Introduction

The uplift of the Qinghai-Tibet Plateau (QTP) dramatically changed the topography of Asia, and the direction of flow of the Yellow River system from west to east is closely tied to this uplift. The Yellow River appeared about 1.7 Ma during phase C of the Qingzang Movement1. 1.2 Ma before present, there was substantial uplift of the Tibetan Plateau2, and tectonic movement of the Kunlun-Yellow River caused the Yellow River to cut the Jishi Gorge and flow into the Linxia-Lanzhou basin3. At that time, the Yellow River basin was only composed of some disconnected lakes4. The uplift of the Tibetan Plateau and river erosion, especially during the Gonghe Movement (0.15 Ma) in the late Pleistocene, greatly promoted the connection of these lakes causing the retrogressive erosion of the Yellow River and downcutting of the river valley in Sanmen Gorge5. Consequently, the Yellow River evolved into a long river flowing from west to east6. The river might have acted as an abiotic physical barrier to gene flow7,8, or it may have geographically facilitated movement and channels of dispersal, which has been observed in Rosa roxburghii9, Rhododendron ripense10 and Terminalia franchetii11, for which frequent gene flow resulted in the genetic homogenization of populations.

The Pleistocene was a time during which dramatic climatic and temperature shifts occurred12. This was especially true during the Mid-Pleistocene Transition (MPT), which lasted from approximately 1.25 million to 700 thousand years ago, and was a period during which the total amount of ice in ice sheets increased globally13,14. The climatic fluctuation in China during the Pleistocene, although not as severe as that in Europe and North America, still occurred during the alternation between glacial and interglacial periods15,16. In China, loess sequences in central China indicated that major shifts occurred toward cooler and drier climates at about 2.4 Ma, 1.2 Ma, and 0.5 Ma, while less pronounced shifts were dated to 1.65 Ma, 0.8 Ma, and 0.2 Ma17,18. Furthermore, marine δ18O records also showed similar shifts19. Three ice core records from the Qinghai-Tibet Plateau revealed that late glacial stage conditions were apparently colder, wetter, and dustier than Holocene conditions20. Additionally, lacustrine sporopollen records indicated that wetter conditions progressed from western China to the east during the early Pleistocene21. These climatic changes can influence the distribution of species because the distributions of some species migrated during glacial periods, and the species survived in refugia, after which many species then re-colonized some areas during the postglacial period22,23,24; alternatively, other species persisted and adapted to the changed conditions25,26, while others became extinct27.

Tamarix austromongolica and Tamarix chinensis are endemic tree species in China28; T. austromongolica is naturally distributed in the upper reaches of the Yellow River, from Qinghai to the ravine region between Shanxi and Shaanxi provinces; T. chinensis, however, mainly occurs in the lower reaches of the Yellow River. The most obvious features of these species include drought, saline, and alkaline tolerance; they are also water tolerant. The small, short-lived seeds in particular require a moist surface soil, short-term precipitation, or may drift with water in the early stage of germination29, which is different from other desert shrubs. As two species that are widely distributed along the coast of the entire Yellow River, T. austromongolica and T. chinensis are very closely related species30. Furthermore, Tamarix is an ancient genus occurring in ancient times in the Mediterranean region31. Therefore, these two species can serve as good candidates for investigating the influence of the Yellow River and paleoclimate on patterns of species distribution.

In this study, we sampled 45 populations along the Yellow River to infer the effects of geological events and climate change during the Pleistocene on two Tamarix species. To better understand the distribution of these two species, we evaluated the following: (1) the relationship between the divergence time of chlorotypes and the paleoclimate, formation, and evolution of the Yellow River; (2) the demographic history of two Tamarix species using genetic data and species distribution models.

Results

Geographic distribution and network structure of haplotypes

Two cpDNA fragments from a total of 382 individuals and nDNA ITS fragments of 45 populations and two out-group individuals were sequenced. The total length of cpDNA sequences was 1640 bp, with 11 chlorotypes identified by nine nucleotide substitution sites and two indels. These 11 chlorotype sequences were submitted to the GenBank database under accession numbers KY621817-KY621842. Chlorotype C10 was unique to population GYJ. The Chlorotype C1 was shared by all of populations, which indicates high migration rates among populations. Interestingly, C2 was found in populations GLZ and SDH, in different species and at different geographical distance between the two populations. The details of chlorotype distribution for each locality are shown in Fig. 1a.

