Patterns of Rotifer Diversity in the Chihuahuan Desert

: Desert aquatic systems are widely separated, lack hydrologic connections, and are subject to drought. However, they provide unique settings to investigate distributional patterns of micrometazoans, including rotifers. Thus, to understand rotifer biodiversity we sampled 236 sites across an array of habitats including rock pools, springs, tanks, ﬂowing waters, playas, lakes, and reservoirs in the Chihuahuan Desert of the USA ( n = 202) and Mexico ( n = 34) over a period of > 20 years. This allowed us to calculate diversity indices and examine geographic patterns in rotifer community composition. Of ~1850 recognized rotifer species, we recorded 246 taxa (~13%), with greatest diversity in springs ( n = 175), lakes ( n = 112), and rock pools ( n = 72). Sampling e ﬀ ort was positively related to observed richness in springs, lakes, rivers, and tanks. Nestedness analyses indicated that rotifers in these sites, and most subsets thereof, were highly nested (support from 4 null models). Distance was positively correlated with species composition dissimilarity on small spatial scales. We predicted species richness for unsampled locations using empirical Bayesian kriging. These ﬁndings provide a better understanding of regional rotifer diversity in aridlands and provide information on potential biodiversity hotspots for aquatic scientists and resource managers.


Introduction
Delineating patterns of species distributions is important for understanding basic and applied questions in biogeography, ecology, and evolutionary biology [1,2]. Species distributions can be used in modeling current communities and in predicting outcomes to both short-term (e.g., acute pollution episodes) and long-term events (e.g., increases in temperature due to climate change). They also inform biogeography and macroecology [3]. Unfortunately, the biogeographic patterns of many small and understudied species have not been well-documented. As members of the Syndermata, rotifers offer a good example of this challenge. While they comprise an important component of freshwater ecosystems and contribute to both the microbial loop and typical aquatic food webs, it is unclear whether their distribution follows ubiquity theory [4,5], or whether they exhibit some level of endemicity [6][7][8][9]. Due to their ability to produce small resting stages that are easily transported by hydrochory [10], were preserved in 4% buffered formalin to view key taxonomic characters. For example, specimens of Lecane and Lepadella were fixed to view characteristics of the lorica, and in some cases trophi were examined using SEM. Keys to the Rotifera used in this study were as follows: Bdelloidea- [81,91,92]; Monogononta- [93][94][95][96][97][98][99][100][101][102][103][104][105]. We identified taxa to species or, if that was not possible, to genus: e.g., Lecane sp. We conducted all of the analyses using the lowest level of identification that we determined. For most specimens, we took voucher images with a SPOT camera and, when possible, voucher specimens were preserved in 70% ethanol and/or 4% buffered formalin. We housed all voucher specimens in UTEP's Biodiversity Collections.

Diversity Indices
To assess diversity of sites we calculated Hill numbers (q) of order 0 (richness, S), 1 (Shannon Index), and 2 (Simpson Index), and Sorensen's Index (SI). Species incidence was characterized at a variety of spatial grains by overlaying 0.1 • , 0.25 • , 1.0 • , 1.25 • , and 2.0 • grids on the site map. We calculated incidence within these grids cell from presence/absence data from each collection site occurring within the boundaries of the grid cell.

Sampling Effort
We tested the relationship between species richness and sampling effort using linear regression in R version 4.0.2 (R Core Team, 2020) for all sites combined, as well as for each habitat type separately.

Indicator Species Identification
We determined indicator species for habitat types by testing for significant associations using the indicspecies package 1.7.8 version in R version 4.0.2 (R Core Team, 2020; https://cran.r-project.org/web/ packages/indicspecies/indicspecies.pdf). This analysis calculates an Indicator Value (IndVal) index to measure the association between species and sites and combinations of sites based on the methods of Dufrene and Legendre [106] and De Caceres et al. [107]. The statistical significance is determined by permutation tests (n = 999).

