More reliable pollen productivity estimates and relative source area of pollen in a forest-steppe ecotone with improved vegetation survey

Abstract

Pollen productivity estimates (PPEs) and the relevant source area of pollen (RSAP) are critical to the reconstruction of past plant abundances. Extended R-value (ERV) models are commonly used for developing PPEs and estimating the RSAP. In this study, in the forest-steppe ecotone of northern China that is sensitive to climate change and significant in vegetation openness reconstruction, pollen data from surface sediment samples of 24 lakes and high-resolution data of vegetation within the lake basins were applied to ERV models to calculate the PPEs of Pinus, Quercus, Betula, Ulmus, Poaceae, Artemisia, Chenopodiaceae, Compositae, and Cyperaceae and the RSAP of different lake sizes. The PPEs of arboreal pollen were higher than herbaceous pollen, and Chenopodiaceae had the highest productivity of herbaceous pollen taxa. The RSAP of lakes with an average lake radius of 400 m was estimated to be 1600 m. Our results suggested that PPEs and RSAP are region specific and that RSAP is also lake size specific even within the same region. Although environmental factors and imperfections in the ERV models may affect the accuracy of results, our study emphasized that a systematic vegetation survey considering zonal and azonal vegetation, the forest and steppe mosaic, and the vertical vegetation belt can help improve the PPEs and RSAP in areas with complicated vegetation distribution, such as the forest-steppe ecotone.

Keywords

ERV models forest-steppe ecotone pollen PPEs RSAP vegetation survey

Introduction

Pollen sequences of lake sediments have been widely used to reconstruct the evolutionary processes of ancient vegetation and environments (Huntley and Prentice, 1988; Prentice and Webb, 1986). Pollen analysis has the primary goal of reconstructing past plant abundances and has been used to quantify the relationship between pollen and vegetation (Theuerkauf et al., 2012). However, the relationship between pollen percentages and plant abundances is not linear (Prentice and Webb, 1986). The percentage of one pollen taxon strongly depends on the percentages of other taxa, which is called the ‘Fagerlind effect’ (Fagerlind, 1952). Besides, plants differ in pollen productivity and pollen dispersal capability, which implies that their abundance may be overrepresented or underrepresented in the pollen record (Soepboer et al., 2007). Hence, despite a great deal of previous research, quantifying the relationship between modern pollen and vegetation remains a challenge in studies of palynology and paleoecology.

Many methods have been proposed and modified for establishing the quantitative relationship between pollen and vegetation. An R-value model was originally proposed as the conversion factor between pollen and vegetation data (Davis, 1963), and a linear model was used for the relationship between pollen flux and vegetation abundance (Andersen, 1970). However, these approaches did not solve the ‘Fagerlind effect’, even as the percentages of pollen data were used more widely. Therefore, extended R-value (ERV) models were proposed (Parsons and Prentice, 1981; Prentice and Parsons, 1983), which expanded and generalized Anderson’s model. These models fit the nonlinear relationship between pollen and vegetation percentages using the maximum likelihood method. The ERV models have been further developed by incorporating two methods of the vegetation distance weighted: Prentice’s model, which simulate the process of pollen deposition at the center of a circular basin from the vegetation source (Prentice, 1985), and the ring source model, which was adopted for pollen deposition over the entire surface of a circular basin (Sugita, 1993). These two methods are species specific and simulate the track of a single pollen grain of each taxa moved by wind using Sutton’s (1953) model. The model assumes that wind speed remains constant and is equal in strength from all directions.

The ERV models have been more widely used to quantify the relationship between pollen and vegetation, produce pollen productivity estimates (PPEs), and describe the relevant source area of pollen (RSAP; Bunting and Hjelle, 2010), especially in Europe (Broström et al., 2008). In total, two pollen sampling strategies have been adopted: (1) moss polster samples have been used for major species in the forests of Finland, Sweden, Swiss, England, Norway, Poland, and northeast China (Baker et al., 2016; Broström et al., 2004; Bunting et al., 2005; Hjelle, 1998; Li et al., 2015; Mazier et al., 2008; Niemeyer et al., 2015; Räsänen et al., 2007; Sugita et al., 1999; Von Stedingk et al., 2008) and for the tundra in the Siberian Arctic (Niemeyer et al., 2015) and the northern China steppe (Li et al., 2011; Xu et al., 2014); and (2) surface lake sediment samples have been used for dominant plant taxa around the lakes on the Swiss Plateau, Denmark, Estonia, Germany, South Africa, and the Tibetan Plateau (Duffin and Bunting, 2008; Nielsen, 2004; Nielsen and Odgaard, 2005; Poska et al., 2011; Soepboer et al., 2007; Theuerkauf et al., 2012; Wang and Herzschuh, 2011). The previous studies were mainly concentrated in forests, especially in Europe, where the winds are relatively weak (Broström et al., 2008). Moreover, ERV models are only concerned with the wind dispersal of pollen and neglect other transport agents.

The vegetation survey methods were quite different among these studies. The three-tier vegetation survey methods organized in concentric rings around the sample sites have been widely used, combining field observations within 0–10 m, existing vegetation maps within 10–100 m, and interpretation of aerial photos within 100–1500 m (Broström et al., 2004). However, the range and resolution of vegetation maps and aerial photos are usually different in different regions (Baker et al., 2016; Duffin and Bunting, 2008; Li et al., 2015; Mazier et al., 2008; Niemeyer et al., 2015; Von Stedingk et al., 2008; Xu et al., 2014). The resolution of some vegetation maps is low, and the different sources of vegetation data among these studies may affect the accuracy and comparability of PPEs and RSAP. Another common method uses only vegetation classification maps and databases around sample sites (Poska et al., 2011; Soepboer et al., 2007; Theuerkauf et al., 2012; Wang and Herzschuh, 2011). Since vegetation survey methods have been shown to have a significant effect on PPEs and RSAP (Bunting and Hjelle, 2010; Twiddle et al., 2012), the most appropriate method for describing vegetation patterns should be applied for improving PPEs and the RSAP.