Figure 1
figure 1

Detailed sampling locality information, chlorotype distribution (a), and ribotype distribution (b) of T. austromongolica (solid circles) and T. chinensis (dotted circles). The color of the pie chart corresponds to the haplotype in the median-joining network. The size of the circles corresponds to the frequency of each haplotype. The base map was downloaded from DIVA-GIS (http://www.diva-gis.org/gdata). The figure was drawn using Diva-GIS v7.5.0 (http://www.diva-gis.org/) and Adobe Illustrator CS5 v15.1.0 (Adobe Systems, Inc.).

Full size image

Based on chlorotypes, we constructed a Median-joining network. One chlorotype (C1) was found in 363 of 382 individuals from the 45 populations and represented the central node of this network; the networks containing clades with characteristic star-like topologies showed that populations that have experienced population expansion. The aligned sequences of ITS in 45 populations were 686 bp; 56 ribotypes (see Fig. 1b) were found among 363 individuals (a total of 726 alleles). Ribotypes R2, R8, R15, R11, R10, and R16 contained a large number of haplotypes; the sequence polymorphisms detected in cpDNA and ITS regions are shown in Supplementary Tables S1, S2.

Genetic diversity and genetic structure

The haplotype and nucleotide diversities inferred from cpDNA sequences show that a low level of variation existed among the populations. Haplotype diversity, Hd, ranged from 0 to1; and nucleotide diversity, π, within populations ranged from 0 to 0.61 × 10−3 (Table 1).

Table 1 Geographical and genetic information inferred from cpDNA sequences for the sampled populations of Tamarix austromongolica and Tamarix chinensis.
Full size table

Both chlorotypes and nuclear ribotypes failed to reflect obvious phylogeographic structure using the STRUCTURE program. CpDNA analyses of molecular variance (AMOVA) indicated that only 0.80% of this variation was observed among T. austromongolica and T. chinensis (Table 2), whereas 96.26% of the variance was intra-populational. For ITS dataset, the AMOVA showed that 75.46% of the variance was within populations.

Table 2 Hierarchical analysis of molecular variance on Tamarix austromongolica and Tamarix chinensis. Note: d.f., degrees of freedom.
Full size table

Time of divergence for chlorotypes

In the time-calibrated chlorotype tree (Fig. 2), all chlorotypes diverged during the Pleistocene. The BEAST analyses placed the origin of T. austromongolica and T. chinensis at 1.85 Ma (95% HPD = 0.70–3.23), which was observed in the GYJ population in this study, suggesting that plants of the genus Tamarix first reached Gansu when migrating from a secondary differentiation center of Xinjiang. The divergence time of C5 (found in NMDK, NMWH, and NXYC) was 1.23 Ma (95% HPD = 0.45–2.13). Chlorotype C1 is the most widely distributed in the Yellow River basin, and its divergence was 0.19 Ma (95% HPD = 0–0.54), which is approximately consistent with the time of MIS 5e-6 and formation of the modern Yellow River, suggesting that climate fluctuations promoted the divergence of chlorotypes of Tamarix.

Figure 2
figure 2

(a) Tamarix austromongolica at site SHQ; (b) Tamarix chinensis at site HFJ; (c) BEAST-generated maximum clade credibility tree of 11 chlorotypes. The length of the blue bars represents 95% highest posterior density, and posterior probabilities are given above the main branches.

Full size image

Demographic analyses

Neutrality tests detected that populations of T. austromongolica have experienced recent demographic expansion based on cpDNA sequences (Table 3). In general, the significantly negative Tajima’s D and Fu’s F values were interpreted as a signal of purifying selection or alternately as demographic expansion. The values of Tajima’s D and Fu’s F inferred from nDNA were negative but statistically insignificant.

Table 3 Demographic estimates for all samples of Tamarix austromongolica and Tamarix chinensis. ns: not significant, *P < 0.05. Note: SSD, sum of squares deviations.
Full size table

The mismatch distributions of both cpDNA and nDNA revealed the occurrence of an expansion of the historical population (Table 3). Sum of squares deviations (SSD) tested the validity of a sudden expansion model based on the SSDs between the observed and expected mismatch. Non-significant values for SSDs indicated that the data did not deviate from that expected under the model of expansion. The raggedness index was calculated similarly, and insignificant raggedness values also indicated populations of both species experienced expansion.

