A LEPIDOCHRONOLOGICAL STUDY OF POSIDONIA OCEANICA: FURTHER ANALYSES ON ANNUAL LEAF PRODUCTION TIME SERIES
 

P. DI DATO*, E. FRESI** AND M. SCARDI*

*Dept of Zoology, University of Bari, Via Orabona 4/A, 70125 Bari, Italy
** Dept. of Biology, University of Rome “Tor Vergata”, Via della Ricerca Scientifica, 00133 Roma, Italy
email: pdidato@mclink.it

Abstract
Lepidochronological analysis is a reliable technique for the identification of annual cycles of leaf and rhizome production in Posidonia oceanica (Potamogetonaceae). Even though several papers have pointed out that periodical patterns in leaf production can be detected, more recent studies have showed that such patterns cannot be found in time series from several Tyrrhenian sites. Runs tests and cross-association procedures have been used to analyze annual leaf production time series from the Villasimius bed (Sardinia). The results clearly showed that the annual leaf production follows a random pattern.

Key-words: lepidochronological analysis, Posidonia oceanica, seagrass production, time series analysis


Introduction
When the leaves of Posidonia oceanica die, the blades detach from their bases, while the bases themselves stay on the rhizome and form a "scale” that slowly decompose (Pergent, 1987). Scale thickness and other anatomic parameters show cyclic variations along the rhizome with annual periodicity (Crouzet, 1981). Depending on both depth and site, the minimum thickness appears in late winter (February-March) and the maximum thickness in autumn (September-October) (Crouzet et al., 1983; Pergent, 1987). The variations in scale thickness and anatomic parameters are not random and are probably caused by the plant response to either endogenous (physiologic and morphogenetic) or exogenous factors (climatic and edaphic) (Pergent, 1987; 1990).
During the last 15 years several papers have showed that lepidochronological analyses are a useful tool in Posidonia ecosystem studies. This technique allows to estimate present and past primary production and to analyse temporal fluctuations of biotic and abiotic factors to which the ecosystem has been exposed (Pergent 1987; 1990). Since scales and rhizomes can persist within the matte for millennia (>4600 years) (Boudouresque et al., 1980), it is possible to retrieve information about the structure and the dynamics of a Posidonia bed up to a quite long time lapse in the past (Boudouresque et al., 1983).
Our study is aims at checking the results obtained in a previous work by  Dolce et al. (1996) about lepidochronological analyses of several Tyrrhenian beds (i.e. Ponza, Ventotene, Palmarola, Giglio, Gulf of Congianus, Villasimius). Those results confirmed the effectiveness of the lepidocronological approach as a tool for estimating primary production, but they also suggested that the interpretation of the variations in scales number per year along the rhizome should be revised, since neither periodicity, nor synchronicity among different rhizomes were detected (see examples in Fig.1).
In particular, in order to check those evidences, we decided to analyze lepidocronological data based on a larger sample of rhizomes from Villasimius (Sardinia), that is one of the sites that were considered in the above cited paper, using the same statistical techniques. We also compare the observed leaf production time series to randomly generated series.


Fig. 1

Hypothesis 1

There is synchronism in the number of scales per cycle among rhizomes from the same site 

Hypothesis 2 

There is no synchronism in the number of scales per cycle among rhizomes from  the same site


Materials and methods
In summer 1993 (July-August) 160 orthotropic rhizomes were collected in 16 stations along south-eastern Sardinia coast (in a 10 to 30 metres depth range).
The scales were detached from the rhizome according to their insertion rank and numbered from the oldest to the most recent one (i.e. the one at the living leaves end). Lepidochronological analysis has been carried out according to the method described by Pergent et al. (1987).
Following the same procedure described by Dolce et al. (1996), we have submitted to runs tests only those rhizomes where more than six complete annual cycles were detected (i.e. 54 out of 160 available rhizomes) and cross-association analysis has been carried out only on the most recent portion of rhizomes with more than eight complete annual cycles (i.e. on 33 of 160 available rhizomes) (see table 1).
 

 
Station
 1a 
 3a 
 4a 
 1d 
 3b 
 4b 
 2d
 4c 
 3d
 1b 
 3c 
 2c 
 4d 
Depth
10
10
10
15
15
15
20
20
20
30
30
30
30
Runs tests
2
2
7
1
6
6
2
6
4
5
3
7
3
Cross-association
1
2
6
0
4
4
1
3
1
4
1
4
2

Table 1. Number of analyzed rhizomes from each sampling station

 

The rationale that supports the choice of runs tests and cross-association analysis is that a qualitative approach to time series analysis is less influenced than a quantitative one by errors in the determination of scale thickness (that, in turn, may cause the shifting of a scale from an annual cycle either to the preceding or the following one).
The runs test is a simple statistical procedure that allows to check whether a series of events has been originated from a completely random process or not. Each "run" is defined as a sequence of observations during which the system does not change its state. Annual leaf production data have been transformed to a binary variable according to the number of produced leaves: greater (1) or smaller (0) than the mean value of all the analyzed rhizomes (7.45 leaves year-1). The null hypothesis for the runs test is that the series of observations does not differ from a randomly generated series, i.e. Ho: Uo = Ue where Uo is the observed number of "runs" and Ue is the number of runs that is expected for a random process. Since the (Uo - Ue)/SEU ratio (where SEU is the standard error of U) is distributed as a standard normal variate and the test is two-tailed, when its absolute value is larger than 1.96 the null hypothesis can be rejected at a 5% level of significance (Davis, 1986).
Cross-association is equivalent to cross-correlation, but it allows to compare sequences of nominal data. To compare the time series of annual leaf production in different rhizomes, each datum in the series was coded by means of a three-state variable, i.e. less than average (<7), average (7) or more than average (>7), according to Dolce et al. (1996). Since cross-association is a chi-square statistics and the length of the time series was limited, a Yates correction was applied.
The null hypothesis states that the two sequences are independent of each other and it can be rejected if the number of matches is larger than the expectation for a random series, i.e. if the chi-square value is larger than 3.84 (at a 5% level of significance). Further details about cross-association, in particular as far as the computations for E and E' are concerned, can be found in Dolce et al. (1996).
Finally, cross-association tests have been carried out not only between couples of different rhizomes, but also between observed rhizomes and randomly generated sequences. The latter were obtained both by means of a random number generator (constrained to the observed distribution of the annual leaf production data) and by randomly shuffling the annual leaf production data for each observed rhizome.

