LAND-SNAIL DIVERSITY IN A SQUARE KILOMETRE OF CRETAN MAQUIS: MODEST SPECIES RICHNESS, HIGH DENSITY AND LOCAL HOMOGENEITY R. A. D. CAMERON 1 , M. MYLONAS 2,3 , K. TRIANTIS 2,3 , A. PARMAKELIS 2,3 AND K. VARDINOYANNIS 2 1
Department of Animal and Plant Sciences, University of Sheffield, Sheffield S10 2TN, UK and Department of Zoology, The Natural History Museum, London SW7 5BD, UK; 2Natural History Museum of Crete, University of Crete, P.O. Box 2208, 71409 Irakleio, Crete, Greece; 3 Department of Biology, University of Crete, P.O. Box 2208, 71409, Irakleio, Crete, Greece (Received 13 May 2002; accepted 11 July 2002)
ABSTRACT 2
The land mollusc fauna of 1 km of Cretan maquis was surveyed by sampling fourteen 400-m2 plots in May 2001, and by resampling six of these in February 2002. Sampling methods were designed to resemble those used in similar surveys of 1-km2 sites in tropical rainforests in Cameroon and Sabah. A total of 27 species were recorded for the site. Slugs and a semi-slug were found only in the 2002 survey. Individual plots were very similar in species richness and composition; the richest plot contained about 85% of the fauna recorded for the whole site. Overall densities were very high. This local homogeneity contrasts with marked heterogeneity over Crete as a whole. These results are contrasted with those from tropical forests, where individual plots vary considerably in richness and composition, where densities are lower and where the total site faunas are larger. Although there are problems associated with differing amounts of sampling error between studies, this contrast is striking and some possible causes are discussed. Further work in tropical forest on limestone may elucidate this.
INTRODUCTION
THE STUDY SITE
Two recent papers (de Winter & Gittenberger, 1998; Schiltuizen & Rutjes, 2001) have assessed the species richness of landmollusc faunas within 1 km2 plots of tropical rainforest. Contrary to earlier claims (Solem, 1984), both these studies revealed high levels of species richness, but also very low densities, and considerable heterogeneity in richness and species composition among smaller plots of 400 m2 within the study sites. Much of the heterogeneity showed no connection with any observed environmental variation among plots. The study in Cameroon by de Winter & Gittenberger (1998) revealed the richest known fauna (97 species) from a single habitat at this scale. It also discussed methodological issues, and suggested a set of protocols for comparative studies. Schilthuizen & Rutjes (2001) used a modified version of these protocols in their study in Sabah, Borneo. While there are many data on land mollusc species richness in specified areas from around the world, there are no replicates of the kind of study carried out in these rainforests. There is good evidence that relationships between densities, species richness and area vary considerably with habitat and region (R. A. D. Cameron, unpublished). It is appropriate to get comparable data from elsewhere. This paper reports on a study of land-mollusc species richness in 1 km2 of maquis vegetation on the island of Crete, using protocols similar to those used in the rainforest studies. We relate our findings to those from the rainforests; because this raises issues of sampling efficiency, we also make comparisons with Tattersfield’s (1996) study of Kenyan rainforest, despite its different aims and scale. We also relate our findings to the fauna of Crete as a whole, and consider the spectrum of size and shape in the site fauna relative to spectra for the site in Cameroon (de Winter & Gittenberger, 1998) and some sites elsewhere (Emberton, 1995; Solem & Climo, 1985).
Apart from cultivations, maquis and phrygana are the predominant vegetation types on Crete (Rackham & Moody, 1996). In choosing a site, we wished to sample in an area in which there was at least 1 km2 of the same vegetation type available as a continuous block, on the same rock formation, without a major altitudinal gradient, and in which all the typical elements of maquis vegetation were widely distributed. We also wished to avoid areas that had been cultivated in the past or that had been subject to recent fires. Previous work indicated that the area just north and west of the peak Stroumboulas, about 12 km due east of the centre of Irakleion, met these conditions. A visit in 1995 had yielded 18 species in 1 ha, above average for Cretan maquis and phrygana generally (Cameron, Mylonas & Vardinoyannis, 2000). Accordingly, we chose a site in this area bounded by a road to the north and by cultivations to the south. The site is relatively flat, about 490 m above sea level, but is very rugged, with many small limestone crags, much loose rock, and small dips and depressions. While subject to grazing and, no doubt, periodical burning, it is clearly uncultivable. The vegetation is typical of wetter Cretan maquis, dominated by Quercus coccifera and Calicotome villosa, with Ceratonia siliqua and Pistacia lentiscus also abundant. There are areas of bare rock, and patches of shrubs in which Phlomis lanata, Coridothymus capitata and Euphorbia acanthothamnus are common. During the course of sampling, an unexpected environmental discontinuity was discovered; rocks in the area including plots C1–C3 (Fig. 1) were naturally impregnated with bitumen; this is characteristic of the Tripolis series limestone. Freshly broken rock gave off a distinctive odour.
