The mid-latitude biodiversity ridge in terrestrial cave fauna

Authors

  • David C. Culver,

  • Louis Deharveng,

  • Anne Bedos,

  • Julian J. Lewis,

  • Molly Madden,

  • James R. Reddell,

  • Boris Sket,

  • Peter Trontelj,

  • Denis White


D. C. Culver (dculver@american.edu) and M. Madden, Dept of Biology, American Univ., 4400 Massachusetts Ave., NW, Washington, DC 20016, USA. – L. Deharveng and A. Bedos, UMR 5202 du CNRS, Museum National d'Histoire Naturelle, 45 Rue de Buffon, FR-75005, Paris, France. – J. J. Lewis, Lewis and Associates, 17903 State Road 60, Borden, IN 47106-8608, USA. – J. R. Reddell, Texas Memorial Museum, Univ. of Texas, 2400 Trinity Ave., Austin, TX 78705, USA. – B. Sket and P. Trontelj, Dept of Biology, Univ. of Ljubljana, Večna pot 111, p.p. 2995, SI-1001 Ljubljana, Slovenia. – D. White, Environmental Protection Agency, 200 SW 35th Street, Corvallis, OR 97333, USA.

Abstract

The world's obligate cave-dwelling fauna holds considerable promise for biogeographic analysis because it represents a large number of independent evolutionary experiments in isolation in caves and adaptation to subterranean life. We focus on seven north temperate regions of at least 2000 km2, utilizing more than 4300 records of obligate cave-dwelling terrestrial invertebrates. In North America, highest diversity was found in northeast Alabama while in Europe highest diversity was found in Ariège, France, and in southeast Slovenia. Based on these regions as well as more qualitative data from 16 other regions, we hypothesize that a ridge (ca 42°–46° in Europe and 34° in North America) of high biodiversity occurs in temperate areas of high productivity and cave density. This may reflect a strong dependence of cave communities on long term surface productivity (as reflected in actual evapotranspiration), because the subterranean fauna relies almost entirely on resources produced outside caves. This dependence may explain the unique biodiversity pattern of terrestrial cave invertebrates.

Obligate cave-dwelling species are known throughout the world, but in spite of more than a century of taxonomic and biogeographic study, large scale patterns of diversity have remained obscure. There are several reasons for this. First, there is the difficulty of sampling in caves. Second, the highly restricted ranges of obligate cave-dwellers means that any one cave is likely to have only a small fraction of the cave fauna of a region. Third, the sampling units, caves, are themselves highly heterogeneous. Caves range from large systems of tens of kilometers such as Postojna Cave in Slovenia and Mammoth Cave in Kentucky, USA, to small crawlways only a few meters in length.

In spite of the sampling difficulties, the obligate cave fauna is an attractive object of study. Caves and other subterranean habitats, compared to most other habitats, present formidable barriers to successful colonization, and to movement between caves once colonization has been successful. Obligate cave dwellers are often relatively easy to distinguish by their loss of eyes and pigment as well as hypertrophy of extra-optic sensory structures (Culver 1982). They come from a wide variety of taxonomic groups – Arachnida, Crustacea, and Insecta among others – and share an evolutionary history of isolation and adaptation in the food-poor, aphotic environment of caves. Patterns in the distribution of the cave fauna are the consequence of repeated, independent invasions, isolation, and adaptation to the cave environment and as such, they should be amenable to rigorous quantitative analysis (Christman and Culver 2001). While movement between caves is possible because caves can be connected by passages much too small for human exploration, the scale of potential dispersal is rarely >100 km and often much less. At scales of 100 km and greater, patterns must be vicariant (in the sense of separate isolation in caves, not dispersalist in the sense of resulting from subterranean movement. Thus, the biogeographic patterns of cave fauna should be clearer than is the case for most groups. Caves also are among the very few habitats on earth where the entire invertebrate community is amenable to analysis, because the number of species they host is very low compared to most non-subterranean habitats. Patterns of biodiversity may therefore embrace all functional groups, allowing less biased interpretation. Nevertheless, because of the paucity of quantitative information about cave faunas, they do not have a place in modern treatments of biogeography. For example, in the recent synthesis of new directions in biogeography (Lomolino and Heaney 2004), discussion of cave faunas is totally absent.

