1We evaluated geographical and climate distributions of 57 vascular plant genera that have disjunct distributions primarily restricted to south-eastern Asia (EAS) and south-eastern North America (ENA). Because clades of each genus in the two regions are considered sister taxa, differences in ecological relationships between the regions reflect independent evolutionary responses to environmental factors and allow a test of ecological conservatism or parallelism.
2The disjunct genera were found to be distributed in 270 grid cells, 163 in EAS and 107 in ENA, each with 3.75° extent in latitude and longitude. Individual disjunct genera occupied larger areas and latitudinal ranges in eastern Asia than in eastern North America, which parallels the larger area and latitudinal range of suitable climate in eastern Asia.
3For all EAS-ENA disjunct genera taken together, the midpoint latitude, northernmost latitude, latitude range and area occupied were all significantly correlated between the two regions. For woody genera, only the correlation for midpoint latitude was significant, whereas for herbaceous genera, all except midpoint latitude were significant.
4A twinspan analysis partitioned the grid cells into four distinctive groups based on presence or absence of the disjunct genera. Grid cells representing these groups in EAS and ENA remained clustered when plotted on axes defined by floristic ordination (non-metric multidimensional scaling) or climate variables (discriminant analysis), showing that the disjunct genera have similar floristic relationships to each other within regions and respond, to a large degree, in parallel to climate variables.
5The lack of correlation in distribution variables among woody genera between the continents in part reflects the greater propensity of eastern Asian woody disjunct genera to extend southwards into subtropical and tropical latitudes.
6Although the disjunct genera exhibit significant conservatism in geographical and climate distribution, reflecting their common evolutionary ancestry, differences between regions in geography, physiography and climate have also acted to diversify environmental relationships between continents.
Understanding the evolution of ecological relationships between organisms and climate or other aspects of the environment can help us to interpret distributions and the responses of populations to climate change (Gaston 2003). Comparing distributions of closely related taxa between ecologically similar but geographically separated regions can provide information about evolutionary lability of environmental relationships or, alternatively, their conservatism. Related taxa presumably share ancestral ecological traits, including geographical extent and tolerance of climate extremes, at least in times past (Jablonski 1987; Ricklefs & Latham 1992; Brown 1995; Webb et al. 2002). The propensity of sister taxa to diverge following disjunction provides insight into the types of species and traits that are more (or less) sensitive to evolutionary modification, as well as whether evolutionary change has resulted from independent diversifying evolution, constraints on distribution due to availability of climate space, or the influence of competing taxa. From a statistical standpoint, each pair of disjunct clades provides an independent assessment of ecological conservatism.
Of approximately 13 500 genera of vascular plants on earth (Mabberley 1997), 57 have disjunct distributions in which some species are restricted to the south-eastern areas of Asia and others to North America. Most are temperate, but several have distribution ranges that extend into tropical areas of the two continents (Li 1952; Qian & Ricklefs 2000; see Appendix S1 in Supplementary Material). The disjunctions within these genera most likely formed when cooling climates in the middle and late Tertiary forced plants, which formerly had continuous distributions across the Bering land bridge, to more southerly temperate regions (Tiffney 1985; Wen 1999; Qian & Ricklefs 2000; Xiang et al. 2000). Fossil data indicate that modern genera had evolved in the North Temperate Zone before temperate forests became fragmented towards the end of the Tertiary (Tiffney 1985; Huntley 1993), but many of these genera disappeared from western North America as mountain building caused climates to become drier during the late Tertiary and Quaternary. European representatives of the eastern Asian (EAS) and eastern North American (ENA) disjunct genera went extinct as a result of late Tertiary climate cooling and Quaternary glaciations (Sauer 1988; Latham & Ricklefs 1993; Manchester 1999; Wen 1999; Svenning 2003).
Species in the EAS-ENA disjunct genera mostly form sister clades in each region (Wen 1999), although some may be paraphyletic and have deeper evolutionary roots in eastern Asia (e.g. Qiu et al. 1995). Their separation between Asia and North America occurred c. 5–25 million years ago, judging from calibrations of the rate of molecular divergence (Wen 1999; Xiang et al. 2000; Donoghue et al. 2001), but most show little morphological diversification between the two regions (Parks & Wendel 1990; Qiu et al. 1995; Wen 1999). Such evolutionary stasis implies stable relationships between plants and key aspects of their environments. If adaptations that influenced distribution were similarly conservative, we would expect to find correlations in the ecological and geographical breadth of contemporary sister lineages between the two continents. Because the EAS-ENA disjunct genera occur primarily in broadleaved deciduous forests in both regions (Li 1952), their two environments are probably correlated, at least to some extent.