The forest-steppe ecotone is widely distributed in northern China and is very sensitive to climate change (Liu et al., 2010). In watersheds in the forest-steppe ecotone, the vegetation pattern is typically a mosaic of forest, steppe, meadow, and halophytic vegetation. From the edge of the water surface to the surrounding mountains, vertical vegetation belts can be found in some lakes (Liu et al., 2008). The variations of vegetation openness under the influence of climate change in the past have great significance in paleoecological studies, and qualitative reconstruction of vegetation openness has received extensive concern (Soepboer et al., 2010; Sugita et al., 1999). Although many studies have attempted to quantify the vegetation openness caused by human activities, for example, in Europe (Abraham and Kozáková, 2012; Broström et al., 2004, 2005, 2008; Poska et al., 2011), few studies have paid attention to climate-induced changes in vegetation openness such as in the forest-steppe ecotone.

ERV models are ideally applied to estimate PPEs in the forest-steppe ecotone in northern China, where wind strength is the major conveyer of pollen grains instead of runoff (Li et al., 2011; Xu et al., 2014). However, vegetation survey methods within a radius of 10 m have also been used in some previous studies in northern China (Li et al., 2011). In addition, other studies have only used surface soil samples (Li et al., 2011, 2015; Xu et al., 2014), which do not provide the proper conditions for deposition and preservation of pollen grains (Soepboer et al., 2007). Consequently, estimates using surface sediment samples from lakes are needed. In this study, we used pollen data from the surface sediment of 24 lakes in northern China and conducted high-resolution vegetation surveys that combined field work and remote sensed imagery. The vegetation survey method was designed based on the vegetation pattern in the forest-steppe ecotone. PPEs for nine major taxa and the RSAP for the different-sized lakes were calculated.

Study area

The forest-steppe ecotone in northern China is cold, windy, and dry in winter. Most of the rainfall is concentrated in the summer (Hao et al., 2014). The distribution of rainfall is nonuniform. According to meteorological data, the mean annual temperature ranges from −3°C to 5.0°C, the mean annual precipitation is 350–400 mm, and the annual average wind speed is 3.0–4.6 m/s. The sub-regions of Xilinguole (42°10′–44°58′N and 112°47′–117°20′E) and Hulunbeier (49°4′–44°30′N and 116°57′–120°36′E) are the two main parts of the forest-steppe ecotone, and they extend southward and eastward to the Daxing’an Mountains (Figure 1). In the Xilinguole sub-region, there are sparse patches of forest dominated by, or in some cases mixed with, Larix, Pinus, Quercus, Populus, Betula, and Ulmus species on steep, shady slopes. The meadow steppe is dominated by several indicative species of Poaceae, such as Stipa baicalensis, Leymus chinensis, and Bromus inermis. Species from Chenopodiaceae, Compositae, and Cyperaceae are also primary elements of the steppe. In the Hulunbeier sub-region, there are sparse forests dominated by Ulmus and Pinus species. In the deciduous shrub community, the common shrub species is Salix linearistipularis. The herb species, such as Leymus chinensis and Stipa baicalensis, and other species from Chenopodiaceae, Compositae and Cyperaceae, are similar to that in the Hulunbeier steppe. In the study area, there are scattered farmlands, which primarily grow maize and wheat.

Figure 1.

The location of the study area in northern China: Xilinguole sub-region (left bottom) and Hulunbeier sub-region (right bottom), and the sample lakes on a vegetation map (modified from Hou, 2001). The main vegetation types are as follows: (1) Larix forest, (2) Pinus forest, (3) temperate deciduous shrub, (4) shrub desert, (5) grasses and forbs meadow, (6) desert steppe, (7) swamp meadow, (8) salt meadow, (9) cold temperate and temperate marsh, (10) cultivated vegetation, (11) Quercus forest, (12) Populus forest, (13) Betula forest, (14) Ulmus forest, and (15) water body.

There are many inland lakes scattered in the study area, so it is possible to conduct a study of PPEs around the lake basins. In the watershed of each lake, the vegetation patterns are complicated and are influenced by many factors. First, climate-determined zonal and soil moisture-determined azonal vegetation form a mosaic. In our study area, the zonal vegetation pattern is a mosaic of forest and steppe. Azonal vegetation includes swamp, halophytic vegetation, and meadows around the lakes and along the river. Second, the altitude range can influence the vegetation patterns. In our study region, areas with large ranges in altitude could support more forests (Hao et al., 2016). Third, the slope aspect has significant influence on vegetation distribution. In our study area, the sparse patches of forest usually grow on steep, shady slopes, and grass typically grows on the sunny slopes (Liu et al., 2002). A sketch map of the complex vegetation patterns typically found around the lakes is shown in Figure 2.

Figure 2.

Sketch map of a typical vegetation pattern around a lake.

Methods

Pollen data

We systematically chose 24 lakes in the forest-steppe ecotone in northern China (Figure 1). The average lake water surface radius was 400 m. Features of individual lakes are listed in Table 1. We recorded the longitude, latitude, and elevation of each sampling site using a global positioning system (GPS). Surface sediment samples were collected at the center of each lake using a sediment sampler. The upper 2 cm of sediment was collected, as did in previous work (Soepboer et al., 2007).

Table 1.

Location, elevation, radius of water, and the water area.