Nestedness
We tested the hypothesis that smaller assemblages of rotifers are nested subsets of larger assemblages based on the habitats in which they are found by using the algorithms implemented in ANINHADO 3.0 (Bangu) [108][109][110]. In this program, the matrix is rearranged (packed) to achieve the densest grouping of species in the habitats [111]. We employed both the Temperature calculator (T • ) and nestedness metrics based on overlap and decreasing fill (NODF) [109], but because the packing is only marginally different, here we report T • . We tested all packed matrices using the 4 null models described by Guimarães & Guimarães [110]. For comparison purposes we also included a meta-analysis of 11 published datasets of rotifers from other biomes including aridlands (n = 5), cold (n = 2), temperate (n = 2), and tropical regions (n = 2). In our previous nestedness study [90] we determined species or habitats to be idiosyncratic when their individual T • was ≥1 SD than the mean of the matrix T • . Since species and site T • often exhibit large variance, we decided to employ a more rigorous criterion, and here we note idiosyncratic species or habitats when their value is ≥2 SD of the mean of matrix T • .

Relationship between Species Richness and Geographic Distance
To determine whether distances between sites were contributing to differences in species composition, we conducted Mantel tests. Geographic distances between sites were estimated using Haversine distances based on GPS coordinates using the R package geosphere 1.5-10 [112]. Bray-Curtis dissimilarity matrices of species composition were constructed using the vegdist function from the R package vegan 2.5-6 [113]. We used Mantel tests, based on Spearman rank correlations, to determine Diversity 2020, 12, 393 5 of 34 whether species composition was related to (1) geographic distances between collection sites, (2) spatial scale (e.g., grid cells size), and/or (3) habitat type.

Prediction of Biodiversity Hotspots
Based on our survey data, we estimated richness throughout the Chihuahuan Desert using empirical Bayesian kriging [114]. Using kriging as a method to predict species richness in unsampled areas has the benefit of illustrating general trends in richness across broad geographic regions. This process uses a probabilistic predictor that models spatial dependence with functions (i.e., semivariograms). A semivariogram model was estimated from the species richness data we obtained in our surveys, and then used that estimate to simulate the richness in unsampled geographic areas. From these newly simulated data, another semivariogram was estimated and evaluated against previous models using Bayes' rule. This process was iterated (n = 100) and the simulated data were used to predict richness at unsampled locations. Richness values were log-empirically transformed (a multiplicative skewing normal score approximation based on the log of our survey richness data) prior to semivariogram fitting. This process ensures that negative richness values are not predicted. Kriging was conducted on species richness at each site and for each grain size.

Species Composition
We identified 246 rotifer species, which represents a substantial portion of known rotifer species, genera, and families (~13, 50 & 77%, respectively) [17,115]. Given that the Chihuahuan Desert comprises only about 0.35% of the global landmass (excluding the poles), it includes a large percentage of known rotifer biodiversity. Species richness ranged from 1 to 44 at a given locality. The site with the highest richness was Laguna Prieta at HTSPHS (S = 44). This site was sampled >20 times during this study. The site with the second highest richness was Lago Colina located in Chihuahua, Mexico (S = 43), but this site was sampled only four times over a 2-year period. Species found in all habitat types (except rock pools) include Brachionus quadridentatus, Cephalodella catellina, Cephalodella forficula, Cephalodella gibba, Colurella obtusa, Euchlanis dilatata, Lecane bulla, Lecane hamata, Lecane luna, and Platyias quadricornis. Lecane quadridentata was found in all habitats except streams.

