran (1953) suggested that, if two populations have a similar density dependence structure, then correlated density independent factors (usually determined by climate) produce synchrony of the two fluctuating populations (‘Moran effect’, Royama 1992). A series of studies published within the past 10 years has shown that spatial synchrony is a general phenomenon (Ranta, Kaitala & Lindström 1998a) not restricted to vertebrates or cyclic species (Mackin-Rogalska & Nabaglo 1990; Hanski & Woiwod 1993; Ranta et al. 1995a; Sutcliffe, Thomas & Moss 1996; Ranta, Kaitala &

694 P. A. Tedesco et al.

© 2004 British Ecological Society, Journal of Animal Ecology, 73, 693–705

Lundberg 1997; Myers 1998; Rusak et al. 1999; Cattadori, Merler & Hudson 2000; Benton, Lapsley & Beckerman 2001; Straile 2002). Most, although not all (Cattadori et al. 1999; Grenouillet et al. 2001), have shown synchrony over a fairly wide geographical area, with the degree of synchrony between populations decreasing with distance. However, identifying synchrony is just the first step; the more difficult task is to identify the mechanism that causes this pattern. Besides the Moran effect, two other mechanisms have been identified as synchronizing factors: dispersal (Ranta et al. 1995a) and more rarely nomadic predators (Ydenberg 1987; Ims & Steen 1990; Ims & Andreassen 2000). Some studies have attempted to disentangle the Moran effect from those of dispersal. In general, both the models and the data suggest that scale is important (Ranta et al. 1995a; Heino et al. 1997; Lande, Engen & Saether 1999; Paradis et al. 1999). At local scales dispersal can be more important than the Moran effect, but at larger scales the Moran effect will be dominant because dispersal ranges are limited and thus the dispersal rates will be negligible. This conclusion is supported by studies on butterfly populations: synchrony at finer spatial scales occurred at distances similar to the dispersal range, whereas at larger scales it was greater than the dispersal range (Sutcliffe et al. 1996). Nevertheless, the two mechanisms are not mutually exclusive and both could be operating at smaller spatial scales, so we cannot really distinguish between the relative importance of each without an experiment or some sophisticated analyses. Demonstrating synchrony therefore requires the manipulation or elimination of one of the synchronizing mechanisms. Grenfell et al. (1998), in a study of the irregular population fluctuations of the sheep populations in the St Kilda archipelago, showed a remarkable synchrony between the populations of two different islands. The role of dispersal was easily dismissed because 3.5 km of Atlantic swell provides a perfect barrier to sheep dispersal. However, the islands are so close together that they will be exposed to similar environmental conditions. Benton et al. (2001) reported that population synchrony was a function of environmental synchrony in an experimental closed system in which populations of a soil mite were isolated and the environmental variations were artificially manipulated. However, these are exceptions. As many naturally synchronized populations will be linked by dispersal, and will also live in correlated environments, it is unlikely that analyses of field data will provide definitive information on the relative roles of correlated noise and dispersal in generating synchrony. Additionally, environmental variation and migration rates are likely to interact (Haydon & Steen 1997; Ranta, Kaitala & Lundberg 1998b; Lande et al. 1999; Kendall et al. 2000), leading to an even greater difficulty in separating the effects of noise and dispersal in field data from synchronized populations. On the other hand, it is generally not possible to apply experimental approaches to many kinds of organisms for

which synchrony has been demonstrated, especially at larger spatial and temporal scales. A particular case where dispersal can be discarded as a synchronizing factor is when two related species display correlated population dynamics (Ranta, Lindström & Lindén 1995b; Lindström, Ranta & Lindén 1996; Myers 1998; Cattadori et al. 2000; Post & Forchhammer 2002). Thus, studying interspecific spatial synchrony as well as its intraspecific component may provide useful cues to identify the underlying synchronizing factor. Population sizes are rarely measured exactly; they usually have to be estimated. While its detrimental effects on model fitting are largely acknowledged (Carpenter, Cottingham & Stow 1994; Link & Nichols 1994; Solow 1995; Shenk, White & Burnham 1998; Solow 1998; Turchin & Ellner 2000; De Valpine & Hastings 2002), sampling error is rarely incorporated explicitly into population models or statistical tests (Link & Nichols 1994; Carpenter et al. 1994; Solow 1998; Fromentin et al. 2001; Harley, Myers & Dunn 2001; De Valpine & Hastings 2002). It is well known that sampling error may affect model parameter estimation and particularly may inflate coefficients of density-dependence (Den Boer & Reddingius 1996, p. 253). Conversely, sampling errors may decrease the observed synchrony between populations because measurement errors bias estimates of correlation coefficients toward zero (Worm & Myers 2003). As a result, neglecting the effect of sampling error may lead to overemphasis of intrinsic factors (e.g. density-dependence) with respect to extrinsic factors (the Moran effect). In this paper, we analysed fluctuations in the population abundance of four fish species in three different river basins in Côte d’Ivoire. Grenfell et al. (1998) dismissed the role of dispersal between two terrestrial (sheep) populations separated by the sea, whereas here we report the geographically opposite case to demonstrate the same process. River systems are very suitable for studying disconnected populations between different catchment basins and assigning synchrony to a unique process, the Moran effect (Cattaneo, Hugueny & Lamouroux 2003). We also analysed spatial synchrony between species to assess whether different species are affected by the same Moran effect. As this study deals with population indices (catches per unit effort), the effect of sampling error on parameter estimation was accounted for. Hydrological variability is known to affect riverine fish population dynamics (Holcík & Bastl 1977; Welcomme 1985, 1989; Laë 1992; Smolders et al. 2000; Grimes 2001; Moses 2001). Thus, the interannual changes in total abundances and in total discharge of rivers have been compared.

Methods As part of a monitoring programme for West African rivers treated with insecticides during the Onchocerciasis

695 Synchrony and the Moran effect in freshwater fishes

Fig. 1. Côte d’Ivoire map showing the location of sampling sites numbered as in Table 1.

