Distribution and habitats of Burnupia trapezoidea (Boettger, 1910) (Gastropoda: Ancylidae) in South Africa

The first comprehensive study of the Ancylidae in South Africa was published by Walker (1923), but the presence of this family in this country was already reported many years earlier (Krauss, 1848). Another important contribution regarding knowledge of this family was made in a monographic survey of the South African non-marine Mollusca (Connolly, 1939). These contributions focused mainly on taxonomic characteristics of the Ancylidae and little information pertaining to the habitat preferences and range of their geographical distribution was made available. Up till the present, 14 species of Burnupia have been reported from southern Africa (Brown, 1994; Appleton 2002; Griffiths et al., 2015). However, Brown (1994) mentions the need for investigating the genetic basis of variation among populations and taxa to understand speciation within the Ancylidae. This view is supported by Appleton (2002), who is of the opinion that revision of this family is needed at both generic and species level. Of all the species of Burnupia currently on record in the database of the National Freshwater Snail Collection (NFSC), a comprehensive account of the distribution and habitat preferences of only two of the most widespread Burnupia species in South Africa, namely B. capensis and B. stenochorias, has yet been published (De Kock and Wolmarans, 2009). The current paper deals with the geographical distribution and habitat preferences of B. trapezoidea, the third most widespread species of this genus in South Africa, as reflected by the records in the database of the NFSC.


INTRODUCTION
The first comprehensive study of the Ancylidae in South Africa was published by Walker (1923), but the presence of this family in this country was already reported many years earlier (Krauss, 1848). Another important contribution regarding knowledge of this family was made in a monographic survey of the South African non-marine Mollusca (Connolly, 1939). These contributions focused mainly on taxonomic characteristics of the Ancylidae and little information pertaining to the habitat preferences and range of their geographical distribution was made available. Up till the present, 14 species of Burnupia have been reported from southern Africa (Brown, 1994;Appleton 2002;Griffiths et al., 2015). However, Brown (1994) mentions the need for investigating the genetic basis of variation among populations and taxa to understand speciation within the Ancylidae. This view is supported by Appleton (2002), who is of the opinion that revision of this family is needed at both generic and species level.
Of all the species of Burnupia currently on record in the database of the National Freshwater Snail Collection (NFSC), a comprehensive account of the distribution and habitat preferences of only two of the most widespread Burnupia species in South Africa, namely B. capensis and B. stenochorias, has yet been published (De Kock and Wolmarans, 2009). The current paper deals with the geographical distribution and habitat preferences of B. trapezoidea, the third most widespread species of this genus in South Africa, as reflected by the records in the database of the NFSC.

