Morphological Differentiation of Betula (Birch) Pollen in Northwest North America and Its Palaeoecological Application Benjamin F

The Holocene 15,2 (2005) pp. 229/237

Morphological differentiation of Betula (birch) pollen in northwest North America and its palaeoecological application Benjamin F. Clegg,* Willy Tinner,** Daniel G. Gavin and Feng Sheng Hu*

(Department of Plant Biology, 265 Morrill Hall, University of Illinois, 505 S. Goodwin Ave, Urbana, IL 61801, USA)

Received 1 June 2003; revised manuscript accepted 12 January 2004

Abstract: Lake sediments from arcto-boreal regions commonly contain abundant Betula pollen. However, palaeoenvironmental interpretations of Betula pollen are often ambiguous because of the lack of reliable morphological features to distinguish among ecologically distinct Betula species in western North America. We measured the grain diameters and pore depths of pollen from three tree-birch species (B. papyrifera, B. kenaica and B. neoalaskana) and two shrub-birch species (B. glandulosa and B. nana), and calculated the ratio of grain diameter to pore depth (D/P ratio). No statistical difference exists in all three parameters between the shrub-birch species or between two of the tree-birch species (B. kenaica and B. papyrifera), and B. neoalaskana is intermediate between the shrub-birch and the other two tree- birch species. However, mean pore depth is significantly larger for the tree species than for the shrub species. In contrast, mean grain diameter cannot distinguish tree and shrub species. Mean D/P ratio separates tree and shrub species less clearly than pore depth, but this ratio can be used for verification. The threshold for distinguishing pollen of tree versus shrub birch lies at 2.55 mm and 8.30 for pore depth and D/P ratio, respectively. We applied these thresholds to the analysis of Betula pollen in an Alaskan lake-sediment core spanning the past 800 years. Results show that shrub birch increased markedly at the expense of tree birch during the ‘Little Ice Age’; this pattern is not discernible in the profile of total birch pollen.

Key words: Betula, birch, pollen, western North America, palaeoecology, ‘Little Ice Age’, Grizzly Lake, Alaska.

Introduction found in tundra, bogs, swamps and disturbed forests (FNA, 1997; Thompson et al., 1999; Viereck and Little, 1986). B. occi- Betula is often a prominent component of arctic and boreal dentalis is a shrubby streamside species (FNA, 1997). Autecolo- pollen records, comprising greater than 40% of the total pollen gies, growth forms and climatic requirements differ greatly assemblages in some modern (e.g., Anderson and Brubaker, among these species. However, their pollen-morphological 1993) or Holocene samples from Alaska (e.g., Hu et al., features are similar, making it difficult to infer palaeoenviron- 1993). The genus is represented by six species in the modern ments from fossil pollen profiles of Betula. flora of the western North American high latitudes, many Numerous studies have been conducted in Europe to achieve of which are known to form hybrids (Flora of North subgeneric resolution of Betula with various degrees of success America Editorial Committee: FNA, 1997). These species are (Eneroth, 1951; Birks, 1968; Prentice, 1981; Caseldine, 2001; B. papyrifera Marshall, B. kenaica Evans, B. neoalaskana Blackmore et al., 2003). Despite some disagreement regarding Sargent, B. occidentalis Hooker, B. glandulosa Michaux and the resolution that can be achieved, these studies demonstrate B. nana Linnaeus subsp. exilis (Sukaczev) Hulte´n. B. papyrifera, that it is possible to separate Betula types using both grain B. kenaica and B. neoalaskana are tree species of boreal forests, diameter and D/P ratio (Birks, 1968; see Caseldine, 2001, for whereasB. glandulosa and B. nana are two common shrub species a comprehensive review). Additionally, Blackmore et al. (2003) discuss several qualitative morphological characters that differ between B. pubescens Ehrhart type and B. nana *Authors for correspondence (e-mail: [email protected]; fshu@life. type pollen, including differences in the vestibulum as well as in uiuc.edu) **Present address: Institut fu¨r Pflanzenwissenschaften, Universita¨t the endopore and ectopore diameters. Similar studies in Bern, Altenbergrain 21, CH-3013 Bern, Switzerland Que´bec, Canada, have shown that the grain diameter can # 2005 Edward Arnold (Publishers) Ltd 10.1191/0959683605hl788rp 230 The Holocene 15 (2005) be used to distinguish B. alleghaniensis Britton from both Canada. We measured the diameter and pore depth of B. papyrifera and B. pumila Linnaeus (Richard, 1970; 1980). pollen grains and calculated the D/P ratio to evaluate This finding was subsequently applied to reconstruct vegeta- these morphological parameters for distinguishing these tional histories in Betula-rich stratigraphies of eastern North species. Our data show that pore depth and D/P ratio America (Richard, 1980; Richard et al., 1982; Marcoux are reliable criteria to separate pollen of tree versus shrub and Richard, 1995). However, tree birch species co-occurring species of Betula in western North America. To demons- with B. alleghaniensis, such as B. papyrifera, cannot be trate that these criteria can help improve palaeoecological separated from the ecologically different shrub species, includ- reconstructions, we then apply these criteria to the pollen ing B. glandulosa and B. nana. analysis of a sediment core from Grizzly Lake (62843?N, In contrast to the studies in Europe and eastern North 144812?W, elevation 720 m above sea level), southeastern America, attempts to separate the western North American Alaska. species of Betula based on pollen morphology have not been successful (Ives, 1977; Edwards et al., 1991). Betula is thus often not identified to species, greatly hampering Methods palaeoenvironmental interpretations in areas where Betula Laboratory analyses is a dominant component of pollen records, such as Alaska Male catkins from 13, 5, 6, 19 and 12 individuals of and adjacent Canada (Anderson and Brubaker, 1993). B. papyrifera, B. kenaica, B. neoalaskana, B. glandulosa and To date the only species-level identifications of Betula in B. nana, respectively, were taken from five herbaria (Table 1). Alaskan paleoecological studies are based on rare occur- We relied on the correct identification of the material by rences of macrofossils that suggest the existence of shrub- the herbaria. Classification of specimens was adjusted to con- birch species during lateglacial times and the arrival/ form to the FNA (1997). B. occidentalis was not included expansion of tree-birch species during the early Holocene because we were unaware of its status as a distinct species at (Hu et al., 1993). the time of sampling. Additionally, the Alaskan Betula type In this study, modern pollen samples of five of the most listed as B. occidentalis may consist of an extensive hybrid common Betula species in the high-latitude regions of western swarm between B. neoalaskana and B. glandulosa (FNA, North America (Figure 1), B. papyrifera, B. kenaica, 1997; Hulte´n, 1968). B. neoalaskana, B. glandulosa and B. nana, were assembled Pollen preparation generally followed Fægri and Iversen from various herbaria (Table 1). These samples represent a (1989). Preparation included the following sequence: 10% wide geographic region throughout their ranges with the HCl at 958C for 10 minutes, 10% KOH at 958C for 10 minutes, majority of specimens deriving from Alaska and adjacent glacial acetic acid rinse, acetolysis (1:9 sulfuric acid and acetic

