Global Subterranean Biodiversity: A Unique Pattern

1. Introduction

Since the 1980s, with the widespread use of the phrase biodiversity [1], the mapping and analysis of biodiversity has excelled at a rapid pace at a range of scales, from small to global. One example of this is the global maps of biodiversity produced by the Worldwide Fund for Wildlife [2]. However, large-scale analyses of subterranean biodiversity—especially the biodiversity of caves and other subterranean habitats, such as soil, epikarst, and the underflow of rivers—are conspicuously absent [3]. There are a number of reasons for this, including difficulties in accessing habitats, incomplete taxonomy, and the dominance of β diversity over α diversity [4,5]. Culver and Sket attempted to circumvent and ignore these problems by concentrating on individual caves and aquifers, as opposed to just regions [6]. Their initial list of 20 caves and aquifers, with 20 or more species limited to subterranean habitats (stygobionts and troglobionts), stimulated interest among researchers in identifying species present in various caves and wells. However, this early attempt and subsequent studies [7,8] remained incomplete, because several large regions remained unexplored and available data of others had not been synthesized. It is why we decided in 2020 to launch a more comprehensive analysis, with the aim to document the richest subterranean biodiversity hotspots at the world scale and to get insight into the understanding of their geographical and ecological pattern. A first special issue of the journal Diversity was published in 2021 on the question [9], and a second one was launched in 2022. The present contribution is the closing paper of these special issues, that synthesizes the current state of our knowledge on these hotspots of subterranean biodiversity. With the completion of the second issue, 12 hotspots of subterranean biodiversity are added to the 14 sites analyzed in the former issue. With two recent cases from the literature, we reach a total of 28 hotspots, that represent almost all subterranean biodiversity hotspots documented thus far at the world scale.

2. Goals

We use here this expanded dataset to synthesize the geographic patterns and formulate global maps of subterranean biodiversity. To this end, we summarize the numbers of species found in the different sites, with 25 or more subterranean specialists, stygobionts and troglobionts. For a better understanding of the observed patterns, we give an overview of the distribution of lower biodiversity spots on earth.

Our second goal is to outline some of the challenges encountered in the analysis, and especially in the comparison of subterranean biodiversity. Challenges linked to sampling unevenness are widespread. We use total species number versus species numbers excluding different zoological groups to evaluate the impact of these biases in between-site comparisons. Challenges are also particularly acute in the definition, both theoretical and practical, for what constitutes a cave (or subterranean) limited species—stygobionts and troglobionts.

3. Results

The current map and list of these 28 sites of high subterranean biodiversity (Figure 1,

Table 1. Physical characteristics of hotspot caves and wells, arranged by increasing latitude.

Country	Cave	Latitude	Longitude	Features
BRA	Areias Cave System	−24.6	−48.7	Karstic
Tropic of Capricorn		−23.5
AUS	Robe River Well 2A	−21.6	115.9	Calcrete/ Hypogene
AUS	Undara Lava Tube System	−18.2	144.5	Volcanic
BRA	Água Clara System	−13.8	−44.0	Karstic
BRA	Igatu Cave System	−12.9	−41.4	Silici-clastic
IDN	Towakkalak System	−5.0	119.6	Karstic
Equator		0
VNM	Hang Mo So	10.2	104.6	Karstic
MEX	Sistema Huautla	18.1	−96.8	Karstic
THA	Tham Chiang Dao	19.4	98.9	Karstic
Tropic of Cancer		23.5
ESP	Cueva del Viento System	28.4	−16.7	Volcanic
ESP	Túnel de la Atlantida	29.2	−13.5	Volcanic
CHN	Feihu Dong	29.2	109.3	Karstic
USA	Comal Springs	29.7	−98.1	Hypogene
USA	San Marcos Artesian Well	29.9	−97.9	Hypogene
BMU	Walsingham Caves	32.3	−64.8	Hypogene
USA	Fern Cave	34.7	−86.3	Karstic
USA	Crystal-Wonder Cave System	35.3	−85.9	Karstic
USA	Mammoth Cave	37.1	−86.1	Karstic
BIH	Vjetrenica Cave System	42.9	18.0	Karstic
ESP	Ojo Guareña System	43.0	−3.7	Karstic
FRA	Baget System	43.0	1.0	Karstic
FRA	Coume Ouarnède System	43.0	0.9	Karstic
ROU	Movile Cave	43.8	28.6	Hypogene
FRA	Lez Aquifer	43.8	3.8	Karstic
FRA	Cent Fonts	43.8	3.6	Karstic
HRV	Lukina Jama-Trojama Cave System	44.8	15.0	Karstic
SVN	Križna Jama	45.7	14.4	Karstic
SVN	Postojna Planina Cave System	45.8	14.2	Karstic

) represent nearly all of the sites known to contain 25 or more stygobionts and troglobionts. Most of these sites are discussed in individual papers in two Special Issues of Diversity, the only exceptions being the Areias Cave System of Brazil [10] and Túnel de la Atlantida of Canary Islands [11]. There are only two sites that are claimed to contain 25 or more species and were not included—Logarček in Slovenia [6] and Sauve Spring in France [12]. The potential occurrence of high-diversity cave faunas in other tropical and temperate regions is briefly discussed further down. We have not included deep soil sites or hyporheic sites unconnected with caves. There are also sites in the hyporheic of rivers with 25 or more stygobionts, the most thoroughly studied being the Danube Flood Plain National Park in Austria [13] and the Rhone River near Lyon [14]. Biodiversity patterns in the hyporheic require a different treatment.

Overall, subterranean hotspot sites have been found on all continents except Africa and Antarctica. The latitudinal range of subterranean hotspots is from 25° S to 45° N, including sites in the tropics, thus far only in the seasonal tropics. There are far fewer sites in the Southern Hemisphere, and none of these are farther south than 25° S. There is a concentration of hotspots at 40 to 50° N latitudes (Figure 2). The 40–50° N cluster generally corresponds to the previously described ridge of high cave biodiversity in Europe [15]. Additionally, there is an almost complete absence of hotspots around the equator; only one, the Towakkalak System in Indonesia, occurs within 10° of the equator.

