Elemental compositions of lichens from Duolun County, Inner Mongolia, China: Origin, road effect and species difference

To assess the response of lichen elemental compositions to road traffic and species difference in the context of high dust input and anthropogenic emissions, two foliose epiphytic lichens (Phaeophyscia hirtuosa, PHh; Candelaria fibrosa, CAf) were sampled near a road adjacent to Dolon Nor Town (Duolun County, Inner Mongolia, China). Twenty elements (Ba, Ca, Cd, Co, Cr, Cu, Fe, K, Mg, Mn, Mo, Na, Ni, P, Pb, Sb, Sr, Ti, V and Zn) in lichen and surface soil samples were analysed using inductively coupled plasma mass spectrometer (ICP-MS). The results demonstrate that lichen elemental compositions are highly influenced by both their natural environment and anthropogenic input. Windblown dust associated with sand dunes and degraded/desertified steppes represents the predominant source of lichen elements. Road traffic can enhance the lichen elemental burden by increasing the number of soil particles. Anthropogenic emissions from the town and road traffic have also led to the enrichment of Cd and Zn in lichens. PHh was higher than CAf in concentrations of 14 terrigenous metals. Both lichens are applicable to biomonitoring of atmospheric element deposition and, in most cases, yield comparable results.


Concentration correlation.
An UPGMA (unweighted pair-group method with arithmetic means) cluster analysis was performed on a correlation matrix of the LC, CAf and PHh data sets separately. The results show that concentration correlations between lichen elements are well presented by cluster analyses, which preserved most of the pairwise distances between the original unmodeled data points (cophenetic correlation coefficient: LC, 0.896; CAf, 0.921; PHh, 0.919).
Generally, the concentration correlations observed in LC (Fig. 2) were closely similar to those in CAf (Fig. 3a) and PHh (Fig. 3b). In all three data sets, elements can be classified into 5 groups at a correlation similarity of 0.55 (Figs 2 and 3). Cd in group G1, Zn in G2 and Cu in G5 had a poor correlation with other elements. G4 contains K, P and Mg (Figs 2 and 3), which are significantly and positively intercorrelated (K-P: r = 0.856 and p = 0.001 in LC, r = 0.896 and p < 0.001 in CAf, r = 0.793 and p = 0.002 in PHh; Mg-P: r = 0.623 and p = 0.001 in LC, r = 0.778 and p = 0.003 in CAf, r = 0.591 and p = 0.003 in PHh). The other 14 elements belong to G3 (Fig. 2).

Concentration difference. Differences in elemental concentrations between the two lichens (tested by
Paired samples T test, α = 0.05) are presented in Table 1. These results show that PHh was similar to CAf in concentrations for 6 elements (Cd, Cr, K, Mo, P and Zn) but was higher than the latter for the other 14 elements (p < 0.05; Table 1). For brevity and clarity, these differences are also presented in Fig. 2.
Differences in elemental concentrations among the 4 plots (P1, 5-10 m off the road; P2, 100 m; P3, 200 m; and P4, 400 m) were tested by Paired samples T test followed by Bonferroni correction (α = 0.0083), and illustrated using heatmap analysis on the range-standardised concentrations (standardised according to Equation 2). These results show that G3 elements were highest near the road, whereas G4 elements (K, P, and Mg) showed an inverse pattern in all lichen data sets (Figs 2 and 3). The difference was insignificant for 5 metals (Na, V, Ti, Cu, and Zn) and significant for 2 metals (Co and Cr) in all three data sets (Figs 2 and 3). A significant difference was found for  Fig. 2). Because an EF of <5 is often evidence of crustal input 5,32 , all these elements are mainly of crustal origin. This conclusion is supported by the Fe:Ti ratios, which if similar between the lichen and soil samples, indicate entrapment of coarse soil particles in lichen thalli 3,28,35 . Fe and Ti are highly positively correlated in all lichen and SS data sets (all r > 0.979, p < 0.001), and the Fe:Ti ratios are not significantly different between lichens and SS (Table 1). Because a good positive correlation between lichen elements can be evidence of a common origin 11,32,34 , the intimate correlations between G3 metals and between G4 elements in lichens is further evidence of terrigenous origin, although these correlations are also a reflection of similar spatial patterns (Figs 2 and  3). Actually, most G3 and G4 elements, such as Fe, K and Ti, are regarded as of crustal origin in Inner Mongolia 32 , even in the Taihang Mountains that experiences heavy air pollution 31,33,34 . Although P had an EF SS of >10 in the lichens (Table 1), its significant positive correlation with K (r > 0.790, p < 0.005) and Mg (r > 0.590 and p < 0.005) in the three data sets of lichens (Figs 2 and 3) suggests that the P enrichment in lichens is most likely a result of active absorption and/or bio-regulation of this essential nutrient rather than evidence of anthropogenic input. . The broken line in the dendrogram shows a correlation similarity of 0.55. In the Correlation column, "*" denote correlation between lichens are significant at p < 0.05. In the Difference column, "ns" and "s" denote insignificant (p > 0.05) and significant (p ≤ 0.05) difference between lichens, respectively. In the Difference among plots columns, different capitalised letters in a row indicate a significant difference in element concentration among plots. Different background colours in each row are the heatmap showing range-standardised concentrations. EF SS denotes enrichment factor normalised to local surface soil with Fe as a reference element according to Equation 1. P1, P2, P3 and P4 denote plots with a distance from the road of 5-10 m, 100 m, 200 m and 400 m, respectively. Cluster analysis was performed on the raw concentrations (n = 24), and the resultant cophenetic correlation index is 0.896. Concentration correlation and difference between lichen species were performed on raw concentrations using Pearson correlation test and Pairedsamples T test, respectively. Concentration differences between plots were evaluated using Paired-samples T tests, followed by a Bonferroni correction (α = 0.0085). The heatmap was generated using Past 3.14 software (Ø. Hammer, Sep. 2016). All tests were performed on Candelaria fibrosa (n = 12) and Phaeophyscia hirtuosa (n = 12).
The high importance of soil contribution to lichen elemental composition is also inferred by the source of Cu. The poor correlation of Cu with terrigenous elements (G3 and G4 elements; Figs 2 and 3) may be partly explained by the fact that Cu is often associated with industrial activities and road traffic 3,11 . A high EF SS of Cu in lichens is expected because samples were collected within 400 m off the road. However, EF SS of Cu is close to the unity with a low variation (CV <21%; Table 1), indicating a predominant terrigenous input superimposed on the road effects.
The high contribution of soil input is clearly due to deposition of wind-blown soil particles on lichen thalli. It is well known that most steppes have been heavily degraded/desertified during the last several decades in the region. The nearest sand dune complex is only 2.5 km north of the sampling sites (Fig. 1). Surface soil in these ecosystems is vulnerable to wind erosion and serves as an important source of frequent sand storms. Atmospheric deposition of these soil particles is a predominant source of lichen elements, as suggested in our recent study in Xilin River Basin, 200 km north of the study area 32 .
The soil contribution to lichen elemental burden appears to have been influenced by road traffic. Concentrations of G3 metals were highest in plots P1 in all three data sets, although multicomparison analyses show that the difference is insignificant for 3 metals (Na, V and Ti) and significant for only 2 metals (Co and Cr) in all three data sets (Figs 2 and 3). Nevertheless, a decreasing trend in concentration of most terrigenous metals with increasing distance from the road can be observed in the plots 5-200 m off the road. The presence of such a pattern has been shown by many relevant studies 3 and is probably due to deposition of terrigenous particles raised by vehicle traffic on lichen thalli. G4 elements show an inverse pattern, characterised by the highest concentrations in plots P4, although the difference is not significant in the PHh data set. This pattern may be a result of better lichen physiology because the three elements are all essential nutrients for plants.
A comparison with relevant studies shows that soil input has greatly increased lichen elemental concentrations in the study area. Concentrations of all terrigenous elements, barring Ca and Mg, are generally higher than or similar to high literature value ranges from various lichens in diverse ecosystems.  Concentration differences between plots were evaluated using Paired-samples T tests followed by a Bonferroni correction (α = 0.0085). The heatmap was produced using Past 3.14 software (Ø. Hammer, Sep. 2016). CAf refers to Candelaria fibrosa (n = 12). PHh refers to Phaeophyscia hirtuosa (n = 12).