© 2004 British Ecological Society, Journal of Animal Ecology, 73, 693–705

Control Programme (OCP) sponsored by the World Health Organization (Lévêque et al. 1988), fish communities were surveyed in several localities. Paugy et al. (1999) showed no detectable influence on fish communities by the insecticides used to control the populations of the aquatic stages of the onchocerciasis vector (a blackfly, Simuliidae, Diptera). Thus, it can be assumed with great confidence that the population dynamics of fish species have not been affected by the OCP. We selected four localities on the basis of the length of the time-series (> 10 years) and their location within the same biogeographical unit, the Eburneo–Ghanean region (Hugueny & Lévêque 1994). The four localities are distributed among three river basins in Côte d’Ivoire (Fig. 1): the Comoé (localities 1, Léraba River at Pont Frontière, and 2, Ganse, monitored 94 and 98 times over 23 and 24 years, respectively), the Bandama (locality 3, Niakaramandougou, monitored 112 times over 24 years) and the Sassandra (locality 4, Semien, monitored 40 times over 12 years). Apart from sites 1 and 2, which are in the same basin, the four sites present the advantage of being isolated from one another. Moreover, sites 1 and 2 are separated by more than 400 km river distance. All four sites belong to the same hydro-climatic zone, the ‘Nordgolf’ (Mahé 1993). Experimental fishing was carried out using sets of gill-nets 25 m long and 2 m deep with various mesh sizes (15, 20, 25, 30 and 40 mm knot to knot) during two consecutive nights (Lévêque et al. 1988). Results are expressed as catch per unit effort (CPUE), which is the number of fish caught in 100 m2 of net per night (18

h to 7 h). Sampling could take place or be efficient only in large pools which fitted the following criteria: more than 1 m deep, more than 20 m wide, more than 500 m long and with little or no current velocity (less than 0·2 m s−1). In addition to their relative homogeneity of size and current velocity, the habitats sampled displayed little variability in the type of substrate, which was muddy with scattered rocks and rarely sandy (de Mérona 1981, D. Paugy, personal observation). In order to minimize missing values and maximize variability in the time-series, we selected fish species with a high abundance and a low number of empty samples. Four species were kept for analysis: Alestes baremoze (Joannis 1835), Characidae; Brycinus macrolepidotus Valenciennes, 1849, Characidae; Schilbe mandibularis (Günther, 1867), Schilbeidae; and Chrysichthys maurus (Valenciennes, 1840), Bagridae. These species represent about 50% of the total samples per sites. Adjacent samples were separated by a few months, and hence the likelihood of finding high differences between two adjacent samples is weak but still possible, because conditions for an efficient sampling may not be fulfilled. Hence, in order to detect and eliminate outliers that can be the result of unfavourable conditions during sampling, some samples with a low CPUE preceded and followed by samples with high CPUE were discarded (about 8% of the samples). Resulting timeseries were log(CPUE + 0·2)-transformed, where 0·2 is the minimal non-zero value found in the series. The sampling scheme was neither regular nor synchronous between localities, preventing a direct evaluation

696 P. A. Tedesco et al.

Table 1. Sampling information summary

Site

Basin

1

Comoé

2

Comoé

3

Bandama

4

Sassandra

Sampling periods

Mean no. of samples per year

1974 −1986 1987−1997 1975 −1986 1987−1997 1974 −1986 1987−1997 1975 −1986

2·4 4·6 3·2 4·7 3·3 4·9 3·2

of spatial synchrony between populations (Table 1). We used instead averages over calendar years for analysing temporal population dynamics and comparisons between sites.

that there is temporal autocorrelation, violating the assumption of serial independence required for most standard inference tests. Different approaches have been proposed to circumvent this problem (Thompson & Page 1989; Pyper & Peterman 1998; Buonaccorsi et al. 2001). The method retained in this paper was to use Monte-Carlo simulations to mimic two time-series changing independently. In this way the distribution of the correlation coefficient is obtained under the null hypothesis of no spatial synchrony. It is thus necessary to fit a population dynamics model to the observed time-series and to estimate its parameters, as now described. In the presence of sampling error, instead of measuring log (CPUE*s ), the true value for the sample s, we measure: log(CPUE s ) = log(CPUE*s ) + θ s

The degree of spatial synchrony within and between species among the different sites was evaluated using cross-correlation analysis with zero time lag (Ranta et al. 1995b; Ranta et al. 1998b; Bjørnstad, Ims & Lambin 1999; Koenig 1999) for annual means of converted time-series. The spatial synchrony pattern of a given species can be synthesized in a site × site correlation matrix. The statistical analysis of such a matrix raises two difficulties. The first is that correlation values cannot be used directly to assess if synchrony is statistically significant because it is likely that there is temporal autocorrelation in times series. The second difficulty is that the cells of a correlation matrix are not statistically independent. For instance, with four sites, if a cell is randomly chosen (say the correlation between sites 1 and 2), only one correlation value remains independent (the correlation between sites 3 and 4) because this is the unique cell that does not share sites with the former cell. Considering the four species together, 4 × 2 independent correlation values can be selected. If none of the species displays spatial synchrony (the null hypothesis) we expect 50% of these values to be positive. The probability of observing a given number of positive correlations can be evaluated by comparison with a binomial distribution (number of trials = 8, probability = 0·5). Thus, to address statistically the combined spatial synchrony of the four species, the four site × site correlation matrices from 1974 to 1986 were subjected to a non-parametric sign test by selecting at random two independent correlation values within each species’ correlation matrix.