Site (full name and abbreviation)	Longitude (E; °)	Latitude (N; °)	Elevation (m a.s.l.)	Radius of water (m)	Water area (ha)
Yuelianghu (ylh)	117.383	42.349	1670	315	31.17
Taiyanghu (tyh)	117.155	42.392	1479	153	7.35
Jiangjunpaozi (jjpz)	117.151	42.591	1489	245	18.86
Zhagasitai (zgst)	116.243	42.606	1360	1162	424.19
Sanggenbei (sgb)	115.833	42.762	1295	906	257.87
Gouxiepaobei (gxpb)	115.801	42.759	1294	554	96.42
Durenaoer (drne)	116.415	43.248	1275	826	214.34
Honggeergaole (hgegl)	115.658	43.337	1174	319	31.97
Shanshuidujiacun (ssdjc)	112.886	42.843	1026	105	3.46
Engehu (egh)	114.816	43.278	1038	197	12.19
Zhananhu (znh)	116.247	42.582	1360	362	41.17
Dazhapang (dzp)	115.464	42.932	1253	180	10.18
Diaoyuhu (dyh)	117.459	44.655	989	98	3.02
Heremuting (hrmt)	117.371	44.637	990	307	29.61
Hakexiang (hkx)	120.160	49.144	637	275	23.76
Hakexiangbei (hkxb)	120.161	49.145	637	275	23.76
Wuquanshan (wqs)	120.544	49.091	703	350	38.48
Bagahu (bgh)	119.266	49.358	591	605	114.99
Chenqiganpao (cqgp)	119.242	49.368	591	430	58.09
Goufeipao (gfp)	118.098	49.466	540	650	132.73
Goufeipaobei (gfpb)	118.099	49.468	538	630	124.69
Dalaidong (dld)	117.131	49.389	660	410	52.81
Mashikeng (msk)	116.970	49.115	633	80	2.01
Dalaisumu (dlsm)	116.952	49.110	631	170	9.08

The samples were treated with HCl and NaOH in the laboratory. The samples were sieved through 180-µm and 10-µm sieves and treated with a heavy liquid solution with a specific gravity of about 2.0 to extract pollen from sediments. Pollens of each sample were identified under an Olympus optical microscope at 400× magnification to a minimum count of 200 grains (Liu et al., 2010), and the pollen count data were input into ERV models.

The pollen-specific fall speeds were calculated following Stoke’s Law (Broström et al., 2004; Gregory, 1973; Sugita et al., 1999) using the average grain size of each taxon; the long and short axes of pollen grains were measured on pollen reference slides. The fall speeds of Pinus, Quercus, Betula, Ulmus, Poaceae, Artemisia, Chenopodiaceae, Compositae, and Cyperaceae were assumed to be 0.039, 0.018, 0.019, 0.022, 0.018, 0.007, 0.009, 0.017, and 0.017 m/s, respectively, as determined in previous studies (Li et al., 2015; Wu, 2011).

Vegetation data

Assuming each lake as the center of its watershed, the images of 24 lake watersheds were downloaded from the Google Earth, which shows recent vegetation with very high resolution (mostly 1 m). We classified the images using supervised classification in the ArcGIS 10.0 program. In total, seven land-cover and land-use types were distinguished from the images: forest, shrubland, grassland, built area, bare land, wetland, and open surface water. We surveyed the vegetation around the lake in eight directions (i.e. N, NE, E, SE, S, SW, W, and NW) from the lake shore, extending out to the vertical vegetation belt of the watershed. The farthest extension was about 3600 m from the lake shore (Figure 3a). We walked along the directions and chose site for vegetation survey when the land-cover type changes according to the field study and classified images. We surveyed grassland and shrubland with plot size of 1 m × 1 m and forest with plot size of 10 m × 10 m. For the same vegetation in the same direction of one watershed, we recorded only one plot for saving time. The location information and the vegetation information of each site were recorded, including the longitude, latitude, elevation, and the distance from the lake shore to the survey site, and the name, height, coverage, and abundance of the plants. The 416 records obtained are shown in Table 2.

Figure 3.

Sketch map of the vegetation survey design: (a) vegetation was surveyed around the lake in eight directions extending out to the vertical belt and (b) vegetation information from the vertical belt to a 5000-m ring from the lake center was obtained using classified remote sensing imagery and buffer overlay analysis.

Table 2.

Vegetation records around 24 lakes.

Site	Radius of water (m)	Number of records	Range of distance from vegetation records to lake center (m)
Site	Radius of water (m)	Number of records	Swamp vegetation or halophytic vegetation	Steppe	Forest
Yuelianghu	315	28	315–400	350–751	426–839
Taiyanghu	153	28	153–198	452–1699	539–1750
Jiangjunpaozi	245	34	266–382	419–3560	840–3635
Zhagasitai	1162	26	1162–1301	1265–1706	1181–1705
Sanggenbei	906	21	906–983	1160–1659	1003–1508
Gouxiepaobei	554	22	554–605	540–1020	560–873
Durenaoer	826	18	830–942	950–1370	850–1100
Honggeergaole	319	17	306–345	368–615	360–650
Shanshuidujiacun	105	14	105–350	350–1071
Engehu	197	10	197–230	321–650	2670
Zhananhu	362	16	362–525	525–1209	558–1209
Dazhapang	180	12	180–260	300–859	1000
Diaoyuhu	98	10	98–125	125–339	181
Heremuting	307	10	307–340	340–653	340–500
Hakexiang	275	15	275–340	340–521	617
Hakexiangbei	275	18	275–325	325–480	830
Wuquanshan	350	10	350–375	530–1721	442–1335
Bagahu	605	11	605–780	796–1834
Chenqiganpao	430	11	430–567	567–1000
Goufeipao	650	7	650–696	700–1581
Goufeipaobei	630	19	630–680	680–1287
Dalaidong	410	20	410–480	480–633
Mashikeng	80	19	80–100	100–350
Dalaisumu	170	20	170–238	230–612