The anthropogenic origin of both Cd and Zn can be inferred from their EF SS > 5.0 (Table 1), which is most likely an indication of a non-crustal or non-local source input 5,32 . The poor correlations of both metals with terrigenous elements also support this inference (Figs 2 and 3). The same result has been observed in a similar ecosystem of the Xilin River Basin, where high EF SS for Cd and Zn and their poor correlation with other elements were attributed to road emissions 32 .
It is known that Cd and Zn are traffic-related metals and often accumulate in lichens near roads 3 . However, a decreasing concentration trend with increasing distance from the road was not observed for Cd and Zn. Actually, concentrations of both metals were rather consistent among plots (Figs 2 and 3) and had a low variability (CV < 11% in all three data sets of lichens; Table 1). This pattern might be due to two sources of input superimposed on the wind-blown soil particle deposition. First, these metals may be emitted from vehicles with a dispersion distance exceeding 400 m, as shown by the enrichment of both metals in lichens 10 km off a highway in the Xilinhot River Basin 32 . Another possible source may be coal combustion in Dolon Nor town, 5 km north of the study sites. The town had a population of approximately 30,000 inhabitants at the time of collection. Coal has been used for many years for domestic heating during the seven colder months (from Oct. 1 to May 1), and released metals could be transported to the lichen thalli with prevailing southward wind in these months. The coal chemical industry (12 km NE of the study sites) and railway (established in 2008, for transportation of coal to industry) are unlikely significant contributors because the industrial operation was not started until September, 2011.
A comparison with other studies shows that Cd and Zn concentrations in our lichens are at the high literature value range or 2-6-fold documented concentrations 1, 11-28, 35, 37-40 . Only a handful of studies, mostly conducted in industrial or urban regions, have documented higher concentrations of Cd and Zn, which are often 1.5-8-fold our data [7][8][9][10] . These results suggest a high anthropogenic input in the region.

Species difference.
It is known that elemental composition in lichens are species-and element-specific 3,32,33 .
This pattern is shown in our study. An insignificant difference between CAf and PHh for concentrations of 6 elements (Cd, Cr, K, Mo, P, and Zn; Table 1, Fig. 2) suggests that the two lichens are comparable in monitoring these elements. Concentrations of the other 14 metals were higher in PHh than in CAf (p < 0.05; Table 1, Fig. 2). A lack of good positive correlation between CAf and PHh in concentration of 4 metals (Na, Ba, V and Ti; r < 0.50, p > 0.05; Fig. 2) indicates that the two lichens are not substitutes for each other in monitoring atmospheric deposition of these metals. In contrast, concentration of 10 metals (Ca, Co, Cu, Fe, Mg, Mn, Ni, Pb, Sb, and Sr) in one lichen can be calculated from the other by linear regression analysis, as indicated by the high positive correlation between the two lichens (r ≥ 0.65, p < 0.01; Fig. 2).
Our results show that lichen elemental compositions in the study area are the result of both natural occurrence and anthropogenic input. Wind-blown dust represents a predominant source of lichen elements, and road traffic can enhance lichen elemental burden by increasing emission of soil particles. Anthropogenic emissions from the town and road traffic have also led to the enrichment of Cd and Zn in lichens. Although PHh was higher than CAf in concentrations of 14 terrigenous metals, both lichens are applicable in biomonitoring atmospheric element deposition and in most cases yield comparable results.

Methods
Study area. The study was conducted in Duolun County, Inner Mongolia, North China. Duolun County is an agro-pastoral ecotone, 200 km north of Beijing (Fig. 1). Dolon Nor town (42°12′N, 116°29′E) is the county seat of Duolun County, with a population of approximately 30,000 inhabitants in 2011. The climate is typical of the Inner Mongolian Plateau with low annual precipitation (380 mm, mostly occurring as rainfall during May-September) and drought and windy episodes during the winter-spring season. Prevailing winds in the area are from NW, W and N in winter and from SW and S in summer. The predominant ecosystems are steppes and sand dunes. Sand dust and fuel combustion are two main factors influencing air quality in the area. Dust storms have been frequent due to steppe degradation and desert expansion in the last several decades. A large amount of fuel has been used for traffic (primarily gasoline) and heating (primarily coal) as the heating season lasts about seven months (from Oct. 1 to May 1). Chemical fertilisers and pesticides have been rarely used, and range fires have been very rare. There were very low, if any, industrial activities prior to 2011. A coal chemical industry was established in 2011, and a railway for transportation of coal to support industry was established in 2008. However, the industrial operation of the coal chemical industry was not started until September 2011.