Results and discussion
Lepidochronological analyses showed that in the Villasimius Posidonia bed the average annual leaf production was 7.45 and that annual leaf production ranged from 4 to 12. Scale thickness ranged from 100 micron to 1300 micron with an average value of 560 micron. Although we collected quite long rhizomes, in only one station more than 20 annual cycles could be counted.
Our results support the conclusions by Dolce et al. (1996) and suggest that the annual leaf production along Posidonia oceanica rhizomes is modulated by a random dynamics (Fig. 2). In fact, our results showed that leaf production data sequences are random within almost every observed rhizome and independent of each other when different rhizomes were compared, even when they were collected in the same site.


 
 Fig. 2

 

A further evidence for a random dynamics was provided by the comparison of observed leaf production data to randomly generated sequences. The results of the statistical tests showed no differences between the two cases (i.e. comparisons among observed data and comparisons between observed data and random sequences), so it is possible to infer that the leaf production in Posidonia oceanica is not a fully deterministic process (see table 2).
   

 
 
Null Hypothesis
Results (p=0.05)
Runs 
Tests
Ho: 
the sequence was generated 
by a random process
 Rejected in 1 out of 54 cases (1.9%)
Cross- 
Association
Ho: 
the two sequences are
independent of each other
 Rejected in 6 out of 528  (1.1%) comparisons among real data series

Rejected in 24 out of 528  (4.5%) comparisons between real and artificial (i.e. randomly generated) data series

Rejected in 10 out of 528  (1.8%) comparisons between real and shuffled data series

Table 2. Results obtained by runs test and cross-association
 

Of course, it could be theoretically regulated by several environmental and physiological factors, but the heterogeneity of the response in different rhizomes, even in the same site and at a very small spatial scale, does not allow to understand which factors are actually relevant.
Since leaf production is a discrete measurement, errors in the determination of the number of scales that belong to the same annual cycle might severely bias the analysis of the time series of leaf production. This is a problem that cannot be easily solved in lepidocrhonological analyses, but using very robust statistical procedures, which rely on a very low level of information (e.g. above/below average annual leaf production), is certainly safer than using a quantitative time series approach.
In conclusion, our results imply that even though scale ranks cannot be considered as time units and that the only reliable information provided by lepidochronological analysis is the length of the rhizome region included between two subsequent minima in scale thickness. Nevertheless, lepidochronological analyses is still to be considered as a very useful tool to determine annual production of Posidonia oceanica in terms of rhizome growth.
 

References

BOUDOURESQUE C.F., GIRAUD G., PANAYOTIDIS P. (1980) Végétation marine de l’île de Port-Cros (Parc National). XIX. Mise en place d’un transect permanent. Tra. Sci. Parc nation. Port-Cros, Fr., 6: 207-221.
BOUDOURESQUE C.F., CROUZET A., PERGENT G. (1983) Un nouvel outil au service de l’étude des herbiers à Posidonia oceanica: La Lépidochronologie. Rapp. P.V. Réun. Commiss. internation. Explor. sci. Médit Monaco, 28 (3): 111-112.
CROUZET A. (1981) Mise en évidence de variations cycliques dans les écailles de rhizomes de Posidonia oceanica (Potamogetonaceae). Trav. sci. Parc nation. Port-Cros, Fr., 7: 129-135.
CROUZET A., BOUDOURESQUE C.F., MEINESZ A., PERGENT G. (1983) Evidence of the annual character of cyclic changes of Posidonia oceanica scale thickness (erect rhizomes). Rapp. P.V. Réun. Commiss. internation. Explor. sci. Médit Monaco, 28 (3): 112-113.
DAVIS J.C. (1986) Statistics and data analysis in geology. Wiley & Sons, Inc., New York: x+646 pp.
DOLCE T, ZIANTONI S., SCARDI M., FRESI E. (1996) Studio lepidocronologico di Posidonia oceanica (L.) Delile in alcuni siti del Mar Tirreno. Atti del VII Congresso Nazionale della Società Italiana di Ecologia. Napoli, 11-14 Settembre 1996, 17: 301-303.
PERGENT G. (1987) Recherches lépidochronologiques chez Posidonia oceanica (Potamogetonaceae). Fluctuations des paramètres anatomiques et morphologiques eds écailles eds rhizomes. Thèse Doctor. Océanol., Univ. Aix marseille II, Fr., 853 pp.
PERGENT G., 1990. Lepodochronological analysis of the seagrass Posidonia oceanica (L.) Delile: a standardized approach. Aquat. Bot., 37 39-54.