Correspondence: R. A. D. Cameron; e-mail:
[email protected]
J. Moll. Stud. (2003) 69: 93–99
METHODS Within the site, fourteen 400 m2 plots were chosen to give geographical coverage, such that each contained all the major © The Malacological Society of London 2003
R. A. D. CAMERON ET AL .
Figure 1. A map of the study site, showing the UTM grid at 100-m intervals in the margins. Note that sample plots are shown to scale.
Kakamega forest, Kenya. Sampling techniques were essentially the same, but plots were 1600 m2 and were distributed within 265 km2 of forest. Abundances were greater and only one species out of 50 was represented by a single specimen. Apparent abundance at our site was far higher. We did not attempt to collect and count all shells found. Some species were so abundant that attempts to collect all of them would have turned searching time into handling time. Furthermore, in this dry, calcareous environment, empty shells may persist intact for many years, and discriminating between ‘fresh’ and ‘long empty’ shells is not always straightforward. In our study 400-m2 plots would each have yielded several hundred shells, had we attempted to collect all those found, and in many cases the numbers would have exceeded a thousand. Only one species, Lindholmiola barbata (Férussac), was represented by a single shell, and it was one of only two species to be found only in one plot out of the 14 searched [the other was the slug Tandonia sowerbyi (Férussac)]. No slugs were found during the May 2001 sampling, but they were found easily in February 2002. Identifications were carried out by M. Mylonas, K. Triantis and K. Vardinoyannis. Nomenclature follows Vardinoyannis (1994). Measurements were made by R. A. D. Cameron on adult shells from the site, or from neighbouring locations. Material collected is held in the Natural History Museum of Crete.
structural elements of the habitat: bare crags and loose rock, gullies and depressions with soil and vegetation, and with scrub cover in the range 25–75% (Fig. 1). Two people searched each plot for 1 h. Two person-hours per plot was also used by Schilthuizen & Rutjes (2001); de Winter & Gittenberger (1998) applied a similar effort, but with some minor variations. As in the rainforest studies, 4 l of litter and topsoil were removed from each plot, taken from patches likely to yield small species. These samples were sieved and searched in the laboratory, discarding material passing through a 0.5-mm mesh. Although Schiltuizen & Rutjes (2001) visited each plot only once, and at the same time of year for all, de Winter & Gittenberger (1998) sampled several plots in two successive years and at different seasons. This added to recorded species richness. In this study, each plot was visited for the first time on 9 or 10 May 2001, and six of them were revisited and resampled in the same way on 15 or 16 February 2002. In addition to following the protocols used in the rainforest studies, our sampling effort, in terms of searching time and volume of litter removed, fits the recommendations of Menez (2001), who conducted a detailed study of sampling efficiency in Mediterranean environments not very different from the one studied here. In both rainforest studies, the authors counted the number of specimens found in each 400-m2 plot, combining the results of searching in the field and litter sorting. Their data cannot be used to estimate absolute densities, but they were able to use them to indicate relative abundance, and to estimate the numbers of species present in their sites but not discovered. Because the numbers of specimens found were low relative to the number of species, many species were represented by single specimens, not just in individual plots, but also for the whole survey. In Sabah, 23 out of 52 species from the plots were so recorded, while for Cameroon, with rather greater abundance, the figures are 9 out of 80 in 1995 and 13 out of 88 in 1996. Because low abundance affects interpretation, we also make comparisons with data from Tattersfield’s (1996) study in
RESULTS Table 1 shows the occurrence of species in each sample plot and, overall, for 2001. A total of 21 species were found. A third of these were found in all 14 plots, and 17 (81%) were found in more than half the plots. The richest plot (A5) was only one species short of the total for the site. Inspection of Table 1 and Figure 1 indicates that the plots on bituminous rock (C1–C3) are species poor, and that the two poorest, C2 and C3, are partially isolated from others by 94
LAND-SNAIL DIVERSITY IN MAQUIS 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 20 1 2 3 4 5 6 7 8 9 30 1 2 3 4 5 6 7 8 9 40 1 2 3 4 5 6 7 8 9 50 1 2 3 4 5 6 7 8 9 60 1 2 3 4 5 6 7 8
Table 1. The occurrence of species in each sample plot, and overall, in 2001. The final column shows the species found in 1 ha within the site in 1995 (see text). SITES SPECIES
B1
B2
B3
Orculella critica
X
X
X
Orculella sp.