The sites are a mixture of individual caves, hydrologically connected caves in a single karst drainage system, karst aquifers, and non-karstic caves including lava tubes and silico-clastic caves. The caves themselves greatly vary in size and depth, but include the world’s longest cave—Mammoth Cave in Kentucky—and one of the deepest—Sistema Huautla in Mexico. On the other hand, many of the caves are less than 1000 m long and only a few meters in depth. Two of the caves are anchialine—tidal caves with a freshwater lens (Geben-Herzberg)—Túnel de la Atlantida (Canary Islands) and Walsingham Caves (Bermuda). The caves and cave systems include both epigenic and hypogenic caves. Epigenic caves (caves formed by falling waters [16]), are organized into subterranean drainage basins [17], which can be delineated by the injection and capture of soluble dyes such as fluorescein [18]. Many of the hotspot caves are epigenic (Table 1), and for some—Água Clara Cave System (Brazil), Vjetrenica Cave System (Bosnia and Hercegovina), Ojo Guareña System (Spain), Coume Ouarnède System (France), Crystal-Wonder Cave System (USA), and Postojna Planina Cave System (Slovenia)—species counts for both the largest caves and the drainage basin are available. For other epigenic caves—Feihu Dong (China), Tham Chiang Dao (Thailand), Towakkalak System (Indonesia), Lukina Jama-Trojama Cave System (Croatia), Sistema Huautla (Mexico), Križna Jama (Slovenia), Fern Cave (USA), and Mammoth Cave (USA)—only data for the caves themselves are available. A few sites are not organized by drainage but instead by their proximity and isolation from other systems, e.g., Undara lava tube system in Australia and Hang Mo So in Vietnam. For other epigenic caves and karst areas—Cent Fonts, Lez aquifer and Baget System (France)— data only for the entire basin are available, with data for individual caves included when available. Baget is of special historical interest as it is the site of extensive studies conducted by R. Rouch, who first suggested that karst basins were the natural units for ecosystem studies [19]. Overall, species counts were used for the entire drainage basin when available.

Some caves are not formed by descending water but ascending water; therefore, they are unconnected to and isolated from surface drainage patterns. Many of these caves are formed by H₂SO₄. The frequency of hypogenic caves is still unclear [16], and many epigenic caves show signs of having hypogenic origins [20]. Two caves—Movile Cave (Romania) and Walsingham Caves (Bermuda)—are hypogenic (Table 1) as are the aquifers associated with the Robe River wells (Australia) and San Marcos Artesian Well (Texas). What was accessible for sampling in these four sites was quite different. Movile Cave is only 240 m long but connected to a much larger, deeper aquifer of between 50 and 100 km² [21]; Walsingham Caves comprise a large number of small caves located in a 4 by 0.5 km isolated band of limestone—the Walsingham Tract [22]. The San Marcos Artesian Well samples the 900,000 km² Edwards/Trinity Aquifer, and the two Robe River wells (about 1 km apart) sample the iron-rich Robe alluvial aquifer (of unknown size) [23].

The overall global pattern (Figure 1) is a concentration of sites in Europe, centered around latitude 40 N. The only exception are the sites in the Canary Islands; there are no sites in continental Africa. In the Americas, there are two small clusters: one is a small cluster of three sites near the intersection of the borders of Alabama, Tennessee, and Georgia, the other one is a small cluster in Brazil, also with three sites.

There are differences in richness among the hotspots (Figure 3), a point that we address in detail in the section on challenges. Here, we note that only three sites have both 25 or more troglobionts and 25 or more stygobionts. All are in the Dinaric karst, which ranges from northeastern Italy to Montenegro: Vjetrenica Cave System (Bosnia and Hercegovina), and Križna Jama and Postojna Planina Cave System (Slovenia). This supports the longstanding claim from Sket [24] that the Dinaric karst is a global center of subterranean biodiversity. The richest site of terrestrial biodiversity is Vjetrenica Cave System, and the richest site of aquatic biodiversity is Walsingham Caves (Bermuda). Only 7 sites out of 28 have 50 species or more when troglobionts and stygobionts are counted together:

• Postojna Planina Cave System, Slovenia (105);
• Vjetrenica Cave System, Bosnia and Hercegovina (93);
• Walsingham Caves, Bermuda (63);
• Križna Jama, Slovenia (59);
• Baget System, France (57);
• San Marcos Artesian Well, Texas (55);
• Ojo Guareña System, Spain (54).

The fauna of many of the sites listed in

Table 2. Numbers of species per cave under different counting options. S, observed number of stygobionts; T, observed number of troglobionts; AOG, aquatic obscure/undersampled groups (Protista, Rotifera, Nematoda, Nemertea, Oligochaeta, Acari, commensals and parasites); MC, microcrustacea (Copepoda, Ostracoda and Syncarida, not including parasites and commensals already counted in AOG); S*, S minus AOG minus MC; TOG, terrestrial obscure/undersampled groups (Acari, Oligochaeta, commensals and parasites); T*, T minus TOG; un, undescribed species; un%, 100*un/(S+T).

Country	Cave	S	T	S+T	AOG	MC	S*	TOG	T*	S*+T*	un	un%	Source
AUS	Robe River Well 2A (a)	43	0	43	11	21	11	0	0	11	21	49	[23]
AUS	Undara Lava Tube System	1	30	31	0	0	1	0	30	31	25	81	[25]
BIH	Vjetrenica Cave System	48	45	93	8	8	32	1	44	76	6	6	[26]
BMU	Walsingham Caves	63	0	63	8	29	26	0	0	26	0	0	[22]
BRA	Água Clara Cave System	8	33	41	0	0	8	1	32	40	30	73	[27,28]
BRA	Areias Cave System	6	22	28	1	0	5	0	22	27	14	50	[10]
BRA	Igatu Cave System	2	35	37	0	0	2	3	32	34	29	78	[29]
CHN	Feihu Dong	4	23	27	0	1	3	1	22	25	14	52	[30]
ESP	Cueva del Viento System (b)	0	42	42	0	0	0	0	42	42	0	0	[31]
ESP	Ojo Guareña System	46	8	54	14	24	8	2	6	14	23	43	[32]
ESP	Túnel de la Atlantida	34	0	34	0	12	22	0	0	22	4	12	[11]
FRA	Baget System	40	17	57	4	27	9	1	16	25	5	9	[33]
FRA	Cent Fonts (c)	43	1	44	2	19	22	0	1	23	4	9	[12]
FRA	Coume Ouarnède System (d)	17	17	34	1	8	8	2	15	23	1	3	[34]
FRA	Lez Aquifer (e)	39	0	39	2	15	22	0	0	22	7	18	[12,35]
HRV	Lukina Jama-Trojama Cave System	16	25	41	0	0	16	2	23	39	20	49	[36]
IDN	Towakkalak System	10	26	36	0	0	10	1	25	35	18	50	[37]
MEX	Sistema Huautla	0	27	27	0	0	0	0	27	27	10	37	[38]
ROU	Movile Cave (f)	13	25	38	3	3	7	1	24	31	3	8	[21]
SVN	Križna Jama (g)	31	28	59	10	10	11	0	28	39	5	8	[39]
SVN	Postojna Planina Cave System (h)	62	43	105	12	29	21	2	41	62	11	10	[40]
THA	Tham Chiang Dao	4	33	37	2	2	0	1	32	32	17	46	[41]
USA	Comal Springs (i, j)	32	0	32	3	2	27	0	0	27	4	13	[42]
USA	Crystal-Wonder Cave System (k)	8	23	31	0	0	8	1	22	30	3	10	[43]
USA	Fern Cave (l)	8	19	27	2	0	6	1	18	23	7	26	[44]
USA	Mammoth Cave (m)	17	32	49	3	2	12	6	26	38	0	0	[45]
USA	San Marcos Artesian Well (j)	55	0	55	8	15	32	0	0	32	16	29	[42]
VNM	Hang Mo So	0	27	27	0	0	0	1	26	26	20	74	[46]