The Extended Bayesian Skyline Plot (EBSP) of T. austromongolica cpDNA indicated a continuous demographic expansion that began around 0.10 Ma (Fig. 3a), and the historical population trend inferred from ITS also showed that T. austromongolica experienced a long period of steady and sustainable growth in its populations (Fig. 3c). The EBSP of T. chinensis cpDNA showed an initial period of stable size followed by an abrupt period of growth in effective population size (Fig. 3b); this change of the effective population size of T. chinensis was also observed in the ITS data sets (Fig. 3d).

Figure 3
figure 3

Past demographic history based on (a) samples of cpDNA from Tamarix austromongolica; (b) samples of cpDNA from Tamarix chinensis; (c) ITS samples from T. austromongolica; and (d) ITS samples from T. chinensis estimated using extended Bayesian skyline plots. In each plot, the x-axis represents the time before present (Ma) and the y-axis represents the effective population size. The dashed and solid lines indicate the median and 95% highest posterior density intervals, respectively.

Full size image

Evolutionary path of T. austromongolica and T. chinensis along the Yellow River

Considering the three scenarios tested with cpDNA sequences, DIYABC analysis software indicated that scenario 2 was the best-supported scenario (Fig. 4), with a posterior probability value higher than those of the other two scenarios (Supplementary Fig. S1). Scenario 2 indicated that Tamarix plants in the lower reaches of the Yellow River came from the upstream giving this scenario more support.

Figure 4
figure 4

Scenarios of Tamarix austromongolica and Tamarix chinensis divergence. (a) scenario 1; (b) scenario 2; (c) scenario 3. Comparison of the scenarios was implemented using the DIYABC software. The map was drawn using Adobe Illustrator CS5 v15.1.0.

Full size image

Species distribution modeling

The model algorithm provided the best results based on the area under the receiver operating characteristic curve (AUC) values; AUC values for these curves varied from 0.941 to 0.971. For both species, the most suitable area climatically (0.54 < P < 1) from the current prediction (601,927 km2) was obviously larger than that from the last glacial maximum (LGM) prediction (564,288 km2) and the last interglacial (LIG, 347,475 km2) predictions; similar patterns were identified for all suitable areas (P > 0.05) (Fig. 5a,b and c). Obviously, climatically suitable areas from prediction expanded during the Quaternary period.

Figure 5
figure 5

Predictions of suitable habitats for Tamarix austromongolica and Tamarix chinensis based on ecological niche modeling using MaxEnt. Predicted distributions are shown for (a) the present time, (b) the last glacial maximum period, about 22,000 years ago, and (c) the last interglacial period, 120,000–140,000 years ago. The base map was downloaded from DIVA-GIS. We reclassified the values in the model output as five adaptability levels; high and low values indicate that the conditions are suitable and unsuitable for the species to occur, respectively. The map was drawn using Diva-GIS v7.5.0 and Adobe Illustrator CS5 v15.1.0.

Full size image

Results from the analysis of variable contributions indicated that three environmental variables, namely, mean temperature of coldest quarter, isothermality, and precipitation of wettest month, were the main climatic factors affecting the current distribution of the two Tamarix species analyzed here; the percentages of their contributions were 26.8%, 16.7% and 16.4%, respectively. This finding showed that wet habitats and temperate zones were probably the main factors that limit the distribution of Tamarix species over time when compared with the effects of other variables.

Discussion

The most important factors that shape the distribution and demography of populations are natural increases in population size (the number of births minus the number of deaths), net migration, and environmental factors, such as water supply and climate conditions32,33,34. T. austromongolica and T. chinensis are perennial species with strong tolerance and adaptability; these species blossom and bear fruit 2–3 years after germination. Therefore, both species produce numerous seeds in spring, summer, and autumn. Population growth rates are also influenced by the rates of migration; the small seeds of Tamarix species have white hairs on one end that enable long-distance wind or water dispersal35, resulting in rapid increases in populations and dispersal along the entire river if environmental conditions are suitable.