Diversity Indices
Of the five most common habitat types, springs had the highest richness (S = 175) while rock pools had the lowest (S = 53) ( Figure 1A). Former cattle tanks also exhibited relatively low diversity (S = 53). In the few rivers (2 rivers, 26 sites) and streams (5 streams, 7 sites), sampled richness was 95 and 26, respectively. When compared to all other sites, springs also had the highest percentage of unique species (34.3%), followed by lakes and tanks (10.5%), playas (9.1%) and finally rock pools (5.7%) ( Table 1). For these systems, Sorensen's Index ranged from 0.36 to 0.54, and most habitats share about 40% of their species (Table 1) with springs and lakes having the most divergent rotifer species communities. Diversity was highest at the largest spatial scale investigated, with the mean diversity for cells at the largest grid size being 48, 35, 27 for q = 0, 1, and 2, respectively. Diversity found for q = 0, 1, and 2 increased at higher spatial grains (r 2 = 0.16, 0.15, 0.12, respectively; p-value < 0.05 for each; Figure 1). The strength of this relationship decreased with increasing Hill number.

Sampling Effort
There was a positive relationship between observed species richness and sampling effort when we included all sites in the analysis, although S is only weakly explained (r 2 = 0.01, p < 0.05; Figure 2). However, when analyzed by habitat type, the relationship was stronger (r 2 = 0.32, 0.17, 0.40, 0.56 for springs, lakes, rivers, and tanks, respectively). Although, in some cases, such as in rock pools, S was weakly explained by sampling effort (r 2 = 0.02, p < 0.05). Playas and streams did not show a significant relationship with sampling effort.

Sampling Effort
There was a positive relationship between observed species richness and sampling effort when we included all sites in the analysis, although S is only weakly explained (r 2 = 0.01, p < 0.05; Figure 2). However, when analyzed by habitat type, the relationship was stronger (r 2 = 0.32, 0.17, 0.40, 0.56 for springs, lakes, rivers, and tanks, respectively). Although, in some cases, such as in rock pools, S was weakly explained by sampling effort (r 2 = 0.02, p < 0.05). Playas and streams did not show a significant relationship with sampling effort. species, T. similis, was present in 24 habitats (~10% of all the sites we studied), including rock pools (n = 18), lakes (n = 4), one pond, and one spring. However, while it also seems to be a rock pool specialist, it was not present in the HTSPHS system. We found T. similis in two rock pool systems of BIBE possessing very different edaphic conditions. In our analysis of the complete dataset several sampling sites (n = 8) were identified as idiosyncratic habitats, but there was no common feature among them: springs (n = 2); lakes and reservoirs (n = 3); ponds (n = 2); cascading pools (n = 1).  Observed species richness (S) as a function of sampling effort in 236 Chihuahuan Desert aquatic sites over 20 years. We shifted some of the data points to reveal their location; some remain obscured by other data points. Lines are linear regressions of the data analyzed separately for each site type. We ranked sampling effort as follows: 1 = 1 sampling event; 2 = 2-5 events; 3 = 6-10 events; 4 = 10-20 events; 5 = >20 events.

Indicator Species Identification
In the indicator species analysis, 144 species were associated with one habitat type, while only 4 species were associated with 6 of the 7 habitat types. Indicator species were identified for all habitat types and some combinations of habitat types ( Table 2). Playas and Lake + Tanks had the most indicator species (n = 5). While two species (C. gibba and L. luna) were indicators of all habitat types except rock pools. Not surprisingly, Hexarthra n. sp. is an indicator species for rock pools. Indicator species with highly significant associations (p < 0.001) include Hexarthra n. sp. with rock pools, Epiphanes brachionus with playa habitats, B. quadridentatus with playa + river + tank habitats, E. dilatata with playa + river + stream + tank habitats, and L. bulla with lake + playa + river + spring + stream habitats. Species that were indicators of combinations of five habitat types include: L. bulla, Philodina megalotrocha, L. luna, and C. gibba.