Notes. (a) one stygophilic species, Tubificidae sp., discarded; two subspecies of humphreysi and their hybrid counted as a single species; (b) the system includes Cueva Felipe Reventón with 38 troglobionts, and Cueva del Viento with 36 troglobionts (erroneously noted 28 in Culver et al. [9]); (c) one additional troglobiont, Laemostenus (Actenipus) oblongus balmae; not counted in the abstract of from Prié et al. [12]; (d) four species discarded (strictly hyporheic species found outside cave); (e) a single aquifer, accessed from several wells close each other; (f) species found in nearby springs and wells have been counted; (g) one stygophilic species, Synurella ambulans, discarded; (h) eleven species discarded (stygophiles or troglophiles, listed as troglobiotic populations of surface species); (i) Table S2 in [42]; (j) aquifer is greater than 2500 km² so single points used; (k) Crystal and Wonder Caves connected by 91 m long surface stream, therefore combined; (l) all taxa considered from the two columns “Historical” and “This Study” of Table 1 in Niemiller et al. [44]; (m) taxa considered from the column “This Study” of Table 1 in Niemiller et al. [45].

are primarily terrestrial or aquatic; nine sites are exclusively one or the other (Table 2). The richer sites tend to be in Europe and North America, but this may in part be due to more thorough data collection, a point that we address below.

4. Challenges

4.1. How to Define Stygobionts, Troglobionts, etc.

Perhaps it is surprising that obligate subterranean dwelling species are difficult to separate from those species that are either transient in caves—in the sense they do not complete their entire life cycle there (sometimes called subtroglophiles [47])—or can survive and reproduce in surface habitats (sometimes called eutroglophiles [47]). Several classifications and terminologies have been proposed [47,48,49,50,51], but for the most part, researchers agree on the core aspects of stygobionts and troglobionts—species that can only survive and reproduce in subterranean habitats, especially caves and aquifers. Some authors [39,40] use the terms stygobionts and troglobionts for species which have subterranean and above-ground populations. These species match the definition of eutroglophiles, and, for practical reasons, particularly with regard to consistency of terminology, were not included in counts. Aside some exceptions [52], a core tenet of speleobiology is that troglobionts and stygobionts often have a convergent morphology known as “troglomorphy” [53], including the reduction of eyes and pigment, increase in size, elongation of the appendages and development of extra-optic sensory structures. A number of troglobionts, however, especially but not necessarily phylogenetically young ones, may display little or no troglomorphy [54]. Subterranean populations of highly variable species show sometimes a reduction or a loss of pigment and eyes, and are considered by some authors as stygobionts or troglobionts. The best known example of this phenomenon is the Mexican cavefish, Astyanax mexicanus, with eyed and eyeless populations [55]. Such cases need to be examined carefully, both to determine if they point to genetically separate taxonomic entities, which seems most likely, and if the cave populations are truly troglomorphic (see below). The taxonomic status of Astyanax mexicanus is disputed, with some authors claiming that the eyeless cave populations are a separate species [56].

The regressive characters (pigment and eye reduction) shown by some species are often invoked by authors to qualify the species as “troglomorphic” and, by extension, troglobiotic; however, these species may not exhibit the set of characters that define troglomorphy in an adaptive sense. Many of them refer to deep-soil life forms, which share eye and pigment regression with troglomorphic cave-restricted species, but not other traits that define troglomorphy. Such species are, in that case, erroneously classified as troglomorphic. Finally, morphological characters of a number of species or morphospecies, especially in the tropics and subtropics, remain undescribed, making it difficult assignment to life forms.

In general, we have followed various authors in their assessment of ecological status, with some exceptions, especially when they disagree on the status of a species. Approaching troglomorphy as a purely morphological qualification and troglobiotic as a purely ecological one would avoid much confusion. Numerous ambiguities and uncertainties persist in the literature, mostly linked to a loose use of terminology and to disputable assignment of species to ecological categories. There is no completely satisfactory terminology for troglomorphic cave populations of species with non-troglomorphic populations outside of caves, like those of Astyanax mexicanus [55]. The statuses of species that have deep-soil species facies, those in Robe River boreholes [23], and of species with deep sea populations, those from Walsingham Caves [22] are disputable because their connection with soil (Robe River) or marine environment (Walsingham Caves) is unknown. These and other problems have led some researchers to reject the terminology or at least reduce its usage [57,58]. On the other hand, redefining troglomorphy and specialization to deep soil (edaphomorphy) based on morphological criteria that are statistically correlated to the occurrence of species in habitats [59] could make such a terminology useful.

Deharveng et al. [46] employed a decision system to determine if a species was troglobiotic in their analysis of Vietnamese caves. These caves are usually short and shallow, with frequent terrestrial and aquatic connections to the exterior, allowing constant inputs of nutrients. This generates cave communities dominated by troglophiles and tramp species, and makes the ecological category assessment of individual species hazardous. They adopted four complementary approaches:

(1)

Morphological inference is based on presence or absence of troglomorphic traits. The presence of a set of convergent troglomorphic traits in most arthropods (eye and pigment reduction combined with appendage and size increases compared to surface relatives) points to obligate cave life. Depigmentation and eye reduction are trends shared by many soil and cave arthropods, and Brignoli [60] stressed that the equation “blind = troglobite” has a limited value. When they are combined with appendage shortening and a decrease in size, they qualify a species as euedaphomorphic [59]. It is only when eye and pigment regression are combined with appendage elongation and an increase in size (or other characters recognized as troglomorphic), that they qualify a species as troglomorphic. Statistically, the correlation of troglomorphic and euedaphomorphic life forms with the ecological categories of troglobiont and edaphobiont is one-way and robust. Where the set of troglomorphic traits is not present, such as in many guano-associated and tropical species, we have to rely on other inferences [61].

(2)

Parallel sampling inference is based on the absence of species outside subterranean habitats, and allows to assign a status of troglobionts to species that do not exhibit troglomorphy (“obligate troglophiles” of Howarth and Wynne [51]). Statistically meaningful data on the occurrence of species, both inside and outside caves, can be extracted from the literature for well-investigated regions. In lesser known areas such as the tropics, sampling in parallel cave and non-cave habitats may allow us to reasonably assess the ecological status of a species, the strength of such an inference being dependent on sampling efforts and on the rarity of the species.

(3)

Taxonomic inference is based on the ecological status of related taxa. Certain groups are known to greatly diversify in subterranean habitats [62,63,64], while others never colonize such habitats. A species from a group which is not prone to underground diversification and lack troglomorphic traits is less likely to be a troglobiont.