Our primary objective was to assess the conservatism of geographical (area and latitude) and climate (temperature and precipitation) distributions among the EAS-ENA disjunct genera of plants. Based on Gaston (1998) and others, most variation in range size within regions occurs at a low phylogenetic level, suggesting considerable lability. Ricklefs & Latham (1992) also found this to be true of the geographical areas occupied by disjunct woody genera, which were uncorrelated between eastern Asia and eastern North America. In contrast, eastern North American and European populations of Fagus have retained closely similar distributions with respect to climate since they were separated 25–10 million years ago (Huntley et al. 1989). This is evident even in newly evolved species in both continents, which suggests that the physiological characteristics that determine species’ distributions are evolutionarily conservative (Huntley et al. 1989). A more recent analysis, based on comparisons of congeners across Europe, eastern Asia and North America (Svenning 2003), also showed that the distributions of cool-temperate tree genera with respect to climate are evolutionarily conservative.
Ricklefs & Latham (1992) found significant correlations between eastern Asia and eastern North America in the latitudinal midpoint of the distribution and, for herbaceous plants, geographical area occupied by the EAS-ENA disjunct genera. However, they did not include climate or other variables that would assess correlations in ecological distribution directly. Furthermore, their floristic database and source of range maps was Li (1952), which is by now out of date. Floristic patterns in China were poorly known before 1952 and, although Li (1952) recorded 500 species in the EAS-ENA disjunct genera, the number has now increased to 770. We therefore re-examine correlations of geographical distribution of EAS-ENA disjunct genera using updated range maps, extending Ricklefs & Latham's (1992) analysis in terms of the quality of the floristic data, the spatial resolution of geographical distributions and the incorporation of climate data with a resolution corresponding to the distributional data. In this study, geographical distribution is quantified with respect to latitude-longitude grid cells of approximately 4° dimension, and climate variables are summarized on the same scale. Following Ricklefs & Latham (1992), we analyse distributions of herbaceous and woody genera separately to determine whether evolutionary changes in distribution differ between groups of plants with different ecological relationships.
Materials and methods
Based on the checklists published by Li (1952), Hong (1993), Wen (1999) and others and current botanical information, we compiled a new checklist that includes 57 Asian-eastern North American disjunct genera of vascular plants (Appendix S1).
The EAS-ENA disjunct genera are distributed between 15° S and 60° N latitude. To document generic floras at a smaller geographical scale within regions, we divided the continental areas into grid cells measuring 3.75° on a side. The area of these grid cells at 20, 30, 40 and 50 degrees of latitude is 1.64, 1.51, 1.34 and 1.12 × 105 km2, respectively. The presence or absence of each of the 57 EAS-ENA disjunct genera in each grid cell was determined according to the distribution maps outlined by Li (1952) and Hong (1993) and additional botanical data (e.g. Flora of North America Editorial Committee 1997; Thompson et al. 1999). Within the region considered, which includes the entire distributions of the 57 disjunct genera, 270 cells (163 cells in eastern Asia and 107 in eastern North America) had at least one of the disjunct genera.
From the distribution maps, we tabulated the total area (km2) covered by the range of each genus in each area, the latitudinal midpoint of its distribution, and the latitudinal range from the northernmost to the southernmost extent of its distribution.
Climate conditions, which include annual precipitation, summer (June through September) precipitation, mean annual temperature, mean January temperature and mean July temperature, were documented for each of the 270 occupied grid cells using the International Institute of Applied System Analysis (IIASA) climatic database (Leemans & Cramer 1991). The IIASA database has been widely used in studies of the plant distribution–climate relationship in eastern Asia and North America (e.g. Monserud et al. 1993; Shao & Halpin 1995). It provides data at 0.5° latitude and 0.5° longitude grid points (Leemans & Cramer 1991); the climate data for all the points included within each 3.75° by 3.75° grid cell were averaged. For each genus, we computed an average for each of the climate variables over all the grid cells it occupied in each continent. Because the northern limits of distributions for many terrestrial plants in the northern hemisphere are thought to represent tolerance of the coldest temperature, we selected, in each continent, five grid cells located at the northern limits of the range of each disjunct genus to characterize these extreme climate conditions.
We conducted two sets of analyses. The first compared the attributes of each disjunct genus in the two regions, using Student t-tests to evaluate the statistical significance of differences in the geographical area and latitude range of the disjunct genera between EAS and ENA. In addition, we conducted analyses of covariance (ancova), with the log10-transformed number of species per genus included as a covariate, to examine the effect of species richness on comparisons of the geographical area and latitude range of genera between the two continents. We used Spearman rank correlation coefficients to quantify the relationship between values for each disjunct genus in EAS and ENA with respect to the midpoint and northern edge latitudes, latitude range, area and climate conditions at the northern edge. We conducted analyses of variance to examine the statistical significance of the differences between the two continents in climate conditions of the geographical areas occupied by the disjunct genera at the latitude-longitude grid dimension of 3.75°.