In addition, the cave fauna itself is of considerable interest. The geographic rarity of cave species makes them especially vulnerable to human activities (Elliott 2005). A number of obligately cave-dwelling species are on the IUCN red list, the U.S. Endangered Species list, and several species are protected in the European Union under the Habitat Directive (Tercafs 2001).

Most quantitative information about large-scale patterns of cave biodiversity comes from two sources. One is lists of the obligate cave fauna for individual countries, many of which are in the multi-authored volume 3 of Encyclopaedia Biospeologica (Juberthie and Decu 2001). While of limited value because of the widely different country areas (and widely different areas of cave regions) and different sampling intensities, these lists do provide an upper limit on known species diversity. The other source lists individual caves especially rich in species. Culver and Sket (2000) compiled a list of all caves known to have twenty or more obligate cave-dwelling species. They found that the 20 caves that met the criteria were almost all in the north temperate zone and that they were concentrated in the Dinaric and Pyrenees Mountains of Europe. In a similar analysis of tropical caves, Deharveng (2005) found that caves with ten or more obligate cave-dwelling species were concentrated in the Indo-Pacific.

We present here the first analysis of the entire obligate terrestrial (i.e. troglobiotic) cave fauna for different regions, scaled so that equal areas can be compared. We focus on the pattern in north temperate caves both because the data are most complete for this region but also because patterns of cave fauna within the temperate zone have largely been ignored, except for the observation that the cave fauna of glaciated areas is depauperate (e.g. Peck and Christiansen 1990).

We analyze seven regions in North America and Europe ranging in size between 2000 and 6300 km2 and supplement this information with faunal lists from another 16 cave regions. Rather than using individual cave faunal surveys as the unit of analysis, we aggregate the data into hexagons of 100 km2 in order to analyze patterns based on complete lists for equal areas. This also has the virtue of decreasing sampling heterogeneity. We supplement this with faunal lists from 16 other regions.

We focus on the following question: what are the patterns of overall species richness in the different regions and how can we account for differences among the regions studied and within the north temperate zone in general?

We limited our analysis to the terrestrial (troglobiotic) fauna because a major component of the aquatic cave fauna, the microcrustacean community, especially copepods and especially those occurring in the upper most layer or “epikarst” of cave-bearing regions, is almost completely unstudied in North America (Pipan and Culver 2005).

Methods and materials

Seven regions were chosen for study. Initially, each region was largely defined politically: 1) the French Pyrenean départements of Ariège (west of Ariège river) and Haute Garonne (east of Garonne river) (called Ariège in the rest of the paper); 2) the French departément of Ardèche in the eastern part of Massif Central; 3) the Slovenian portion of the Dinaric Mountains; 4) Monroe, Greenbrier, and Pocahontas Counties in the Appalachian Mountains of West Virginia; 5) Jackson, Madison, and Marshall Counties in the Interior Low Plateau of Alabama; 6) Travis and Williamson Counties in the Balcones Escarpment of Texas; 7)Crawford, Harrison, Lawrence, Monroe, Orange, Owen, and Washington Counties in the Pennyroyal Plateau of Indiana. Within these political boundaries, blocks of more or less continuous limestone (separated by no more than 15 km of non cave-bearing rock) were used. Operationally, this resulted in the removal of isolated regions only in Ariège.

All seven of these sites are areas of high subterranean diversity on a continental scale. The Dinarides and Ariège contain “hotspot” caves (Culver and Sket 2000), and Ardèche is a well studied cave region on a plateau, contrasting with the more mountainous Ariège site. The four North American sites are in the three most diverse cave regions – the Interior Low Plateaus, the Appalachians and the Balcones Escarpment (Culver et al. 2003).

The area of the cave-bearing regions ranged from 2000 (Texas) to 6300 km2 (Slovenia); the number of caves in each region sampled was >120 and the number of records of troglobionts was >350. (We use the word troglobiont rather troglobite to indicate obligate cave-dwelling terrestrial species on semantic grounds [see Sket 2004 and the Oxford English Dictionary].) In fact, we know of no other datasets from other regions on earth that meet these criteria.