In the ArcGIS 10.0 program, concentric ring buffers from the lake shore to 5000 m from the lake center were set to a width of 50 m (Figure 3b). The covers of major taxa for each concentric ring were obtained through the following steps: (1) from the lake shore to the vertical belt (determined according to remote sensing images), the mean percentage coverage of each taxa in each 50-m-wide concentric ring was calculated using data recorded in eight aspects, and (2) from the vertical belt we surveyed farthest to 5000 m from the lake center, vegetation composition of each vegetation type was obtained by combining the classified images with vegetation survey data and the Vegetation Atlas of The People’s Republic of China (1:1,000,000; Hou, 2001). The mean area of each vegetation type in each 50-m-wide ring was calculated in ArcGIS, and the percentage coverage of major taxa in each ring was calculated.

ERV models

The ERV models can describe the relationship between pollen and vegetation using the maximum likelihood function to estimate PPEs and RSAP (Sugita, 1993, 1994). The correlation between pollen data and vegetation abundance will improve as the area of surveyed vegetation increases around a lake until it reaches a certain distance, where the correlation between pollen and vegetation will not improve further. Therefore, the area from the edge of the lake to the relevant source distance is called the RSAP (Soepboer et al., 2007). The ERV models can calculate the likelihood function score, which will fluctuate with increasing area and tend to be a constant when it reaches the RSAP. This can also allow us to develop stable PPEs (Broström et al., 2008).

In this study, we used an ERV Analysis.v1.2.3 program developed by Sugita (unpublished). There are three sub-models in the program. These sub-models have different site factors to adjust for the use of the proportional pollen or vegetation data. Models 1 and 2 use pollen data and vegetation percentages (Parsons and Prentice, 1981; Prentice and Parsons, 1983), whereas pollen percentages and plant abundance data expressed in absolute units (e.g. such as cover or biomass per area) are used in model 3 (Sugita, 1994). The different assumptions of the background pollen values are the factors underlying the differences between the three sub-models. Since we used the percentage data, we chose sub-models 1 and 2. In addition, the program requires input data, such as plant abundance and pollen assemblage. Other factors are also taken into account, such as lake radii (Table 1), pollen fall speed, and wind speed, which we set to the empirical value of 3 m/s. The vegetation distance weighting methods were specified. There are four distance weighting methods: d⁻¹, d⁻², Prentice’s (1985) model, and the ring source model (Sugita, 1993). We used the ring source model, which is more appropriate for lake samples.

The PPEs of major taxa, Pinus, Quercus, Betula, Ulmus, Poaceae, Artemisia, Chenopodiaceae, Compositae, and Cyperaceae, were calculated by ERV models with the dataset from the 24 lakes, and Poaceae, which had a PPE of 1, was chosen as the reference taxa. In addition, in order to know the effects of basin size, eight lakes were selected from the southern sub-region: ylh, tyh, jjpz, hgegl, zgst, sgb, gxpb, and drne (Table 1), where their distribution is concentrated, so that the climate and vegetation composition were similar, and basin size was the only variable. The average radius of all of the eight lakes was 550 m. They were divided into two sub-groups, with four lakes in each sub-group. The average radius of ylh, tyh, jjpz, and hgegl was 250 m, and they were categorized into the small lake sub-group. The average radius of zgst, sgb, gxpb, and drne was 850 m, and they were categorized into the large lake sub-group. The RSAPs for the three different lake size groups were calculated using four major herbaceous taxa in the sub-region, Poaceae, Artemisia, Chenopodiaceae, and Cyperaceae, since the number of taxa should be less than the number of lakes (Soepboer et al., 2007).

Results

Pollen assemblages

We identified 31 taxa of pollen, including eight arbors, two shrubs, and 21 herbs from the sediment of the 24 lakes. The pollen assemblages were dominated by herb pollen, with less arbor and shrub pollen. Poaceae, Artemisia, Chenopodiaceae, Compositae, and Cyperaceae were the major herb pollen taxa, and Pinus, Quercus, Betula, and Ulmus were the major arboreal pollen taxa (Figure 4).

Figure 4.

Pollen percentage diagram for major taxa in the lake samples.

PPEs

The PPEs of nine selected major taxa and their standard deviation are shown in Figure 5. The results of sub-model 1 were chosen since the results of sub-models 1 and 2 were similar, and it was difficult to judge which one was better (Prentice and Parsons, 1983). The PPEs of arboreal pollen were commonly higher than herbaceous pollen. Quercus (58.05 ± 9.54) and Pinus (20.07 ± 6.96) have the highest productivities of the arboreal pollen taxa, followed by Ulmus (9.44 ± 3.50) and Betula (1.16 ± 0.43). Chenopodiaceae (50.49 ± 3.56) had the highest productivity of herbaceous pollen taxa, even higher than some arboreal taxa, while Artemisia (1.29 ± 0.27) had similar productivity to Poaceae. Compositae (0.19 ± 0.20), and Cyperaceae (0.01 ± 0.01) had the lowest PPEs.

Figure 5.

PPEs with standard deviation for nine major taxa from the 24 lakes in the forest-steppe ecotone of northern China.

RSAP

For the 24 lakes in northern China, the likelihood function scores in relation to the distance from the lake shore are shown in Figure 6. The ERV models were used to calculate the curves with the pollen and vegetation data from the 24 lakes. We can infer that the curve reaches its asymptote at a distance of 1600 m (Figure 6). So, the RSAP of lakes with an average lake radius of 400 m was estimated to be 1600 m from the lake shore. In other words, the plants farther than 1600 m from the shores of lakes with a radius of 400 m may be irrelevant to the pollen assemblages in the lake sediment.