A second set of analyses compared attributes of the grid cells in each region grouped by a global ordination based on the floristic composition of the disjunct genera in each grid cell. First, we divided the 270 grid cells into groups based on floristic relationships, using two-way indicator species analysis (twinspan; Hill 1979). twinspan is a hierarchical, divisive clustering technique based on dividing a reciprocal averaging ordination space; it has been widely used in analysing ecological data at both small scales (e.g. sample plot data, Legendre & Legendre 1998) and at broad scales (e.g. data at a regional scale, FAUNMAP Working Group 1996). Floristic data subjected to twinspan were the presence or absence of each disjunct genus in each of the 270 grid cells. The twinspan divided the grid cells into four groups (clusters) at the second division level and eight groups at the third division level. We used the four-group division in further analyses because this produced two northerly groups of grid cells that were broadly distributed in both continents. The third and fourth groups of grid cells represented subtropical and tropical extensions of different sets of genera in the two continents, but which nonetheless could be compared with each other with respect to climate variables.
Secondly, we ordinated the grid cells in both continents according to their floristic composition using a global non-metric multidimensional scaling (NMDS; Minchin 1987). We then used stepwise regression to evaluate the relationship between the NMDS axes and the climate variables within each continent to determine the degree to which differences in the geographical distributions of the disjunct genera can be related to climate. NMDS is well suited to floristic data (McCune & Mefford 1999) because it makes no assumptions about the data and provides a robust solution. NMDS optimizes the rank-order correspondence between grid-cell distances in the ordination space and floristic dissimilarities of the disjunct genera between grid cells. NMDS was run on the presence or absence data matrix of the disjunct genera (i.e. 57 genera by 270 grid cells) without rotating the axes, using the Sørensen index as a measure of distance. The maximum number of iterations was set at 200. Three axes provided an optimal solution to the ordination. The means of stress were 46, 23 and 16, respectively, for dimensions 1, 2 and 3. A Monte Carlo test (n = 30) indicated that all the three dimensions were significant (P < 0.05). We also conducted a discriminant analysis (DA, Dillon & Goldstein 1984) to examine the degree to which the division of the 270 grid cells by the twinspan, which resulted solely from the floristic data for the grid cells, matches the pattern of climate data over each region. The DA used the five climate variables as predictor variables and the twinspan division as the grouping variable. The percentage of the grid cells correctly assigned to a twinspan cluster indicates the strength of the association of the distributions of the disjunct genera with the climate variables.
We used systat version 7 (Wilkinson et al. 1992) for statistical tests (e.g. t-test and analyses of variance and covariance), correlation analysis and discriminant analysis, and used PC-ORD version 4 (McCune & Mefford 1999) for other multivariate analyses (i.e. twinspan and NMDS). Details on statistical analyses are explained in the Results section in connection with each particular test or analysis.
correlation in geographical distribution between eas and ena
Comparisons of generic distributions
For woody and herbaceous genera taken together, the pattern of generic richness in relation to latitude is similar in EAS and ENA (Fig. 1). For example, the highest diversity of the EAS-ENA disjunct genera occurred in temperate areas between 25° and 45° N in both continental regions. The largest number of disjunct genera in any single grid cell was 43 in EAS and 50 in ENA.
The disjunct genera each occupy, on average, significantly larger areas in EAS than in ENA (30.9 × 105 vs. 21.5 × 105 km2 per genus, t = 3.12, d.f. = 56, P = 0.003). This difference in area is roughly proportional to the relative number of grid cells occupied by all the disjunct genera in eastern Asia (163) and eastern North America (107). When woody and herbaceous genera were compared separately between the two continents, differences in occupied area remained significant among woody genera (33.6 × 105 vs. 18.8 × 105 km2 per genus, t = 3.29, d.f. = 29, P = 0.003), but not among herbaceous genera (27.7 × 105 vs. 24.6 × 105 km2 per genus, t = 0.90, d.f. = 26, P = 0.4). Latitude ranges of the disjunct genera in eastern Asia exceeded those in eastern North America for both growth forms: 24.5° vs. 17.0°, t = 4.15, d.f. = 56, P < 0.001 for all genera; 26.2° vs. 17.1°, t = 2.90, d.f. = 29, P = 0.007 for woody genera; 22.6° vs. 16.9°, t = 3.65, d.f. = 26, P = 0.001 for herbaceous genera. However, when the log10-transformed number of species per genus was included as a covariate in analysis of covariance, region was not a significant effect for woody genera (F1,57 = 0.92, P = 0.34 for area, F1,57 = 0.36, P = 0.55 for latitude range), and region was only marginally significant for latitude range among herbaceous genera (F1,51 = 0.02, P = 0.90 for area, F1,51 = 4.24, P = 0.05 for latitude range). Thus, it would appear that the disjunct lineages have diversified in terms of number of species and geographical range in proportion to the area of suitable environmental conditions in each continent.