© 2004 British Ecological Society, Journal of Animal Ecology, 73, 693–705

- As emphasized above, spatial synchrony cannot be tested by checking the significance of the coefficient of correlation between two time-series because it is likely

eqn 1

where θs is the sampling error. For simplicity, the θs’s are supposed to be distributed normally (θs ∼ N(0, η )), mutually independent and independent of the catches. We defined Yt as the mean of the log of the observed catches of the year t and for simplicity we assumed that the number of samples per year (n) is constant: n

Yt =

∑ log(CPUE )/ n s

eqn 2

s =1

In the same way, we defined Xt as follows: n

Xt =

∑ log(CPUE )/ n * s

eqn 3

s =1

Combining eqn 1, eqn 2 and eqn 3 leads to: Yt = Xt + δt

eqn 4

with Var(δ) = Var(θ)/n = ψ

eqn 5

We assumed that interannual population dynamics may be approximated by a simple autoregressive model, the stochastic Gompertz model, as defined by Dennis & Taper (1994): Xt+1 = aXt + b = εt

eqn 6

where a (< 1) is a coefficient of density-dependence, b is a constant and εt is a random variable that represents the stochastic variation in the process. The error term εt is assumed to be normally distributed (εt ∼ N(0, φ)) and iid (identically but independently distributed). Given the simple assumptions that have been made and adding that we neglect, for a given species, differences between sites in the parameters of interest, four parameters are needed for modelling a time-series of catches within a site: b, a, φ and ψ. Fortunately, as the correlation between two variables does not depend on their means, b can be neglected as long as we are interested in

697 Synchrony and the Moran effect in freshwater fishes

testing for spatial synchrony. The sampling error variance cannot be estimated with the data available from Côte d’Ivoire samples. This value was estimated from data of very similar river systems sampled by the same technique and having the same (Alestes baremoze and Brycinus macrolepidotus) or closely related (Schilbe intermedius and Chrysichthys auratus) species that we analysed here. These data come from the same World Health Organization (WHO) Onchocerciasis Control Program sampling protocol, but for countries neighbouring Côte d’Ivoire − Mali and Guinea – and from the years 1985 to 1988. Separation of CPUE by set of gill-nets for each sample allowed us to estimate a sampling error. As we are interested in the sampling error that is made with two sets of gill-nets, we constituted pseudo-samples and their replicates in the following way. First, two sampling dates were chosen at random. A set of gill-nets from the first sampling date was chosen at random and the corresponding CPUE is computed. The same procedure was conducted for the second date. The two resulting CPUEs were summed to obtain a pseudo-sample (C1) produced with two set of gill-nets. The CPUE corresponding of the remaining sets of gillnets were summed and constitute the replicate (C2 ) of C1. Given the assumptions that have been made in eqn 1, the sampling error variance, η, is estimated using the following relationship: Var [ log(C1) − log(C2)] = Var [θ1 − θ2] = 2η.

© 2004 British Ecological Society, Journal of Animal Ecology, 73, 693–705

Knowing η and in turn from eqn 5, ψ, two parameters remain to be estimated (a, φ). To do this we used covariance structure analysis (see Appendix I) by fitting the observed variance – covariance matrix between Yt+1 and Yt to the theoretical one deduced from eqn 4 and eqn 6. To test for spatial synchrony between two populations of the same species, simulations were performed to mimic the processes described by eqn 4 and eqn 6, without the constant terms. Given the estimated values for a and φ, we first simulated a one-order autoregressive process having the same length as the observed series by randomly drawing the error process value at each time step within a normal distribution of zero mean and variance φ. The same procedure was used independently for the second site. In both cases, the initial value generating the series was fixed at zero. Values drawn independently within a normal distribution of zero mean and variance ψ were added to each value of the two time-series to mimic sampling error. Finally, the correlation between the two series was computed, eventually by excluding values that are missing in the observed time-series. A sample of 1000 simulated correlation values provided the expected distribution under the null hypothesis that there is no Moran effect between populations against which the statistical significance of the observed value is assessed. We compared spatial synchrony between species in the same way, using the estimates of a, φ and ψ obtained for each species.

In order to minimize the number of missing values and sampling error variance on the mean we included only years that had been sampled at least three times (n = 3 in eqn 2, eqn 3 and eqn 5). If there were more than three samples in a year, three samples were drawn at random. Unfortunately, this approach cannot be applied to the whole data set because of the variability in the number of samples per year. Restricting the analysis to years that were sampled at least three times greatly reduced the sample size for most time-series. However, comparisons between sites 2 and 3 could be conducted with 19 years instead of 24 years and hence the modelling approach was carried out for these two sites (see the data in Appendix II). The knowledge of the sampling error variance allows the estimation of the Moran effect, the true correlation between the Xs (the true log of catches), by using the following formula (Worm & Myers 2003): Corr ( Xat , Xbt ) = Corr (Yat , Ybt )

[va/(va + ψ )] [vb/(vb + ψ )]

where Ya is the observed value for site a, Xa the true value and va is Var(Xa).

To assess the relationship between regional changes in hydrology and the fish population dynamics observed in this study we used the data of Mahé (1993), who computed regional indices of hydrology over different areas in West Africa. Côte d’Ivoire rivers belong to the so-called ‘Nordgolf ’ region that includes the coastal basins ranging from the Cavally in the west to the Oueme in the east. The regional index used is the total discharge (in m3 s−1) of rivers belonging to the region. To have a regional index of fish abundance, log(CPUE) were summed across sites and species for each year to produce an annual total catch of the four species analysed. As hydrological data are not available after 1989, the analysis was restricted to the period 1975–86 where the four localities were surveyed.

Results An inspection of the population dynamics graphs of annual mean abundance of the four species suggests a rather high degree of synchrony (Fig. 2). The time-series of the four species showed a peak from ∼1979–81 and a trough from ∼1992–94, before the abundances increased until the end of the series (Fig. 2). In order to compare the synchrony pattern between pairs of sites related to distance, we focused on years 1974–86 to produce comparable correlation coefficients between site 4 and the others. In such a comparison, no decrease in synchrony with intersite geographical distance was observed for any of the species (Fig. 2). We also noticed that synchrony

698 P. A. Tedesco et al.

Fig. 2. Annual mean time-series for the different species (a, b, c and d). For each species the relationship between synchrony and distance between sites is showed by correlograms where the triangle shows () the comparison between sites of the same basin (sites 1 and 2). Synchrony was calculated for time-series from 1974 to 1986 in order to compare synchrony between site 4 and others.

© 2004 British Ecological Society, Journal of Animal Ecology, 73, 693–705

between the two sites from the same catchment basin did not stand out from the other correlations. Mean cross-correlations showed notable differences between the first (1974 – 86, four sites sampled) and the second period (1987– 97, three sites sampled) (Table 2). Spatial synchrony during the second period was higher, even when site 4 was excluded from the first period. The four site × site correlation matrices from 1974 to 1986 were subjected to a non-parametric sign test. Considering only the independent correlations between sites 1 and 2 and between sites 3 and 4, there is only one observed negative correlation within these matrices (between sites 1 and 2 for Alestes baremoze, see Fig. 2).