In August 18, 2011, samples were collected in a Populus forest. The forest is 5 km south of Dolon Nor town, 2.5 km south of a sand dune complex, and 12 km SW of coal chemical industry (Fig. 1). The forest was dominated by P. cathayana and P. davidiana, planted during 1980's and 1990's. The terrain is rather flat with an elevation of 1,268-1,274 m.
Sampling strategy. Three sampling lines, perpendicular to a north-south provincial road crossing the forest, were set at an interval of 800 m. At each sampling line, 4 plots 10 × 10 m 2 in size at a distance from the road of 5-10, 100, 200 and 400 m were selected. At each plot, two epiphytic lichens (PHh and CAf) were collected at a height of 1.0-2.0 m above ground from 5-10 trees to obtain 15-30 thalli and stored in paper bags. These species were selected because they are dominant in the sampling area. To collect soil samples, 3 plots 50 × 50 m 2 in size were randomly selected along each sampling line. At each plot, 10 SS samples (0-10 cm) were randomly collected using a soil gauge, mixed to represent average element composition, and stored in polyethylene bags.
Sample preparation and chemical analysis. Lichen thalli were carefully cleaned under a microscope, and then were oven-dried at 70 °C for 72 h to a constant weight. Samples were not washed before mineralisation to avoid the leaching of ionisable forms or loss of material, as suggested by the relevant studies 6, 32, 34 . SS samples were thoroughly cleaned and dried using the above method. All samples were ground and homogenised using a grinding mill equipped with Tungsten Carbide jars (Retsch MM400; Retsch GmbH, Haan, Germany).
Samples (200-300 mg) were mineralised in a mixture of HNO 3 and H 2 O 2 for lichens, and in a mixture of HNO 3 , HF, HCl, and HClO 4 for SS samples. The concentration of 20 elements (Ba, Ca, Cd, Co, Cr, Cu, Fe, K, Mg, Mn, Mo, Na, Ni, P, Pb, Sb, Sr, Ti, V and Zn) was determined using an inductively coupled plasma mass spectrometry (ICP-MS; Agilent 7700X; Agilent Technologies, Tokyo, Japan) at the Hebei Geological Laboratory. Analytical quality control was assured using a series of standard reference materials: IAEA-336 (Portuguese Lichen, issued by the International Atomic Energy Agency), GBW10014 (cabbage), GBW10015 (spinach) and GBW10052 (green tea) for lichens, as well as GBW07451, GBW07452 and GBW07457 for SS (all materials mentioned above were issued by the Institute of Geophysical and Geochemical Exploration, Chinese Academy of Geological Sciences). Methods have been published elsewhere 31 where "Fe" and "El" are the chosen reference element (Fe) and the element under consideration, respectively. The subscripts and the square brackets denote the sample type and concentrations, respectively.
Range-standardisation of elemental concentration. The raw concentration was range-standardised according to Equation (2)  where "[El]" and "[El]′" denote the raw elemental concentration and the range-standardised values, respectively. The subscripts "min" and "max" denote minimum and maximum raw concentrations, respectively.

Statistical analyses.
A one-sample Kolmogorov-Smirnov test was performed on the raw concentration distribution of each element. Because of the normal data distribution, parametric analyses were used in subsequent analyses. Concentration correlations between lichen elements were tested using a Pearson correlation analysis and cluster analysis. The cluster analysis was performed using the UPGMA criterion to construct the hierarchical tree and correlation distance to measure dissimilarities. The heatmap was produced on the range-standardised concentrations. All statistical analyses mentioned above were performed using the Past 3.14 software (Ø. Hammer, Sep. 2016). Concentration differences between lichen species and among plots were tested using Paired-samples T test on the raw concentrations. Bonferroni corrections were used in the case of multicomparison. These analyses were performed using SPSS 13.0 (SPSS Inc., Chicago, IL, USA). Diagrams in Fig. 1 were prepared using Inkscape software 0.91 (Free Software Foundation Inc., USA; http://www.inkscape.org/).