From the standpoint of assessing long-term ecological conservatism of disjunct lineages, correlations in distribution between the continents provide the essential information. For woody and herbaceous genera combined, the latitude midpoints of the disjunct genera extended over a wider latitude range, both to the north and to the south, in eastern Asia than in eastern North America (Fig. 2). Correlations calculated for the midpoint latitude, northernmost latitude, latitude range and area of the disjunct genera between the two regions were all significant (rs ranging from 0.307 to 0.595, P < 0.05), with midpoint latitudes yielding the strongest correlation (Table 1). These relationships remained stable in multiple regressions that included the number of species in each region as an independent variable.
Table 1. Spearman rank correlation coefficients for geographical and climate variables of the eastern Asian-eastern North American disjunct genera of vascular plants in the two continents
Spearman rank correlation coefficients (rs)
Overall (n = 57)
Woody (n = 30)
Herbaceous (n = 27)
Correlations for geographical variables were calculated based on metrics obtained from all 270 grid cells. Correlations for climate variables based on the average value for the grid cells occupied by each disjunct genus of the 89 grid cells of the twinspan cluster B. This cluster comprises most of the distributions of most of the genera and identifies the core area of all these genera.
Latitudinal distribution is closely related to growth form in both eastern Asia and eastern North America. Woody genera were distributed in significantly lower latitudes than herbaceous genera with respect to both their midpoint latitudes (26.1 ± 9.4° vs. 34.9 ± 4.6°, t = −4.6, d.f. = 55, P < 0.001 in EAS; 33.0 ± 4.8° vs. 37.3 ± 2.5°, t = −4.3, d.f. = 55, P = 0.001 in ENA) and the northern edges of their distributions (39.2 ± 7.5° vs. 46.2 ± 5.9°, t = −4.0, d.f. = 55, P < 0.001 in EAS; 41.6 ± 4.5° vs. 45.7 ± 4.7°, t = −3.38, d.f. = 55, P = 0.001 in ENA). For woody genera, the correlation between the two continents was insignificant for northern edge, latitude range and area, but significant for midpoint of latitude (Table 1). This pattern for woody genera contrasts with that for herbaceous genera, in which northernmost latitude, latitude range and area were significantly correlated between regions (Table 1). The mean latitudes of the northern edges did not differ between the two continents in either woody (t = −1.52, d.f. = 29, P = 0.14) or herbaceous genera (t = 0.45, d.f. = 26, P = 0.66).
Comparisons of twinspan grid cell clusters
The twinspan analysis divided the 270 grid cells containing the EAS-ENA disjunct genera into four clusters at the second division level (Fig. 3). The two southern clusters (i.e. clusters C and D), located in tropical regions, were primarily unique to one of the two continental regions; all of the 49 cluster-C grid cells were restricted to EAS and 38 of the 39 cluster-D grid cells were restricted to ENA. The grid cells of these two clusters each had relatively few EAS-ENA disjunct genera (6.8 ± 4.1 in cluster C and 3.1 ± 2.5 in cluster D). For the most part, different EAS-ENA disjunct genera have extended their ranges and diversified in tropical latitudes in each region (11 unique to EAS, five unique to ENA, and 11 in both regions; see Appendix S1). The cluster-B grid cells (53 in EAS and 36 in ENA), located in temperate regions in both continents, include the major concentration of the disjunct genera (compare Fig. 3 with Fig. 1). All 57 disjunct genera were distributed in at least some of the grid cells of this cluster in both continental regions. On average, each of the grid cells in this cluster had 28.5 ± 9.6 disjunct genera. The latitudinal ranges of the cluster-B grid cells were similar in the two continental regions (Fig. 3). Cluster A had 60 grid cells in EAS and 33 in ENA. Most of the grid cells in this cluster were located in cold climates to the north and dry climates to the west, and represent extensions of the main distributions of 24 and 28 disjunct genera, respectively, from their distributional centres. The richness of the disjunct genera in the grid cells of this cluster was low, averaging only 4.7 genera per grid cell.
In the NMDS ordination, the grid cells of cluster B were generally clustered tightly on all three facets of the three-dimensional NMDS ordination, indicating that the EAS-ENA disjunct genera in those grid cells are similar to a large degree (Fig. 4a). The grid cells of clusters C and D also formed tight clusters in a similar region of the NMDS space, which reflects the extension of these clusters towards wet, tropical environments in both continents. However, the grid cells of clusters C and D were also displaced with respect to each other, especially on NMDS axis 3, reflecting the different genera that have tropical extensions in each continent (see Appendix S1). The grid cells of cluster A, which was shared by EAS and ENA, were distributed around the periphery of the grid cells of cluster B in all the three facets of the NMDS ordination. This pattern is generally in accordance with the geographical distribution pattern of these grid cells (compare Fig. 4a with Fig. 3), and results from the fact that different subsets of genera have extended their distributions towards colder vs. drier climates.