X
Pleurodiscus sudensis
X
Granopupa granum
X
Rupestrella rhodia
X
Rupestrella phillippii
X
Mastus olivaceus
X
X
X
X
X
X
Albinaria spratti
X
X
X
X
X
X
Albinaria hippolyti
X
X
X
X
X
Truncatellina rothi
X
X
D1
D2
D3
X
X
X
X
X
X
X
X
X
A1
A2
A3
A4
X
3
21
X
X
X
X
X
10
71
X
X
X
X
X
X
X
X
X X
Cecilioides acicula
X
X
A5
X
X
X
X
C1
X
C2
C3
Total
%
1995 X
X
X
9
64
X
X
X
14
100
X
5
36
X
X
X
X
X
X
9
64
X
X
X
X
X
X
X
7
50
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X X
X
14
100
X
11
79
X
X
14
100
X
X
7
50
X X
Eopolita protensa
X
X
X
X
X
X
X
X
X
X
X
X
X
X
14
100
Vitrea contracta
X
X
X
X
X
X
X
X
X
X
X
X
X
X
14
100
X
X
X
X
X
X
5
36
X
Vitrea clessini Helicopsis bathytera
X
X
X
X
X
X
X
X
X
X
10
71
X
Trochoidea mesostena
X
X
X
X
X
X
X
X
X
X
X
X
X
X
14
100
X
Metafruticicola noverca
X
X
X
X
X
X
X
X
X
X
X
X
X
X
14
100
X
1
7
Lindholmiola barbata
X
Eobania vermiculata
X
X
X
Helix aperta
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
Helix aspersa
X
X
X
X
X
X
X
X
X
X
X
X
X
TOTAL
17
14
15
18
13
18
13
16
17
17
20
13
10
9
12
86
10
71
13
93
21
X X 18
study site, but in the latter, faunas of smaller plots are on average 2.5 times as rich as those of Sabah. By contrast, the Kenyan data indicates relative uniformity, given the greater distances between plots. Even there, however, there is a greater contrast between individual plots and the overall area than in our data. Table 4 presents estimates of z, the slope of the logarithmic species/area relationship:
cultivations. No other geographical or environmental relationships are apparent. Table 1 also shows the results of searching 1 ha on the site in 1995, using a similar amount of effort, and employing the same techniques. The 1-ha site was near plots A1 and A4, not on bituminous rock. Of the three species found in 2001, but not in 1995, two, Granopupa granum (Draparnaud) and L. barbata, were patchily distributed, but the other, H. aperta Born, was widespread, but not abundant. Table 2 shows the effects of resampling six plots in February 2002. Six species were added to the site total; four of these are slugs (Deroceras and Tandonia species) and one, Daudebardia rufa, is a semi-slug. All would normally be underground in May. In the richest plot, A5, the 2002 snail fauna is identical to that of 2001. In the others, there are small variations between years, but no evidence of a significant overall shift in the diversity of snails. Out of 126 presence/absence cells for snails (6 sites 21 species), 23 (18%) show changes between years. Pleurodiscus sudensis (Pfeiffer), found in all six sites in 2001, occurred in only three in 2002, while Monacha syriaca (Ehrenberg) was recorded five times in 2002, but not at all (anywhere on the site) in 2001. It was recorded in 1995 on the edge of cultivations just outside the 1-ha sample site. Apart from adding slugs, the effect of combining samples from both years is a marginal increase in snail species richness and a reduction in the variation between sample plots. Table 3 presents analyses of these results, together with comparative data from the rainforest studies. Whittaker’s index of diversity, I, is given by S/, where S is the total number of species in the fauna considered, and is the mean number of species per plot (Whittaker, 1975; Cameron, 1995). By comparison with the 1 km2 rainforest series, even when the latter are subdivided on the basis of ecological distinctions between plots (de Winter & Gittenberger, 1998), faunas from plots in the maquis are very uniform and similar, and are even more so if plots on bituminous rock are excluded. Even allowing for considerable differential sampling error, the contrast with Sabah is extreme; the Sabah site probably holds at least three times as many species as our