For each region, a list of records of described troglobiotic species and subspecies was developed by regional experts (France by Bedos and Deharveng, Slovenia by Sket and Trontelj, Indiana by Lewis, Texas by Reddell, and Alabama and West Virginia by Culver). The Ardèche data was compiled mainly from the work of Balazuc (1986) and the Alabama dataset largely came from the extensive published records of Peck (1989, 1995). Although there are undescribed species known from all seven regions, we used only records of described species. The propensity to describe geographic variants as subspecies varies from country to country so we did most of our analysis with species rather than subspecies. For each cave with a troglobiont, the latitude and longitude were recorded.

Hexagons of 100 km2 were generated using a process analogous to recursive diamond subdivisions of an icosahedron (White 2000). Such hexagons offer several advantages in spatial statistical analysis, including minimum shape and area distortion and an equal number of equidistant neighbors for all non-edge hexagons (White et al. 1992). Each species record was then associated with a particular hexagon using ArcMapTM software (Environmental System Research Inst., Redlands, CA, USA).

Species accumulation curves were generated using the incidence model of species accumulation curves proposed by Colwell et al. (2004) in the software package EstimateS ver. 7 (Colwell 2004). This allows direct comparison of accumulation curves along with their confidence intervals. The total species estimator, SChao2 is

image

where Sobs is the observed number of species, Q1 is the number of species occurring in a single hexagon, and Q2 is the number of species occurring in two hexagons (Colwell and Coddington 1994).

Additional information on total numbers of troglobionts was obtained for six other karst regions in North America (Culver et al. 2003) as well as England (Proudlove 2001), Switzerland (Strinati 1966), Slovakia (Juberthie et al. 2001), Czech Republic (Bosak and Vasátko 2001), Belgium (Tercafs 1994), Quercy Causses in France (Lebreton 1986), Portugal (Da Gama and Afonso 1994), Bassin Parisien in France (Juberthie and Ginet 1994), Sardinia (Guéorguiev 1977), Spanish Cantabria and Valencia (Bellés 1987), the Italian Appenines (Latella pers. comm.), and Bosnia and Hercegovina (Sket et al. 2004).

Results

Regional characteristics

A total of 4365 records of troglobiotic species in 1664 caves were included in the analysis (Table 1). The average number of records per cave, a measure of sampling intensity per cave, was rather constant in the seven areas, ranging from a low of 2.23 in Texas to a high of 3.54 in West Virginia with a Coefficient of Variation of 15.5%. On the other hand, the number of records per hexagon, a measure of thoroughness of geographic coverage was two to three times higher in Ariège and Texas than the other regions, with a Coefficient of Variation of 46.1%.

Table 1.  Basic biological and geological characteristics of the seven regions.
 SloveniaArdècheAriègeWest VirginiaAlabamaIndianaTexas
  1. 1A record is the occurrence of a troglobiont in a cave.

  2. 2 Each region was classified for Pleistocene history by whether it was close (50 km) to mountain glaciers (MG) and whether or not it was close (100 km) to the continental ice sheet (CI).

Number of caves sampled359142327109174229324
Number of records1987360803367501623724
Number of species122409529864642
Pleistocene history2MGMGMG  CI 
Fragmentation       
of karstLowLowHighHighLowLowHigh
Mean latitude46°N45°N43°N38°N34°N39°N30°N
Approximate no. of caves4000550180090023002000900
No. of 100 km2 hexagons56223027424420
Mean no. of caves/hexagon71.4325.0060.0033.3354.7645.4545.00
Mean no. of records/cave2.752.542.463.372.882.722.23
Mean no. of records/hexagon17.6316.3626.7713.5911.9314.1636.20

All of the regions have a continuous terrestrial history dating at least from the Miocene with the opportunity for cave development dating from that time as well. Because of erosional processes, individual caves are not that old, rarely being more than a few million years old (White 1988). However, cave formation is a more or less continuous process with new voids forming below the old, eroded rock layers (Gabrovšek 2002), so that caves may have been present in a region for longer than the lifetime of an individual cave. The regions do differ with respect to their proximity to Pleistocene continental and montane glaciers (Table 1). Except for Indiana, North American regions were not geographically proximate to any glaciers while all three European sites were. Regions also differed in fragmentation of the cave-bearing (karst) areas. Texas, Ariège, and West Virginia were fragmented although in the case of West Virginia it is the least fragmented part of the fragmented karst of the Appalachians (Barr 1967). Finally, Slovenia, Ariège and Alabama were especially cave-rich, with over 50 caves per hexagon (0.5 km−2) (Table 1).