The scores of NMDS axis 3 and latitudes of the grid cells of cluster B were highly positively correlated in both continental regions (r = 0.90, P < 0.001 in EAS; r = 0.91, P < 0.001 in ENA). This suggests that the composition of the disjunct genera in the grid cells of cluster B changes in parallel along a latitudinal gradient in the two continental regions, which is in agreement with the significant correlations in the latitude midpoints of the 57 disjunct genera between EAS and ENA as discussed above. This is also consistent with the placement of clusters C and D relative to cluster B on NMDS axis 3.
correlation in climate between eas and ena
Comparison of generic distributions
Correlations in all the five climate variables at the northern boundaries between EAS and ENA were not significant for woody disjunct genera but, except for summer precipitation, they were significant for herbaceous disjunct genera (Table 2). On average, mean annual temperature, mean January temperature and annual precipitation at the northern boundaries of the disjunct genera were higher in ENA than in EAS (Table 2).
Table 2. Spearman rank correlation coefficients and t-tests for the means of the selected climate variables of northern limit latitudes of the eastern Asian (EAS)-eastern North American (ENA) disjunct genera of vascular plants between the two continents
Correlation coefficients (rs)
Mean ± SD (n = 57)
Number of genera: n = 30 for woody plants, n = 27 for herbaceous plants.
As discussed above, the distributions of the EAS-ENA disjunct genera in the grid cells of clusters A, C and D represent the expansions of their distributions from the centres of the disjunct genera in the cluster-B grid cells. We will focus on the cluster-B grid cells in further analyses because this cluster comprises the largest part of the distributions of most of the genera and identifies the core area of all the EAS-ENA disjunct genera, having the highest generic diversity. The averages of each of the five climate variables for the cluster-B grid cells occupied by each disjunct genus were significantly correlated between EAS and ENA (P < 0.05) (Table 1). When woody and herbaceous genera were considered separately, the means of the climate variables were not significantly correlated between the two regions for woody genera, except for summer precipitation, but they were significantly correlated for herbaceous genera, except for mean July temperature (Table 1).
Comparisons of the regions occupied by disjunct genera
Annual precipitation did not differ significantly (P > 0.05) between the grid cells of EAS and those of ENA, whether all the grid cells or only those in cluster B were considered (Table 3). In contrast, summer precipitation was significantly higher, particularly among cluster B cells, and temperatures were significantly lower, although less so in cluster B cells, in EAS than in ENA, reflecting the summer monsoon climate of eastern Asia (Table 3).
Table 3. Descriptive and anova statistics of the selected climate variables for the grid cells in which the eastern Asian (EAS)-eastern North American (ENA) disjunct genera are distributed
Mean ± SD
All grid cells
No. of grid cells
Annual precipitation (mm)
1305.38 ± 915.49
1203.05 ± 607.96
Summer precipitation (mm)
618.98 ± 450.09
518.64 ± 285.00
Mean annual temperature (°C)
10.52 ± 12.37
15.23 ± 9.28
Mean January temperature (°C)
−0.95 ± 20.64
6.59 ± 15.25
Mean July temperature (°C)
20.40 ± 5.67
23.17 ± 4.42
Cluster-B grid cells only
No. of grid cells
Annual precipitation (mm)
1245.0 ± 513.4
1115.0 ± 226.6
Summer precipitation (mm)
739.3 ± 333.7
446.1 ± 143.7
Mean annual temperature (°C)
10.8 ± 6.0
13.9 ± 6.5
Mean January temperature (°C)
−0.6 ± 8.6
3.1 ± 9.6
Mean July temperature (°C)
21.2 ± 5.4
24.4 ± 3.6
relationship between distribution and climate in eas and ena
Although the grid cells in both continents were ordinated with respect to their floristic composition in the same analysis, the relationship between the NMDS axes and climate variables differed somewhat in detail between the two continents (Table 4). Precipitation varied along all three NMDS axes in eastern Asia, but did not enter into the stepwise regressions in eastern North America. Temperature gradients along the NMDS axes were greater in ENA than in EAS. The strongest correlation with climate variables appeared for NMDS 3 (R2 = 0.76, 0.88), for which increasing values were associated with higher summer temperatures and greater seasonality of temperature. The significant association of this axis with higher summer precipitation in eastern Asia reflects the different climate patterns in eastern Asia, which are associated with gradients in temperature across the continent, and for which the association with NMDS 3 might be fortuitous.