Figure 1 Ranges of ﬁve common Betula species in high-latitude North America. Figure modiﬁed from FNA (1997). Benjamin F Clegg et al.: Morphological differentiation of Betula pollen in North America 231

Table 1 Herbarium specimens of Betula used for pollen measurements

Species Specimen ID Sample location (as indicated on herbarium sheets) Grain diameter Pore depth (avg. and s in mm) (avg. and s in mm)

1 B. papyrifera 1191-2 Cultured 25.229/1.74 3.549/0.39 1 9476 Stockade Beaver Creek, Weston County, Wyoming 19.569/0.99 2.819/0.30 1 16370 Near Derinda, Illinois 23.099/1.29 3.289/0.33 1 8409 Creek S of Hanover, Illinois 22.199/1.26 3.159/0.29 1 N/A [IV] Rogers Park, Chicago, Illinois 21.809/1.50 3.129/0.33 1 N/A [V] Edgewater, Chicago, Illinois 21.849/1.48 2.949/0.32 2 3392272 Montcalm Co., Michigan 23.229/1.01 3.189/0.31 2 2793485 Carberry, Manitoba, Canada 22.559/1.65 3.099/0.38 2 1596572 Eardley Parish, Gatineau Co., Ottawa, Canada 25.169/1.86 2.789/0.17 3 R78 Kobuk River, Alaska 19.979/1.54 2.839/0.34 4 ALTA78351 65840?N, 128850?W, Mackenzie Valley, Canada 25.439/2.10 3.099/0.33 5 58718 North Tongass National Haines, Alaska 25.379/1.79 3.249/0.27 1 N/A [I] Columbia R., 6 miles from Newport, Stevens Co., 24.049/2.75 3.369/0.56 Washington

5 B. neoalaskana 58746 15 miles NW of Portage near Chugach Nat. Bird P., 21.209/1.63 2.599/0.24 Alaska 5 57892 Northwest Alaska 21.629/1.38 2.509/0.34 5 V.93763 68842?N, 134808?W, Reindeer Station, Caribou Hills 17.649/1.10 2.689/0.41 5 58730 Fairbanks, Alaska 22.719/1.72 2.789/0.28 5 58726 Glenn Hwy near mile 80, Mentasta Pass, Alaska 23.119/1.70 2.699/0.33 5 58729 Central Alaska 21.859/1.71 2.659/0.31

1 B. kenaica 3927-3 Cultured 22.169/1.54 3.299/0.34 5 594 Juneau, Alaska 24.859/1.74 3.159/0.19 5 18367 Matanuska, Alaska 20.479/1.23 3.169/0.37 5 38462 59817?N, 158835?W, Aleknagik, Bristol Bay, Alaska 20.749/1.26 2.439/0.25 5 58756 60837.5?N, 149831?W, mile 45 Seward Hwy, L. Summit 20.889/1.27 2.919/0.26