Taxonomic characteristics

The average number of species per genus is quite constant (Table 2), with a Coefficient of Variation of only 20.9%. We believe that this indicates, at the species level, that there are no great differences in taxonomic distinctness used to separate species, i.e. splitters versus lumpers. In fact, the lowest and highest ratios of species per genus were both in North America rather than on the two continents (Table 2). On the other hand, at the subspecific level, there are considerable differences in the number of “excess” subspecies described, ranging from zero per species in Alabama, Texas and West Virginia to 0.89 in Ariège, with a coefficient of variation of 94.6% (Table 2). We think that much of this difference is indeed due to differences in taxonomic practice, with a clear North American bias against naming subspecies (e.g. Wilson and Brown 1953). There are also large differences in the number of monotypic genera found in each region. Two regions – Ariège and Slovenia – were richest in the number of monotypic genera (Table 2).

Table 2.  Summary of species, subspecies, and generic richness for the seven study regions. Monotypic genera listed include subgenera (subgeneric name is given in parentheses).
 AlabamaArdècheAriègeIndianaSloveniaTexasWest Virginia
Number of species864095461224229
Additional subspecies0118575801
Number of genera30214123481619
Species/Genus2.871.902.322.002.562.631.53
Ssp./Spp.0.000.280.890.150.490.000.03
Monotypic generaSpeleobia AlabamocreagrisSpeotrechusPeltonychia(Arbasus) Gesciella Scotoniscus Speonomus(Metaspeonomus) Antrocharis Paraspeonomus Rhagidia (Deharvengiella)SpelobiaParastalita Strasseria Bathysciotes Leptodirus Rectipenis SphaerobathysciaTexamaurops TexoreddelliaSpelobia Horologion

As is characteristic of cave fauna in general (Gibert and Deharveng 2002), it is a highly biased sample of the surface-dwelling invertebrate fauna (Table 3). In the regions studied here, arachnids, beetles, springtails, and millipedes predominate in the caves. However, there are major differences among the regions, and these differences are highly significant (log-likelihood χ2=92.9, DF=18, p<.0001).

Table 3.  Taxonomic distribution of troglobionts of four major groups in the study areas. Row percentages are given in parentheses. Other less speciose groups (e.g. Gastropoda and Diplura) are present in the study areas.
RegionArachnidaDiplopodaCollembolaColeoptera
Slovenia25(24)16(15) 8(8)55(53)
Ardèche10(29) 6(17)13(37) 6(17)
Ariège12(14) 4(5)13(16)54(65)
West Virginia10(38) 3(12) 6(23) 7(27)
Alabama41(50) 7(9) 6(7)28(34)
Indiana16(37) 9(21)11(6) 7(16)
Texas21(62) 2(6) 2(6) 9(26)

Species accumulation curves and species richness estimates

Species accumulation curves based on incidence functions (Colwell et al. 2004) show what appears is a very striking pattern (Fig. 1). Four sites (Ardèche, Indiana, Texas, and West Virginia) have virtually identical species accumulation curves lower than the other three. The other three sites (Alabama, Ariège, and Slovenia) have consistently higher species accumulation curves, although they do not cluster as strongly as the other four. The most appropriate comparison is not the final point on the curve but rather at a point of equal number of sampled hexagons (Fig. 2). Four regions are closely grouped – Ardèche, Indiana, Texas, and West Virginia, with total species richness in 20 hexagons of between 25 and 40. Alabama has a higher number at 49 and Ariège and Slovenia have yet higher numbers at around 70 to 80. All curves except Alabama do share one feature – namely they have begun to level off but have not yet reached an asymptote (Fig. 1). The anomalous non-asymptotic curve for Alabama is largely due to the very high number of single cave endemics (>35) in this area (Christman et al. 2005) which means that as more hexagons are added more species continue to be added.

Figure 1.

Species accumulation curves. Incidence functions of number of troglobionts plotted against the number of 100 km2 hexagons, following Colwell (2004) and Colwell et al. (2004). Standard errors at 20 hexagons are shown in Fig. 2.

Figure 2.

Species richness estimates. Means and standard errors of estimates of species numbers (Mao-Tau estimates) and total species richness (Chao2) for all seven regions for 20 hexagons of 100 km2 area.