Figure 6.

(a) Likelihood function score plot and RSAP for the 24 lakes with a radius of 400 m in the forest-steppe ecotone of northern China. Likelihood function score plot and RSAP for the lakes with different radii from a small region in southeastern Inner Mongolia; (b) RSAP for the eight lakes with a radius of 250 m; (c) RSAP for the four lakes with a radius of 550 m; and (d) RSAP for the four lakes with a radius of 850 m.

In order to know how basin size affects the RSAP, we calculated the RSAP of the sub-region group with eight lakes with different average radii, the small lake sub-group with four lakes, and the large lake sub-group with four lakes to analyze the relationship between lake radius and RSAP. The likelihood function scores, which fluctuate with increasing distance from the lake edge, are shown in Figure 6. We can see from the three curves that the RSAPs of lakes with radii of 250, 550, and 850 m are 1200, 1900, and 2700 m, respectively. We found that RSAP increased with increasing radius. So, changes in lake size have significant influences on RSAP.

Discussion

Our estimates for Ulmus, Artemisia, and Compositae are lower, while Chenopodiaceae was notably higher than the results in typical steppe using surface soil samples (Xu et al., 2014). PPEs of major herb taxa and tree taxa have been obtained in Europe, South Africa, Siberia, and China (Baker et al., 2016; Broström et al., 2008; Duffin and Bunting, 2008; Li et al., 2015; Niemeyer et al., 2015; Wang and Herzschuh, 2011; Xu et al., 2014). Table 3 shows the comparison of PPEs obtained in this study, as well as the results in other regions with different vegetation types. Comparing the results from previous studies, it became apparent that the results of PPEs were not consistent in different regions (Broström et al., 2008; Soepboer et al., 2007). Our estimates for Quercus and Chenopodiaceae were higher than the results in other regions, while those for Betula, Artemisia, and Compositae were lower, and the PPEs of other taxa were within a reasonable range of values in different regions with different vegetation types.

Table 3.

PPEs and vegetation information in previously published studies and this study.

Study area	Vegetation zone	Major vegetation types	Sampling type	Pinus	Quercus	Betula	Ulmus	Poaceae	Artemisia	Chenopodiaceae	Compositae	Cyperaceae	Reference
Finland	Northern boreal zone	Pine woods, birch woods, mires	Moss polsters	2.50		6.30		1.00				0.001	Räsänen et al. (2007)
Central Sweden	Forest-tundra ecotone	Coniferous woodland, birch woodland, alpine tundra	Moss polsters	30.70		2.73		1.00				1.16	Von Stedingk et al. (2008)
Estonia	Hemi-boreal forest zone	Woodland, agricultural land	Lake sediments	4.71	7.09	1.75		1.00	3.44				Poska et al. (2011)
Southern Sweden	Southernmost nemoral zone/boreo-nemoral zone	Woodland, cultivated fields, pasture	Moss polsters	5.70	7.60	8.90	0.80	1.00			0.25	0.65	Broström et al. (2004)
Poland	Białowieża forest: the transition between the boreal and the nemoral biogeographic zones	Evergreen spruce woods, deciduous trees	Moss polsters	51.38	44.76	30.80		1.00					Baker et al. (2016)
Tibetan Plateau	High alpine vegetation zone	Subalpine shrub, alpine meadows	Lake sediments					1.00	2.08	5.44		1.03	Wang and Herzschuh (2011)
Northern China	Typical steppe zone	Temperate grassland	Moss polsters				11.53	1.00	11.21	6.74	0.39	0.94	Xu et al. (2014)
Northern China	Forest-steppe ecotone	Sparse patches of deciduous forest, temperate steppe	Lake sediments	20.07	58.05	1.16	9.44	1.00	1.29	50.49	0.19	0.01	This study

PPE: Pollen productivity estimate; ERV: extended R-value.

PPEs were estimated using ERV sub-model 1 and Poaceae as reference taxa, except for the study in typical steppe in northern China, for which ERV sub-model 2 was applied.

The inconsistencies in PPEs can be caused by the differences of plant species in different regions (Matthias et al., 2012). The PPE for Quercus in this study was 58.05, much higher than the values in Sweden, Estonia, and Poland (Baker et al., 2016; Broström et al., 2004; Poska et al., 2011). This may be related to the different species of Quercus since pollen taxa can usually be identified in the genus level. In Europe, there were oak trees mostly Quercus robur, while there were trees of Quercus mongolica in this study. Besides, the PPE of Betula in this study was 1.16, slightly lower than that in Finland, Sweden, and Estonia (Broström et al., 2004; Poska et al., 2011; Räsänen et al., 2007; Von Stedingk et al., 2008). It may be because there were trees of Betula platyphylla and Betula davurica in forest-steppe ecotone, while the birches were mostly Betula pendula and Betula pubescens in Europe.