log10 number of species a z log10 area for various comparisons within Crete, and within each of the rainforest sites. They are all spot estimates, based on species counts in only two sizes of area. Values of the intercept are not informative here. In Crete, data for areas larger than the study site come from maps in Vardinoyannis (1994) and refer only to snails. As with other general survey data (Cameron, 2001), all areas on Crete were not subject to equal sampling intensity, and the 100 km2 area including the study site may be under-recorded. Nevertheless, there is a consistent pattern: at small scales, values of z are very low indeed, but at larger scales they steepen rapidly. No data is available to extend the 1 km2 rainforest sequences beyond the study sites. Within the sites, values of z are consistently higher than those of the maquis site. While this merely presents the data in Table 3 in a different way, it facilitates comparison with other studies using different areas (see discussion below). Table 5 shows the distribution of some shell size and shape parameters for the fauna of the study site, with comparative data from the Cameroon site and other rich sites given by Emberton (1995), and Solem & Climo (1985). No data are available for the Sabah site. The Cretan fauna resembles that from Kentucky, the only other temperate zone fauna listed, in the general distribution of size parameters. It lacks the large numbers of small, litterdwelling species found in the moist tropical and subtropical sites. It shares with Kentucky and the Cameroon rainforest a pronounced bimodality in shell shape, which is not found in the 95
R. A. D. CAMERON ET AL . consider the effects of sampling error. Gotelli & Colwell (2001) recommend the use of accumulation or rarefaction curves to get comparable estimates of species richness, and the tropical rainforest studies referred to provide estimates of actual species richness based on such methods. As individuals were not counted in our study, rarefaction techniques were not possible. Individuals were, however, very numerous, and the high degree of consistency between plots within years, and of the same plots between years and seasons (at
New Zealand or Madagascan sites; bimodality is, indeed, at its most extreme in the Cretan site.
DISCUSSION Sampling error and the comparability of data Given the great differences in the patterns of local diversity shown in some of the comparisons we make, it is pertinent to
Table 2. The occurrence of species in the six plots sampled both in May 2001 and in February 2002. X occurrence in 2001; Y occurrence in 2002. Slugs and the semi-slug named in bold.
SPECIES
Total
Total
Total
2001
2002
both
B1
B3
A2
A3
A4
A5 XY
1
1
1
XY
XY
XY
XY
XY
XY
6
6
6
Orculella sp.
XY
XY
XY
XY
X
XY
5
5
6
Pleurodiscus sudensis
XY
X
X
XY
X
XY
6
3
6
Truncatellina rothi Orculella critica
Granopupa granum
X
XY
XY
XY
X
XY
4
4
6
Rupestrella rhodia
XY
X
XY
XY
XY
XY
6
5
6
Rupestrella phillippii
XY
XY
XY
XY
XY
XY
5
6
6
Mastus olivaceus
XY
XY
XY
XY
XY
XY
6
6
6
Albinaria spratti
XY
XY
XY
XY
XY
XY
6
6
6
Albinaria hippolyti
XY
XY
XY
XY
XY
XY
6
6
6
X
XY
XY
XY
XY
3
4
5
Cecilioides acicula Eopolita protensa
XY
XY
XY
XY
XY
XY
6
6
6
Vitrea contracta
XY
XY
XY
XY
X
XY
6
5
6
X
XY
2
1
2
0
2
2
Vitrea clessini Daudebardia rufa
XY
Tandonia cretica
XY
XY
XY
Tandonia sowerbyi
XY
Deroceras rethimnonensis
XY
XY
XY
XY
Deroceras lasithionensis
XY
XY
XY
XY
Helicopsis bathytera
XY
XY
XY
XY
Trochoidea mesostena
XY
XY
XY
XY
Metafruticicola noverca
XY
XY
XY
Monacha syriaca
XY
XY
Eobania vermiculata
XY
XY
XY
0
2
2
0
1
1
XY
0
6
6
XY
0
5
5
XY
XY
6
6
6
XY
XY
6
6
6
XY
XY
XY
6
6
6
XY
XY
XY
0
5
5
XY
XY
XY
6
6
6
XY
Helix aperta
XY
XY
XY
XY
XY
XY
4
6
6
Helix aspersa
XY
XY
XY
XY
XY
XY
6
6
6
TOTAL 2001, snails
17
15
16
17
17
20
20
TOTAL 2002, snails
17
16
18
19
15
20
TOTAL SNAILS
18
19
19
19
20
20
TOTAL OVERALL
21
22
21
21
23
23
21 21 26
26
Table 3. Data on species richness and diversity for the study site, and for the rainforest sites in Sabah (Schilthuizen & Rutjes, 2001), in Cameroon (de Winter & Gittenberger, 1998), and in Kenya (Tattersfield, 1996). Note that at the Sabah site, nine more species were found outside the plots. For Cameroon, F V (forest floor and vegetation) gives data for all plots in which both were searched, and B (boulders) represents searches restricted to large boulders and the vegetation on them. No counts of specimens were made in the maquis, but values would be in the high hundreds (see text). Maquis (all)