Nestedness
We evaluated nestedness in rotifers from the 236 Chihuahuan Desert aquatic habitats at several levels: (1) the completed dataset; (2) by habitat type (lakes, playas, tanks, springs, cascading pools, and rock pools); (3) by geospatial scale (0.1 • , 0.25 • , 1.0 • , 1.25 • , and 2.0 • ). As a comparison, we completed a meta-analysis on data from 11 published studies that examined rotifer assemblages from other biomes (see above). We report results of these analyses in Table 3 and summarized them below. Table 3. Comparative statistics of nestedness among selected studies based on presence/absence data of rotifer species. (See Table A1 for an explanation of the sites, including the abbreviations used here.).    The complete dataset exhibited nestedness, with support from 4 null models (p < 0.001). At this scale, only two idiosyncratic species (identified as those with a T • ≥ 2SD above the mean matrix T • = 2.55): Hexarthra n. sp. and Trichocerca similis. Of these two species, Hexarthra n. sp. [89] had the most restrictive distribution. It was confined to a group of 25 isolated rock pools at HTSPHS, indicating that it is a rock pool specialist. (See also the discussion below on rock pools.) The other idiosyncratic species, T. similis, was present in 24 habitats (~10% of all the sites we studied), including rock pools (n = 18), lakes (n = 4), one pond, and one spring. However, while it also seems to be a rock pool specialist, it was not present in the HTSPHS system. We found T. similis in two rock pool systems of BIBE possessing very different edaphic conditions. In our analysis of the complete dataset several sampling sites (n = 8) were identified as idiosyncratic habitats, but there was no common feature among them: springs (n = 2); lakes and reservoirs (n = 3); ponds (n = 2); cascading pools (n = 1).
We also examined a portion of the dataset that included only BIBE habitats in which one pool cascaded into another (n = 40). In that analysis two species (Epiphanes daphnicola and T. similis) and one habitat (a pool surrounded by lush vegetation) possessed idiosyncratic T • . Since the edaphic conditions of these pool habitats are different, we separated them by location (n = 5) to explore whether they exhibited unique species distributions. In the Cattail Spring pools (n = 12) four species (C. obtusa, Lecane pyriformis, Proales cryptopus, and Tripleuchlanis plicata) and one small pool isolated from the main flowage yielded idiosyncratic T • . Surprisingly in Ernst canyon, none of the 16 species or 12 rock pools proved to be idiosyncratic. Tuff canyon pools (n = 6) also possessed no idiosyncratic species and only one idiosyncratic habitat (one small pool). In the rock pool flowage of the Window Trail pools (n = 10 sites) one species (L. pyriformis) and one habitat (a small tinaja nearly filled with small rocks and sediment, surrounded by plants) possessed idiosyncratic T • . The rock pools at HTSPHS yielded no idiosyncratic species. However, as noted above Hexarthra n. sp. was found in all sites except for two artificially enlarged, sheltered rock pools. Those rock pools were also possessed idiosyncratic T • . In a separate study of six artificial rock pools (mesocosms) placed at HTSPHS, only one species (Lecane nana) had an idiosyncratic T • . Interestingly, this species was not found in natural habitats of HTSPHS during our extensive sampling effort (n > 20 for most sites over 20 years).
Nestedness was evident across all five geospatial scales (0.1 • , 0.25 • , 1 • , 1.25 • , and 2.0 • ), with support from 4 null models (P < 0.001) at each scale. A total of 38 idiosyncratic species were identified in the geospatial analysis and of these eight were identified at more than one spatial scale: Brachionus plicatilis; Brachionus variabilis; Cephalodella cf. misgurnus/pachyodon; Euchlanis calpidia; Paradicranophorus sordidus; P. vulgaris; T. similis; and Wulfertia ornata. Ten regions were identified as idiosyncratic across the five geospatial grids. No obvious pattern of habitats emerged from the scale analysis.

