Distribution of abundance across the range in eastern North American trees

Helen T Murphy1, Jeremy VanDerWal and Jon Lovett-Doust

1 Author for correspondence

Department of Biological Sciences

University of Windsor

Windsor Ontario N9B 3P4

Canada

Phone: 1-519-253 3000 ext 2722

Fax: 1-519-971 3609

Email: [email protected]

Running title: Macroecology of eastern North American trees.

1 ABSTRACT

Aim We analyzed spatial datasets of abundance across the entire, or near entirety of the geographic ranges of 134 tree species to test macroecological hypotheses concerning the distribution of abundance across geographic ranges.

Location Our abundance estimates come via the USDA Forest Service Forest

Inventory and Analysis Eastwide Database, which contains data for 134 eastern North

American tree species.

Methods We extracted measures of range size and the spatial location of abundance relative to position in the range for each species to test four hypotheses: (a) species occur in low abundance throughout most of their geographic range; (b) there is a positive interspecific relationship between abundance and range size; (c) species are more abundant in the centre of their range; and (d) there is a bimodal distribution of spatial autocorrelation in abundance across a species range.

Results Our results demonstrate that (a) most species (85%) are abundant somewhere in their geographic range; (b) species achieving relatively high abundance tend to have larger range sizes; (c) the widely held assumption that species exhibit an

‘abundant-centre distribution’ is not well supported for the majority of species; we suggest ‘abundant-core’ as a more suitable term; and (d) there is no evidence of a bimodal distribution of spatial autocorrelation in abundance.

Main Conclusions For many tree species, high abundance can be achieved at any position in the range, though suitable sites are found with less frequency toward range edges. Competitive relationships may be involved in the distribution of abundance across tree ranges and species with larger ranges (and possibly broader niches) may be affected more by biotic interactions than smaller ranging species.

2 Keywords abundance, distribution, range, abundant-centre, macroecology, spatial autocorrelation, occupancy

3 INTRODUCTION

Increased understanding of landscape- and regional-scale processes in the distribution and abundance of organisms has led to the relatively recent emergence of macroecology (Gaston & Blackburn, 2000), and an awareness of patterns and processes that are otherwise indistinguishable at smaller scales. Macroecology involves the ecology of “wide expanses of space, long periods of time and large numbers of taxa” (Blackburn & Gaston, 2003), and addresses patterns of distribution and abundance of species at geographical spatial scales and evolutionary time scales

(Brown, 1995). Thus complex relationships between variables, such as species range size and individual abundances, and location in the range and abundance, underpin most macroecological theories. Whereas distribution and abundance have been the major emphases of investigation thus far, here we focus on a different perspective: the distribution of abundances across geographic ranges.

Tree-dominated communities are among the richest in ecological structure and biotic diversity, in both above- and below-ground aspects. How large numbers of competing species manage to coexist remains a major unresolved problem in community ecology and it is particularly acute for plants, given that most plants require remarkably similar resources and acquire them in similar ways (Silvertown,

2004). Here we use available spatial datasets of abundance (in the form of phytosociological ‘importance values’) and range boundaries for each of 134 eastern

North American tree species (representing 29 taxonomic families) (see Table S1 in

Supplementary Material) to test four major macroecological hypotheses: (a) most species occur in low abundance throughout most of their geographic range; (b) there is a positive interspecific relationship between abundance and range size; (c) species

4 are more abundant in the centre of their range; and (d) there is a bimodal distribution of spatial autocorrelation in abundance across a species range. We discuss the results in terms of problems of plant coexistence and the role of both abiotic and biotic factors in regulating species distribution and abundance.

When abundance distributions have been examined over entire geographic ranges (or nearly entire ranges) it has generally been observed that species exist in low abundance at most sites and at high abundance in only a few sites (Brown et al.,

1995). Furthermore, Murray & Lepschi (2004) showed that the majority of species that are locally rare in open forest communities in Australia were abundant at other locations within their geographic range (existing as ‘somewhere abundant’ species); thus sampling at any given point in a landscape should capture most species at low abundance and a few species in high abundance.