Maquis 2001
Maquis
Maquis
2001
Ex C
2002
2001+2
No. of 400 m2 plots
14
11
6
6
36
24
7
27
Total species, S
21
20
26
26
52
95
54
50 23.4
Mean no. of species/plot, Standard error Whittaker’s I
Sabah (all)
Cameroon
Cameroon
Kenya
(F+V)
(B)
(1600 m2)
15.0
16.2
20.2
21.8
6.1
26.6
23.6
0.84
0.67
0.70
0.4
0.58
1.86
2.34
0.9
1.40
1.23
1.29
1.19
8.52
3.57
2.29
2.10
Range, species per plot
9-20
13-20
18-23
21-23
2-14
12-45
15-31
13-33
Mean % of plots occupied per species
71
81
78
84
12
27
43
48
Mean no. of specimens per site
*
*
*
*
Standard error
96
10.8
78.9
78.3
137
1.05
9.8
13.4
n.a.
LAND-SNAIL DIVERSITY IN MAQUIS 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 20 1 2 3 4 5 6 7 8 9 30 1 2 3 4 5 6 7 8 9 40 1 2 3 4 5 6 7 8 9 50 1 2 3 4 5 6 7 8 9 60 1 2 3 4 5 6 7 8
In both the rainforest studies at the 1-km2 scale, however, using the same protocols, this is not the case. Numbers of specimens retrieved were low overall (374 from 36 plots in Sabah, 2654 from 36 plots in Cameroon), and substantial proportions of the species found were represented by single specimens. Both de Winter & Gittenberger (1998) and Schilthuizen & Rutjes (2001) were aware of the effects of sampling error in their data, and they assess its effects on their site inventories. In Cameroon, estimates indicate that the true site fauna contains 103–108 species, as against 80 found in 1995, 88 in 1996 and 97 overall. A single-season study of 16–20 plots may thus underestimate species richness by around 30%. The discrepancy is worse in Sabah, where a 36-plot survey in one season may underestimate by 40% or more (52 species found in plots against estimates in the range 81–88). In the maquis, a winter sampling of only six plots yielded 26 of the 27 species known from the site. These discrepancies will be worse at the level of individual plots, especially in Sabah, where it is clear that the protocols are inadequate for producing inventories. Given a mean of only 10.8 specimens per plot, and the recorded total of 52 species, the minimum value of Whittaker’s index obtainable is 4.8, when each species present in a plot is represented by a single specimen. In Cameroon, however, there is evidence that sampling error does not explain all the greater heterogeneity between plots relative to that in maquis. de Winter & Gittenberger (1998) conclude that the richest plots at their site might contain up to 75% of the site fauna (compared to a recorded maximum of 46%). This is still below the proportion found in the richest maquis plot (85%), and, indeed, the average maquis plot contains about 75% of the site fauna. Data on the numbers of specimens and species found in the Cameroon plots shows that there is a positive relationship between the two until about 80–100 specimens are found. Above this abundance, plot species richness varies between 22–45 species independently of sample size. Where plots were sampled in successive years, mean aggregate species richness per plot was 42.2 and the maximum 51. Among samples made on boulders, the heterogeneity of samples is much less than among the standard plots, even though the mean number of specimens per sample is the same (78.3 on boulders, 78.8 in plots). Greater homogeneity is achieved when sampling is restricted to a recognizable and permanent ecological feature. The greater heterogeneity in standard plots reflects real differences and the true mean species richness of plots is likely to be well below 75% of the site fauna. In this context, it is worth noting the much greater homogeneity in Tattersfield’s (1996) data from Kenya, despite the much larger area over which sample plots were distributed. Litter samples were the same as in the other studies, but sample plots were four times as large. An average of 137 specimens per plot were found in plots from natural forest stands.