Relationship between Species Richness and Geographic Distance
Mantel tests showed a significant correlation between distance and species composition for grid cell sizes below 1.25 • . The effect became progressively larger at smaller grid cell size, being the most substantial at cell size 0.1 • (p = 0.01) and the least significant at the largest grid cell size (2 • ; p = 0.1). Species composition in springs demonstrated no significant correlation with distance at any spatial scale investigated. In contrast, playa species composition showed significant correlations with distance at all grain sizes. Tank composition was significant at all grain sizes with the exception of 0.25 • . All other habitats showed significant correlation at small grain sizes, but little correlation at large grain sizes (See Table 4). Stream sites were too few (n = 3) to adequately assess using Mantel tests, and thus were not analyzed as a separate habitat. Table 4. Mantel correlation coefficients (r) between Haversine geographic distances and Bray-Curtis dissimilarity values for rotifer communities between sites (n) at each grid size investigated. Habitat types were then analyzed separately, with the exception of streams due to low number of samples (n = 3 at grid size 0.1 • ).

Region
Mantel r Statistic P-Value n

Prediction of Biodiversity Hotspots
Generally, patterns of predicted species richness were similar among the spatial scales investigated (Figure 3). At smaller scales, localized hotspots of richness are apparent within the Chihuahuan Desert.
At the site level, 0.1 • and 0.25 • grid cell sizes, predicted species richness was highest in a band spanning from the southern Chihuahuan Desert northward along the western border to the El Paso/Juarez area, and a band spanning from Guadalupe Mountains National Park (TX) to Balmorhea State Park (TX), with low predicted richness along the Rio Grande in this area. When we excluded the site level, a band of high predicted richness exists from Samalayuca across the Rio Grande to Balmorhea State Park, each with localized hotspots (Figure 3B,C). Cuatro Ciénegas showed high richness at most scales ( Figure 3B-D). At grid cell sizes >0.25 • , distinct hotspots are less apparent ( Figure 3D). At these higher scales, local hotspots are more difficult to resolve due to the lower number of grid cells present within the Chihuahuan Desert (n = 24 for 1 • grid cells).
While our choice of collection sites was pragmatic and based on accessibility, sampling >225 diverse habitats over a 20-year period, with many sites visited multiple times, this study comprises an extensive survey. Due to its thorough nature, our analysis of Chihuahuan Desert aquatic systems offers additional insight to the understanding diversity of rotifers in aridlands, and it offers testable predictions regarding the presence of biodiversity hotspots at a regional level.
Among habitats, rotifer species richness was highest in springs (n = 175) and lowest in rock pools (n = 53) followed closely by tanks and playas (n = 57, 66, respectively). This difference in diversity may reflect the relative stability of these habitats in terms of hydroperiod and/or connectivity with other sites. For example, the ephemeral rock pools at HTSPHS are unique in character from all other rocky basins examined in our study. All of the HTSPHS rock pools have nearly identical edaphic conditions, and the Hexarthra found in these pools was identified as a strong indicator species for rock pools ( Table 2). For rotifers, the use of the indicator species concept has been used mostly in regard to water quality [99]; thus, our application is somewhat unique. It should be noted that some species have been highly associated with acidic habitats (e.g., Cephalodella hoodi [156], Cephalodella acidophila [157], Keratella taurocephala [158]), and function as indicators. The five species with significant indicator values associated with a combination of five habitat types (L. bulla, P. megalotrocha, L. luna, and C. gibba) possess wide ecological tolerances. Another implication is that these morphospecies likely represent cryptic species complexes [159,160] (see below).
Locations we identified possessing high predicted richness generally overlap the proposed wetland priority sites for the Chihuahuan Desert [25]. However, we found low richness in the Rio Grande and at aquatic sites in White Sands National Park (NM). Several priority areas were sparsely sampled in our study (i.e., the Apachean and the Meseta central subregions); making the predicted richness within these regions less reliable. However, some unusual outcomes occurred at various spatial scales. At our smallest scale (e.g., site level;) some areas that contain highly sampled locations yielded low overall predicted richness. For example, at HTSPHS large numbers of ephemeral rock pools are in close proximity to more speciose playas such as Laguna Prieta, the site with the highest richness in our survey (n = 44). The low diversity of these rock pools decreased our predicted richness for the entire area at the smallest spatial scale. At the 0.1 • grid size, the low diversity rock pools and high diversity playas of HTSPHS are combined, resulting in a hotspot on the kriging map. We found similar scenarios at Cuatro Ciénegas (Mexico), BIBE (TX) and Bottomless Lakes State Park (NM). At the largest spatial scale (grid size 1 • ), the pattern seemed to be more influenced by sampling intensity.