The Chao2 estimate (SChao2) of total species richness (including unsampled species) shows two distinct groups (Fig. 2) – a group at the lower end of values (ca 50) of Ardèche, Texas, Indiana, and West Virginia, and a “hotspot” group ranging from 113 (Ariège) to 130 (Slovenia). In this case, Alabama is clearly clustered with Ariège and Slovenia.

Discussion

From an evolutionary point of view, several unifying generalities about cave fauna emerged in the twentieth century. In particular, the morphology of cave animals was shown to be strongly convergent, both in its reductions (eyes and pigment in particular) and in its increases (appendage lengthening and increase in extra-optic sensory structures), as a result of varying combinations of natural selection and neutral mutation coupled with genetic drift (see Christiansen 2005).

Generalities about the ecology and biogeography of cave fauna have been far more elusive. We do know that, relative to surface habitats, species richness is much lower, the typical pattern in caves where there is no light and little food. For example, in a single Ariège forest litter sample of 250 cm3, there are, on average, more species of Collembola (16.5) than in all of the caves of Ariège (13) (Deharveng and Lek 1995). It has also been recognized that the ranges of troglobionts and stygobionts are small compared to surface-dwelling species (Lamoreaux 2004), although the reasons remain contested. It may be the result of the extreme age of a relict fauna, the absence of significant subsurface dispersal following colonization, or restricted subsurface dispersal accompanied by speciation (Christman et al. 2005). What has not been emphasized is the potential general interest in biogeographic patterns of subterranean fauna.

The results of the analysis of the patterns presented here allow us to propose some new generalizations about the biogeography of the terrestrial cave fauna, in particular, patterns of species richness.

Of the seven intensively sampled regions, there are three that consistently had higher species richness than the others, irrespective of how it was measured (Fig. 2) – northeastern Alabama, the Slovenian Dinarides, and the Ariège.

Although there is not sufficient data to analyze other regions to the extent done here, it is possible to fill in the map of subterranean biodiversity beyond the seven regions intensively studied. This is relatively easy to do in the case of areas with low diversity. If the total reported number of troglobionts from a region >2000 km2 is less than fifty or so (see Figs. 1 and 2), it cannot be a hotspot. Under this criteria, most North American cave regions cannot be hotspots (Culver et al. 2003), including the Ozarks, the Black Hills, the Guadalupe Mountains, and the Driftless Area of the upper Midwest (Fig. 3A). Even the area around Mammoth Cave in Kentucky, USA, is highly unlikely to be a hotspot. While Mammoth Cave has 26 troglobionts, the most of any U.S. cave (Culver and Sket 2000), this comprises most of the troglobiotic fauna of the immediate region, and is well below 50.

Figure 3.

A. Map of species richness patterns of North American troglobionts. Major karst areas of eastern and central United States are shown in light gray. While there are many caves in the western United States and Canada, there are no large karst areas, and no areas of rich fauna. The open triangles are areas with few if any troglobionts, the gray triangle are areas with <50 species, usually much <50. The gray circles are three areas analyzed in this study with <50 species in 5000 km2 of area or less. The black circle is the diversity hotspot in northeast Alabama. The boundary of the Pleistocene ice sheet is shown as a solid line. A pair of dashed lines indicates the hypothesized position of the high diversity ridge. B. Map of species richness patterns of European troglobionts. The open triangles are areas with few if any troglobionts, the gray triangles are areas with <50 species, usually <50, and the gray circle is Ardeche, with <50 species in 5000 km2 of area or less. The black circles are the diversity hotspots in Slovenia and Ariege. Black triangles are other possible diversity hotspots. The boundary of the Pleistocene ice sheet is shown as a scored solid line. A pair of dashed lines indicates the hypothesized position of the high diversity ridge.

Likewise there are many areas of Europe with cave areas >2000 km2 with <50 species. These include karst areas in Belgium, England, Switzerland, Slovakia, the Czech Republic, and the French Quercy Causses and Bassin Parisien to the north, as well as Portugal, Sardinia, Spanish Valencia, and the Italian Appenines to the south (Fig. 3B).

Are there other hotspots? Guéorguiev (1977) has suggested that, in the Dinaric Mountains, the troglobiotic richness reaches it apex, not in the western part in Slovenia, but to the southeast in southeastern Hercegovina and adjoining parts of Montenegro and Croatian Dalmatia (see also Sket et al. 2004). In Europe, other sites in the Pyrenees, the Lessianian mountains in northeast Italy, as well as Cantabria in Spain hold some promise.