(4)

Barcoding inference is based on genetic divergence between populations and species. Within a troglophilic or stygophilic species, molecular analyses may characterize populations that live in caves as different from those that live outside [65], leading to split the original species into cave-restricted and non-cave-restricted lineages or species. Barcoding may conversely lead a species to lose its ecological status of cave-restricted if it is shown to be molecularly inseparable from another species which is not cave-restricted.

4.2. Taxonomic Completeness

As is to be expected with a rare, elusive fauna, sampling for the majority (if not all) of the sites listed in Table 2 is incomplete, and the extent of incompleteness varies among sites. A few sites, most notably Postojna Planina Cave System (Slovenia), Vjetrenica Cave System (Bosnia and Hercegovina), and Mammoth Cave (USA) have a centuries-old history of biological study. Others, especially those in the tropics and subtropics, have a decade long or less history of biological study. Lukić et al. [36] reported that almost all information for the fauna of the Lukina Jama-Trojama Cave System in Croatia dates from the 1990s or later. The same is true for Fern Cave in Alabama [44], Hon Chong in Vietnam [46], and the Água Clara Cave System in Brazil [27,28]. Sampling effort is therefore a major determinant of the richness of species in these caves. We estimated sampling effort by comparing numbers of species with the date of the first listing of the fauna. Other measures, such as total number of publications, are difficult to estimate due to difficulties in defining publication about the site, for example, does it include monographic taxonomic studies where the species in question are a small part of the study? When date of first faunal list and number of species are compared, the regression accounts for approximately one-third of the total variance in species number (Figure 4). This gives pause to more biological interpretations (see below).

Some groups, e.g., beetles and amphipods, have been searched for and studied in most hotspot sites. For other taxonomic and ecological groups, this is not the case. Among aquatic species, these include parasites and commensals, Protista, Rotifera, Nematoda, Nemertea, Oligochaeta and water mites. For example, subterranean-limited Protista have been reported from Planina Postojna Cave System [40] and Walsingham Caves [22], but have been searched for in few of the other caves (but see [66]). Only three contributions report on subterranean rotifers—Ojo Guareña [32], Robe Valley [23] and Cent Fonts [12,67]—and only for the Robe Valley is it claimed that the rotifers are stygobiotic. The data on obligate subterranean aquatic species summarized in Table 2 are “corrected” for this discrepancy by eliminating species of these groups. A second source of bias in the counting of obligate subterranean aquatic species is microcrustacean fauna—Copepoda, Ostracoda, and Syncarida. For sites with aquatic fauna, at least a few of these species have been reported, but a rich source of these species is the epikarst [68] that has rarely been sampled. In the two caves where it has been sampled, 15 copepod species were discovered in Postojna Planina Cave System [69] and 9 copepod, ostracod, and syncarid species were discovered in Ojo Guareña [32,70] (Table 2). Species numbers with copepods, ostracods, and syncarids deleted are also shown in Table 2.

Although the variability of the level of study of the terrestrial species is similar to that of the aquatic fauna, we were less successful in correcting for this bias. Higher level categories often do not have global distributions [71]; therefore, absence is often the result of this rather than lack of collecting. Acari is one group that has been inconsistently studied among hotspot caves. Mites are rare in caves that lack guano, and this is largely due to the undersampling of their preferred habitats, i.e., cave soils and organic matter. In most cases, they also remain undescribed or unidentified. The only cave with more than two species reported as troglobionts is Mammoth Cave (Table 2), which contains six species. Five of the six Mammoth Cave species were first described in the 19th century, but these data are problematic since several of these species have not been found since [45]. Coleoptera are the best studied, but, while investigated in all hotspot caves, have a biased geographic distribution, being most common in the Holarctic and least common in dry and tropical areas, while the reverse is globally true for Arachnida (

Table 3. Relative percentages of troglobionts among arachnids (excluding mites) and beetles in hotspot caves where terrestrial cave fauna includes more than 10 troglobionts. Caves are ranked by decreasing values of the ratio Coleoptera/Arachnida (ratio Co/Ar). N, species number.

Country	Cave System	Arachnida		Coleoptera		Ratio
		N	%	N	%	Co/Ar
FRA	Baget System	2	22.2	7	77.8	3.50
FRA	Coume Ouarnède System	2	25.0	6	75.0	3.00
BIH	Vjetrenica Cave System	10	43.5	13	56.5	1.30
USA	Crystal-Wonder Cave System	5	50.0	5	50.0	1.00
CHN	Feihu Dong	5	50.0	5	50.0	1.00
SVN	Križna Jama	7	50.0	7	50.0	1.00
SVN	Postojna Planina Cave System	10	52.6	9	47.4	0.90
ESP	Cueva del Viento System	16	55.2	13	44.8	0.81
USA	Mammoth Cave	10	55.6	8	44.4	0.80
ROU	Movile Cave	8	61.5	5	38.5	0.63
HRV	Lukina Jama-Trojama Cave System	5	62.5	3	37.5	0.60
USA	Fern Cave	6	66.7	3	33.3	0.50
AUS	Undara Lava Tube System	10	71.4	4	28.6	0.40
BRA	Areias Cave System	6	75.0	2	25.0	0.33
VNM	Hang Mo So	8	80.0	2	20.0	0.25
BRA	Água Clara Cave System	9	81.8	2	18.2	0.22
THA	Tham Chiang Dao	10	83.3	2	16.7	0.20
BRA	Igatu Cave System	13	86.7	2	13.3	0.15
IDN	Towakkalak System	11	91.7	1	8.3	0.09
MEX	Sistema Huautla	18	100.0	0	0.0	0.00

) [59].

5. Coldspots, Low-Diversity Spots and Undersampled Spots

While patterns of high diversity are emerging at a global level, the distribution of remaining subterranean diversity on Earth has been less thoroughly scrutinized. A better knowledge of the spatial patterns of lower subterranean biodiversity would help our understanding of hotspot patterns, making clearer where coldspots are located, as well as predicting where sites of high diversity may be expected to be found in future investigations. Rough outlines of our knowledge on these lower biodiversity sites are summarized below.

5.1. Where Are the Coldspots?

Large regions where troglobionts and stygobionts are absent or rare have been extensively documented in many parts of the world, especially Canada [72], Germany [73], and Poland [74]. Their biodiversity is assumed to have been largely depleted by Quaternary glaciations. Therefore, troglobionts are restricted to a few refugium massifs [74,75] that remained free of ice during glaciations. Stygobionts are more diversified, having survived under the ice caps [75] or having recolonized from areas unaffected by glaciations.

As expected and confirmed by emerging patterns, hotspots of subterranean diversity are all located within the southern and the northern areas affected by glaciations. However, these are small sites in a very large area comprising diverse environments, where it is usually difficult to know if local low diversity and coldspots are real or the result of undersampling. This uncertainty is reflected in considerable ecological and geographical sampling gaps. Understanding gap distribution would aid our understanding of patterns with high richness.