The tendency for individual species to show positive relationships between abundance and range size (or occupancy) has been widely reported for a variety of taxa and appears to be remarkably robust over multiple spatial scales and across a variety of measures (Gaston et al., 2000). At least eight major mechanisms have been described, both biological and artifactual, that may account for positive interspecific abundance-range size relationships (Gaston et al., 1997). These include sampling artifacts and phylogenetic non-independence, individual species attributes and population dynamics. Studies of interspecific abundance-occupancy relationships in plants, although generally supporting the positive relationship have tended to focus on narrowly defined habitat types, or on abundance in only a small proportion of the species’ range (Gotelli & Simberloff, 1987; Collins & Glenn, 1990).

5

The ‘abundant–centre distribution’ describes the widely held assumption in biogeography and macroecology that species abundances tend to be greater toward the centre of the geographical range and lower in the periphery (Brown, 1984; Sagarin

& Gaines, 2002). In an influential paper Brown (1984) argued that local abundance reflects how well a particular site meets a species’ physiological and ecological requirements. Brown suggested that spatial autocorrelation in these axes

(representing dominant dimensions of the niche) could underpin the abundant-centre distribution. Increasing the distance from the optimal site should decrease the probability of a site meeting the multidimensional needs of that species (Brown,

1984). Thus there are three main predictions of the abundant-centre distribution: (1) abundance should be higher in sites closer to the centre of the range; (2) species should occupy more sites towards the centre of the range; and (3) abundance and occupancy should decline linearly towards the edge of the range.

Brown et al. (1995) reported that abundance exhibited a distinctive bimodal pattern of spatial variation within the geographic ranges of four passerine birds. Spatial autocorrelation analysis produced two peaks, one at short distances (corresponding to proximal sites within the geographic range), and one at the maximum distance

(corresponding to sites at opposite ends of the geographic range). Thus sites close together were more likely to have similar abundances irrespective of where they were located in the range, and sites at the range periphery tended to have consistently low abundance, producing the second peak at maximum distances. This second peak in spatial autocorrelation at maximum distances implies that species respond similarly to all aspects of the range edge.

6

Critical examination of macroecological theories in the past has suffered from a paucity of good quality datasets (Sagarin & Gaines, 2002; Mathius et al., 2004) and analyses have often been based on inadequate range size information, restricted spatial coverage of abundance data, and/or limited taxonomic extent (Sagarin &

Gaines, 2002; Kotze, 2003; Blackburn et al., 2004). Fortunately, a growing availability of large scale geographic information system (GIS) datasets and sophisticated tools for spatial analysis has enabled patterns of distribution of abundance within species’ ranges to be examined in more detail. The spatial data on both abundance and distribution which we use here has been collected and compiled in a consistent way for all species in the dataset. Furthermore, the complete distribution of species’ abundances within the geographic range is known for most species. Thus many of the problems and biases relating to limited, inadequate or unreliable data previously identified in macroecology studies are avoided (see e.g.,

Sagarin & Gaines, 2002; Blackburn et al., 2004). However, our use of importance values rather then absolute abundance does have some limitations which are discussed further below.

METHODS

Abundance estimates (phytosociological importance values) were available at the

Eastwide Database from U.S.D.A. Forest Service Forest Inventory and Analysis data

(Prasad & Iverson, 2003), and comprise >100,000 plots and records for c. three million trees in thirty-seven states. Importance values are calculated for each species as:

7 IV(x) = 50*BA(x) / BA(all species) + 50*NS(x) / NS(all species)

where x is a particular species at a plot, BA is basal area, and NS is number of stems

(summed for overstory and understory individuals). In monotypic stands, the IV would reach the maximum of 100. Importance value scores are averaged for 20 x 20 km cells and are available for the U.S. east of the 100th meridian (total of 8407 cells,

3.36 million km2) (Fig. 1).

Importance values effectively reduce detailed datasets for individual species in local stands to standardized values, depending on the total number of tree species present, and are produced as composite values representing both a species’ density and relative dominance, compared to other species in that community. Thus importance values represent relative abundance rather than absolute abundance. The use of relative abundance measures has some limitations that should be kept in mind when patterns of abundance are compared across sites that differ in overall vegetation density. For example, an IV of 100 in a cell indicates a monoculture regardless of whether the cell contains 100 trees or just one tree (a highly unlikely scenario since it is forested plots that are surveyed). Similarly, a species with a low IV value may still be quite abundant but occur in a very high density site (thus its relative abundance is low). In the absence of absolute abundance data, in this analysis we compare relative abundance across sites with the assumption that a species occurs, on average, in sites with similar overall density and standing biomass across its range. Indeed, Enquist & Niklas (2001) have shown that in tree-dominated communities standing biomass varies little with respect to latitude, species number, or tree species composition.