MATERIALS AND METHODS
Burnupia species are relatively inconspicuous freshwater limpets. They are usually attached to solid objects on the substratum or marginal and aquatic vegetation in a wide variety of water body types. However, they need to be specifically looked for during surveys. Up until the late 1980s, Health personnel of government institutions and municipalities made substantial contributions to the number of samples in the NFSC. These personnel received special training regarding sampling techniques during workshops organised by staff of the former Snail Research Unit at the Potchefstroom University for Christian Higher Education. All persons involved in these surveys were provided with a printed collector's form compiled by staff of the former Snail Research Unit. The locality of the collection had to be entered and various alternatives regarding the type and conditions of the habitat had to be underlined, according to the judgment of the collector at time of collection. From1990 up until the present, additions to the NFSC have come mainly from collections made during a number of surveys conducted by scientists and students from the Potchefstroom Campus of the North-West University, where the NFSC is housed.
Identification of the specimens was done by following the diagnostic criteria laid down by Walker (1923), Connolly (1939 and Oberholzer and van Eeden (1969), and was based largely on shell and radula characteristics; however, origin of the samples was also taken into account. Details of the biotopes of all samples of B. trapezoidea, dating from 1956 to 2016, which could be located on a 1:250 000 topo-cadastral map series of South Africa, were extracted for analysis from the NFSC database.
The 121 loci ( 1 / 16 square degrees) in which the collection sites were located were distributed in preselected intervals of mean annual air temperature and rainfall to illustrate the frequency of occurrence of this species within specific intervals. Statistica 12, Nonparametrics, 2 x 2 Tables, McNemar Fischer exact software were used to calculate Chi square values to determine the significance in difference between the frequency of occurrence in, on, or at the range of options for each variable, such as type of water body, type of substratum or temperature interval. An effect size (Cohen, 1977) was calculated for all the different variables discussed in this paper. The effect size is an index which measures the degree of discrepancy between the frequency distribution of a given species in the set of alternatives of a given variable such as water bodies, as compared to the frequency of distribution of all other mollusc species in the database in the set of alternatives for the same variable (Cohen, 1977). According to this author, values for this index in the order of 0.1 and 0.3 indicate small to moderate effects respectively, while values in the order of 0.5 and higher indicate practically significant large effects. A more detailed account of the significance and interpretation of specific values calculated for this index in a given situation is given in our earlier publications (De Kock and Wolmarans, 2005a;b).To determine the importance of the role played by each of the variables recorded in the database on the occurrence, numbers and geographical distribution of B. trapezoidea, a Random Forest analysis (Liaw and Wiener, 2002; R Core Team, 2013) was done. The mean decrease in accuracy caused by a variable is determined during the out of bag error calculation phase. The more the accuracy of the Random Forest decreases due to the exclusion (or permutation) of a single variable, the more important that variable is deemed and therefore variables with a large mean decrease in accuracy are more important for classification of the data.

RESULTS
The 445 samples of B. trapezoidea which could be pinpointed on a topo-cadastral map series fell within the 121 loci depicted in Fig. 1.
This species was collected in 10 of the 14 different waterbody types represented in the database. The largest numbers of samples came from rivers and streams; however, the six samples collected in springs represented a slightly higher percentage (1.74%) of the total number of times molluscs were reported from this type of water body in the database ( Table 1).
The frequency of occurrence in rivers and streams did not differ significantly (χ = 1.46, df = 1; p > 0.05); however, Chi square values indicated that the frequency of occurrence in  these two water body types differed significantly (p < 0.05) from most of the other types of water body. An effect size of 0.44, which is deemed as moderate to large, was calculated for water bodies. The majority of samples were collected in perennial habitats for which the water conditions were described as clear and fresh, while equal numbers of samples were reported from slow-running and standing water ( Table 2). The frequency of occurrence in perennial water did not differ significantly from seasonal water (χ = 0.83, df = 1; p > 0.05). A large value was calculated for the effect of current speed, while both perennial and seasonal water, as well as turbidity, had only a small effect and salinity no effect at all ( Table 2). The majority of samples were reported from water bodies with predominantly stony substrata (Table 3) and the frequency of occurrence on this type of substratum differed significantly (p < 0.05) from all other types.
More than 80% of the samples were collected at sites which fell within the interval of mean annual temperature ranging from    (Table 4) and the frequency of occurrence differed significantly from sites falling within the other two intervals in which samples were collected (p > 0.05). More than 70% of the samples were reported from sites falling within the interval of mean annual rainfall ranging from 601 to 900 mm (Table 4). A moderate and a moderate to large effect value were calculated for rainfall and temperature, respectively (Table 4). The Random Forest analysis (Fig. 2) selected water bodies, the presence or absence of marginal and aquatic vegetation, and type of substratum as the most important variables that played a significant role in determining the presence and numbers of specimens of B. trapezoidea in a specific area. In contrast, according to this analysis, temperature had a moderate effect, while current speed and turbidity played only a minor role in this respect.