Assuming a binomial distribution with a probability of 0·5 and 8 trials, the probability of observing a number of positive values higher than or equal to 7 is 0·035. As the number of positive correlations is significantly higher than expected, the null hypothesis of no spatial synchrony is thus rejected for the four species combined.

Using data from Mali and Guinea, the estimated variance of the error due to fishing with two sets of gill-nets appeared very similar between species (from 0·067 to

699 Synchrony and the Moran effect in freshwater fishes

Table 2. Mean cross-correlations excluding correlation between sites 1 and 2 (located within the same basin). Correlations for each species and for total time-series, period from 1974 to 1986 with and without site 4, and period from 1987 to 1997

Species

Mean correlation coefficient for total time-series

Mean correlation coefficient for time-series 1974 – 86

Mean correlation coefficient for time-series 1974 – 86 without site 4

Mean correlation coefficient for time-series 1987–97

A. baremoze B. macrolepidotus S. mandibularis C. maurus

0·38 0·48 0·45 0·43

0·17 0·29 0·37 0·26

0·05 0·14 0·46 0·17

0·67 0·65 0·73 0·75

Table 3. Estimates of parameters a, ψ and φ (see text and Appendix I) Species

A. baremoze

B. macrolepidotus

S. mandibularis

C. maurus

φ ψ a

0·0453 0·07 0·849

0·0039 0·07 0·984

0·0863 0·07 0·684

0·0620 0·07 0·715

Table 4. Results of the parametric bootstrapping test within and between species. Mean cross-correlations from regularized timeseries between sites 2 and 3 and the statistical significance shown by a P-value. No P-value for B. macrolepidotus is given because of a lack of reliable estimates of parameters from Table 3. Intraspecific correlation coefficients corrected for sampling error are also given

© 2004 British Ecological Society, Journal of Animal Ecology, 73, 693–705

r

r corrected for sampling error

Within-species A. baremoze B. macrolepidotus S. mandibularis C. maurus

0·61 (P = 0·01) 0·46 0·64 (P = 0·002) 0·61 (P = 0·007)

0·87 0·88 0·83 0·93

Between species A. baremoze–B. macrolepidotus A. baremoze–S. mandibularis A. baremoze–C. maurus B. macrolepidotus–S. mandibularis B. macrolepidotus–C. maurus S. mandibularis–C. maurus

0·55 0·65 (P < 0·001) 0·58 (P = 0·001) 0·48 0·58 0·58 (P < 0·001)

0·071). To simplify, we used a single value of 0·07 for ψ for the four species ( Table 3). Harley et al.’s (2001) covariance analysis including sampling error variance, ψ, allowed us to produce maximum likelihood estimates of parameters a and φ (Table 3). For B. macrolepidotus the estimated value of parameter a was close to 1 and led to an inconsistent estimation. As a result, no Monte-Carlo simulations were conducted for this species because of a lack of reliable estimation of parameters. From Fig. 3, the assumption of linearity for the autoregressive models seems realistic, because there is no clear discrepancy from a linear relationship between population growth rate and abundance. The relation was linear and negative in all species (Fig. 3). Finally, synchrony between regularized time-series from sites 2 and 3 was tested. We observed high correlation coefficients (r ∼ 0·6) and significant P-values for within-species synchrony as well as between-species

synchrony (Table 4). When removing the effect of sampling noise, spatial synchrony between species reached 0·88 on average (Table 4). There was no indication of less synchrony between species than within species.

According to Fig. 4, the regional discharge index is positively correlated with total log(CPUE) in the following year (r = 0·54). This suggests that interannual variability in recruitment and/or interannual survival linked to hydrology is one of the factors that can lead to a Moran effect. Because from 50% to 87% of the population abundances of the four species were caught by the smaller gill-nets mesh size (15 mm), it is likely that changes are driven mainly by the recruitment of young fish.

700 P. A. Tedesco et al.

Fig. 3. Relationship between the observed annual population growth rate Yt+1 – Yt and the observed log population abundance Yt for each population of fish species and for regularized time-series (see text for the definition of variable Y ).

Fig. 4. ‘Nordgolf ’ regional discharge index (log m 3 s −1) at year t − 1 from Mahé (1993) related to the summed abundance (log CPUE) of the four studied species in the following year t.

© 2004 British Ecological Society, Journal of Animal Ecology, 73, 693–705

When spatial synchrony between sites 2 and 3 were computed by removing the effect of sampling noise, high (0·87) or very high (0·93) values were obtained (Table 4). According to the Moran theorem, these correlations should on average equal the correlation of the synchronizing climatic factor(s) between the two sites. The correlation between annual discharges at the mouths of the Comoé river (site 2) and the Bandama river (site 3) is 0·84 (0·88 on a log-scale) over the period 1974–86 (from data in Mahé 1993), thus highly similar to the averaged intraspecific spatial synchrony corrected for sampling error (0·88). This observation is consistent with the probable role of hydrology as one of the main synchronizing factors.

Discussion In this study we identified significant synchrony in the population dynamics of four fish species between different catchment basins that cannot be connected by dispersal, suggesting the action of a climatic synchronizing factor (the Moran effect). This conclusion is strengthened by the interspecific synchrony that was observed because dispersal cannot be the synchronizing factor between non-conspecific populations. When separating time-series in two periods, from 1974 to 1986 and from 1987 to 1997, a higher degree of synchrony was found for the second period both within and between species. The most likely explanation for