least for snails) both indicate that single samples can be expected to retrieve a very high proportion of the actual fauna. The protocols adopted follow a tradition of successful inventory work in temperate forests (Wäreborn, 1969) and Menez (2001) has shown that they yield reliable estimates of species richness in Mediterranean habitats similar to those studied here. When two samples from different seasons, but from the same plot, are combined there is a striking uniformity of fauna between plots in the maquis; 17 out of 21 species of snail are recorded in all six plots and a further two in five out of the six. Table 4. Spot estimates of z, the slope of the logarithmic species/area curve, for various pairs of areas in and beyond the study area in Crete, based on snails only, and within the rainforest study sites. Mean, Max and Min refer to recorded species richness. For details of the 9-ha site in Cameroon, see de Winter & Gittenberger (1998). Comparison
z
From
To 2
2
Crete
Cameroon
Sabah
Kenya
Mean 400 m
1 km
0.04
0.168
0.297
–
Max 400 m2
1 km2
0.006
0.100
0.188
–
Min 400 m2
1 km2
0.108
0.268
0.438
–
1 ha
1 km2
0.03
–
–
–
1 km2
–
0.067
–
–
Mean 1600 m
265 km2
–
–
–
0.063
Max 1600 m2
265 km2
–
–
–
0.036
Min 1600 m2
265 km2
–
–
–
0.113
1 km2
100 km2
0.04
–
–
–
100 km2
400 km2
0.27
–
–
–
400 km2
1600 km2
0.27
–
–
–
1600 km2
8700 km2
0.38
–
–
–
9 ha 2
Table 5. Percentages of the snail species recorded from various sites of 1 km2 or less, falling into specified size and shape categories. Semi-slugs are excluded. Width and height refer to maximum shell dimensions, and H/W is the ratio of the two measurements. Data for Kentucky forest and Madagascar rainforest from Emberton (1995), for New Zealand scrub forest from Solem & Climo (1985), and for Cameroon from de Winter & Gittenberger (1998). Note that Emberton (1995) reanalyses the New Zealand data on a smaller, genuinely sympatric data set, but does not give data on all parameters. Proportions do not differ significantly from those given in Solem & Climo (1985). 5 mm
5–10 mm
10–20 mm
20 mm
No. species
Width Crete
50
5
32
13
22
Kentucky
40
20
25
15
42
New Zealand
84
15
1
0
81
Madagascar
66
13
6
15
52
Cameroon
67
18
8
7
86
Scale and species richness
Height Crete
27
41
14
18
22
Kentucky
47
36
10
7
42
New Zealand
91
8
1
0
81
Madagascar
56
17
13
13
52
Cameroon
40
36
8
16
86
0.8
0.8–1.2
1.2–1.6
1.6–2.4
2.4
H/W Crete
45
9
0
14
32
Kentucky
71
2
5
17
5
New Zealand
83
9
4
4
0
Madagascar
33
25
17
15
10
Cameroon
16
21
3
41
16
Cameron (2001) has shown that in land mollusc faunas the relationship of the slope (z) of the logarithmic species/area curve to scale varies with latitude, and with climates past and present. Within the temperate zone, z remains low (0.06–0.12) over a huge range of scales (1 m2 to 100,000 km2 or more), when richest, rather than average sites are considered at the smallest scales. In Mediterranean, and in non-arid subtropical and tropical regions, however, z becomes much steeper at larger scales (from, perhaps, 1000 km2 upwards), reflecting both a wider range of habitats and small-scale biogeographical differentiation. There are thus global variations in the relationship between local and regional diversities (Srivastava, 1999). Local, or site diversities may show much less variation across regions than 97
R. A. D. CAMERON ET AL . diversity at larger scales. At such local levels, variation in species richness with area should reflect ecological rather than biogeographic factors. As de Winter & Gittenberger (1998) point out, it is reasonable to regard species found within a 1 km2 site as sympatric; within that area, richness at smaller scales should reflect the grain of structural and nutritional resources. Sampling at a range of scales up to 1 km2 should enable us to make ecological comparisons across regions and habitats, and to refine the rather awkward distinction between sympatric and mosaic diversity made by Solem (1984). The longer-term objective is to determine the assembly rules for molluscan faunas. Although limited to two scales only, our results for Cretan maquis show a very high level of homogeneity between plots; the grain of the environment is such that most, if not all, significant elements of the habitat can be found in a single plot and, when they are present, they are occupied. We have no doubt, though it requires verification, that at much smaller scales this would cease to apply, and there would be much greater variation in richness and composition between plots. At scales of 1 m2 or less, temperate forest faunas show considerable heterogeneity, although the richest plots contain most of the diversity found in much larger areas (Nekola & Smith, 1999). Once beyond the bounds of the study site, z