4 B. glandulosa ALTA78324 65838?N 128813?W, Mackenzie Valley, Canada 21.899/1.38 2.459/0.25 2 1644172 Quebec, Canada 20.769/1.70 2.069/0.25 2 2385843 Pamiagdluk Kuˆngmint, Greenland 19.739/1.41 1.829/0.25 3 N/A [1] 57848?N, 122853?W, mile 207 Alaska Hwy 19.309/1.21 2.339/0.34 3 N/A [2] 16 miles N of Tetlin Jct., Alaska 19.839/1.46 2.169/0.24 3 N/A [3] Â/60 miles SW of McKinley Park St, Alaska 20.449/1.24 2.339/0.20 3 N/A [4] 8 miles N of Kimberley, British Columbia, Canada 20.949/1.26 2.549/0.28 3 N/A [5] 12 miles N of Donald, British Columbia, Canada 20.849/1.27 2.509/0.19 3 N/A [6] Lolo Hot Springs, Missoula Co., Montana 19.869/1.10 2.149/0.27 3 R73 Kobuk River, Alaska 20.889/1.21 2.319/0.25 5 V. 108328 Lake Nishlik, Wood Tikchik State Park, Alaska 19.509/1.41 2.599/0.40 5 V. 101486 59802?N, 158828?W, Lake Nunavanguluk 19.029/1.21 2.209/0.22 5 87901 63834?N, 148846?W, west side of Carlo Mt 22.129/1.86 2.229/0.30 5 86166 68850?N, 143830?W, north of Ambresvajun Lake 18.129/1.59 2.359/0.40 5 V. 93862 68842?N, 134808?W, Reindeer Station, Caribou Hills 22.989/1.03 2.399/0.33 5 5108 67815?N, 141840?W, Howling Dog, Alaska 23.009/1.10 2.699/0.32 5 78883 60855?N, 138843?W, St Elias Mts NW of Slims R. 21.769/1.78 2.679/0.32 5 58677 62820?N, 146830?W, Lake Louise, Alaska 21.109/1.49 2.379/0.41 5 V. 118625 61814?40??N, 149834?00??W, Chuchach Mountains 21.349/1.10 2.569/0.24

1 B. nana 3105 Yukon territory, Canada 19.749/2.16 2.199/0.30 1 19740 Near Granite Pass, Big Horn Mts, Wyoming 19.939/1.64 2.409/0.39 1 24433 62838?N, 136848?W, mile 151, Alaska Hwy to Dawson 19.459/1.11 2.009/0.21 4 ALTA41128 69825?N, 132852?W, Tuktogaktuk, Canada 20.049/1.36 2.109/0.21 5 V.96558 66834?N, 164825?W, Cape Espenberg, Kitluk R. area 20.699/1.52 2.379/0.21 5 V.116898 65838?N, 146846?W, Lime Rock, Rocky Mts 21.639/1.54 2.209/0.28 5 9232 68805’/12’N, 165832?/47?W, Ogotoruk Creek Drainage 23.259/1.10 2.759/0.32 5 V.127902 61807?N, 155836?W, Sparrevohn LRRS Air Force St 16.109/1.14 1.839/0.24 5 590 Eureka Rh. on Glenn Hwy, Alaska 19.919/1.88 2.409/0.34 5 V.124767 64808?N, 159825?W, Nulato Hills, mid. Umalakleet val. 19.419/0.92 2.269/0.25 3 R 163 Koyokuk River, Alaska 20.969/1.19 2.289/0.22 3 R 231 Agiak Lake, Alaska 22.539/1.57 1.879/0.19

1Herbarium of the University of Illinois at Urbana-Champaign. 2Herbarium of the St Louis Botanical Garden. 3Herbarium of the University of Washington. 4Herbarium of the Canadian Forest Service. 5Herbarium of the University of Alaska Museum. 232 The Holocene 15 (2005)