701 Synchrony and the Moran effect in freshwater fishes

© 2004 British Ecological Society, Journal of Animal Ecology, 73, 693–705

the apparent change through time of the degree of synchrony is the irregular sampling frequency in the data (from one to six samples per year). Irregularity produced annual means with different sampling error rates for each year. Because sampling error decreases with the number of samples averaged, years with a few samples will have a higher sampling error and hence a reduced apparent synchrony. Between years 1978 and 1981 our time-series often had only two samples per year, while other years had four or more samples (Table 1). This could explain the difference in the degree of synchrony between periods. Theoretical models suggest that adding dispersal to the Moran effect increases spatial synchrony. In our study, the pair of sites located within the same basin did not have a higher synchrony than pairs of sites located in different basins. There was an exception, however, for Brycinus macrolepidotus (Fig. 2). Nevertheless, the distance between the two sites belonging to the same basin is > 400 km along the river, reducing the likelihood of a strong effect of dispersal on synchrony in this case. Spatial synchrony between related species has been already reported for many taxa (Ranta et al. 1995b; Lindström et al. 1996; Myers 1998; Cattadori et al. 2000; Post & Forchhammer 2002). In fact, if climatic perturbations are responsible for spatial synchrony, species sharing similar demographic strategies will react in the same way to environmental variations. Actually, interspecific spatial synchrony is stronger evidence of a Moran effect than intraspecific synchrony is because there is no possible effect of dispersal. The difficulty with interspecific comparisons is that the assumption of identical density-dependent dynamics underlying the Moran theorem does not hold. As a result, the strict equality between environmental correlation and population size correlation is not expected, even if spatial synchrony in population dynamics is nevertheless expected. Our results show almost identical spatial synchrony within and between species (0·58 and 0·57), suggesting that despite some differences between species in population dynamics parameters (Table 3), these differences are not sufficient to cause a serious bias when interspecific correlations are computed. It is thus likely that the species studied are responding in the same way to the same Moran effect. The occurrence of a Moran effect strongly suggests that climate has an impact on population dynamics. Nevertheless, identifying which climatic factors are involved in the Moran effect is not an easy task because the climatic factor actually causing synchrony need not to be the same every year. Hydrology (or precipitation) is a priori a good candidate for a Moran effect in freshwater fish because hydrological variability is known to affect riverine fish population dynamics (Holcík & Bastl 1977; Welcomme 1985, 1989; Laë 1992; Smolders et al. 2000; Grimes 2001; Moses 2001). In particular, a strong correlation between some measures of flood intensity and catches in subsequent years is observed frequently in intertropical fisheries (Welcomme 1985,

1989; Laë 1992; Smolders et al. 2000; Moses 2001) because floods affect many species of fish through improved breeding, growth and survival as flood level and duration increase. Our results show a positive correlation between the regional discharge index and the total log(CPUE) in the following year, and therefore recruitment appears to be the main factor affecting changes in abundances. This is consistent with previous studies on temperate fish that have shown that recruitment is the parameter most affected by regional fluctuations in weather (Grenouillet et al. 2001; Cattaneo et al. 2003). Our results did not show a decrease of the degree of synchrony with increasing distance. A clear decrease in synchrony with distance is generally resulting from dispersal (Ranta et al. 1998b; Cattadori et al. 2000). As there is no connection between most sites studied here, our result is not surprising. The absence of any spatial pattern in population synchrony suggests that the underlying synchronizing factor, climate, is homogeneous over the considered area. This is the case for hydrology, because the three rivers considered here belong to the same hydrological region, the ‘Nordgolf’ (Mahé 1993). The Moran theorem states that the correlation between population size is the same as that between the environmental stochasticity. When computing synchrony by removing the sampling noise, the environmental and population correlations are nearly identical and confirm the role of hydrology as a synchronizing factor. A study on the scales of correlation for recruitment in fish populations (Myers, Mertz & Bridson 1997) concluded that climate is a more important synchronizing factor in the marine environment than in freshwaters. According to this study the correlation scale for recruitment in marine species is typically 500 km, whereas for freshwater species it is less than 50 km. However, this conclusion is based on only five freshwater species, some of them in lakes. Spatial synchrony in recruitment has been recently observed at a large spatial scale (> 100 km) for European riverine species (Grenouillet et al. 2001; Cattaneo et al. 2003). Our results demonstrate a largescale (160 km) spatial synchrony for at least three species of freshwater fish. Clearly, more studies are needed to improve knowledge of the occurrence of spatial synchrony in riverine fish population dynamics. Besides the Moran effect and dispersal, a third hypothesis has been suggested for explaining spatial synchrony in population dynamics: predation by nomadic predator(s) (Ims & Steen 1990). This hypothesis assumes that a predator species is capable of rapidly tracking the local prey densities by dispersing between sites. Because of the large spatial scale involved in the present study, no predator species can fit this role, particularly as it would have to be terrestrial or aerial to move freely between sites. However, spatial synchrony may be caused by human predation. Unfortunately, there are no fishery statistics allowing the comparison of interannual variation in fishing effort between Côte d’Ivoire rivers, so there is no way of testing for a synchronizing

702 P. A. Tedesco et al.

© 2004 British Ecological Society, Journal of Animal Ecology, 73, 693–705

role of fisheries. It should be mentioned that site 2 is located within a nature reserve in which fishing is prohibited and nevertheless the temporal catch pattern is there similar to the other sites. Estimating spatial synchrony and testing its statistical significance depend on the model chosen to mimic the observed times series of population abundances. If the model is not realistic or if the parameters are not estimated accurately, misleading results may be obtained. The time-series modelling retained in this study may seem complex with regard to approaches that have been used commonly for studying spatial synchrony (but see Cattadori et al. 2000). However, it cannot be simplified without losing realism because sampling noise cannot be neglected. The fact that no reliable parameter estimations could be obtained for B. macrolepidotus underlines the difficulties that are raised by the use of this model. In the case of B. macrolepidotus, a slight reduction in the value of the sampling variance (from 0·07 to 0·06) leads to stable estimates of the required parameters. This suggests that correct estimation of the sampling variance is a crucial step. The fact that sampling variances were estimated in other regions than Côte d’Ivoire and in some cases for other species than the ones analysed here raises the possibility that these estimates were confounded by interspecific and geographical variation. However, as the estimates were highly similar between species and between regions, it can be concluded that interspecific and geographical differences are not a point of concern. From the data in Table 4, the mean of the observed values of spatial synchrony is 0·58 between site 2 and site 3, but increases to 0·88 if the sampling noise is removed. As this value matches almost perfectly the correlation observed at the same scale between hydrological variability, there is no reason to believe that the variance in the sampling error has been greatly under or over estimated. Great emphasis has been given to non-linearity in population dynamics as a factor obscuring the relationship between population synchrony and environmental correlation (Grenfell et al. 1998; Blasius & Stone 2000; Grenfell, Finkenstädt & Wilson 2000; Greenman & Benton 2001). As the Gompertz model provides a good linearization of the fish data analysed (Fig. 2), non-linearity is not a point of concern in the present study. Intersite differences in population parameters of the same species may also lead to population synchrony that differs from the environmental correlation (Peltonen et al. 2002). To keep the number of parameters low with regard to sample size, populations of the same species have been considered homogenous in their population dynamics. The fact that interspecific spatial synchrony does not differ from intraspecific spatial synchrony suggests that differences in population parameters do not greatly affect spatial synchrony. It is likely that sampling error occurred in most of the populations that have been analysed so far in the search for spatial synchrony and that the Moran effect has been underestimated to an unknown degree. This can be problematic if values of spatial syn-