Although tree density in plots likely increases from northern to southern latitudes (Enquist & Nicklas, 2001), averaging of individual plot IV to 20 x 20 km cells, and further averaging of IV’s in distance from edge-of-range classes, where each distance class includes cells in latitudes both north and south of the centre of the range (see below), minimises issues of variation in plot density.

8

Between 1971 and 1977, Elbert Little, Chief Dendrologist with the U.S.D.A. Forest

Service published a series of maps of tree species ranges based on inventory lists, detailed forest surveys, field notes and herbarium specimens (Prasad & Iverson,

2003). These published (and now digitized) maps have become the standard reference for most U.S. and Canadian tree species ranges. We calculated two indices of range size: the number of cells for which an IV was recorded for the cell (representing its

‘area of occupancy’) for species for which essentially the entire range is encompassed by the IV study area; and also via the number of cells within the range area

(representing the ‘extent of occurrence’ of the species).

‘Centre of range’ may be taken to mean the point from which the species originated and dispersed to all other points, or it may mean the geographical centre of the current distribution. Either way it is a difficult point (both figuratively and literally) to identify precisely for many species. We do not define a priori the centre of a species range. Instead we measure the distance from each cell to the nearest edge-of-range.

In this case ‘edge-of-range’ is the nearest boundary as per Little’s range maps. The total range size for each species was separated into ten even distance-from-edge classes to allow for comparison of mean abundance values at various distance classes.

Thus distance class 1 represents cells in the highest 10% of the range of distances from the edge (i.e., points closest to the ‘centre’) for a given species, and class 10 represents those in the lowest 10% of distances from the edge (i.e., points closest to the edge).

9 The distance between the centre of each occupied cell and every other occupied cell was calculated, and for each species we produced spatial autocorrelograms (Brown et al., 1995) of Moran’s I (covariance between points at a given distance divided by variance of all points), for all pairs of points at ten spatial ‘lags’ (i.e., the maximum distance between occupied cells divided into ten even distance increments). The value of Moran’s I is a rough analogue to the Pearson correlation coefficient and generally varies between 1 and −1. Positive autocorrelation in the data translates into positive values of I, negative autocorrelation into negative values, and values close to zero represent no spatial autocorrelation (Legendre & Legendre, 1998).

Data were extracted from the IV and range maps using ArcGIS 9.0 (ESRI Inc,

Redlands CA). All statistics were conducted using SPSS 12 for Windows (SPSS Inc,

Chicago IL). One-way ANOVAs with Tukey’s post-hoc tests were used to test for significant differences between average IV values (log transformed) at increasing distance classes from the centre to the edge of range.

In order to control for possible effects of phylogeny on the results, we performed analyses on all species and then also following serial removal of species in each of the two largest taxonomic orders: Fagales (forty-seven species) and Pinales (twenty-four species). Removal of the largest orders did not change the general pattern of results for any of the analyses, therefore only the results with all species included are reported here.

RESULTS

10 About half of the 134 species in the study have relatively small ranges (<1.25 million km2). The frequency distribution of range sizes was distinctly log-normal (Fig. 2).

The mean number of overlapping ranges in any cell in the study area was 43.3 (SE

0.166, range 1-73). The average number of tree species per cell was 18.5 (SE 0.095, range 1 – 44), with an average IV in each cell of 0.720. There was a weak negative correlation between the number of species occupying a cell and the average IV of those species (r = -0.160, p<0.001, n = 8282).

Abundance-occupancy relationships

Most tree species occurred in relatively low abundance throughout most of their range

(IV <15 across 97% of the range) (see Table 1), while some 50% of species reached high abundance (IV >40) in some part of their range. Only 25 species recorded an IV

>81, indicating at least one cell of very high abundance somewhere within the range, and only one of those species (slash pine, Pinus elloittii) had an IV >81 over more than 1% of its range (Table 1). On average, 60% of cells within a species’ range were unoccupied by it. Fully fifteen percent of species never achieved an IV >15 in any cell, anywhere in the range (Table 1).