What do the subterranean hotspots in Alabama, Ariège, and Slovenia have in common? What they don't have in common is easier to enumerate. They are not all European sites, as the single cave hotspot list of Culver and Sket (2000) might suggest. Their proximity to Pleistocene glaciation is not consistent (Table 1), ranging from near (Ariège) to far (Alabama). However important the Pleistocene may have been in isolating animals in caves (Peck 1984), it is of no help explaining the location of hotspots except that hotspots do not occur in glaciated areas. The three hotspot regions also do not share a particular “overrepresentation” of the same taxonomic group. For example, Alabama has an “excess” of Pseudoscorpionida and Ariège has an “excess” of Collembola. Finally, they do not share a particular fragmentation of karst, fragmentation in this context meaning the inverse of geographic continuity of the cave-bearing limestone. Ariège is a highly fragmented cave region and the other two are not. So the hypothesis that geological structure imposes a constraint on species richness (e.g. Barr 1967) is not sustained at this scale.

What the three areas have in common is twofold. First, the density of caves in these three regions is higher than in any of the other regions studied. The number of caves per se is probably not important but it is a surrogate for the amount of available habitat. The greater the amount of available habitat, the potentially larger the cave populations are, and thus extinction rates should be lower. An increased density of caves (and habitat) may also provide greater opportunities for colonization of caves as well. This is in line with previous findings that number of caves was the best available predictor of regional species richness in the U.S. (Culver et al. 2003), and at a smaller scale, that the number of caves in a U.S. county and the number of caves in adjoining counties was a good predictor of number of troglobionts (Christman and Culver 2001), and that the amount of cave passage in a region of several hundred km2 was a good predictor of troglobiont hotspot caves in Slovenia (Culver et al. 2004). The more interesting question, in a way, is why these regions have such high density of caves. In the case of the Dinaric Mountains in Slovenia, high cave densities result in part from a long history of cave development in the massive thick limestone with a series of cycles of cave formation and erosion (Mihevc 2001). So, it may be that the hotspot regions are actually older, at least in the sense of the length of time cave habitats have been available in the region.

The second feature that the three regions share is that they and other potential hotspots occur along a very narrow latitudinal band – ca 42–46°N in Europe (Fig. 3B) and 33–35°N in North America (Fig. 3A), an apparent ridge of high subterranean biodiversity. Of course, latitude per se does not determine diversity (Hawkins and Diniz-Filho 2004). We suspect that the important variable is long-term high productivity. In Europe, where the ridge is better defined, both to the north and south, cave regions have lower biodiversity. South of the hotspots is a region of significantly drier climate, e.g. Iberian Peninsula and southern Italy. To the north mean annual temperature declines with the increase in latitude. This pattern suggests that European hotspots are associated with areas of long-term high productivity, as measured by high temperature and rainfall. In western Europe, the southern extent of the ridge is defined by mountain ranges (Pyrenees and Alps). In the Balkan Peninsula, the ridge is not well defined, primarily because we do not know enough about the geographic distribution of species richness in the southern margin, especially Greece. Greece is rich in species (Sket et al. 2004) but whether there is a concentration of species in any 2000 km2 area is unclear. It is interesting to note that the southern border of the biodiversity ridge also corresponds to the range limit of many small mammals (Kryštufek and Griffiths 1999).

In North America, the hypothesized ridge is very short, disappearing by mid-continent (Fig. 3A), and lacks any topographic correlates. As is the case with Europe, it is associated with areas of high rainfall and temperature – the southern most karst area with significant terrestrial habitat in the humid east. The karst areas in Texas and New Mexico to the west are significantly drier and available evidence (Culver et al. 2003) indicates that the ridge does not continue anywhere in western North America (Fig. 3A). The karst area of Florida to the south, the only cave area south of the northeast Alabama region, has almost no troglobionts and relatively little terrestrial habitat, consisting largely of permanently or seasonally flooded passages. Perhaps even more importantly, it was submerged during the Pleistocene. To the north are the karst regions of Indiana and West Virginia with lower long-term productivity, because of lower temperatures.