The difference in vegetation patterns is also an important influencing factor to PPEs (Bunting and Hjelle, 2010; Twiddle et al., 2012), and standard vegetation survey methods are needed (Bunting et al., 2013). In our study area, the vegetation pattern was complicated, as indicated by a zonal and azonal vegetation complex, a vertically distributed vegetation complex, and a slope-distributed vegetation complex. These are all important for the calculation of PPEs. The zonal vegetation dominates in the watershed. The azonal vegetation, such as swamp vegetation and halophytic vegetation, is usually distributed near the lake shore (Liu et al., 2008). The vertical belt at high elevations is usually forest, which has pollen that spreads farther than herb taxa. The slope vegetation usually forms many patches differing in different directions, which complicates the vegetation pattern. The vegetation patterns in the watershed might strongly affect the estimation of PPEs. For example, the PPE of Chenopodiaceae was notably higher in this study than the values obtained from the typical steppe in Inner Mongolia as well as alpine vegetation (meadow, steppe, and desert) in Tibetan Plateau (Wang and Herzschuh, 2011; Xu et al., 2014). It may be because the Chenopodiaceae plants were distributed in azonal vegetation, such as salty meadow dominated by Suaeda glauca growing on saline and alkaline soil near the water, and made great contributions to the pollen of the lake sediments, leading to a high PPE for Chenopodiaceae (Liu et al., 2008). Since the vegetation patterns have significant influence on the PPE results, we did the vegetation survey at a large scale and in eight directions, using high-resolution remote sensing images, so we could capture the details of the complicated vegetation patterns. Systematic vegetation surveys can help improve PPEs and determination of RSAP in watersheds with complicated vegetation distribution.

In addition, different sample materials can cause differences in PPEs (Mazier et al., 2008; Soepboer et al., 2007). Moss polsters and lake sediments are the common sample materials, and the age they present could be quite different (Räsänen et al., 2004; Soepboer et al., 2007). We also have to stress that the ERV models themselves have some imperfections, which can cause uncertainties in PPEs. First, the models need the input vegetation data to be from lakes with the same radius. Vegetation around lakes with different radii must be transformed to the mean radius by stretching or compressing. This affects the accuracy of the vegetation data. So, it is preferable to use lakes of the same size, but it is not easy to find those lakes in practice. Second, ERV models are only concerned with pollen grains transported by wind, while other conveyers of pollen grains, such as runoff, are not taken into account.

RSAP estimates of different samples, such as moss polsters (Li et al., 2015; Mazier et al., 2008; Räsänen et al., 2007; Von Stedingk et al., 2008) and lakes of different sizes (Matthias et al., 2012; Niemeyer et al., 2015; Poska et al., 2011; Soepboer et al., 2007; Sugita, 1994; Sugita et al., 1999; Theuerkauf et al., 2012), have been obtained in previous studies. Table 4 shows the comparison of RSAP estimates obtained in this study, as well as the results in other regions with different vegetation types. From these studies, it can be concluded that the RSAP results simulated for basins with the same size are not consistent in different regions. The RSAP in Estonia (Poska et al., 2011) was between 1500 and 2000 m for lakes with a radius of approximately 100 m, and the RSAPs in other studies (Matthias et al., 2012; Niemeyer et al., 2015; Theuerkauf et al., 2012) were even larger, from 7000 m to 25 km for medium lakes. For the studies using moss polsters, there are also great discrepancies between them, for example, the RSAP values in Northern Siberian Arctic and Western Norway were 10 and 10.5 m (Bunting and Hjelle, 2010; Niemeyer et al., 2015), while the values were 1000 and 2000 m in northern China steppe and forest (Li et al., 2015; Xu et al., 2014).

Table 4.

Overview of RSAP in previously published studies and this study.

Study area	Vegetation zone	Sampling type	Average radius of basin (m)	RSAP (m)	Reference
Northern Siberian Arctic	Arctic and alpine tundra	Moss polsters	1	10	Niemeyer et al. (2015)
Western Norway	Heathland	Moss polsters	1	10.5	Bunting and Hjelle (2010)
England	Atlantic	Moss polsters	1	50–150	Bunting et al. (2005)
Switzerland	Pasture woodlands	Moss polsters	1	300	Mazier et al. (2008)
Central Sweden	Forest-tundra ecotone	Moss polsters	1	500	Von Stedingk et al. (2008)
Eastern Scotland	Pine woodland	Moss polsters	1	300–500	Twiddle et al. (2012)
Finland	Northern boreal zone	Moss polsters	1	1000	Räsänen et al. (2007)
Northern China	Typical steppe zone	Moss polsters	1	1000	Xu et al. (2014)
Northeastern China	Forest	Moss polsters	1	2000–2500	Li et al. (2015)
Estonia	Hemi-boreal forest zone	Lake sediments	100	1500–2000	Poska et al. (2011)
Tibetan Plateau	High alpine vegetation zone	Lake sediments	100	2200	Wang and Herzschuh (2011)
Central Europe	Forest	Lake sediments	100	7000	Theuerkauf et al. (2012)
Northern Siberian Arctic	Arctic and alpine tundra	Lake sediments	100	25000	Niemeyer et al. (2015)
Norway	Boreo-nemoral and southern boreal vegetation zones	Lake sediments	135	900–1100	Hjelle and Sugita (2012)
Eastern Germany	Forest	Lake sediments	140	7000	Matthias et al. (2012)
Swiss Plateau	Nemoral	Lake sediments	190	800	Soepboer et al. (2007)
Southern African	Arid savanna and bushveld	Lake sediments	200	600–900	Duffin and Bunting (2008)
Northern China	Forest-steppe ecotone	Lake sediments	400	1600	This study

RSAP: relevant source area of pollen.

The average basin radii of moss polsters were set to 1 m.

These inconsistencies in RSAP estimations among different studies can be caused by many factors. Our study emphasized that the patterning of the vegetation within a landscape can affect the RSAP estimates, as suggested by previous studies (Broström et al., 2005; Nielsen and Odgaard, 2005; Soepboer et al., 2007). The RSAP can be influenced by vegetation patch size: larger patches cause a larger RSAP (Bunting et al., 2005; Calcote, 1995). For example, the variation of RSAP in Northern Siberian and northeastern China may be caused by the different vegetation patterns in arctic tundra and temperate forest (Li et al., 2015; Niemeyer et al., 2015). In our study, as mentioned above, the vegetation patterns are complicated, and the vegetation survey method was applied to obtain more details regarding the vegetation patches in order to obtain a more reliable RSAP. Besides, the modeling of RSAP also depends on sampling types and pollen taxa composition (Niemeyer et al., 2015).