anhydride) at 958C for 3 minutes, 10% KOH at 958C for 5 min- smallest for B. glandulosa and B. nana (2.359/0.36 mm and utes, fuxin stain for 2 minutes, 95% and 100% ethanol rinse, 2.229/0.36 mm, respectively). Mean grain diameter is larger tert-butanol rinse, and mounting in silicon oil. Thirty pollen for B. papyrifera (23.039/2.52 mm) than for all other species, grains from each specimen were measured with a semi- and it varies little among B. kenaica, B. neoalaskana, B. glan- automated procedure. A digital image of each grain was dulosa and B. nana (21.829/2.15 mm, 21.359/2.35 mm, obtained under 1000/ magnification with a Cool SNAP 20.719/1.86 mm and 20.309/2.26 mm, respectively). The D/P digital camera. Measurements of grain diameter and pore ratio is largest for B. glandulosa and B. nana (8.999/1.44 † depth on each image were made with MetaView software. and 9.339/1.59, respectively), intermediate for B. neoalaskana Grain diameter was defined as the distance from the inner (8.199/1.42) and smallest for B. papyrifera and B. kenaica margin of the pore wall to the outer margin of the opposite (7.509/1.13 and 7.379/1.09, respectively). Mean pore depth wall with the pollen grain in polar view (Ives, 1977). Pore and grain diameter are positively correlated for all 55 indivi- depth was defined as the distance from the outer margin of duals (r/0.64, PB/0.001; Figure 2). However, for each the sexine to the inner margin of the nexine across the vestibu- species considered individually, this correlation is weak lum (Ives, 1977; H.J.B. Birks, personal communication 2000). (B. papyrifera:r/0.53, P/0.06) or insignificant (all other Preparation of the fossil material from Grizzly Lake fol- species: rB/0.5, P/0.1). lowed the same method used for preparing herbarium samples, For pore depth, the majority of the overall variance is with the addition of an HF step to dissolve silicates (Fægri and explained by the among-species effect (53.9%, PB/0.001), Iversen, 1989). Digital images of Betula pollen grains were and the remaining variance is mainly due to the within- taken with a Sanyo digital camera under 1000./ Measure- individuals rather than among-individuals effect (30.1% and ments of grain diameter and pore depth on each image were 16.0%, respectively). Pairwise comparisons show that mean made with a semi-automated procedure in Quantimet500# pore depth differs significantly for all species pairs except for (copyright by Leica Cambridge Ltd, 1992). Betula pollen was B. kenaica versus B. papyrifera, B. kenaica versus B. neoalas- separated for 25 pollen grains per sample by applying pore- kana, and B. nana versus B. glandulosa (Table 3). The largest depth and D/P-ratio thresholds established with the her- difference in pore depth occurs between B. papyrifera barium reference material. and B. nana (9% overlap) or B. glandulosa (13% overlap) (Table 3; Figure 3). No statistical difference of the pore depth Statistical analyses exists between the two shrub-birch species and between two The relationship between the pollen pore depth and grain pairs of tree-birch species: B. papyrifera versus B. kenaika diameter for all Betula species was assessed with the Pearson and B. kenaica versus B. neoalaskana. correlation coefficient. Only the mean values of the 30 grains For grain diameter, only a small portion of the total vari- per individual were used for this calculation, because replicate ance in grain diameter is explained by the among-species effect measurements within individuals are not independent samples. (18.3%), and the majority of variance is due to the among- and We used a nested ANOVA to test the null hypothesis that within-individuals effects (45.7% and 36.0%, respectively) the means of each of the three pollen parameters (pore depth, (Table 2). Pairwise species comparisons show that mean grain grain diameter and D/P ratio) do not differ among species diameter differs significantly only for two species pairs: (Sokal and Rohlf, 1981). Nested ANOVA partitions total vari- B. papyrifera versus B. glandulosa and B. papyrifera versus B. ance to three levels of effects: among species, among individuals nana. The overlap is large for these two pairs (47% and 43%, within species, and within individuals. It tests whether variance respectively) and even larger for all other species comparisons is significantly greater among species than among individuals (65/98%). within species. The variance within individuals is equivalent The variance of the D/P ratio is similar for the among- to the unexplained variance of the model. We also compared species (26.9%) and the among-individuals (24.0%) effects, with the majority being within-individuals effects (49.1%) the three pollen parameters for the percentage of vari- (Table 2). Pairwise species comparisons show that mean D/P ance explained by each effect. ANOVAs were run with SPSS ratio differs significantly for both B. kenaica versus B. nana 10.0 software (univariate general linear models procedure). or B. glandulosa, and B. papyrifera versus B. nana or B. glan- We then tested the null hypothesis that the means of each of dulosa (Table 3). The largest difference in the D/P ratio occurs the three pollen parameters do not differ between each pair between B. papyrifera and B. nana (33% overlap of pore-depth of the five Betula species. We used nested ANOVA as described distributions) or B. glandulosa (39% overlap). No statistical above but included only two species at a time. We report difference of the D/P ratio exists between each pair of the the F-statistic and P-values with levels of significance for the three tree species, between the two shrub-birch species, and among-species effect. Significance is determined with the Bon- between B. neoalaskana and either of the shrub-birch species. ferroni-corrected P-value of 0.05 for 10 pairwise comparisons

(i.e., each pairwise comparison with PB/0.005 is significant). In addition, to help visualize the similarity of pollen para- Discussion meters we described the percentage of overlap of size distributions using the Mann-Whitney U statistic. We scaled the U Morphological comparisons of tree- and statistic with a multiplier of 2/(n1n2), where n1 and n2 are shrub-birch pollen the number of pollen grains measured for each species. The A separation of the two shrub-birch species B. nana and scaled U statistic ranges from 0 (complete separation of size B. glandulosa is not possible because of a lack of statistical distributions) to 100 (complete overlap of size distributions). difference for all three parameters (Table 3). The three tree species show variation in some pollen-morphological characters. However, the differences occur along a gradient, and only Results the endmembers B. papyrifera and B. neoalaskana are significantly different from each other. Because neither species can Mean pore depth is largest for B. papyrifera and B. kenaica be separated effectively from B. kenaica, any attempt to separ-

(3.119/0.40 mm and 3.019/0.40 mm, respectively; errors are ate tree birch species in regions where B. kenaica occurs or had