We grouped species into four categories based on their abundance distributions: (1) everywhere sparse (IV always <15) (sensu Murray et al., 1999), (2) somewhere abundant (max IV between 15-40), (3) somewhere high abundance (max IV between

41-80), (4) somewhere very high abundance (max IV between 81-100) (see Table S1 in Supplementary Material). Abundance category was related to range size (extent of occurrence) (Fig. 3), with Category 1 species (everywhere sparse) having the smallest ranges and Category 4 the largest. In addition our results indicate a weak but

11 significant positive correlation between mean IV (range 0.001 – 21.8) and range size

(estimated as extent of occurrence) for all 134 species (r = 0.232, p<0.01), and between mean IV and occupancy (number of cells occupied) (r = 0.202, p <0.05) for the 106 species for which > 95% of the range is encompassed by the study area.

There were a couple of outliers in the dataset, i.e., species with relatively small geographic ranges and high average importance values. Loblolly pine (Pinus taeda) and Slash pine (Pinus elliottii) both have range sizes in the lower 40% of the dataset, but had the highest mean IVs (20.6 and 21.8). Both are important commercial pine species and occur in natural and intensively managed stands throughout their ranges

(Burns & Honkala, 1990).

The abundant-centre distribution

Three patterns of distribution of abundance from centre to edge of range are evident among the four abundance categories. Abundance in Category 1 species (everywhere sparse) remains relatively constant, and low, until about mid-way between the centre and edge of range, before gradually declining (Fig. 4a). Category 2 species

(somewhere abundant) show a steady gradual decline from centre to edge of range

(Fig. 4b). Categories 3 (somewhere high abundance) and 4 (somewhere very high abundance) species both show a peak in abundance at intermediate distances between centre and edge of range (Fig. 4c, d). When only the 106 species for which abundance data is available for >95% of the range are considered, the overall patterns remain very similar. Of those 106 species, 42 (40%) had significantly higher (p

<0.05) abundance values at points closest to the centre of the range (distance class 1) than at the edge (class 10). Distance class 1 was, however, the most abundant class for only 11 (10%) of these species. Distance class 4 had most species having an

12 abundance significantly greater than the edge (48 species of 106, 45%). In all, 73 species (of 106, 69%) had a significantly higher mean abundance value somewhere other than the edge (class 10), and of these, 64 (87%) had their highest abundance value at a distance between classes 1 and 6. Many species (53 of the 73) had significantly higher abundances (than the edge) at more than one category between 1 and 7. Hence for most species the distribution pattern could be more accurately described as an ‘abundant-core’ distribution rather than an abundant-centre, with core habitat surrounding the centre and extending 60-70% of the way towards the edge of range. Fully 42 species (40%) showed no significant difference in mean abundance between any of the distance classes, and so do not support the abundant-centre distribution at all.

The average number of cells occupied declined linearly from the centre to the edge of range. At distance class 1 the average number of cells unoccupied was 45%, at class

5 it was 50% and at class 10 it was 70%. In our dataset, unoccupied cells within the range of each species are recorded as a zero IV. We removed all the unoccupied cells to determine if abundance was still significantly lower at edges in sites where a species does occur. Fig. 4 includes the distribution of abundances from centre to edge of range with unoccupied cells removed. Category 1 species show little variation in abundance across the range (Fig. 4a) and Category 2 species still show a gradual decline from centre to edge (Fig. 4b). Category 3 (Fig. 4c) and 4 species (Fig. 4d) continue to show a peak in abundance at intermediate distances, however the average abundance at the classes closest to the edge do not decline below that at class 1 (as they do when all cells are considered). Seventy-three species (of 106, 69%) had no distance class having a significantly higher abundance than the edge. Distance class 1

13 was only significantly higher than class 10 for 17 species (16%), and again distance class 4 had the most species (20, 19%) with a significantly higher abundance than at the edge. Furthermore, 24 species (23%) had at least one distance class with a significantly higher abundance than the centre (class 1), and 9 species (8%) had a higher abundance at class 10 than at class 1.

A gradual decline in the average number of co-occurring tree species from centre to edge of range can be seen (Fig. 5a-d). On average, a species coexists with fewer species as the distance from the centre of the range increases. Highest numbers of co- occurring species are found at latitudes between 31° and 39° (Fig. 6a). Category 1 and 2 species tend to have highest abundance at latitudes between 34° and 39° (Fig.