DISCUSSION
Burnupia trapezoidea was described by Boettger in 1910 from the type locality in Botswana (Brown, 1994), and its occurrence in South Africa was reported 2 years later by Walker (1923  Provinces can most probably be attributed to the paucity of perennial water bodies in these areas and the fact that Burnupia spp. in general have poor abilities to overcome conditions of desiccation prevalent in these types of water body (Brown, 1994). Brown (1994) records this species from Kamanyab in Namibia and Ngamiland in Botswana and is of the opinion that it probably occupies much of the range known for this genus, which is in accordance with the considerable overlap of its distribution with those of B. capensis and B. stenochorias reported by De Kock and Wolmarans (1909). Brown (1994) also mentions that B. trapezoidea is 'widely thought to occur only in Africa, but reports of its presence in the southern part of South America cast doubt on this. ' The large and moderate to large effect values which were calculated for types of water bodies and types of substrata, respectively, suggest that these variables played a significant role in establishing the documented geographical distribution of this species in South Africa, a conclusion also substantiated by the results of the Random Forest analysis (Fig. 2). These findings are also in accordance with the results reported for B. capensis and B. stenochorias by De Kock and Wolmarans (2009). The results of the Random Forest analysis also indicated that temperature and turbidity played a minor role in determining the presence or absence of B. trapezoidea in specific areas, findings also supported by the moderate to small effect values calculated for these two variables, respectively (Tables 2 and 4). Statistical analyses indicated that temperature only played a moderate role in the presence or absence of this species in a specific area, which is in accordance with its geographical distribution which shows that it was collected in aquatic biotopes located in some of both the coldest and warmest areas in South Africa (Fig. 1).
The findings that the majority of samples of B. trapezoidea were collected in rivers and in biotopes with stony substrata is in accordance with the results of a project currently in progress to assess the ecological health of the Marico River (North West Province) and the Crocodile River (North West and Limpopo Provinces), which is being conducted by the authors and postgraduate students of the Potchefstroom Campus of the North-West University (Kemp et al., 2016). During these surveys, many specimens of this species were found on stones in in-current and out-of-current biotopes. However, specimens were also found in aquatic and marginal vegetation biotopes, which also substantiate the results depicted in Fig. 2. The importance of biotope availability for macroinvertebrate communities in South African rivers has already been reported by Dallas (2007) and Odume et al. (2015). The finding that B. trapezoidea can utilise benthic, as well as vegetation biotopes and is widely distributed in South Africa, makes it an excellent candidate for the assessment of the bioavailability of metals in river systems in South Africa. It is known that freshwater molluscs can accumulate metals in their soft tissue and also incorporate various mineral and organic substances in their shells (Carrell et al., 1987;Pourang, 1996;Korniushin, 2000). Laboratory experiments also indicated that B. stenochorias could possibly be exploited as an ecotoxicological indicator in wholeeffluent toxicity testing (Davies-Coleman and Palmer, 2004).
According to Lydeard et al. (2004) the largest number of documented extinctions, out of all taxa, are reported for the molluscs. Furthermore, it is estimated that the rate of extinction of freshwater organisms could be as high as 4% per decade (Dudgeon et al., 2006), and knowledge with regard to the conservation status of South African freshwater molluscs is poor (Herbert, 1998). With regard to its conservation status, B. trapezoidea is listed as 'data deficient' in the revised edition of the ICUN Red Data List, but Seddon et al. (2011) are of the opinion that habitat loss and degradation could pose a threat to Burnupia spp. in general. In view of the fact that members of this genus are sensitive to desiccation and are most frequently found in streams and rivers, global warming could have a negative impact on their survival. A significant decline in species diversity and numbers of gastropods due to climatic warming was indeed reported by Mouthon and Daufresne (2006) in an assessment of the Sâone River in France. However, in the extensive on-going biodiversity survey of the two rivers and selected tributaries in the North-West Province mentioned earlier, many specimens of B. trapezoidea are still being found at new localities (Kemp et al., 2016), and this species could, therefore, probably be considered for listing as not endangered. In view of its relatively wide distribution, largely sessile, benthic lifestyle, and established ability of molluscs to accumulate various substances, including pollutants, in their soft material and shells, it is recommended that the feasibility to utilise B. trapezoidea as an indicator of river ecosystem health in South Africa should be explored.