Of the 17 different ways we examined nestedness in the Chihuahuan Desert sites, only three did not exhibit nestedness. The rock pools of Tuff Canyon had no support from the null models; Window Trail Canyon had support from only two; and the artificial rock pools (mesocosms) had support from only one model. These results are not surprising as the basins within of each of these systems are quite similar: Tuff Canyon (basalt larva and tuff deposits); Window Trail (limestone); Mesocosms (plastic basins filled with artificial pond water). This indicates that, for nestedness to be present, the inclusive habitats must possess environmental heterogeneity, and if nestedness were not present, we would expect the species assembly to be random within the habitats [161,162].
In the 18 ways that we analyzed nestedness in our Chihuahuan Desert dataset, we recorded a large number of species to be idiosyncratic (n = 59;~24%). These species are those, that within the context of the data, contributed disproportionately to the overall matrix temperature; i.e., their occurrence is, therefore, unexpected in that nested group (Table 3). It is notable that most of the idiosyncratic species are generally considered cosmopolitan or having broad environmental tolerances. Our analyses also show that rotifer assemblages are correlated with distance at smaller spatial scales but are more homogenous at the regional level (Table 4). Other papers have reported similar patterns in multiple studies analyzing species assemblages or populations of a single species [147,160,[163][164][165][166][167]. Thus, our results seem to support the Baas Becking Principle-"Everything is everywhere, but, the environment selects"-the ubiquity hypothesis [168]. That is, organisms with small dispersal stages (<1 mm) are easily, and widely, dispersed, but arrival does not necessarily guarantee persistence in a habitat [169].
We know that in rotifers, community structure may result from a combination of their high dispersal capacity and their ability to create resting egg banks [5,170]. These two traits can lead to the monopolization of local habitats if the initial colonization and subsequent production of an egg bank leads to rapid adaptation and then to the exclusion of other species. This construct has been named the monopolization hypothesis [171,172]. Thus, at small spatial scales, monopolization leads to high dissimilarity among sites, as may be the case of rock pools and springs in our study (lowest v. highest species richness). However, the high dispersal capability of rotifers may lead to increasing community similarity at larger spatial scales. In general, community composition of organisms with high dispersal ability are less impacted by geographic distances than those with low capacity. Local edaphic conditions, including the arrival sequence, ultimately selects the composition of assemblages that endures.
At larger spatial scales, a greater degree of habitat heterogeneity is present within each region, resulting in a reduction of assemblage differences among regions because of shared habitat types occurring within the larger geographic areas. We have previously reported that rotifer assemblages are more homogenous at the regional level, thereby supporting the relative cosmopolitan nature of dominant rotifer species [57]. However, there can be significant associations between local environmental parameters and species assemblages [53]. Here we report that Chihuahuan Desert spring assemblages were not correlated with distance at any spatial scale investigated. This may be due to the unique edaphic conditions present in each habitat. This was seen in T. similis, which was found in a series of small to large rock pools lying along an erosional channel of Cretaceous limestone in Ernst canyon (n = 12 sites) [173], as well as in Tuff canyon (n = 6 sites) where the rocks pools are arrayed in a channel of eroded basalt lava and tuff deposits [174].
We note that our estimate of richness is likely underestimated, as we could not identify some specimens to species; this is especially true for the Bdelloidea. In addition, it is well known that many traditional species of rotifers are, in fact, complexes of cryptic species [175,176]. For example, two species common in our samples, E. dilatata and B. plicatilis, are comprised of at least 4 and 15 separate lineages, respectively [145,147]. Two of the four newly described species of the E. dilatata complex occur in the Chihuahuan Desert [147]. During the surveys undertaken for this study, they were all recorded as E. dilatata. Finally, several new species are pending formal description.
Our research identified rotifers that exhibited distribution patterns at two extremes: either widely or narrowly distributed. Five species were widely distributed: i.e., being present in 50 or more of the sites we sampled. These species were E. dilatata, L. bulla, L. luna, L. patella, and P. megalotrocha. The perception in the literature is that species with wide distributions have few specific growth requirements. However, as noted above some of these species may represent cryptic species complexes: E. dilatata [147], L. bulla [56], P. megalotrocha [177], and L. luna (Walsh, unpubl. data). On the other hand, some species were narrowly distributed. In our collections we found 70 species only once (e.g., Asplanchna intermedia, Brachionus rotundiformis, Cephalodella dentata, Filinia limnetica, Synchaeta tremula). These species may possess rigorous requirements for growth, be poor dispersers, and/or poor competitors, in each case restricting their distributions.
In addition, we did not sample all sites evenly. We sampled some sites only once at one station, while we sampled others >20 times and from multiple stations/microhabitats within the waterbody. We showed that for sites at BIBE, increased sampling effort increased the number of species recovered even up to seven collections [88]. Similarly, among all sampled habitat types, sampling effort increased richness found, although this relationship was weakest in rock pools, possibly due to their low diversity.