The divergence time of C10 in the GYJ population was dated to 1.85 Ma. The paleo-magnetic dating of Liupanshan loess, near the GYJ population, showed the first grade Yellow River terraces reached to 1.8 Ma36, and the divergence time of C5 (1.23 Ma) coincided with the Kunlun-Yellow River Tectonic Movement. Therefore, our data provide biological evidence for the approximate geological age of the Yellow River. Based on the divergence time of chlorotype C1 (0.19 Ma), the demographic histories of T. austromongolica and T. chinensis could be traced back to 0.15 Ma and 0.02 Ma, respectively. As mentioned earlier, the Gonghe Movement started from around 0.15 Ma, which resulted in a substantial uplift of the QTP and promoted the formation of modern Yellow River37,38; therefore, it appears that the expansion of these two species was closely matched to the formation of the modern Yellow River.

It is not easy to interpret why the expansion of T. chinensis began around 0.02 Ma. T. austromongolica and T. chinensis are morphologically similar, and both species flower from May to October. Previous studies showed that these two species are closely related phylogenetically30,31; therefore, we speculate perhaps T. chinensis is a species that evolved and diverged from T. austromongolica along the Yellow River, which could explain why the species expanded relatively late. Overall, the valleys of the Yellow River are not only known as cradles of China’s ancient civilization but also provide moist habitat for seed germination and plant growth. Clearly, the Yellow River has been revealed to be the main driving force for demographic expansion of T. austromongolica and T. chinensis.

These two species experienced demographic expansion during the late Pleistocene, which was a period of multiple glacial-interglacial cycles and dramatic environmental changes. Numerous studies have suggested that environmental changes associated with the Pleistocene climate (e.g., sea-level fluctuations) played roles in determining both the origin and distribution of living organisms39, such as mammals40, other vertebrates41, insects42, and plants43; however, the responses of different species to these changes have been diverse and depended on their habitat suitability and environmental requirements44. For example, neotropical savanna tree species experienced the greatest expansion during the LIG and a retraction during the LGM45. Two tropical tree species, Erythrophleum ivorense and E. suaveolens, experienced demographic bottlenecks during the last glacial period46, whereas a temperate tree frog species, Hyla sarda, underwent range expansion mostly during the last glacial phase47.

T. austromongolica and T. chinensis are temperate species endemic to China; although this region was not directly affected by extensive ice sheets, it also experienced severe climatic change throughout the Quaternary. Also, during the LGM, the low thermal conditions in the Eurasian continent produced temperatures that were 4 °C–16 °C lower than today48. These low temperatures affected the distribution and evolution of plants49,50, such as the temperate tree Pteroceltis tatarinowii51 and the cool-temperate deciduous tree Quercus mongolica52, which retreated southward and then re-colonized the previously northern region post-glacially. Ostryopsis davidiana, which is a temperate deciduous shrub species in northern China, maintained multiple refugia in the northern and southern regions of the Qing Mountains rather than having survived only in the south53.

Unlike previous studies, we found that T. austromongolica and T. chinensis experienced demographic expansion during the late Pleistocene. The EBSP analysis and predictions of paleo-distribution models under past climatic conditions consistently suggested that the populations of these two species not only survived but also experienced a period of range expansion. One interpretation of this is that Tamarix is a temperate deciduous species and shows a strong tolerance to cold and drought54,55, with killing temperatures for T. ramosissima, T. chinensis, and hybrids ranging from −33 °C to −47 °C56; this type of adaptation was probably important during past range expansions of these populations.

Our genetic data showed that the haplotype diversity value for the entire dataset averaged 0.11, while nucleotide diversity averaged 0.057 × 10−3; meanwhile, Hd of the GLZ and NMDK sites were 0.56 and 0.60, respectively, and the populations from the lower reaches seemed to have lower values. Population genetic theory predicts that colonization can result in a decrease in genetic diversity, known as the founder effect, and there is a loss of diversity in newly established populations due to a small number of founders57,58,59.

Various researchers have experimentally investigated the effects of recent colonization events on genetic diversity in both herbs and birds. Recently, D’Andrea et al. studied the molecular biogeography of Lactuca serriola. They found that significantly lower genetic diversity characterized the newly colonized parts of the range of this species when compared with historical populations; this confirmed the importance of founder effects60. Similar results have been reported in Geranium carolinianum; genetic diversity patterns across China have revealed that reduced diversity has resulted from successive founder events during range expansion61. However, in contrast, long-distance dispersal maximized the evolutionary potential for invasive European starlings (Sturnus vulgaris) because multiple introductions associated with successful dispersal strategies may lead to relatively high genetic diversity, especially when introductions occurred from different regions62.