1 SD), intermediate for B. neoalaskana (2.659/0.33 mm), and the potential to occur in the past would result in high Benjamin F Clegg et al.: Morphological differentiation of Betula pollen in North America 233

a fairly clear separation of tree- and shrub-birch pollen. In contrast, grain diameter is not a reliable parameter for separating tree- and shrub-birch pollen. The differences are significant between B. papyrifera and the shrub-birch species

but insignificant for any other tree/shrub pairs (Table 3). The D/P ratio is less effective than the pore depth but shows a clearer differentiation between tree- and shrub-birch pollen

than the pollen diameter. All tree/shrub comparisons except B. neoalaskana versus the shrub-birch species are statistically different for the D/P ratio (Table 3). The distinct pore-depth difference between tree- and shrub- birch pollen has not been reported previously, because most attempts to separate birch pollen have focused only on the grain diameter (Richard, 1970; 1980; Prentice, 1981; Edwards, et al., 1991). However, our grain-diameter results are in agreement with previous investigations. On the basis of grain diameter, Edwards et al. (1991) also found that the three species B. neoalaskana (then classified as B. resinifera (Regel) Britton), B. nana and B. glandulosa are indistinguishable from one another. Our finding of significant differences in the grain diameter between B. papyrifera and the two shrub-birch species supports Ives (1977), who found ‘observable differences’ in the grain diameter of these species but noted that Figure 2 Scatter plot of pore depth versus grain diameter. Each these differences are insufficient for a good separation. point represents the average of 30 pollen grains from each However, Ives (1977) found no difference among B. papyrifera individual. and the two shrub-birch species based on the D/P ratio, which suggests a stronger correlation between pore depth and grain diameter in his measurements. uncertainties on the identification. However, every tree-birch species is significantly different from every shrub-birch species Threshold values for separating tree- and with respect to pore depth. We thus group the species into shrub-birch pollen ‘shrub birches’ and ‘tree birches’ hereafter. Differentiating pollen from tree- and shrub-birch species can Pore depth is the most powerful parameter for distinguish- be accomplished by the application of a simple threshold value ing between pollen from tree-birch and shrub-birch species, to the pore depth. A similar threshold value for the D/P ratio as indicated by the highly significant differences (PB/0.005) may be helpful in some situations (see next section below). To for all tree/shrub comparisons (Figure 3; Table 3). In calculate these thresholds, we divided the data for all species addition, the ANOVA of all species shows that pore depth is into size classes in increments of 0.05 mm for the pore depth the only parameter for which the majority of variation is and 0.05 for the ratio parameter. Because the pollen- explained by among-species differences (Table 2), resulting in morphological criteria do not differ between B. nana and

Figure 3 Distributions of individual pollen parameters (pore depth, grain diameter and ratio of grain diameter to pore depth) for (above) individual Betula species and (below) species groups (tree versus shrub). 234 The Holocene 15 (2005)

Table 2 Nested ANOVA of three pollen parameters measured on 55 individual specimens from ﬁve Betula species (B. nana, B. glandulosa, B. neoalaskana, B. kenaica and B. papyrifera)

Dependent variable Source df SS MS Variance component % variance F

Pore depth 33.98; Among species 4 213.574 53.394 0.166 53.9 PB/0.001 17.00 Among individuals within species 50 78.570 1.571 0.049 16.0 PB/0.001 Within individuals 1595 147.436 0.092 0.092 30.1 Total 1649 439.580 0.267 0.307 100.0 Grain diameter 5.05 Among species 4 1769.225 442.306 1.135 18.3

P/0.002 39.14 Among individuals within species 50 4379.188 87.584 2.844 45.7 PB/0.001 Within individuals 1595 3568.849 2.238 2.238 36.0 Total 1649 9717.262 5.893 6.217 100.0 11.95 Among species 4 978.272 244.568 0.717 26.9 PB/0.001 Ratio 15.64; Among individuals within species 50 1023.121 20.462 0.638 24.0 PB/0.001 Within individuals 1595 2086.320 1.308 1.308 49.1 Total 1649 4087.713 2.479 2.663 100.0