chrony are compared with values of climatic correlation with the aim of assessing the likelihood of a Moran effect (for instance Koenig 2002; Peltonen et al. 2002). The concern about global climatic change has raised the problem of the mismatch between ecological and climatic modelling scales (Root & Schneider 1995; Walther et al. 2002). Within the foreseeable future, even the highest resolution three-dimensional general circulation models (GCM) suitable for integration over 50 or more years will not have a grid with nodes much less than 100 km apart (Root & Schneider 1993). Most ecological research occurs on scales far smaller than that. As a result it is not well known if ecological processes can be described appropriately, or are homogeneous in their effects, at spatial scales matching GCM’s resolution. Identifying a Moran effect acting on populations dynamics over such spatial scales, as in the present study, provides useful information in this context by confirming the regional impact of climate on population processes.

Acknowledgements This work was made possible through invaluable participation and the input of many individuals. The World Health Organization (WHO) Onchocerciasis Control Programme (OCP) has given the main part of the financial support for the aquatic monitoring. We are very grateful to the West African fishermen for realizing the 24-year data sets. We also thank Elimane Diop for providing some of the data used for estimating sample error variance and two anomymous referees for improving the quality of this manuscript. We thank Crane Rogers for his corrections as an Anglophone. The financial support which made this work possible came from a PhD fellowship of the IRD (Institut de Recherche pour le Developpement) within the Research Unit 131.

References Benton, T.G., Lapsley, C.T. & Beckerman, A.P. (2001) Population synchrony and environmental variation: an experimental demonstration. Ecology Letters, 4, 236 – 243. Bjørnstad, O.N., Ims, R.A. & Lambin, X. (1999) Spatial populations dynamics: analysing patterns and processes of population synchrony. Trends in Ecology and Evolution, 14, 427 – 432. Blasius, B. & Stone, L. (2000) Non linearity and the Moran effect. Nature, 406, 846 – 847. Buonaccorsi, J.P., Elkinton, J.S., Evans, S.R. & Liebhold, A.M. (2001) Measuring and testing for spatial synchrony. Ecology, 82, 1668 – 1679. Carpenter, S.R., Cottingham, K.L. & Stow, C.A. (1994) Fitting predator–prey models to time series with observation errors. Ecology, 75, 1254 – 1264. Cattadori, I.M., Hudson, P.J., Merler, S. & Rizzoli, A. (1999) Synchrony, scale and temporal dynamics of rock partridge (Alectoris graeca saxatilis) populations in the Dolomites. Journal of Animal Ecology, 68, 540 – 549. Cattadori, I.M., Merler, S. & Hudson, P.J. (2000) Searching for mechanisms of synchrony in spatially structured gamebird populations. Journal of Animal Ecology, 69, 620–638.

703 Synchrony and the Moran effect in freshwater fishes

© 2004 British Ecological Society, Journal of Animal Ecology, 73, 693–705

Cattaneo, F., Hugueny, B. & Lamouroux, N. (2003) Synchrony in brown trout, Salmo trutta, population dynamics: a ‘Moran effect’ on early-life stages. Oikos, 100, 43 – 54. De Valpine, P. & Hastings, A. (2002) Fitting population models incorporating process noise and observation error. Ecological Monographs, 72, 57 – 76. Den Boer, P.J. & Reddingius, J. (1996) Regulation and Stabilization Paradigms in Population Ecology. Chapman & Hall, London. Dennis, B. & Taper, M.L. (1994) Density dependence in time series observations of natural populations estimation and testing. Ecological Monographs, 64, 205 – 224. Fromentin, J.-M., Myers, R.A., Bjørnstad, O.N., Stenseth, N.C., Gjosaeter, J. & Christie, H. (2001) Effects of densitydependent and stochastic processes on the regulation of cod populations. Ecology, 82, 567 – 579. Greenman, J.V. & Benton, T.G. (2001) The impact of stochasticity on the behaviour of non linear population models: synchrony and the Moran effect. Oikos, 93, 343 – 351. Grenfell, B.T., Finkenstädt, B.F. & Wilson, K. (2000) Nonlinearity and the Moran effect − reply. Nature, 406, 847. Grenfell, B.T., Wilson, K., Finkenstädt, B.F., Coulson, T.N., Murray, S., Albon, S.D., Pemberton, J.M., Clutton-Brock, T.H. & Crawley, M.J. (1998) Noise and determinism in synchronized sheep dynamics. Nature, 394, 674 – 677. Grenouillet, G., Hugueny, B., Carrel, G.A., Olivier, J.M. & Pont, D. (2001) Large-scale synchrony and inter-annual variability in roach recruitment in the Rhône River: the relative role of climatic factors and density dependent processes. Freshwater Biology, 46, 11 – 26. Grimes, C.B. (2001) Fishery production and the Mississippi River discharge. Fisheries, 26, 17 – 26. Hanski, I. & Woiwod, I.P. (1993) Spatial synchrony in the dynamics of moth and aphid populations. Journal of Animal Ecology, 62, 656 – 668. Harley, S.J., Myers, R.A. & Dunn, A. (2001) Is catchper-unit-effort proportional to abundance? Canadian Journal of Fisheries and Aquatic Sciences, 58, 1760 – 1772. Haydon, D. & Steen, H. (1997) The effects of large- and smallscale random events on the synchrony of metapopulation dynamics: a theoretical analysis. Proceedings of the Royal Society of London, Series B, 264, 1375 – 1381. Heino, M., Kaitala, V., Ranta, E. & Lindström, J. (1997) Synchronous dynamics and rates of extinction in spatially structured populations. Proceedings of the Royal Society of London, Series B, 264, 481 – 486. Holcík, J. & Bastl, I. (1977) Predicting fish yield in the Czechoslovakian section of the Danube River based on the hydrological regime. Internationale Revue der Gesamten Hydrobiologie, 62, 523 – 532. Hugueny, B. & Lévêque, C. (1994) Freshwater fish zoogeography in west Africa: faunal similarities between river basins. Environmental Biology of Fishes, 39, 365 – 380. Ims, R.A. & Andreassen, H.P. (2000) Spatial synchronization of vole population dynamics by predatory birds. Nature, 408, 194 –196. Ims, R.A. & Steen, H. (1990) Geographical synchrony in microtine population cycles: a theoretical evaluation of the role of nomadic avian predators. Oikos, 57, 381 – 387. Kendall, B.E., Bjørnstad, O.N., Bascompte, J., Keitt, T.H. & Fagan, W.F. (2000) Dispersal, environmental correlation, and spatial synchrony in population dynamics. American Naturalist, 155, 628 – 636. Koenig, W.D. (1999) Spatial autocorrelation of ecological phenomena. Trends in Ecology and Evolution, 14, 22 – 26. Koenig, W.D. (2002) Global patterns of environmental synchrony and the Moran effect. Ecography, 25, 283 – 288. Laë, R. (1992) Influence de l’hydrologie sur l’évolution des pêcheries du delta central du Niger, de 1966 –1989. Aquatic Living Resources, 5, 115 – 126.