Table 4. Significant variables in stepwise regressions of each of the non-metric multidimensional scaling (NMDS) axes with respect to five climate variables in each continent (EAS, n = 163; ENA, n = 107)
Bold type, P < 0.0001; italic type, P < 0.001; normal type, P < 0.05.
Discriminant analysis (DA) based on the five climate variables separated the four clusters of the 270 grid cells derived from the twinspan (Wilks’λ = 0.162; F = 45.1, d.f. = 15 723, P < 0.001) (Fig. 4b). The DA correctly classified about 73% of the 270 grid cells into the four twinspan clusters. Of the 89 grid cells in cluster B, 66 (74%) were correctly classified. For the 23 grid cells ‘wrongly’ classified by the DA, 17 (74%) were assigned into cluster A, which belongs to the same first division as cluster B in the twinspan analysis, and the other six into clusters C and D.
The spatial scale used in the present study in assessing the conservatism of geographical and climate distributions among the eastern Asian-eastern North American disjunct genera of plants was at the latitude-longitude grid dimension of 3.75°. This grid represents an intermediate grain size compared with those used in many previous studies investigating large-scale relationships between taxon diversity and geographical and climatic variables, which usually range from 1° by 1° to 10° by 10° latitude-longitude grids (e.g. Rahbek & Graves 2001). Although the climate data are available at a finer scale than we have used in recording geographical distribution, finer resolution of distribution, such as that used by Rahbek & Graves (2001), was not warranted because of the coarseness of the distribution maps, interpolation of the climate data between stations more widely spread than 0.5°, and increased pseudoreplication in statistical analyses owing to spatial autocorrelation. Positive spatial autocorrelation is a common feature of ecological data, which can reduce the power of statistical tests because the close proximity of some pairs of observations reduces their independence as samples (Legendre & Fortin 1989). This is a problem associated with most, if not all, studies based on spatial data, but it is unlikely to affect the results and conclusions of this study, including, for example, those in Table 2. The statistical tests in Table 2 provide general comparisons of climate variables between the two continents. For those tests that are not significant, the inflated degrees of freedom resulting from positive spatial autocorrelation have no consequence. For those tests whose F-values are about 10 or higher, even if the number of observations were reduced to 20, the tests would still be significant at P < 0.01. For the two tests that are marginally significant in Table 2 (F = 4.2 and 5.4), positive spatial autocorrelation might affect the interpretation of the tests. However, because their significance is at a marginal level, we interpreted these tests conservatively anyway.
Phylogenetic relationship is a concern when traits are fixed early in the evolution of a clade and are inherited by the descendants of ancestors showing these traits. We assessed whether such relationships posed a potential problem in this analysis by nested analysis of variance (anova) based on a hierarchy of taxonomic levels (Derrickson & Ricklefs 1988; Bell 1989). Although other methods have been developed based on phylogenetic trees, adequately supported hypotheses for the relationships among most angiosperm clades are not available (Soltis et al. 2000), and so tree-based approaches offer no advantage over taxonomic approaches at this point. We used the orders and families of the Angiosperm Phylogeny Group (2003), which are based on the best molecular phylogenetic information presently available, to estimate the distribution of variance among genera within families, among families within orders, and among orders within flowering plants (see Appendix S2). For each of the traits (latitude and climate variables) considered in this analysis, the proportion of the variance at the lowest level varied between 49 and 99%. None of the family or order-level effects was significant in nested models using general linear models. We randomly trimmed the data set to obtain a balanced design that permitted statistical appraisal of variation at each level in the nested hierarchy using regular nested anova. In this analysis, 32–100% of the variation was at the level of genera within families, and only one variance component at a higher level (northernmost latitude in eastern Asia) was significant. Thus, most of the variance in the traits we considered resides at the lowest taxonomic level of our study, obviating the need for phylogenetic corrections in the analyses.
The primary purpose of our analyses was to determine the correlation in geographical distribution of disjunct genera between eastern Asia and eastern North America and the degree to which these correlations reflect similar distributions of the disjunct genera with respect to climate variables in each continent. The key results of the analyses can be summarized as follows. (i) Individual disjunct genera occupied larger areas and latitudinal ranges in eastern Asia than they did in eastern North America, in parallel with the larger area and latitudinal range of suitable climate in eastern Asia. (ii) Area and latitude of lineages on one continent were correlated with those on the other continent, but the correlations applied primarily to herbaceous genera, which in general had more northerly distributions than woody genera. (iii) In spite of the generally cooler and more seasonal (monsoon) climate of eastern Asia, the distributions of disjunct genera were similarly related to climate variables in both continents.