5.2. Lower Biodiversity Habitats

Some habitats widespread on earth are devoid of hotspots. It may be that they really are low biodiversity habitats or that they are rarely sampled, at least from a biodiversity perspective. Aside from local and special microhabitats, such as hypotelminorheic habitats [76], several of these habitats are extensively distributed—marine caves [77], anchialine caves [78], littoral interstitial habitats [79,80], phreatic and hyporheic of rivers [81], and MSS (milieu souterrain superficiel) and scree in all kinds of rocks [82].

Anchialine. Most hotspots in our survey are filled with or connected to freshwater. Only two are related to anchialine habitats with a marine/freshwater boundary (Túnel de la Atlantida and Walsingham Caves). While much less frequent than freshwater caves, they have a global distribution, and are relatively common in some regions, such as the Yucatan Peninsula and the Mediterranean Sea [83]. The high biodiversity of marine and anchialine caves is often emphasized at a regional level, especially in the Mediterranean Sea [79]. However, strictly marine caves appear to have few stygobionts [77].

Littoral interstitial. There is a dichotomy between freshwater interstitial and littoral interstitial habitats at the level of faunal composition, but not at the level of habitat characteristics, which are similar in most respects except for salinity, nor at the level of habitat continuity, as freshwater and littoral interstitial are largely adjoining [80,84]. Contrary to freshwater interstitial habitats that largely contribute to the biodiversity of several hotspots in our survey, littoral interstitial habitats are absent, despite their global importance. None of the few faunistic datasets of these last habitats currently available in the literature explicitely points to hotspot, but it has been demonstrated that they have a rich and very distinctive fauna, which differs from the freshwater interstitial fauna at high taxonomic level. Delamare Deboutteville [79] reports for instance 95 species in the interstitial of a 300 m long beach of Southern France (Canet-Plage).

Deep Phreatic. Among the hotspots documented in these two Special Issues of Diversity, three are clusters of wells that penetrate phreatic waters—Edwards Aquifer (Texas), Robe River (Australia), Lez aquifer (France) (Table 2, Figure 1). These are, to our knowledge, the only deep wells that have been sampled.

Hyporheic. The alluvia of rivers are not included in this study but certainly contain a rich fauna. Danielopol and colleagues [13,81] showed that some European sites (Danube, Rhine and Rhône rivers) hosted a rich interstitial fauna. The Sava River in central Europe [85] and the Flathead River in Montana can be added to this list [86]. For example, the Grand Gravier site along the Rhône River near Lyon, France, yielded 30 species of stygobionts [14]. Although sampling difficulties may be technically challenging, more sampling with Bou Rouch pumps is a promising way to find new hotspots. The challenge here is also to define site units to be compared.

MSS and screes. Although widespread in temperate regions and known to host obligate subterranean fauna [3,87], we do not know of any hotspots of troglobiotic biodiversity in these habitats. Published lists for single sites usually include a few troglobiotic species [82,88,89,90]. Given serious sampling difficulties and the small number of investigated sites, especially in regions known to harbor rich biodiversity, it cannot be ruled out that future investigations in such sites may provide more troglobionts.

Epikarst. Individual epikarst drips may have up to ten species of stygobiotic copepods, and all the drips in a cave may have up to 15 species [69,91]. On average, about five stygobiotic epikarst copepods are found per cave [91]. Not all species are found in epikarst pools, and thus specialized drip collectors are needed. Only a few epikarst sites have been thoroughly sampled in Slovenia, Italy, Romania, Spain, and the United States [92].

Hypotelminorheic habitats. They harbor even fewer species—four or less in seeps in the upper Potomac basin near Washington, D.C. [93,94]. The spatial extent of these habitats is difficult to assess, but they are locally common and completely unconnected with caves and other deeper subterranean habitats.

5.3. Lower Biodiversity Sites and Regions

While significant progress has been made over the past several decades, especially in the tropics [71,95,96,97,98], many large cave areas remain unsampled or undersampled. Even in Europe, sampling is highly uneven [99]. Because of the rarity of species and difficulty of collecting them, multiple trips are needed to obtain a more or less complete list of species [95,97]. As a result, many species, even in well-studied areas, are known only from a handful of specimens, sometimes even one. The few studies on sampling completeness using species accumulation curves, indicate that sampling may require 100 or more caves to contact 70 percent of the fauna [100]. One feature of sampling completeness is that it is not just in low-diversity areas that additional species are expected to be discovered. For example, Zagmajster et al. [100] have shown that more new species of beetles are expected to be found in high-diversity rather than in low-diversity sites. In the same line, new troglobionts among cave beetles of China are mostly described from karsts that were already known to be the richest in cave species [101].

With the possible exception of Mammoth Cave, Vjetrenica Cave System, and Postojna Planina Cave System, authors of articles that focus on individual caves indicated that more species remain to be discovered. Is it possible that the hotspot caves of Table 2 have just been more thoroughly sampled than other caves? While there certainly remain hotspot sites to be discovered, we believe that hotspot species and their mapped distributions are roughly representative of real patterns, even if collections need to be completed for several habitats and regions.

In support to this hypothesis, a number of caves that have been well studied are rich in cave restricted fauna, but clearly not hotspots, and sometimes do not even host any troglobiont. Therefore, their global distribution may help to reinforce and finetune the hotspot patterns. We highlight a few of these below, as well as pointing out that entire regions are unlikely to yield hotspot caves, including the extensive cave system occurring in once glaciated Europe and North America.

5.3.1. Lower-Biodiversity Spots in Africa

Two hotspots have been documented in Africa, both located in the Canary Islands, but none have been recognized in the rest of the continent. The northern fringe of Africa between the Sahara and the Mediterranean Sea, which belongs faunistically to the Mediterranean basin, has no hotspot of subterranean biodiversity, in contrast to the Northern Mediterranean basin where the density of hotspots is the highest in the world [15] (Table 2). The Djurdjura massif in Algeria, that was the subject of intensive investigations conducted 110 years ago by Peyerimhoff (in [102]), is probably the richest of North Africa in cave-restricted fauna, with about 29 species, often troglomorphic, linked to cold caves and snow pits. These karstic features are densely distributed in the massif, but remain generally unconnected, and individual caves have no more than 12 troglobionts.

In sub-Saharan Africa, where karst is limited to some extent, several caves have been significantly sampled for subterranean biodiversity, especially in the Congo Republic, Democratic Republic of Congo, Kenya, Madagascar, South Africa and Tanzania [103]. Phreatic habitats were particularly studied in Somalia [104]. Most of these sites are usually very rich in troglophilic and guano-dependent species, but none has provided more than 20 strictly subterranean species. The Wynberg Cave System in South Africa (with 19 species considered as cave-restricted) is the richest one in sub-Saharan Africa [105], while Kulumuzi cave in Tanzania has a similar number of troglobites, but most with an uncertain ecological status [106].