6b) while Category 3 and 4 species have their lowest abundance is the same area (Fig.

6c). Category 1 and 2 species show another peak in abundance at latitudes between

44° and 47°. Category 3 and 4 species peak in abundance at the highest latitudes (45-

49°) of the study area. The latitudinal abundance pattern of Category 1 and 2 species mirrors the latitudinal pattern in average number of species per cell, whereas Category

3 and 4 species have lowest abundances at latitudes where number of co-occurring species is greatest.

Spatial autocorrelation in abundance

We found evidence of spatial autocorrelation in IV for points close together, i.e., the first spatial lag, for all species (Fig. 7). Moran’s I remained close to zero or slightly negative for most of the intermediate spatial lags for Categories 1, 2 and 3 species, while the maximum lag (10) was the most negative. Category 4 species showed negative spatial autocorrelation for all spatial lags beyond the second.

14

DISCUSSION

The proportion of ‘everywhere-sparse’ species in our dataset (15%) is generally consistent with percentages reported for plants in open forest (9%) and open- woodland (9%) in Australia (Murray et al., 1999; Murray & Lepschi, 2004).

Similarly, Rabinowitz et al. (1986) concluded that most British plants could be found in high abundance at some locations in the range. Moreover, the fact that everywhere-sparse species tend to have small geographic ranges is consistent with the widely reported positive abundance-occupancy relationship in plants (Gotelli &

Simberloff, 1987; Collins & Glen, 1990; Guo et al., 2000).

The widely held assumption that species exhibit an ‘abundant-centre distribution’ is not well supported for the majority of eastern North American tree species. Despite the intuitive appeal of mechanisms lending theoretical support to the abundant-centre distribution, our results are not inconsistent with other empirical analyses of the distribution of abundance in plant species and in a variety of other organisms. Prince et al. (1985) recorded population size of prickly lettuce, Lactuca serriola, at 500 sites along a transect from its southern to northern range boundary in Britain and found no decline in abundance or population density towards range edges. Sagarin and Gaines

(2002) recently reviewed 145 separate tests of the abundant-centre distribution in 22 empirical studies of a variety of taxa (including molluscs, fish, insects, birds, mammals and plants). Among the striking results, these authors found that only 39% of the tests supported the hypothesis, and that all but two studies inadequately sampled the species range. Most studies relied on a small number of points and severely undersampled the range edges.

15

McGill (2005) has recently begun to develop a more quantitative description of the distribution of abundances about a range. He recognized such structural measures of a distribution as its continuity (abundances varying in a smooth continuous fashion); the pattern of peaks, drops and tails; occurrences of unimodality of peak abundances; and centeredness of the peak, as well as other estimates of symmetry. Based on North

American breeding bird distributions, McGill rejects the relatively long-standing claim that species distributions of abundance have a Gaussian (normal) distribution pattern across the range. He argues compellingly for a consensus position that he describes as the ‘peak-and-tail structure of abundance’ across a species range. The major departure from a Gaussian model involves the absence of any peak abundance centeredness, symmetry, and unimodality, all of which McGill showed to be unsupported by the distributions of avian abundance data.

Abundance patterns in trees have long been presumed to reflect primarily abiotic factors (landscape-level, often latitudinally-related), and to a lesser extent biotic effects (involving competition, and successional status) and historical factors (such as forest management and clearing events). However recent empirical evidence points to a more prominent role for biotic factors interacting with abiotic factors in regulating species distribution patterns (Schwarz et al., 2003). Depending upon their location and habitat, some plant species can reach densities of 40 m-2 in grassland, or nearly

300 tree species ha-1 in tropical forests, and the question of how stable coexistence occurs among these species remains unresolved, though a great range of theories are available (Silvertown, 2004). Silvertown (2004) usefully reviewed these and derived four tests from competition theory with which to test the available hypotheses.

16 Silvertown concludes that despite the recent emergence of powerful unified neutral theories of Hubbell (2001) and Bell (2001) (proposing that species are competitively equivalent and niche differences are essentially irrelevant, and that regional species diversities are governed by a drift-like process), recent studies suggest that plants do segregate along a number of environmental niche axes, including gradients of light, soil moisture and root depth, and that partitioning of soil nutrients occurs in conjunction with species-specific microbial symbionts.