B. glandulosa, the size-class distributions of these species were reference material. If the tree-type pollen derives only from pooled, resulting in a composite shrub-birch distribution. B. papyrifera and B. kenaica, 91.5% of the tree-type pollen Similarly, because the size-class distributions do not differ would be correctly classified, compared to only 58.3% if significantly between B. kenaica and B. papyrifera or between B. neoalaskana were the only source of tree-birch pollen. In B. kenaica and B. neoalaskana, B. kenaica could be merged comparison, the D/P-ratio threshold correctly classifies with either of the other two tree-birch species. However, pore 79.7% of the shrub-birch pollen and 58.7% of the tree-birch depth differs significantly between B. neoalaskana and pollen of our reference material. If the tree-type pollen derives B. papyrifera. Thus, pooling the size-class distributions of all only from B. papyrifera and B. kenaica, 71.3% of tree-type pol- three tree-birch species may bias the composite distribution. len would be correctly classified, compared to only 46.1% if We created a composite distribution by first pooling B. papyr- B. neoalaskana were the only source of tree-birch pollen. The ifera and B. kenaica and then averaging this combined distri- threshold values remain the same if calculated specifically for bution with that of B. neoalaskana. This approach is based B. neoalaskana and the shrub-birch species. Therefore the on the facts (1) that B. papyrifera and B. kenaica show no sig- low percentage of correctly classified tree-birch pollen where nificant differences in any of the measured parameters, (2) that B. neoalaskana is the only tree-birch species is not a conse- B. papyrifera and B. kenaica belong to the same North Amer- quence of the assumptions made when combining the three ican clade (B. kenaica is probably derived from B. papyrifera; tree-birch species. FNA, 1997), whereas B. neoalaskana belongs to a circumpolar complex of birch species more distantly related to either of the Size variations related to laboratory treatments other two tree species (FNA, 1997), and (3) that the present and depositional environments geographic range of B. neoalaskana overlaps extensively The reliable differentiation of pollen types using size mea- with the combined ranges of B. papyrifera and B. kenaica surements depends on the ability to rule out, or account (Figure 1). Hence pollen grains of B. neoalaskana are expected for, potential changes in pollen-morphological parameters. to co-occur with either of the other two tree-birch species. By Reitsma (1969) addressed various chemical environments weighting the size-class distributions in this way, the composite that can affect the grain diameter. He concluded that soak- tree-birch distribution avoids a priori assumptions as to which ing pollen in 958C water swelled the grains, that HF shrunk tree-birch species to expect in a given region. the grains, and that acetolysis caused an abrupt expansion To minimize the probability of a false identification, we followed by slow shrinkage, depending on the duration of chose the threshold value that minimizes the sum of the areas the acetolysis treatment. Various other chemicals, such as under the shrub-type distribution above that value and under 10% KOH, only have an effect after acetolysis treatment. the tree-type distribution below that value (Figure 3). The Thus, it is important for any size-based differentiation of threshold values are 2.55 mm and 8.30 for the pore depth pollen types that the pollen preparation strictly follows and the D/P ratio, respectively. Pollen grains with pore depths a standardized procedure to allow comparisons of measure- smaller than 2.55 mm or D/P ratios larger than 8.30 are thus ments among laboratories (e.g., PALE, 1994; Bennett and classified as shrub birch, whereas those with values larger than Willis, 2001). 2.55 mm or smaller than 8.30 are classified as tree birch. The morphology of fossil pollen may also change in The pore-depth threshold correctly classifies 80.1% of the response to variations in the natural chemical environment shrub-birch pollen and 74.9% of the tree-birch pollen of our in lake sediments, such as the pH difference of gyttja versus Benjamin F Clegg et al.: Morphological differentiation of Betula pollen in North America 235

Table 3 Comparison of pore depth, grain diameter and ratio of diameter to pore depth of ﬁve species of Betula. Upper values (in bold) are the percent overlap of the size distributions as calculated by a scaled Mann-Whitney U statistic. The lower values are F statistics and P values from a nested ANOVA testing for between-species difference in mean pore depth, grain diameter, and ratio parameter. Shaded areas highlight comparisons among tree birch species (upper left) or between shrub birch species (lower right). See Table 1 for sample sizes

Dependent Variable Taxa B. kenaica B. neoalaskana B. glandulosa B. nana

Pore depth B. papyrifera 89% 35% 13% 9% 0.64 22.95 90.56 86.24 (P/0.435) (PB/0.001)* (PB/0.001)* (PB/0.001)* B. kenaica 48% 22% 16% 7.23 29.63 29.59 (P/0.025) (PB/0.001)* (PB/0.001)* B. neoalaskana 51% 36% 10.18 15.46 (P/0.004)* (P/0.001)* B. glandulosa 77% 2.27

(P/0.143) Grain diameter B. papyrifera 70% 65% 47% 43% 1.42 2.99 16.11 12.93 (P/0.251) (P/0.102) (PB/0.001)* (P/0.002)* B. kenaica 98% 73% 65% 0.17 2.44 2.48

(P/0.693) (P/0.132) (P/0.136) B. neoalaskana 78% 72% 0.88 1.28 (P/0.358) (P/0.274) B. glandulosa 90% 0.51

(P/0.480) Ratio B. papyrifera 93% 69% 39% 33% 0.16 4.631 29.67 29.36 (P/0.698) (P/0.046) (PB/0.001)* (PB/0.001)* B. kenaica 64% 35% 29% 3.32 15.16 14.38 (P/0.101) (PB/0.001)* (P/0.002)* B. neoalaskana 67% 57% 4.25 5.63 (P/0.051) (P/0.031) B. glandulosa 86% 1.01

(P/0.323)