On the whole, Africa appears to be relatively poor in subterranean biodiversity, with the exception of the Canarian hotspots.

5.3.2. Lower-Biodiversity Spots in Southern Tropical Asia and the Pacific

In lowlands of tropical Asia and the Pacific, about 12 caves have been documented in the literature as having more than 15 cave-restricted species. Three of them were treated as hotspots in the two Special Issues of Diversity, as they host more than 25 such species (Towakkalak System in Indonesia, Tham Chiang Dao in Thailand and Hang Mo So in Vietnam). Nine other caves scattered in the lowlands of the same region have between 15 and 24 cave-restricted species: Batu Caves in Malaysia [107] and Saripa Cave in Sulawesi, Indonesia [37] with 24 species, Clearwater Cave in Sarawak [108] with at least 22 species, Ganxiao Dong in Southern China [30] and the Sangki System in Sumatra, Indonesia [71] with 20 species, Ma San Dong in Southern China [71] with 17 species, Batu Lubang in Halmahera, Indonesia [71] and Tham Thon in Laos [71] with 16 species, Tham None in Laos [71] with 15 species. Several other caves which have been well sampled are much less rich than those cited above. With the exception of the Siju caves in Meghalaya, India which have been studied in detail for more than one century and has only 10 cave-restricted species [109], these low-diversity caves seem to be located in oceanic islands. The lava tubes of Hawaii are not particularly diverse, despite having a well-known, highly distinctive fauna [110]. Culver and Pipan [3] reported only 37 troglobionts from all of the Hawaiian Islands, although there are undoubtedly a number of undescribed species [111]. Only eight troglobionts have been recorded in the longest lava tube in the world, Kazamura Cave in Hawaii [111]. By contrast, the Canary Islands, off the coast of Africa, aside from having two lava tube hotspots, are generally rich in cave fauna [112]. The well documented Fapon Cave in the karst of Santo island in Vanuatu has no more than four unambiguous troglobionts [113] and can thus be considered a coldspot for cave-restricted fauna. A clear common feature of these lowland caves is the wide occurrence of guano and the impressive abundance and diversity of its associated fauna, with a number of species difficult to assign to the traditional ecological categories used for temperate cave fauna [114].

Biological data for caves above an elevation of 500 m are extremely limited in Southern Asia and the Pacific, but valuable datasets exist for two cave systems above 2000 m asl., located in the highlands of Papua New Guinea: Atea Kananda, which host 13 potential troglobionts and a few uncertain stygobionts [115], and Selminum Tem, with at least 19 troglobionts and 5 stygobionts [116]. Their fauna differs widely from that lowland caves of the region, being more similar to that of temperate caves, as shown by the near absence of bats and guano-associated species, the absence of some groups of arachnids (amblypygids, schizomids), and the presence of highly troglomorphic species of beetles. Given that several groups of species collected in these caves have not yet been studied [117], their species richness is likely at the currently recognized level of hotspots in tropical Asia.

The pattern of cave-restricted species richness described above for tropical Southern Asia and the Pacific could be driven by the combined effect of the microclimate (which determines the presence of bats, swiftlets and guano) and of the geological history (which accounts for the length of karst isolation from potential sources of colonizers). Whether patterns in American caves of humid tropics match those of the Old World tropics remains to be explored, but the relatively large amount of data available notably for Cuba, and to a lesser extent for Venezuela [61], does not indicate a rich fauna of cave-obligate species.

5.3.3. Low-Biodiversity Spots in the Temperate Zone

Coldspots and Glaciations. Caves of polar areas, roughly north of 50 °C and south of 45 °C in New Zealand and South America, are relatively well studied [72,118,119,120]. All are very poor in cave-restricted species, especially in troglobionts. Wind Cave in South Dakota, in glaciated North America, and with passages of nearly 250 km, has only two reported troglobionts and no stygobionts [121]. Knight [122], in an extensive review of the aquatic fauna of Swildon’s Hole in the Mendip Hills of England, found no stygobionts or troglobionts. In North America, caves even 100 km south of Pleistocene glaciations in the Appalachians have depauperate fauna [123], but those in the Interior Low Plateaus of Indiana have rich fauna [124], both attributed to the effects of the Pleistocene. A similarly low level of cave obligate biodiversity is documented for the Northern Alps, where cave-restricted fauna are assumed to have been extirpated during glaciations. Würm glaciers cover most parts of these mountains, for instance, in the Swiss and Savoy Alps [125], which harbor very few cave-restricted species [126]. As a rule, all caves located in areas of the Northern Hemisphere that were glaciated or affected by permafrost during quaternary glaciations have a low number or are devoid of obligate subterranean taxa [127]. Most terrestrial cave-restricted species have not been able to recolonize following deglaciation, while it has been much easier for a significant number of aquatic interstitial species to do so, which today may be found farther north [127].

Between the Biodiversity Ridge and the Southern Limits of Glaciations in Europe. South of the limits of areas strongly affected by glaciations and the northern limit of the biodiversity ridge described by Culver et al. [15], there are vast territories where cave-restricted species may be present, but where the caves that have been sampled to some extent, are at most moderately rich. For example, the Carpathians, the second largest mountain range in Europe after the Alps, have been investigated for its cave fauna for more than a century. Its richest cave in Romania, aside from Movile [21], according to a recent published inventory [128], has only 16 cave-restricted species (3 troglobionts and 13 stygobionts). In the Jura range, northwest of the Alps, 22 cave-restricted species were reported in Grotte du Pissoir, France, as a result of intensive sampling over several years [129].

Lower Diversity Areas in the European Biodiversity Ridge. All the well-documented sites of the biodiversity ridge did not provide rich cave fauna. For example, the large cave at Predjama Castle in Slovenia is less than 10 km from Postojna Planina Cave System, a hotspot cave (Table 2), and has been the subject of inventories for over 10 years [130]. Only 11 troglobionts have been reported here, compared to 43 from Postojna Planina Cave System. Such moderately rich sites within the ridge of biodiversity are not uncommon. They may be explained by local site characteristics, such as a narrower range of habitats. More interestingly, large regions on the ridge have well investigated biodiversity spots that are only moderately rich. The most obvious is the Alpine range, which spans a large section of the ridge, even in its southwestern part that was weakly affected by glaciations. Alpine caves documented so far have a subterranean biodiversity lower than those of the west of the ridge (the Pyreneo-Cantabric range) and those of its eastern part (Dinarides). The richest subterranean biodiversity spot documented this far in the Alps is the well-studied Arena Cave in the Lessinian Mountains of Italy (24 obligate subterranean species, of which 16 are troglobionts and 8 stygobionts) [131]. None of the other caves investigated in the French Alps contains more than 20 cave-restricted species [132].