Our data also implicate a role for biotic factors interacting with abiotic agents.

Though we do not have all possible species represented in our dataset, the results suggest that species generally compete with fewer other tree species as range edges are approached (Fig. 6). The fact that cells closest to the centre of a species range have, on average, the most other species, and cells closest to the edge of range have the fewest suggests that the species in our dataset respond similarly to environmental gradients that result in a general decline in habitat suitability. This is not surprising since the geographic range-centres of many of the species are located in relative proximity to one another. For example, the geographic centres of the ranges of over half (76) of the species can be contained in an area covering only 16% of the entire study area (and located approximately in the middle of it (see Fig. 1). This location is also a focus for ecoregional richness and overall species richness (Ricketts et al.,

1999).

The potentially increased competition appears to impact species abundance patterns in different ways. Whereas for Category 1 and 2 species, abundance declined even as the number of potentially competing species declines, for Category 3 and 4 species

17 abundance initially increased as the number of co-occurring species declines. These latter, medium-to-large range and somewhere-high-to-very-high abundance species may be initially taking advantage of a release from competition with overlapping but more specialised species, to increase their abundances. As range edges draw nearer, however, average abundance gradually declines despite further decreasing numbers of co-occurring species. Wider-ranging species (i.e., Categories 3 and 4) should be less likely to respond to subtle changes in environmental gradients across the geographic range since they can, in theory, tolerate a wider amplitude of conditions (Stevens,

1989). They may therefore be regulated more by biotic factors like competition, across a large part of their range. Species having narrower tolerances (and smaller ranges) may more readily show decreases in abundance over even minor changes in environmental gradients.

Loehle (1998) reported a tradeoff between growth rate and freezing tolerance for 22 species of North American trees and suggested that, as a result, northern hemisphere trees are out-competed by trees with faster growth rates at their southern range limits.

Furthermore, Pither (2003) found that the lowest January mean daily minimum temperature accounted for 78% of the variation in range size in North American tree and shrub species. Pither speculated that the widest-ranging, cold-tolerant tree species potentially trade off between cold tolerance strategies and competitive ability, and are generally out-competed in regions that are consistently warm and moist, such as the species-rich south-eastern United States. Indeed our results indicate that the somewhere-high to very-high-abundance (and generally wider ranging) species in our dataset have their lowest abundance in latitudes between 31° and 38° in the face of

18 high numbers of likely competing species, while the everywhere-sparse to somewhere-abundant (and smaller range) species peak in abundance in this area.

It appears that for many tree species high abundance can be achieved at any position in the range, though suitable sites are found with less frequency toward range edges.

Tests of the abundant-centre distribution that only use data from sites where the species is present therefore may not accurately reflect the real pattern of abundance, as it represents both individual population size and the density of populations.

Furthermore, these spatial autocorrelation results indicate that species may not respond to all range edges in the same way, that is, abundance values at opposite directions of the range tend to be relatively dissimilar to each other. The latitudinal pattern of abundance suggests that wider-ranging species achieve relatively high abundance at northern range limits compared with southern limits. Studies that address only one aspect of the range edge may therefore produce spurious results.

To our knowledge no other studies have considered such a large sample of ecologically important organisms with relative abundances over the entire or near- entirety of the geographic range, from a consistent and comparable dataset. Although our results provide some support for the ‘abundant-centre’ distribution in that species were never more abundant at edges and the proportion of sites occupied is higher in central parts of the range, it was also true that areas closest to the centre are not always the most abundant areas within the range. Rather, peaks of abundance often occur at intermediate areas between the centre and edge of range, though these peaks usually occur closer to the centre than to the edge. Furthermore species may reach relatively high abundances at peripheral parts of the range, however, their occurrence

19 there is less frequent than in central parts of the range. Our results suggest that species with large ranges are more likely to show an ‘abundant-core’ distribution and that abundance of these species may be more affected by factors other than abiotic gradients across large parts of their range, though they still experience reduced abundance as range edges are approached. Competitive relationships may be involved in the distribution-abundance patterns across tree ranges.

20 REFERENCES

Bell, G. (2001). Neutral macroecology. Science, 293, 2413-2418.

Blackburn, T.M. & Gaston, K.J. (2003). Macroecology: concepts and consequences.

Blackwell Science, Oxford.