for areas of Crete steepens rapidly; this Mediterranean contrast with wetter temperate zone faunas is typical (R. A. D. Cameron, unpublished). At site level, diversity is rather more than half that found in rich temperate forests, but at larger scales patterns diverge; Crete, at about 8700 km2, has a richer fauna than that of Great Britain at about 241,000 km2. As yet, we lack comparative data on the degree of homogeneity within temperate forest sites over the same range of area (R. A. D. Cameron and B. M. Pokryszko, unpublished), but even at larger scales the value of Whittaker’s I remains modest (Cameron, 1995). At 1 km2, the maquis fauna contains only 20–30% of the number of species found in the rainforests sampled at this scale. The greater homogeneity of the maquis fauna, however, means that differences at 400 m2 are much less marked. Even allowing for gross sampling error, it is remarkable that, at this scale, the maquis contains on average 2.5–3.0 times as many species as the Sabah rainforest. Even in comparison with Cameroon or Kenya, where sampling error is less acute, maquis holds about 70% of the number of species found in tropical forest. The obvious explanation of these differential scale effects is that the grain of relevant environmental variables is coarser in the tropical forests. Against this, however, neither de Winter & Gittenberger (1998) nor Schilthuizen & Rutjes (2001) found clear associations between species richness and composition and identifiable variation in the environment. In both studies, apparent densities were far lower than in the maquis. While this creates sampling problems, discussed above, it also raises the possibility that many species in the rainforest may have a fluctuating metapopulation structure (Hanski, 1999), perhaps reflecting the transitory nature of molluscan microhabitats; substantial patches of newly created suitable habitat may remain unoccupied for some time following a localized extinction. The existence of such transient and shifting sub-populations in theory permits the coexistence of species that might otherwise compete (see below). At the practical level, it reduces the clarity of individual associations of species with components of the environment. Work on a greater range of environments, and especially on tropical sites with exposed limestone, will clarify these problems, and reveal the extent to which there are more niches available in lower latitudes (M. Schilthuizen, unpublished).
is a little stronger for rare instances involving particular speciespairs (Boycott, 1934). A telling point against lies in the numerous cases in which very closely related and morphologically similar species co-exist. Even when morphological features, such as shell shape, are analysed in relation to way of life, rather than to competition, the evidence suggests that phylogenetic constraints as well as selection influence shape distribution within faunas (Emberton, 1995). The markedly bimodal distribution of shell shape in this maquis fauna is typical of many others (Cain, 1977), and both very tall and very flattened forms may be suited to rocky environments (Heller, 1987). However, a similar pattern in the Cameroon rainforest may reflect phylogenetic constraint on the dominant Streptaxidae, which are all high spired. de Winter & Gittenberger (1998) could find no correlation between shell shape and observed way of life, unlike Heller (1987) for the fauna of Israel, mostly from habitats not dissimilar to maquis. The only hint we can find of interactive constraints on species richness lies in the ratio of higher taxa to species and this is, inevitably, dependent on the assumption that equivalent criteria are used in defining all taxa at that level. In the maquis fauna, 27 species are distributed among 11 or 12 families, the latter figure applying if the old Helicidae (Helicoidea) are split into smaller families. In Cameroon, 97 species are distributed among 12 families. In Sabah, 61 species are distributed among 14 families. Differences in familial diversity are very slight compared to those at species level. A significant number of genera in both rainforest sites are represented by three or more species, though many of these are not fully described. Thus, although there are some congeneric species pairs in the maquis fauna, the species living there are, on average, more distantly related and morphologically distinct than those in the rainforests. They occur at higher densities, and in apparently more continuous populations, creating more opportunities for competitive interaction. Data from many more environments, with a wide range of densities and species richness, are required to detect any consistent regularities of this kind.