Conclusions
Understanding the biogeography of rotifers remains an important problem. Indeed, the general perception that they do not have a biogeography remains largely untested. Rousselet was the first to pose this idea; he argued that " . . . the Rotifera enjoy a cosmopolitan distribution which is not limited to continents, but extends to all places on the surface of the earth where suitable conditions prevail" [15]. This view, which presaged that of Baas Becking, had been the prevailing view until challenged by several researchers [4,8,169,178]. Yet a large part of the question of whether rotifers possess a biogeography remains rooted in three issues. (1) There is a rotiferologist effect-that the distribution of rotifers indicates more the distribution of researchers, and the habitats that they survey, than the rotifer species themselves [179]. (2) Currently, there are few venues where researchers can receive training in rotifer taxonomy and identification [180]. Thus, identification is often limited to easily recognized species. (3) Recently researchers have come to the realization that cryptic speciation is widespread within the phylum [145,147,181,182] (see also above). Thus, reports of a species from distant locations that are identified based solely by morphological characters may be insufficient to consider them as identical. Emerging science on cryptic speciation suggests that they may be genetically distinct enough to warrant the designation of separate species. Examples of previously unrecognized morphological and ecological differences in the B. plicatilis complex [145], among other species [159], support this contention. Until these issues are, to a large degree, settled, an adequate test of whether rotifers fit the ubiquity hypothesis is not possible.
Thus, our research effort addresses three important aspects in understanding species distributions and biogeography. We covered a broad geographic range, provided a long-term study, and used repeated sampling of sites. Thus, it is not surprising that our study yielded a large number of species. Supporting our previous study that focused on a smaller geographic region (i.e., BIBE), here, we found that sampling effort was positively correlated with rotifer richness in more permanent habitats (e.g., lakes, springs, rivers) and in anthropogenic tanks. In addition, for some sites our efforts spanned seasons and years. Our predictive maps show that it is probable that additional rotifer species remain undiscovered in the Chihuahuan ecoregion. They also give guidance for focusing efforts, as well as for conservation prioritization. Additional diversity also may be revealed by molecular applications such as DNA sequencing to delineate cryptic species and environmental sequencing of water and sediments to find rare species and/or to sample habitats during desiccated periods. In conjunction with environmental data (e.g., water quality data, land use patterns), our findings also can be used to determine ecological drivers of rotifer species assemblages.