Towards the east. The European Biodiversity Ridge was recognized till the eastern Dinarids [15]. Further east, the ridge could be now extended to the Movile Cave hotspot which is located in the latitudinal range of other hotspots. The absence of documented hotspot east of Movile does not allow to extrapolate, but caves hosting 15–20 cave-restricted species are known in the Caucasus, which remains much less known than the regions included in the ridge. This suggests that, in the future, the ridge may be shown to continue much further east.

6. Discussion

6.1. The Emerging Global Pattern and Its Causes

The distribution of hotspot caves shown in Figure 1 is emphatically not one of high tropical diversity with a decline in richness towards the poles. The distribution of hotspot caves is also different in the Nearctic and the Palearctic. With the exception of sites in the Canary Islands, Palearctic hotspots are clustered along 40° N, a ridge of high subterranean biodiversity previously noted [15,133]. They point out that the ridge is the area of highest secondary productivity in Europe. Such a ridge of high biodiversity does not occur in North America, but there is a small area near the combined border of Tennessee, Alabama, and Georgia of similarly high biodiversity and presumed high secondary productivity [15,43,44]. More generally, a difference between Europe and North America is that European mountain ranges are often oriented east–west while North American mountains are north–south in orientation. Mountain range orientation has important implications for the effect of the Pleistocene and other glaciations on faunal migrations. This may partly explain why there is a high-diversity ridge in Europe but not in North America. The relationship between the Pleistocene glaciations and the distribution of cave fauna is that glaciation may be a major driver of the extinction, isolation and subsequent speciation of surface-dwelling terrestrial invertebrates in caves [134,135]. However, it is by no means certain that the Pleistocene is an important driver of either isolation or speciation. Based on molecular clock determinations, many subterranean lineages are considerably older than the Pleistocene (e.g., [136]).

Several types of subterranean sites are more likely to be hotspots than others. They are as follows:

• Phreatic aquifers. Relatively few aquifers have been sampled, usually in wells or springs. Five of these sites are on the hotspot list—San Marcos Artesian Well (Texas), Comal Springs (Texas), Robe River (Australia), Lez aquifer (France), Cent Fonts (France), and Baget System (France). The first three sites are also sites of chemoautotrophy, which acts to increase the resource base of subterranean communities.
• Sites with known chemoautotrophy, including Movile Cave and Walsingham Caves.
• Lava tubes. Canarian lava tubes and Australian lava tubes, which occur very close to the surface, have a high species richness once again possibly due to increased resources, including tree roots [137].

If this pattern proves to be robust, then a major determinant of cave biodiversity is available organic matter. Of course, the availability of organic matter is a complicated issue in itself, and may be dependent on details of topography (e.g., rugosity [138]), temporal distribution of rainfall, vegetation and disturbance.

On the other hand, it is not the entire explanation. The richest sites, those in the Dinaric karst, are, as far as we know, not particularly rich in organic matter relative to the rest of the world. However, the Dinaric karst has several unique features:

• It is next to the Mediterranean Sea, and the marine fauna of the Mediterranean was a source of colonists of subterranean sites, particularly during the Messinian Salinity Crisis.
• It is a region of high annual rainfall, relative to the rest of Europe. Additionally, temperatures are high for that latitude of the Dinarides. Therefore, productivity is higher.

Another exception to this pattern are tropical karsts of the humid tropics, where the best investigated caves (e.g. Niah Cave, Batu Caves and Mulu caves in Asia, Kulumuzi and Shimoni caves in Eastern Africa, or caves in Cuba, Guatemala, Venezuela in central America), particularly rich in guano, are not biodiversity hotspots for troglobionts in the traditional sense, as documented above. It can therefore be hypothesized that the nature of the organic matter available may be as important as its amount.

Some comparison among sites is possible, but first, corrections need to be made to account for differences in taxonomic coverage for different sites and in the proportion of undescribed species. The microcrustacean fauna, especially in epikarst, is often quite rich but it has only been studied in a few sites (e.g., Ojo Guareña); therefore, we eliminated all micro-crustacea (Ostracoda, Copepoda, Syncarida) for further analysis. We did likewise with those few parasites and commensals as well as several aquatic groups that have only been sporadically reported or described, including Protista, Oligochaeta and Nemertina. Finally, we eliminated Acari, which have not been described or studied in most caves, at least in the last hundred years (e.g., Mammoth Cave).

The resulting estimates (Table 2, Figure 5) are clustered into three groups. First, there are low-diversity sites, ones that primarily consist of micro-crustaceans—Ojo Guareña System (Spain) and Robe River Well 2A (Australia). Second, the great majority of caves (23 in all) range in species numbers from 22 to 40, suggesting a wide range of macroscopic stygobiotic and troglobiotic fauna. Among these 23 caves, seven have a rich aquatic and terrestrial fauna (at least 10 species in each ecological category): Coume Ouarnède and Baget Systems, Movile Cave, Towakkalak System, Mammoth Cave, Lukina Jama-Trojama Cave System, Križna Jama. The third group includes the three richest caves—Cueva del Viento, Postojna Planina Cave System and Vjetrenica Cave System—which have 42 to 76 cave-restricted species. The Cueva del Viento from Canary Islands is exceptional in its exclusively terrestrial fauna, which is clearly as rich in troglobionts as the two world richest hotspot caves—Postojna Planina Cave System and Vjetrenica Cave System—caves that anchor in a sense the two ends of the Dinarides.

6.2. Weighting Species Value in Conservation of Subterranean Sites

Weighting the importance of a species in conservation is common, and there are several types of weighting that are particularly relevant to subterranean site conservation. The first is that obligate subterranean-dwelling species are weighted more than others. In this review, we have ignored non-obligate species for the most part, although some authors of articles focusing on individual caves have included lists of species that maintain permanent populations in subterranean sites—eutroglophiles and stygophiles. As a practical matter, lists of troglophiles and stygophiles are less readily available. For example, no list of troglophiles and stygophiles is available for the well-studied Mammoth Cave since 1968 [139]. A second widely used weighting takes into account the number of occurrences of species. For example, if each species is to have an equal weight overall, and it occurs in n sites, each occurrence is given a weight of 1/n. This is especially important for subterranean fauna for its high levels of single-site endemism. In the eastern U.S., 211 of 467 troglobionts were single-cave endemics [140]. A related weighting is that of extent of the range. A species may be common within a very narrow range, such as many Cambalopsidae in tropical Asia [141], or widespread, such as the copepod Acanthocyclops hispanicus Kiefer, 1937 in Southern Europe. A fourth weighting is that of abundance. Some species are common in the sites where they are found, and others are extremely rare. In some cases, a species is known from only one or two specimens, such as the milliped Euzkadiulus sarensis (Mauriès, 1970) from Grotte de Sare in Pyrenees, and troglobionts of many tropical caves, such as Eostemmiulus coecus Mauriès, Golovatch & Geoffroy, 2010 from Hang Mo So in Vietnam. Distribution disjunction may be another weighting option, with species of disjunct geographical distribution being given a greater weight [96]. Related to this is the possibility of providing extra weighting to type localities. A fifth weighting is that of phylogenetic distinctness. The subterranean fauna is replete with examples of monotypic genera and even supra-generic taxa. Among these are Glacicavicolini from the USA, with its unique species, Glacicavicola bathyscioides Westcott, 1968 (Coleoptera), or the Collembola Bessoniella procera Deharveng & Thibaud, 1989, the only species of the subfamily Bessoniellinae, known from a few caves of a small Pyrenean massif. More generally, a proxy of phylogenetic distinctness is the number of supra-specific taxa. A sixth possible weighting is to give species with extreme morphological modifications (troglomorphy of Christiansen [53]) more weighting than a less modified species, as suggested by Gallão & Bichuette [97].