Taken together, different introductions and dispersal strategies contributed different amounts of genetic diversity through the introduction of different numbers of individuals or individuals carrying different amounts of diversity. It is well known that genetic diversity is linked to adaptive potential63; low genetic diversity in introduced populations may lower their abilities to adapt and hence increase the risk of extinction in novel environments. Thus, to maintain adaptive potential and minimize the risk of extinction, the founder populations with relatively high levels of genetic diversity need to be conserved.

Conclusions

The main goal of the present study was to understand how the two widespread East Asian temperate deciduous tree species, T. austromongolica and T. chinensis, responded to climatic fluctuations in the Pleistocene and prehistorical geological events affecting the Yellow River. Based on a wide geographic range of samples and by combining information on cpDNA and nDNA sequences, we dated the divergence time of chlorotypes to the Pleistocene and reconstructed the demographic expansion history of T. austromongolica and T. chinensis throughout the Yellow River valley based on EBSP, a distribution model, and ABC analyses. The results suggested that the populations of these two species experienced a period of range expansion. Additionally, the occurrence of the expansion of these two species was closely matched to the formation of the modern Yellow River.

Materials and Methods

Population sampling

A total of 382 individuals were sampled from 45 populations including 147 individuals of 21 populations of T. austromongolica and 235 individuals in 24 populations of T. chinensis, during September 2015 to September 2016, across eight provinces along the Yellow River. Of these locations, SFY and SWS were located along the Fenhe River, the second largest tributary of the Yellow River. Additionally, in Shandong Province, T. chinensis is common in alkaline and saline soil areas along the coasts of the Bohai Sea; as a result, the sample points here are denser geographically than those of other samples. Moreover, two individuals of Myricaria bracteata and Myricaria paniculata were collected in Gansu and Qinghai, for use as outgroups in the analysis. Samples of fresh leaves were dried using silica gel; the location of each population, including longitude, latitude and elevation, were recorded with a GPS unit. Table 1 provides detailed information on the sample locations. The voucher specimens were deposited in the herbarium of the Ecological Laboratory of Henan Agricultural University (HEAC).

DNA extraction, PCR amplification, and sequencing

Total genomic DNA was extracted using a Plant Genomic DNA Extraction Kit (Tiangen, Beijing, China) according to the manufacturer’s protocol and stored at −20 °C. TrnL-F64, rps1665 and ITS66 were selected; DNA amplifications were performed in a gradient PCR system (Biometra, Germany) with the following cycling conditions: 95 °C (5 min); 38 cycles of 95 °C (30 s), 56 °C (30 s), 72 °C (1 min 50 s), and then 72 °C (10 min); conditions only differed based on the annealing temperatures (58 °C for rps 16). We carried out PCR amplification in 20 μL reaction volumes for each individual; PCR products were checked on 1.5% agarose gels and purified with the TaKaRa MiniBEST Agarose Gel DNA Extraction Kit (Dalian, China); all PCR products were subsequently sequenced on an ABI 3730 DNA Sequence Analyzer at the Beijing Genomics Institute (Beijing, China), sequencing with forward and reverse primers in all individuals. Sequences trnL-F (KJ729796.1) and rps16 (KJ729745.1) of Reaumuria soongarica were downloaded from the US National Center for Biotechnology Information (http://www.ncbi.nlm.nih.gov/).

Nucleotide diversity and genealogy

Sequence contig was done using Seqman with the DNAstar program (DNAstar, Madison, WI, USA); all the variance sites were carefully checked optically in Chromas. For nDNA ITS sequences, if double peaks occurred in the same position and the weakest signal reached 1/3 of the strongest signal, we considered the site to be heterozygous, and inferred their phases using the PHASE algorithm in DnaSP 5.1067. Haplotype diversity (Hd) and nucleotide diversity (π) inferred from cpDNA sequences within populations were all done in DnaSP. The genealogical topologies of chlorotypes and ribotypes were constructed using the program Network Version 5.0 with a median-joining model.

Genetic structure

To determine whether the populations were structured, genetic differentiation among and within populations was calculated using STRUCTURE 2.3.468 and Arlequin ver3.569; the significance of AMOVA was tested based on 1000 permutations.