*Bonferroni-corrected PB/0.05. marl (Aario, 1941; Praglowski, 1966). Should such differences in the region (FNA, 1997; Figure 1). A detailed pollen record in the chemical environment of fossil pollen substantially alter from this site spans the last 800 years based on six 210Pb and pollen-grain sizes, a standardized laboratory procedure would six 14C dates, encompassing the ‘Little Ice Age’ (LIA) period not eliminate the uncertainty of identification. Prentice (1981) (ÂAD/ 1480/1825). circumvented the problem of size changes by applying the Gor- Pooling all measured fossil Betula pollen grains from Grizzly don-Prentice maximum likelihood method. This method ident- Lake for each of the parameters shows two clearly separated ifies the most likely proportion of species in a mixed pollen modes in the pore-depth distribution, and a less clear but sample based on the differences observed in their known distri- nevertheless distinctly bimodal distribution of the D/P ratio. butions. An application of this method to the pore depth of In contrast, the grain diameter has a unimodal distribution fossil birch pollen in North America may be suitable if size (Figure 4). We infer that the pore-depth modes at 1.8 mm changes are suspected to have occurred within sediments. and 2.9 mm represent shrub-type and tree-type pollen, respect- However, the simplest solution is to use the D/P ratio, which ively. The mode at 1.8 mm is smaller than for the shrub-type remains largely unaffected by size changes (B.F. Clegg, unpub- reference material at 2.2 mm, whereas the mode at 2.9 mmis lished data). Thus, the D/P ratio can nevertheless be used as a in good agreement with that of the tree-type reference material verification of pore-depth result, even though the D/P ratio is at 2.8 mm. A plausible explanation for shifts in the modes is less effective at distinguishing shrub- versus tree-birch pollen. that the birch populations around Grizzly Lake probably represent only a subset of our large, geographically diverse, refer- Application to the Grizzly Lake pollen record ence material. This explanation is supported by the narrower A fossil pollen record from Grizzly Lake, SE Alaska (Tinner distributions in both pore depth and D/P ratio of the fossil and Hu, 2001) is particularly useful to demonstrate the utility pollen when compared to the modern reference material of a birch-pollen differentiation. The lake lies near the upper (Figures 2 and 3). However, the D/P-ratio modes of the Griz- elevational limit of B. neoalaskana in the study area (Hulte´n, zly Lake pollen at 7.0 and 9.6 are in good agreement with the 1968), and B. kenaica and shrub-birch species are also found respective modes of the B. kenaica/B. papyrifera-group and 236 The Holocene 15 (2005)

Figure 4 Pooled data of fossil Betula pollen grains from Grizzly Lake for each of the three parameters. Dashed lines in the pore-depth and ratio parameters represent the threshold values 2.55 mm and 8.30, respectively, calculated using the modern reference material. shrub-type herbarium material. This observation suggests that from oxygen-isotopic data from Farewell Lake, central at least part of the difference in the pore depth between refer- Alaska, interpreted as representing as ‘Little Ice Age’ cold ence and fossil material may be chemically induced. The event (Hu et al., 2001). The tree-birch elevational range prob- robustness of the ratio parameter illustrates its value for verify- ably lowered during the LIA, whereas shrub birches that were ing pore-depth results. adapted to cold climatic conditions and fire disturbance In the Grizzly Lake record, undifferentiated Betula pollen increased in abundance. These vegetational fluctuations are fluctuates between 18 and 29% of total pollen (Figure 5). It not discernible without separating birch pollen into shrub exhibits no general trends during the past 800 years, apart and tree types. Thus, the application of our differentiation cri- from low values centred at AD 1725. In contrast, the differen- teria of birch pollen to the fossil-pollen analysis at Grizzly tiation of Betula pollen using the pore-depth threshold reveals Lake clearly demonstrates the potential to greatly enhance striking stratigraphic variations in tree- and shrub-types. Prior palaeoecological interpretations from fossil pollen records with to AD 1480 the abundance of tree-type pollen (12/18%) Betula as a major constituent. slightly exceeds that of shrub-type pollen (c. 12%). Between AD 1480 and 1825, shrub-type pollen increases markedly to a maximum of c. 22%, and tree-type pollen decreases to a mini- Acknowledgements mum of c. 3%. The low percentages of undifferentiated Betula centred at AD 1725 represent the combined effect of depressed We thank Ruth Beer for measuring Betula pollen for the tree- and shrub-type pollen. Following AD 1825, tree-type pol- Grizzly Lake proﬁle. The Herbaria of the Canadian Forest len (c.12/18%) occurs again in higher abundance than shrub- Service, Missouri Botanical Garden, Natural Resources type pollen (c.6/13%). The D/P ratio confirms the pore-depth Canada, University of Alaska Museum, University of Illinois, trends, although the pattern appears more muted. The general University of Washington, Linda Brubaker, Mary Edwards, agreement between pore depth and D/P ratio suggests that the Wyatt Oswald and Dave Seigler kindly provided Betula pollen observed patterns are not artifacts of chemically induced size samples. We would like to express our special thanks to John changes. Birks for inspirational discussions, Pierre Richard for very help- The period of shrub-type Betula dominance between AD ful correspondence and encouragement, and Alan Batten for

1480 and 1825 coincides with a cooling of 1.7/ 8C as inferred Betula taxonomical advice. Comments from Lynn Anderson,

Figure 5 Partial pollen diagram from Grizzly Lake, SE Alaska, based on W. Tinner et al. (unpublished data). The temperature proﬁle is based on coupled bulk-carbonate and ostracode d18OdatainHuet al. (2001) and is given as D8C relative to the twentieth-century mean. Pollen percentages are based on the total pollen. Data are presented for equal time intervals (50 years per data point). Benjamin F Clegg et al.: Morphological differentiation of Betula pollen in North America 237