The results for basins with different sizes are also different, even in the same region. RSAP is also strongly influenced by basin size (Hjelle and Sugita, 2012; Soepboer et al., 2007). For example, for the moss polsters whose basin size can be about 1 m, the RSAPs were mostly between 50 and 500 m – much smaller than that for lakes (Table 4). In our study, the RSAPs of lakes with radii of 250, 550, and 850 m were 1200, 1900, and 2700 m, respectively. We found that RSAP increased with the increase in water surface radius, implying that changes in lake size have significant influences on RSAP.

Moreover, the number of taxa may affect the RSAP results. Nielsen (2005) found that the RSAPs were larger when using 4 taxa than when using 11 taxa. While in our study, the RSAPs (1900 m for the lakes with a radius of 550 m) calculated using four taxa were comparable with that using nine taxa (1600 m for lakes with a radius of 400 m). The number of taxa may not change the RSAP results.

Conclusion

PPEs for major taxa in the forest-steppe ecotone of northern China were obtained using lake sediment samples and high-resolution vegetation data. We found that the PPEs of arboreal pollen were higher than herbaceous pollen, and Chenopodiaceae had the highest productivity of herbaceous pollen taxa. In addition, PPEs of the same taxa can differ in different study regions for the same vegetation type, which means PPEs are region specific. RSAPs are also region specific and are lake size specific even in the same region. The inconsistencies of PPEs and RSAP results in different regions can be caused by many factors such as plant species, vegetation pattern, sampling type, and basin size. We suggested that vegetation survey designed methods based on the vegetation pattern can obtain more details of vegetation abundance, which help improve the PPEs and RSAP in the forest-steppe ecotone.

Footnotes

Funding

This research was supported by grants from NSF of China (41530747 and 41325002).

References

Abraham

Kozáková

(2012) Relative pollen productivity estimates in the modern agricultural landscape of Central Bohemia (Czech Republic). Review of Palaeobotany and Palynology 179: 1–12.

Andersen

(1970) The relative pollen productivity and pollen representation of north European trees, and correction factors for tree pollen spectra. Danmarks Geologiske Undersogelse II 96: 1–99.

Baker

Zimny

Keczynski

et al . (2016) Pollen productivity estimates from old-growth forest strongly differ from those obtained in cultural landscapes: Evidence from the Bia owieza National Park, Poland. The Holocene 26(1): 80–92.

Broström

Nielsen

Gaillard

et al . (2008) Pollen productivity estimates of key European plant taxa for quantitative reconstruction of past vegetation: A review. Vegetation History and Archaeobotany 17(5): 461–478.

Broström

Sugita

Gaillard

(2004) Pollen productivity estimates for the reconstruction of past vegetation cover in the cultural landscape of southern Sweden. The Holocene 14(3): 368–381.

Broström

Sugita

Gaillard

et al . (2005) Estimating the spatial scale of pollen dispersal in the cultural landscape of southern Sweden. The Holocene 15(2): 252–262.

Bunting

Hjelle

(2010) Effect of vegetation dada collection strategies on estimates of relevant source area of pollen (RSAP) and relative pollen productivity estimates (relative PPE) for non-arboreal taxa. Vegetation History and Archaeobotany 19(4): 365–374.

Bunting

Armitage

Binney

et al . (2005) Estimates of ‘relative pollen productivity’ and ‘relevant source area of pollen’ for major tree taxa in two Norfolk (UK) woodlands. The Holocene 15(3): 459–465.

Bunting

Farrell

Broström

et al . (2013) Palynological perspectives on vegetation survey: A critical step for model-based reconstruction of Quaternary land cover. Quaternary Science Reviews 82: 41–55.

10.

Calcote

(1995) Pollen source area and pollen productivity-evidence from forest hollows. Journal of Ecology 83: 591–602.

11.

Davis

(1963) On the theory of pollen analysis. American Journal of Science 261: 897–912.

12.

Duffin

Bunting

(2008) Relative pollen productivity and fall speed estimates for southern African savanna taxa. Vegetation History and Archaeobotany 17(5): 507–525.

13.

Fagerlind

(1952) The real significance of pollen diagrams. Botaniska Notiser 105: 185–224.

14.

Gregory

(1973) The Microbiology of the Atmosphere. London: L. Hill.

15.

Hao

Liu

(2016) Pollen-detected altitudinal migration of forests during the Holocene in the mountainous forest–steppe ecotone in northern China. Palaeogeography, Palaeoclimatology, Palaeoecology 446: 70–77.

16.

Hao

Liu

Yin

et al . (2014) Varied responses of forest at its distribution margin to Holocene monsoon development in northern China. Palaeogeography, Palaeoclimatology, Palaeoecology 409: 239–248.

17.

Hjelle

(1998) Herb pollen representation in surface moss samples from mown meadows and pastures in western Norway. Vegetation History and Archaeobotany 7(2): 79–96.

18.

Hjelle

Sugita

(2012) Estimating pollen productivity and relevant source area of pollen using lake sediments in Norway: How does lake size variation affect the estimates? The Holocene 22(3): 313–324.

19.

Hou

(2001) Vegetation Atlas of The People’s Republic of China (1:1000 000). Beijing, China: Science Press.

20.

Huntley

Prentice

(1988) July temperatures in Europe from pollen data 6000 years before present. Science 241(4866): 687–690.