Ricklefs & Latham (1992) found a significant correlation of midpoints of latitudes between disjunct genera in the two continental regions (r = 0.42). The updated distributions used in the present study resulted in a similar, but higher correlation (r = 0.60). For all EAS-ENA disjunct genera taken together, correlations between the continents also were significant for the latitude at the northern edge of the range, the range of latitudes between the northern and southern edges, and the area occupied by each genus. However, these patterns of correlation differed between herbaceous plants, for which all the geographical components, except for midpoint of latitude, were significantly related, and woody plants, for which only the midpoint of latitude had a significant correlation between regions.
Thus, the EAS-ENA disjunct genera present two related but contrasting phenomena. The more straightforward pattern is the parallelism in the distributions of sister taxa isolated from common ancestral distributions on different continents, in many cases over periods of many millions of years through times of extreme climatic and physiographic change. The general lack of such parallelism among woody taxa, compared with herbaceous taxa, is more difficult to place in context, but might be related to differences between the two growth forms in the length of disjunction time, accessibility to tropical areas, sensitivity to environmental change, and evolutionary stasis (Ricklefs & Latham 1992).
Time of disjunction
Area parallelism in herbaceous disjunct genera together with the fact that these genera occur farther to the north than woody disjunct genera suggests that the persistent correlations in geographical range of the disjunct congeners might be due to later separation of herbaceous disjunct genera compared with woody genera. This hypothesis is marginally supported by divergence times estimated from genetic distance: 8.5 ± 6.1 × 106 years for herbaceous genera (n = 7) and 17.8 ± 5.6 × 106 years for woody genera (n = 6) (t = 2.8, P < 0.05) (Parks & Wendel 1990; Wen & Jansen 1995; Lee et al. 1996; Schnabel & Wendel 1998; Donoghue et al. 2001). Further resolution of the timing of disjunction and its relationship to the degree of correlation between traits in descendant lineages awaits further analyses of genetic distance.
Connection to the tropics and geographical heterogeneity
The woody members of the disjunct genera are generally distributed more to the south than are herbaceous ones in both eastern Asia and eastern North America. A smaller area of tropical and subtropical climate in eastern North America compared with eastern Asia might have weakened the correlation among the geographical ranges of the more southerly distributed woody disjunct congeners between the two regions. Several of the woody disjunct genera in eastern Asia extend their ranges southward into the tropics. For both continents combined, 18 of 30 woody genera extend into the grids of clusters C or D, whereas only nine of 27 herbaceous genera do (G statistic (Sokal & Rohlf 1981): G = 4.1, P < 0.05). Moreover, different woody genera exhibit more southerly distributions in Asia and eastern North America, which would tend to reduce the correlation between the areas occupied in each continent. Most of the 30 woody disjunct genera occur between 25 and 30° N latitude in eastern Asia, often at mid- to high elevations. Those latitudes in eastern North America are in the Gulf of Mexico except for the low-lying Florida Peninsula. Although Florida extends southwards to approximately 25° N latitude, nine of the 30 woody disjunct genera in North America do not extend south of 30° N and only four genera reach the southern part of the Florida Peninsula (Wunderlin 1998). Several of the woody disjunct genera that are absent from southern Florida (e.g. Carya, Gelsemium and Nyssa) occur in more southerly latitudes, generally at higher elevations, in Central America (e.g. Gelsemium extends southwards to 15° N), but few reach northern South America, possibly due to the connection of North America and South America being relatively recent and restricted in area. Whether or not this factor might also underlie the non-parallelism in climate at the northern distribution ranges between the two continents, the inability of the disjunct genera to extend their distributions to more southerly latitudes in the Americas apparently did result in a relative northerly ‘shift’ of the centres of the distribution ranges in eastern North America compared with eastern Asia.
Sensitivity to environmental change and evolutionary stasis
Correlations between regions may differ for woody and herbaceous members of the EAS-ENA disjunct genera because each is sensitive to different scales of spatial heterogeneity when they respond to change in the environment (Ricklefs & Latham 1992). The herbaceous disjunct genera mostly inhabit temperate mesic forests, and they probably evolved in association with woody disjuncts under the same range of climates. However, herbaceous taxa are smaller and presumably they can persist in smaller microhabitat or edaphic spaces compared with woody taxa, especially trees. When rapid change in the environment causes the boundaries of niche space to shift, the relatively smaller niches of herbaceous taxa may be shifted from one site to another but they may still persist in the same area. The larger niche spaces of woody taxa are more likely to be altered by general environmental change, especially climate change. Accordingly, herbaceous taxa would face less directional selective pressure and exhibit greater evolutionary stasis or conservatism in ecological attributes compared with woody taxa. Among the EAS-ENA disjuncts, herbaceous genera appear to be ecologically more specialized and evolutionarily more conservative than are woody genera (Ricklefs & Latham 1992; Guo & Ricklefs 2000). Thus, although climate and other aspects of the environment have changed during the past millions of years (Wolfe 1979), herbaceous taxa are so specialized (Antos 1988) that the narrowly circumscribed parameters of their niches may persist over a large geographical area in spite of changing composition of microhabitat patches locally (Ricklefs & Latham 1992). The strong parallelism in area of geographical range for herbaceous genera is consistent with the hypothesis that the extant herbaceous disjunct genera have exhibited stable ecological requirements and tolerance over evolutionary time. This hypothesis should be tested by finer scale comparisons of distribution and adaptation in herbaceous and woody plants.