Stephen Blackmore, Robert Clegg, Paul Henne, Jason Lynch, Hu, F.S., Ito, E., Brown, T.A., Curry, B.B. and Engstrom, D.R. David Nelson, Jian Tian and an anonymous reviewer greatly 2001: Pronounced climatic variations in Alaska during the last two improved the manuscript. This research was supported by a Uni- millennia. Proceedings of the National Academy of Sciences of the versityof Illinois SURE Undergraduate Fellowship to BFC, a United States of America 98, 10552/56. Hulteń, E. 1968: Flora of Alaska and neighboring territories. Swiss NSF Postdoctoral Fellowship to WT, and a Packard Fel- Stanford: Stanford University Press. lowship in Science and Engineering and US NSF grant OPP Ives, J.W. 1977: Pollen separation of three North American 01/08702 to FSH. This is US NSF PARCS contribution 217. birches. Arctic and Alpine Research 9, 73/80. Marcoux, N. and Richard, P.J.H. 1995: Ve´ge´tation et fluctuations climatiques postglaciaires sur la coˆte septentrionale gaspe´sienne, References Que´bec. Canadian Journal of Earth Sciences 32, 79/96. PALE 1994: Research protocols for PALE: palaeoclimates of arctic lakes and estuaries. Past Global Changes (PAGES) Workshop Aario, L. 1941: Die gro¨ßenstatistische Analyse der Betulapollen in Report, 53 pp. Torfproben. Geologische Rundschau 32, 612/26. Praglowski, J. 1966: On pollen size variations and the occurrence Anderson, P.M. and Brubaker, L.B. 1993: Holocene vegetation and of Betula nana in different layers of bog. Grana Palynologica 6, climate histories of Alaska. In Wright, H.E. Jr, Kutzbach, J.E., 528/43. Webb, T. III, Ruddiman, W.F., Street-Perrott, F.A. and Bartlein, Prentice, I.C. 1981: Quantitative birch (Betula L.) pollen P.J., editors, Global climates since the last glacial maximum, separation by analysis of size frequency data. New Phytologist Minneapolis: University of Minnesota Press, 386/400. 89, 145/57. Bennett, K.D. and Willis, K.J. 2001: Pollen. In Birks, H.J.B., editor, Reitsma, T. 1969: Size modifications of recent pollen grains under Tracking environmental change using lake sediments, Dordrecht: different treatments. Review of Paleobotany and Palynology 9, Kluwer, 5/32. 175/202. Birks, H.J.B. 1968: The identification of Betula nana pollen. New Richard, P.J.H. 1970: Atlas pollinique des arbres et de quelque Phytologist 67, 309/14. arbustes indige`nes du Que´bec. III. Angiospermes (Salicaceés, Blackmore, S., Steinmann, J.A.J., Hoen, P.P. and Punt, W. 2003: Myricaceés, Juglandaceés, Corylaceés, Fagaceés, Ulmaceés). Le The northwest European pollen flora 65: Betulaceae and Naturaliste Canadien 97, 97/161. Corylaceae. Review of Paleobotany and Palynology 123, 71/98. ____1980: Histoire postglaciaire de la ve´ge´tation au sud du lac Caseldine, C. 2001: Changes in Betula in the Holocene record from Abitibi, Ontario et Que´bec. Geógraphie Physique et Quaternaire / Iceland a palaeoclimatic record or evidence for early Holocene 34, 77/94. hybridisation. Review of Paleobotany and Palynology 117, 139/52. Richard, P.J.H., Larouche, A. and Bouchard, M.A. 1982: A˚ ge de la Edwards, M.E., Dawe, J.C. and Armbruster, W.S. 1991: Pollen size de´glaciation finale et histoire postglaciaire de la ve´ge´tation dans la of Betula in northern Alaska and the interpretation of late Quater- partie centrale du Nouveau-Que´bec. Geógraphie Physique et narvegetationrecords.Canadian Journal of Botany 69, 1666/72. Quaternaire 36, 63/90. Eneroth, O. 1951: Investigations of the possibility of differentiating Sokal, R.R. and Rohlf, F.J. 1981: Biometry. New York: W.H. the pollen of different species of Betula in fossil material. Freeman. Geologiska fo¨reningens i Stockholm fo¨rhandlingar 73, 343/405. Thompson, R.T., Anderson, K.H. and Bartlein, P.J. 1999: Atlas of Fægri, K. and Iversen, J. 1989: Textbook of pollen analysis. relations between climatic parameters and distributions of Chichester: Wiley. important trees and shrubs in North America. US Geological Flora of North America (FNA) Editorial Committee 1997: Flora Survey Professional Paper 1650 A and B. of North America north of Mexico, (volume 3). New York: Oxford Tinner, W. and Hu, F.S. 2001: Responses of fire and vegetation to University Press. Little-Ice-Age climatic change in boreal Alaska. The Ecological Hu, F.S., Brubaker, L.B. and Anderson, P.M. 1993: A 12 000 year Society of America (ESA) 86th Annual Meeting, 221. record of vegetation change and soil development from Wien Viereck, L.A. and Little, E.L. Jr 1986: Alaska trees and shrubs. Lake, central Alaska. Canadian Journal of Botany 71, 1133/41. Fairbanks: University of Alaska Press.