This explanation is plausible because there is a connection between surface productivity and the amount of food available in the cave. In nearly all caves, there is no significant chemoautotrophy so that all available food in the cave results from surface productivity. Of course, a major difference with most surface fauna is that the troglobiotic fauna has little or no ability to move in response to changing patterns of productivity. Nevertheless, productivity may well be a major determinant of extinction rate. The hypothesized ridge of high subterranean biodiversity may correspond to regions where productivity remained high and did not experience major decreases in recent geological times (no extremely dry episodes like in southern Europe and no extremely cold episodes like in northern Europe).

It is possible that this ridge continues through Asia. There are potential subterranean hotspots in the karst regions of western Caucasus in Georgia and possibly the Tien Shan Mountains in Kyrgyzstan. Little known biologically (Gvozdetski et al. 1994), they hold promise based on their relationship to productivity maps, and the presence of major cave regions (Klimchouk 2004a,b). To the east of Kyrgyzstan to near the eastern coast of the Asian continent, the climate is probably too dry to support subterranean hotspots. Shikoku Island in Japan, with over 80 troglobiotic species of trechine beetles and a rich Collembola fauna (Ueno 2001), is somewhat south (33°N), and in line with the North American site.

Although we have not analyzed data on tropical caves in this study, our findings may help explain the pattern in the tropics as well. Little quantitative data are available about tropical cave biodiversity, but two generalities emerge. The number of troglobionts in those caves studied is generally less than in hotspot caves in north temperate areas (Deharveng 2005). Second, if all species, not just troglobionts, are counted, tropical terrestrial communities are at least as diverse and probably more diverse than temperate caves, especially if bat guano communities are included (Deharveng and Bedos 2000). The relative scarcity of tropical troglobionts may stem from the fact that fewer species have become isolated in caves, perhaps because the effects of the Pleistocene were less profound in the tropics. The inability of cave animals to migrate long distances (hundreds of kilometers or more) means that the arid zone (the horse latitudes) that generally separates the tropics from moist north temperate zones is a barrier to the migration of troglobionts into what may well be a favorable subterranean environment. This, coupled with the relatively infrequency of local isolation of species in the tropics, the causes of which are still unclear, may well explain the tropical pattern.

It is worth noting some of the similarities with the surface biota. The two major apparent determinants of biodiversity for the subterranean terrestrial fauna – productivity and habitat availability – are the same as many other faunas (Hawkins et al. 2003, Turner and Hawkins 2004). The importance of cave density may represent an increase in speciation rate, in line with Hubbell's (2001) fundamental biodiversity number, θ=2ρAν, where θ is the fundamental biodiversity number, ρ is density of individuals, A is area, and ν is speciation rate. Long term regions of stable productivity should have high densities of animals in caves, suggesting that ρ is the important variable in the case of the terrestrial cave fauna. Area was controlled for in this study, but cave density may affect speciation rate.

We think an expansion of intensively rather than extensively studied cave regions will ultimately provide the explanation for this ridge of diversity. Does it circle the globe? Is it only present in areas of extraordinarily high cave density (and perhaps a longer history of available habitat)? If it circles the globe, then regions like the eastern Balkans, the Caucasus, the Tien Shan Mountains, and perhaps Shikoku Island, are part of this unique ridge of subterranean species richness. The pattern stands in contrast with surface patterns of which we are aware but the subterranean pattern is rooted in the role of productivity, and the consistency of productivity in subterranean habitats.

Subject Editor: John Spence.

Acknowledgements

Work on this project was made possible by a temporary appointment of DCC to the Museum National d'Histoire Naturelle, Paris. We are grateful to many colleagues who provided both data and ideas, especially Daniel Fong (Washington, DC), Horton Hobbs (Springfield, OH), Leonardo Latella (Verona, Italy), Tanja Pipan (Postojna, Slovenia), Slavko Polak (Postojna, Slovenia), Valerio Sbordoni (Rome, Italy), Katie Schneider (College Park, MD), and Fabio Stoch (Rome, Italy). Annie Bauby (Paris) and Marie-Claude Souqual (Paris) helped with data entry, and Charles Gers (Toulouse, France) and Arnaud Faille (Paris) helped with checking data quality. DCC was supported by a grant from the Cave Conservancy Foundation (Glen Ellyn, VA). This article has been subjected to the U.S. Environmental Protection Agency's peer and administrative review and approved for publication. Approval does not signify that the contents reflect the views of the Agency, nor does mention of trade names or commercial products constitute endorsement or recommendation for use.

Ancillary