climate and geographical distribution
The distribution limits of terrestrial plants frequently correspond to particular climate conditions (Ricklefs & Miller 1999; Gaston 2003), and the EAS-ENA disjunct genera are no exception. Among many climate variables, temperature and precipitation are considered major limiting factors to most terrestrial plant species (Brown & Lomolino 1998). Accordingly, it is not surprising that the average conditions within the distributions of the disjunct genera and conditions at the northern edges of their ranges are correlated between the two continents, just as their geographical ranges are. Strong correlations in climate conditions were also found in a larger sample of cold-temperate congeneric trees between Europe, eastern Asia and North America (Svenning 2003).
The correlation of geographical patterns of the EAS-ENA disjunct genera between the two continental regions, particularly the central areas of their distributions, can be predicted by climate variables to a large degree. The distributional centres of each group of grid cells identified by twinspan formed tight clusters within which the grid cells of the two continental regions were intermingled in both NMDS and discriminant analyses (DA). Departures from strong correlations in some geographical attributes, especially for woody genera, may to some extent result from the poorer success of climate variables in predicting the edges of the geographical ranges of disjunct genera at the grid scale of this study. The microclimate conditions under which individuals of disjunct genera grow at their distributional edges may be more similar to the optimal climate conditions at their distribution centres than the regional climate conditions at the edges of their distribution ranges. Thus, the climate data that we used may be unable to resolve geographical patterns at the edges of the distributions of the disjunct genera.
On average, temperature and precipitation at the northern boundaries of the disjunct genera are higher in eastern North America than in eastern Asia. This difference can arise in several ways. First, more cold-tolerant species among the disjunct congeners might have gone extinct in eastern North America, where Pleistocene glaciations were more extensive than in eastern Asia (Pielou 1991). During the Last Glacial Maximum (18 000 year BP), boreal forest or ice sheets occupied much of the current range of the disjunct genera in North America (Williams et al. 2000). Possibly, some species of the disjunct genera that were primarily distributed in the northern part of eastern North America prior to Pleistocene glaciations and were unable to track their shifting environments might have gone extinct (Latham & Ricklefs 1993; Jackson & Weng 1999).
Secondly, some eastern North American species that survived from the last glaciation might have adapted to altered local environments, for example towards warmer temperatures during deglaciation. If this were the case, one would expect that species in eastern North America would be distributed in areas of lower elevations and warmer climates compared with congeneric species in eastern Asia. One such example is Liriodendron, a well-known EAS-ENA disjunct tree genus. This genus has two extant species, L. chinense in eastern Asia and L. tulipifera in eastern North America. The Asian species occurs in widely scattered populations distributed in specific habitats, usually at mid- to high elevations. In contrast, the American species is broadly distributed at low elevations and often becomes weedy over much of its range (Parks et al. 1983).
Thirdly, the northern edges of the ranges of the disjunct genera might not have reached climatic equilibrium following glaciations in eastern North America. Wright (1964), Davis (1976) and Ricklefs & Latham (1992), among others, suggested that this might explain variation in the ranges of woody taxa. However, several palaeoecologists (e.g. Webb 1986, 1988; Huntley & Webb 1989; Margaret B. Davis, personal communication) have rejected this idea, arguing that plants migrated quickly enough following the retreat of the glaciers to maintain their ranges in climatic equilibrium. Webb and his coworkers (Webb 1988; Overpeck et al. 1992) argue that much of eastern American vegetation before 9000 year BP had modern analogs and that the modern mixed forest began to develop 6000 years ago. This does not, however, necessarily mean that the northern edges of the ranges of all the North American plants have reached climatic equilibrium. For example, a recent study (Johnstone & Chapin 2003) shows that the northern distribution of Pinus contorta var. latifolia is not in equilibrium with current climate.
Our general conclusion is that strong distributional and climatic parallels between disjunct lineages in eastern Asia and eastern North America argue that adaptations governing the relationship of plants to their environments, climatic conditions in particular, are generally conservative. However, differences in the climate histories and geographical distributions of habitat types within regions may have caused variations in distribution that reduced correlations between regions. It is not surprising in this analysis that the observed patterns of distribution of disjunct genera reflect both phylogenetic inertia in the adaptations of plants to their environments and the special history and geography of each region.
We thank three anonymous referees and David J. Gibson for insightful comments on the manuscript. We thank Changhui Peng and Guofan Shao for climate data.