Lande, R., Engen, S. & Saether, B.-E. (1999) Spatial scale of population synchrony: environmental correlation versus dispersal and density regulation. American Naturalist, 154, 271–281. Lévêque, C., Fairhurst, C., Abban, K.E., Paugy, D. & Curtis, M.S. (1988) Onchocerciasis control programme in West Africa: ten years monitoring of fish populations. Chemosphere, 17, 421 – 440. Lindström, J., Ranta, E. & Lindén, H. (1996) Large-scale synchrony in the dynamics of capercaillie, black grouse and hazel grouse populations in Finland. Oikos, 76, 221–227. Link, W.A. & Nichols, J.D. (1994) On the importance of sampling variance to investigations of temporal variation in animal population size. Oikos, 69, 539 – 544. Mackin-Rogalska, R. & Nabaglo, L. (1990) Geographical variation in cyclic periodicity and synchrony in the common vole, Microtus arvalis. Oikos, 59, 343 – 348. Mahé, G. (1993) Les Écoulements Fluviaux Sur la Façade Atlantique de l’Afrique. ORSTOM, Paris. de Mérona, B. (1981) Zonation ichthyologique du bassin du Bandama (Côte d’Ivoire). Revue D’hydrobiologie Tropicale, 14, 63 – 75. Moran, P.A.P. (1953) The statistical analysis of the Canadian lynx cycle. II. Synchronisation and meteorology. Australian Journal of Zoology, 1, 292 – 298. Moses, B.S. (2001) The influence of hydroregime on catch, abundance and recruitment of the catfish, Chrysichthys nigrodigitatus (Bagridae) and the bonga, Ethmalosa fimbriata (Clupeidae) of southeastern Nigeria’s inshore waters. Environmental Biology of Fishes, 61, 99 – 109. Myers, J.H. (1998) Synchrony in outbreaks of forest Lepidoptera: a possible example of the Moran effect. Ecology, 79, 1111 –1117. Myers, R.A., Mertz, G. & Bridson, J. (1997) Spatial scales of interannual recruitment variations of marine, anadromous, and freshwater fish. Canadian Journal of Fisheries and Aquatic Sciences, 54, 1400 –1407. Paradis, E., Baillie, S.R., Sutherland, W.J. & Gregory, R.D. (1999) Dispersal and spatial scale affect synchrony in spatial population dynamics. Ecology Letters, 2, 114–120. Paugy, D., Fermon, Y., Abban, K.E., Diop, M.E. & Traoré, K. (1999) Onchocerciasis control programme in West Africa: a 20-years monitoring of fish assemblages. Aquatic Living Resources, 12, 363 – 378. Peltonen, M., Liebhold, A.M., Bjørnstad, O.N. & Williams, D.W. (2002) Spatial synchrony in forest insect outbreaks: roles of regional stochasticity and dispersal. Ecology, 83, 3120– 3129. Post, E. & Forchhammer, M.C. (2002) Synchronization of animal population dynamics by large-scale climate. Nature, 420, 168 – 171. Pyper, B.J. & Peterman, R.M. (1998) Comparison of methods to account for autocorrelation in correlation analyses of fish data. Canadian Journal of Fisheries and Aquatic Sciences, 55, 2127 – 2140. Ranta, E., Kaitala, V. & Lindström, J. (1998a) Spatial dynamics of populations. Modelling Spatiotemporal Dynamics in Ecology (eds J. Bascompte & R. V. Solé), pp. 47 – 62. Princeton University Press, Princeton. Ranta, E., Kaitala, V., Lindström, J. & Lindén, H. (1995a) Synchrony in population dynamics. Proceedings of the Royal Society of London, Series B, 262, 113 –118. Ranta, E., Kaitala, V. & Lundberg, P. (1997) The spatial dimension in population fluctuations. Science, 278, 1621– 1623. Ranta, E., Kaitala, V. & Lundberg, P. (1998b) Population variability in space and time: the dynamics of synchronous population fluctuations. Oikos, 83, 376 – 382. Ranta, E., Lindström, J. & Lindén, H. (1995b) Synchrony in tetraonid population dynamics. Journal of Animal Ecology, 64, 767 –776.