ACKNOWLEDGEMENTS Steven Roberts and Maria Panagiotidou helped us with the sampling. Peter Tattersfield discussed sampling error and methods with us. Menno Schilthuizen made valuable comments on the manuscript, and told us of work in progress.
REFERENCES BOYCOTT, A.E. 1934. The habitats of land Mollusca in Britain. Journal of Ecology, 22: 1–38. CAIN, A.J. 1977. Variation in the spire index of some coiled gastropod shells, and its evolutionary significance. Philosophical Transactions of the Royal Society of London, Ser. B, 277: 377–428. CAMERON, R.A.D. 1995. Patterns of diversity in land-snails: the effects of environmental history. In: Biodiversity and Conservation of the Mollusca (A. C. van Bruggen, S. M. Wells & T. C. M. Kemperman, eds), 187–204. Backhuys, Leiden. CAMERON, R.A.D. 2001. Some species/area relationships in the British land mollusc fauna and their implications. Journal of Conchology, 37: 337–348. CAMERON, R.A.D., MYLONAS, M. & VARDINOYANNIS, K. 2000. Local and regional diversity in some Aegean land-snail faunas. Journal of Molluscan Studies, 66: 131–142. DE WINTER, A.J. & GITTENBERGER, E. 1998. The land-snail fauna of a square kilometer patch of rainforest in southwestern Cameroon: high species richness, low abundance and seasonal fluctuations. Malacologia, 40: 231–250. EMBERTON, K.C. 1995. Land-snail community morphologies of the highest-diversity sites of Madagascar, North America, and New
Assembly structure, shell morphology and higher taxa In general, evidence that land mollusc faunas are structured by competition is very weak (Solem, 1984), although the evidence 98
LAND-SNAIL DIVERSITY IN MAQUIS 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 20 1 2 3 4 5 6 7 8 9 30 1 2 3 4 5 6 7 8 9 40 1 2 3 4 5 6 7 8 9 50 1 2 3 4 5 6 7 8 9 60 1 2 3 4 5 6 7 8
Zealand, with recommended alternatives to height-diameter plots. Malacologia, 36: 43–66. GOTELLI, N.J. & COLWELL, R.K. 2001. Quantifying biodiversity: procedures and pitfalls in the measurement and comparison of species richness. Ecology Letters, 4: 379–391. HANSKI, I. 1999. Metapopulation ecology. Oxford University Press, Oxford. HELLER, J. 1987. Shell shape and land-snail habitat in Mediterranean and desert fauna. Biological Journal of the Linnean Society, 31: 257–272. MENEZ, A. 2001. Assessment of land-snail sampling efficacy in three Mediterranean habitat types. Journal of Conchology, 37: 171–176. NEKOLA, J.C. & SMITH, T.M. 1999. Terrestrial gastropod richness patterns in Wisconsin carbonate cliff communities. Malacologia, 41: 253–269. RACKHAM, O. & MOODY, J. 1996. The making of the Cretan landscape. Manchester, Manchester University Press. SCHILTHUIZEN, M. & RUTJES, H. A. 2001. Land-snail diversity in a square kilometre of tropical rainforest in Sabah, Malaysian Borneo. Journal of Molluscan Studies, 67: 417–423.
SOLEM, A. 1984. A world model of land-snail diversity and abundance. In: World-wide snails (A. Solem & A.C. van Bruggen, eds), 6–22. E.J. Brill/W. Backhuys, Leiden. SOLEM, A. & CLIMO, F.M. 1985. Structure and habitat correlations of sympatric New Zealand land-snail species. Malacologia, 26: 1–30. SRIVASTAVA, D.S. 1999. Using local-regional richness plots to test for species saturation: pitfalls and potentials. Journal of Animal Ecology, 68: 1–16. TATTERSFIELD, P. 1996. Local patterns of land-snail diversity in a Kenyan rain forest. Malacologia, 38: 161–180. VARDINOYANNIS, K. 1994. Biogeography of land-snails in the south Aegean island archipelago. PhD Thesis, University of Athens (In Greek). WÄREBORN, L. 1969. Land molluscs and their environments in an oligotrophic area in southern Sweden. Oikos, 20: 461–479. WHITTAKER, R.H. 1975. Evolution of species diversities in land communities. Evolutionary Biology, 10: 1–68.
99