21.

Bunting

et al . (2011) Pollen-vegetation-climate relationships in some desert and desert-steppe communities in northern China. The Holocene 21(6): 997–1010.

22.

Nielsen

Zhao

et al . (2015) Pollen production estimates (PPEs) and fall speeds for major tree taxa and relevant source areas of pollen (RSAP) in Changbai Mountain, northeastern China. Review of Palaeobotany and Palynology 216: 92–100.

23.

Liu

Wei

Liu

et al . (2008) Determinants of pollen dispersal in the East Asian steppe at different spatial scales. Review of Palaeobotany and Palynology 149(3–4): 219–228.

24.

Liu

Cui

(2002) Holocene history of desertification along the woodland-steppe border in Northern China. Quaternary Research 57(2): 259–270.

25.

Liu

Yin

Zhu

et al . (2010) How did the forest respond to Holocene climate drying at the forest–steppe ecotone in northern China? Quaternary International 227(1): 46–52.

26.

Matthias

Nielsen

Giesecke

(2012) Evaluating the effect of flowering age and forest structure on pollen productivity estimates. Vegetation History and Archaeobotany 21(6): 471–484.

27.

Mazier

Broström

Gaillard

et al . (2008) Pollen productivity estimates and relevant source area of pollen for selected plant taxa in a pasture woodland landscape of the Jura Mountains (Switzerland). Vegetation History and Archaeobotany 17(5): 479–495.

28.

Nielsen

(2004) Modelling pollen sedimentation in Danish lakes at c. ad 1800: An attempt to validate the POLLSCAPE model. Journal of Biogeography 31(31): 1693–1709.

29.

Nielsen

Odgaard

(2005) Reconstructing land cover from pollen assemblages from small lakes in Denmark. Review of Palaeobotany and Palynology 133(1): 1–21.

30.

Niemeyer

Klemm

Pestryakova

et al . (2015) Relative pollen productivity estimates for common taxa of the northern Siberian Arctic. Review of Palaeobotany and Palynology 221: 71–82.

31.

Parsons

Prentice

(1981) Statistical approaches to R-Values and the pollen-vegetation relationship. Review of Palaeobotany and Palynology 32: 127–152.

32.

Poska

Meltsov

Sugita

et al . (2011) Relative pollen productivity estimates of major anemophilous taxa and relevant source area of pollen in a cultural landscape of the hemi-boreal forest zone (Estonia). Review of Palaeobotany and Palynology 167: 30–39.

33.

Prentice

(1985) Pollen representation, source area, and basin size: Toward a unified theory of pollen analysis. Quaternary Research 23: 76–86.

34.

Prentice

Parsons

(1983) Maximum likelihood linear calibration of pollen spectra in terms of forest composition. Biometrics 39(4): 1051–1057.

35.

Prentice

Webb

(1986) Pollen percentages, tree abundances and the Fagerlind effect. Journal of Quaternary Science 1(1): 35–43.

36.

Räsänen

Hicks

Odgaard

(2004) Pollen deposition in mosses and in a modified ‘Tauber trap’ from Hailuoto, Finland: What exactly do the mosses record? Review of Palaeobotany and Palynology 129: 103–116.

37.

Räsänen

Suutari

Nielsen

(2007) A step further towards quantitative reconstruction of past vegetation in Fennoscandian boreal forests: Pollen productivity estimates for six dominant taxa. Review of Palaeobotany and Palynology 146: 208–220.

38.

Soepboer

Sugita

Lotter

(2010) Regional vegetation-cover changes on the Swiss Plateau during the past two millennia: A pollen-based reconstruction using the REVEALS model. Quaternary Science Reviews 29(3–4): 472–483.

39.

Soepboer

Sugita

Lotter

et al . (2007) Pollen productivity estimates for quantitative reconstruction of vegetation cover on the Swiss Plateau. The Holocene 17(1): 65–77.

40.

Sugita

(1993) A model of pollen source area for an entire lake surface. Quaternary Research 39: 239–244.

41.

Sugita

(1994) Pollen representation of vegetation in Quaternary sediments: Theory and method in patchy vegetation. Journal of Ecology 82: 881–897.

42.

Sugita

Gaillard

Broström

(1999) Landscape openness and pollen records: A simulation approach. The Holocene 9(4): 409–421.

43.

Sutton

(1953) Micrometeorology: A Study of Physical Processes in the Lowest Layers of the Earth’s Atmosphere. New York: McGraw-Hill Education.

44.

Theuerkauf

Kuparinen

Joosten

(2012) Pollen productivity estimates strongly depend on assumed pollen dispersal. The Holocene 23(1): 14–24.

45.

Twiddle

Jones

Caseldine

et al . (2012) Pollen productivity estimates for a pine woodland in eastern Scotland: The influence of sampling design and vegetation patterning. Review of Palaeobotany and Palynology 174: 67–78.

46.

Von Stedingk

Fyfe

Allard

(2008) Pollen productivity estimates from the forest–Tundra ecotone in west-central Sweden: Implications for vegetation reconstruction at the limits of the boreal forest. The Holocene 18(2): 323–332.

47.

Wang

Herzschuh

(2011) Reassessment of Holocene vegetation change on the upper Tibetan Plateau using the pollen-based REVEALS model. Review of Palaeobotany and Palynology 168(1): 31–40.

48.

(2011) Relationship between pollen and vegetation based on ERV model in Hulunbeier steppe. PhD Thesis, Peking University.

49.

Cao

Tian

et al . (2014) Relative pollen productivities of typical steppe species in northern China and their potential in past vegetation reconstruction. Science China Earth Sciences 57(6): 1254–1266.