With the above weighting schemes, there is an implicit weighting of species richness for the simple reason that more species will result in higher weights for richer sites; however, there are exceptions. If only single-site endemics are given any weight, then the result is that more isolated sites are weighted higher, and richer caves in larger areas, but with low endemism, are given lower weights. The same may happen if phyletically isolated species, which are often geographically isolated, are given more weight. Valuing geographic or phyletic isolation makes sense biologically, but it might be more effective to consider species richness and this kind of weighting separately than in combination.

Simple species count is the approach that was adopted in the different papers of the Special Issues of Diversity. It is classically used to evaluate site biological diversity, especially for conservation purpose, as done for instance at a large scale in Europe [142] or Brazil [143]. The different options of species weighting mentioned above would probably rank subterranean hotspots differently than the basic approach based on species count. It is obviously an important issue for site selection in a conservation perspective, that would deserve deeper investigations.

6.3. Vulnerabilities and Threats

It has been argued that subterranean organisms, as a consequence of less variable environmental conditions in the subterranean realm, are more susceptible to many types of environmental perturbation, including global warming [144], and the tone of many papers is that subterranean habitats and their fauna are delicate and vulnerable [145].

Factual evidence on the vulnerability of aquatic fauna is ambiguous and generally lacking [146], but Mammola et al. [147] provide at least a scenario for considering the effect of climate change on subterranean fauna. Subterranean fauna is certainly vulnerable to climate change, but this threat is not as immediate as some others, with regard to causing extinction in the short term. There is a growing body of evidence for short-term variations in subterranean microclimate, including daily and annual temperature cycles in caves [148], yet it is remarkable how little is known about long-term fluctuations in temperature in subterranean habitats, including changes over the last few decades [149,150].

Several authors have listed the threats and vulnerabilities of subterranean fauna [51,151]. They vary from site to site and from species to species. For example, the complete destruction of limestone hills for cement production in Vietnam is an immediate and irreversible threat to all subterranean communities, that is leading several microendemic species to extinction [45]. Hotspot sites that have been documented in these two Special Issues of Diversity are likewise vulnerable due their particularly rich endemic fauna, and call for vigilance. Their vulnerability is dependent on the size and isolation of karst or hydrogeological units available for the cave-restricted fauna. In this regard, tropical tower karsts with hills scattered on non-limestone terrain are especially vulnerable to local limestone exploitation [46,151]. Additionally, cave-restricted species are vulnerable to a variety of environmental perturbations, at an extent that is unknown in most cases. Danielopol & Marmonier [81] demonstrated however that some groundwater crustacean species, such as Proasellus slavus (Remy, 1948), are “regulators”, able to maintain more or less unaltered activity independently of variable environmental conditions. More generally, cave-restricted species are clearly sensitive to desiccation, but contrary to what is sometimes stated in the literature, many seem to be able to cope with large ranges of temperatures [84].

Cave animals may be more fragile at the individual level than most surface organisms in the face of certain chemical or climatic disturbances [152]. However, they are usually protected in cave habitats from the most severe disturbances that affect surface habitats [153], including, to some extent, from climate warming. It may be the reason why cave communities subsist almost unaltered in regions where surface environment and fauna have been severely degraded by deforestation or other land-use changes. This certainly also occurred at the geological scale, and may explain why the proportion of relictual taxa of various ages is much higher in subterranean than in surface habitats, as can be easily inferred from biodiversity inventories involving cave and non-cave species [154].

6.4. Protection Strategies

There are a number of protection strategies available for subterranean habitats, the most prominent of which are site protection, acquisition by government agencies (e.g., Mammoth Cave National Park), and legislation (e.g., European Habitat Directive). Site protection of course depends on whether subsurface sites are explicitly protected or, in fact, inadvertently protected. For example, many caves in the mountains of Montana are protected since they are located in the Bob Marshall Wilderness Complex of Flathead National Forest. The Lez aquifer is protected because the aquifer is used for the water supply of the municipality of Montpelier. The overall efficacy of protection varies from site to site and country to country, and is dependent on resources committed to education and to protection enforcement. It is worth noting that several sites on the hotspot list are on government-owned land. They include the following:

• Fern Cave (National Wildlife Refuge);
• Igatu Cave System (Chapada Diamantina National Park);
• Lukina Jama–Trojama Cave System (Velebit National Park);
• Mammoth Cave (National Park);
• Ojo Guareña System (National Monument);
• Movile Cave (owned by the municipality of Mangalia);
• Tham Chiang Dao (Chiang Dao Wildlife Sanctuary);
• Towakkalak System (Bantimurung-Bulu Saraung National Park);
• Vjetrenica Cave System (owned by the municipality of Ravno).

Other forms of cave protection involve conservation through private ownership and show caves. In the United States and elsewhere, non-governmental organizations have been created for the purpose of protecting caves by acquiring cave entrances and surrounding properties. The largest such organization in the U.S., the Southeastern Cave Conservancy, owns 32 cave preserves with more than 170 caves. Show caves, such as Križna Jama, Tham Chiang Dao, and Cueva del Viento, are protected because of the commercial value of the intact cave.

Few if any of these protections are complete. Portions of the aquifer and cave may be outside the protected area, or devoted to touristic visit, and protection strategies themselves are often inadequate. Some of the most recent examples come from the strategies to protect cave-dwelling bats. A common technique for bat protection is the creation of a gate at the cave entrance to prevent human access. However, some bat species, such as Myotis grisescens (Howell, 1909), the gray bat, are sensitive to gates and have difficulty passing through, making them more vulnerable to predators such as snakes and owls [155]. The species Miniopterus schreibersi Kuhl, 1817, emblematic of conservation efforts in Europe, is also gate-sensitive. The effectiveness of gating caves for bat protection was determined to be inconclusive based on a meta-analysis of 21 case studies [156]. Gates placed externally to the cave entrance are in any case less of a deterrent to bats. Some gates also restrict the access of other small mammals, impeding the flow of organic matter into caves, and thus negatively impacting the terrestrial cave community. Rigid general rules of protection seem to be ineffective; instead, a demonstration of real risks and solutions is required.