Divergence time of chlorotypes

Regarding the fossil record of Tamarix, Kräusel70 believed that Tamarix records in Egypt were from the Lower Oligocene. Researchers in China71,72 found Tamarix fossils from Gansu Province dated to the Oligocene Baiyanghe Fm and Huoshaogou Fm; all these showed the Tamarix plant appeared before the Oligocene. Tamaricaceae and Frankeniaceae formed the closest sister clade with this node dated to 43–30 Ma73; Zhang et al. estimated an age of about 70 Ma in the light of woody families origin74. Tank et al. suggested an age for this clade of 49.7 Ma75 and ca 53.8 Ma in Magallón et al.76. Ultimately, 53.8 Ma was chosen cautiously as the family root for our molecular dating.

To relate genetic differentiation found among chlorotypes to Pleistocene events, the divergence time was estimated using BEAST Version 2.3.277; the best fit nucleotide substitution model (GTR+G) was selected with Modeltest Version 3.778 in conjunction with PAUP* Version 4.0b1079, Model selection was conducted based on Bayesian Information Criterion. Relaxed clock log-normal was implemented and the mutation rate of 0.9 × 10−9 substitutions per site per year for cpDNA noncoding regions was used to calibrate the tree80. We ran the Markov chain Monte Carlo (MCMC) chain for 10 million generations, with a sampling every 1,000 generations. We used Tracer software to visualize and check for convergence to a stationary distribution and for high effective sampling size values (ESSs > 200); the first 10% of trees were discarded as burn-in in TreeAnnotator Version 2.3.2, and the resulting trees were drawn in FigTree Version 1.4.2.

Demographic history

We used Tajima’s D and Fu’s Fs to infer historical demographic processes. We also conducted mismatch distribution analyses based on both SSDs and the Raggedness index81 with parametric bootstrapping (1000 replicates) to estimate population expansion.

To investigate the potentially relatively complex effects of population size dynamics, we retraced the demographic history of Tamarix with the EBSPs82 in BEAST 2.3.2. We applied a general time-reversible substitution model and strict clock rate 0.9×10−9substitutions per site per year; the weights for EBSP operators and the initial value were adjusted to improve MCMC mixing. We used a scale factor of 0.5 for cpDNA because only the female cpDNA contributes to the effective population size, and then the MCMC chain was run for 10,000,000 iterations.

Evolutionary path of T. austromongolica and T. chinensi

To identify source populations and the colonization patterns of Tamarix along the Yellow River, further analysis of the evolutionary paths of T. austromongolica and T. chinensis was inferred from an ABC using DIYABC software Version 2.1.083. Two groups were defined based on T. austromongolica and T. chinensis species, and three evolutionary scenarios were developed and tested: (a) scenario 1, in which T. austromongolica and T. chinensis of size N1 and N2 have diverged t generations in the past from an ancestral population of size N1 + N2; (b) scenario 2, in which T. austromongolica derived from ancestral population at t2, T. chinensis derived from T. austromongolica at t1; (c) scenario 3, in which T. chinensis derived from ancestral population at t2, T. austromongolica derived from T. chinensi at t1. We generated 3, 000 000 simulated datasets per scenario and chose uniform prior distributions for effective population sizes. To confirm the validity of using ABC to analyze our data, a posterior probability of scenario was assessed and finally obtained the best-supported scenario for comparison.

Species distribution modeling

To predict the area of distribution for T. austromongolica and T. chinensis in the past climate based on habitat suitability, ecological niche modeling was conducted with the Maximum Entropy algorithm (MaxEnt, ver. 3.3.3k)84. Except for our 45 sampling sites, 15 reliable distribution sites were selected for prediction from the National Specimen Information Infrastructure (NSII, www.nsii.org.cn). The latitude and longitude of each T. austromongolica and T. chinensis collection site in MaxEnt are given in Supplementary Table S3. Nineteen bioclimatic variables of current and LGM (about 22ka, 2.5 arc-minute) and LIG (about120–140 ka, 30 arc-second) were obtained from the Worldclim database (www.worldclim.org); these 19 variables included temperature and precipitation and together are considered particularly pertinent to species distributions. The random test percentage was set to 25%, and the maximum number of iterations was set to 1000. To assess the quality of the generated distribution models, a value representing the AUC was performed.