704 P. A. Tedesco et al.

© 2004 British Ecological Society, Journal of Animal Ecology, 73, 693–705

Root, T.L. & Schneider, S.H. (1993) Can large-scale climatic models be linked with multiscale ecological studies? Conservation Biology, 7, 256 –270. Root, T.L. & Schneider, S.H. (1995) Ecology and climate: research strategies and implications. Science, 296, 334 –341. Royama, T. (1992) Analytical Population Dynamics. Chapman & Hall, London. Rusak, J.A., Yan, N.D., Somers, K.M. & McQueen, D.J. (1999) The temporal coherence of zooplankton population abundances in neighbouring North-temperate lakes. American Naturalist, 153, 46 – 58. Shenk, T., White, G.C. & Burnham, K.P. (1998) Samplingvariance effects on detecting density dependence from temporal trends in natural populations. Ecological Monographs, 68, 445 – 463. Smolders, A.J.P., van der Velde, G., Roelofs, J.G.M. & Guerrero Hiza, M.A. (2000) El Nino caused collapse of the Sabalo fishery (Prochilodus lineatus, Pisces: Prochilodontidea) in a South American river. Naturwissenschaften, 87, 30 – 32. Solow, A.R. (1995) Fitting population models to time series data. Ecological Time Series (eds T.M. Powell & J.H. Steele), pp. 20 – 27. Chapman & Hall, New York. Solow, A.R. (1998) On fitting a population model in the presence of observation error. Ecology, 79, 1463 –1466. Straile, D. (2002) North Atlantic Oscillation synchronises food–web interactions in central European lakes. Proceedings of the Royal Society of London, Series B, 269, 391 – 395.

Sutcliffe, O.L., Thomas, C.D. & Moss, D. (1996) Spatial synchrony and asynchrony in butterfly population dynamics. Journal of Animal Ecology, 65, 85 – 95. Thompson, K.R. & Page, F.H. (1989) Detecting synchrony of recruitment using short, autocorrelated time series. Canadian Journal of Fisheries and Aquatic Sciences, 46, 1831–1838. Turchin, P. & Ellner, S.P. (2000) Living on the edge of chaos: population dynamics of Fennoscandian voles. Ecology, 81, 3099 – 3116. Walther, G.-R., Post, E., Convey, P., Menzel, A., Parmesan, C., Beebee, T.J.C., Fromentin, J.-M., Hoegh-Guldberg, O. & Bairlein, F. (2002) Ecological responses to recent climate change. Nature, 416, 389 – 395. Welcomme, R.L. (1985) River Fisheries, Fisheries Technical Paper 262. FAO, Rome. Welcomme, R.L. (1989) Review of the present state of knowledge of fish stocks and fisheries of African Rivers. Canadian Special Publication of Fisheries and Aquatic Science, 106, 515 – 532. Worm, B. & Myers, R.A. (2003) Meta-analysis of cod–shrimp interactions reveals top-down control in oceanic food webs. Ecology, 84, 162 – 173. Ydenberg, R.C. (1987) Nomadic predators and geographical synchrony in microtine population cycles. Oikos, 50, 270– 272. Received 19 May 2003; accepted 15 December 2004

705 Synchrony and the Moran effect in freshwater fishes

Appendix I Given eqn 4 and eqn 6, the theoretical variance– covariance matrix between catches at t + 1 and catches at t is: Var(Yt + 1 )

∑ = Cov(Y ,Y

t

t +1

)

Cov(Yt + 1,Yt ) Var(Yt )

φ aφ +ψ 2 2 1 a 1 a − − = a φ φ ψ + 2 1 − a 2 1−a

As ψ is estimated independently, there are two unknown parameters (a, φ). These remaining parameters are estimated by using a maximum likelihood approach. The basic procedure is to fit the observed variance-covariance matrix (S ) to the theoretical one. The fitting function that is minimized to find the maximum likelihood estimates is: −log | ∑ | + tr( S ∑ ) − log | S | − d −1

It should be noted that because a is constrained to be less than one, unregulated populations following a random walk (a = 1) cannot be modelled properly with this approach.

where ‘tr’ is the matrix trace and d is the dimension of S. To do this we used the function ‘nlminb’ from the S+ statistical package.

Appendix II The time-series (Yt) used to fit the variance-covariance matrices are presented in the following table. Under the assumption that a species has the same dynamic in different sites, both sites 2 and 3 were joined for each species. A. baremoze

© 2004 British Ecological Society, Journal of Animal Ecology, 73, 693–705

B. macrolepidotus

S. mandibularis

C. maurus

Years

Site 2

Site 3

Site 2

Site 3

Site 2

Site 3

Site 2

Site 3

75 76 77 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97

1·4911 1·4471 1·4400 1·1781 1·5252 1·8147 1·3821 1·2503 0·9859 1·2677 1·4126 0·4806 1·0638 0·4473 0·3037 0·5308 0·9905 1·6819 0·9890

2·0198 1·8845 1·1368 1·3418 1·8827 1·0616 0·9687 0·7843 0·9821 0·8742 0·8781 0·9514 0·4183 0·3842 0·5199 − 0·0769 0·9620 0·9755 1·1511

1·0919 0·5262 0·4664 0·5555 0·4838 0·8948 1·0773 0·4177 0·6484 0·6190 0·1185 − 0·1298 − 0·1906 0·0249 − 0·5176 0·2443 0·4924 0·7727 0·6621

0·3220 0·7442 0·9769 0·6418 1·0858 0·5910 0·5579 0·5421 0·6337 0·8687 0·6992 0·2294 − 0·0628 0·2212 0·2678 0·2174 0·0744 0·8074 1·0493

1·2499 1·4287 1·3277 1·0923 1·3170 0·9633 1·0887 1·0013 1.0316 0·9602 − 0·1298 0·7983 0·0182 − 0·2078 − 0·5176 − 0·2766 0·7973 1·3263 1·3044

0·8673 1·0859 1·1968 1·0415 0·8593 0·5860 0·7191 − 0·0628 0·6113 0·6386 0·3593 0·6546 0·5082 − 0·3518 0·5049 − 0·3222 0·5314 0·7460 0·2576

0·6200 0·8494 0·4925 0·7162 1·2838 1·1893 − 0·4073 0·7588 0·5769 0·9717 0·7600 0·8863 0·1965 0·0122 − 0·0382 0·0454 0·5379 1·2715 0·8456

0·2557 0·3770 0·6997 0·4125 0·7817 0·5146 0·4719 0·6095 0·5221 0·2174 0·4025 0·2582 − 0·4396 − 0·1298 − 0·2582 − 0·5176 0·5677 0·7183 0·7510