Blackburn, T.M., Jones, K.E., Cassey, P. & Losin, N. (2004). The influence of spatial resolution on macroecological patterns of range size variation: a case study using parrots (Aves: Psittaciformes) of the world. Journal of Biogeography, 31, 285-293.

Brown, J.H. (1984). On the relationship between abundance and distribution of species. American Naturalist, 124, 255-279.

Brown, J.H. (1995). Macroecology. University of Chicago Press, Chicago.

Brown, J.H., Mehlman, D.W. & Stevens, G.C. (1995). Spatial variation in abundance.

Ecology, 76, 2028-2043.

Burns, R. M. & Honkala, B.H. (1990) Silvics of North America: 1. Conifers; 2.

Hardwoods. USDA Agriculture Handbook 654. Washington, DC.

Collins S.L. & Glenn, S.M. (1990). Effects of organismal and distance scaling on analysis of species distribution and abundance. Ecological Applications, 7, 543-551.

21 Gaston, K.J. & Blackburn, T.M. (2000). Pattern and Process in Macroecology.

Blackwell Science, Oxford.

Gaston K.J., Blackburn, T.M., Greenwood, J.J.D., Gregory, R.D., Quinn, R.M. &

Lawton, J.H. (2000). Abundance-occupancy relationships. Journal of Applied

Ecology, 37, 39-59.

Gaston, K.J., Blackburn, T.M. & Lawton J.H. (1997). Interspecific abundance-range size relationships: an appraisal of mechanisms. Journal of Animal Ecology, 66, 579-

601.

Gotelli N.J. & Simberloff D. (1987). The distribution and abundance of tallgrass prairie plants – a test of the core-satellite hypothesis. American Naturalist, 130, 18-35.

Guo, Q., Brown, J.H. & Valone, T.J. (2000). Abundance and distribution of desert annuals: are spatial and temporal patterns related? Journal of Ecology, 88, 551-560.

Hubbell, S.P. (2001). The Unified Neutral Theory of Biodiversity and Biogeography.

Princeton University Press, Princeton.

Kotze, D.J., Niemelä, J., O’Hara, B. & Turin H. (2003). Testing abundance-range size relationships in European carabid beetles (Coleoptera, Carabidae). Ecography, 26,

553-556.

22 Legendre P. & Legendre L. (1998). Numerical Ecology. Developments in

Environmental Modelling. Elsevier, Amsterdam.

Mathius, P.V.C., Mendonça, C.V., Rangel, T.F.L.V.B. & Diniz-Filho, J.A.F. (2004).

Sensitivity of macroecological patterns of South American parrots to differences in data sources. Global Ecology and Biogeography, 13, 193-198.

McGill, B.J. (2005). Structure of abundance across a species range: synthesis, evidence and mechanisms. In revision (Draft available at /www.brianmcgill.org).

Murray, B.R. & Lepschi, B.J. (2004). Are locally rare species abundant elsewhere in their geographical range? Austral Ecology, 29, 287-293.

Murray, B.R., Rice, B.L., Keith, D.A., Myerscough, P.J., Howell, J., Floyd, A.G.,

Mills, K. & Westoby, M. (1999). Species in the tail of rank-abundance curves.

Ecology, 80, 1806-1816.

Pither, J. (2003). Climate tolerance and interspecific variation in geographic range size. Proceedings of the Royal Society of London Series B, 270, 475-481.

Prasad, A.M. & Iverson, L.R. (2003). Little's Range and FIA Importance Value

Database for 135 Eastern US Tree Species, U.S.D.A. Forest Service, Delaware, URL http://www.fs.fed.us/ne/delaware/4153/global/littlefia/index.html.

23 Prince, S. D., Carter, R. N. & Dancy, K. J. (1985) The geographical distribution of prickly lettuce (Lactuca serriola): II Characteristics of populations near its distribution limit in Britain. Journal of Ecology, 73, 39-48.

Rabinowitz, D., Cairns, S. & Dillon, T. (1986). Seven forms of rarity and their frequency in the flora of the British Isles. In: Conservation Biology: the Science of

Scarcity and Diversity (ed. by Soulé, M.E.), pp 182-204. Sinauer Associates, Inc.,

Sunderland, Massachusetts.

Ricketts, T. H., Dinerstein, R., Olsen, D.M., Loucks, C.J., Eichbaum, W., DellaSala,

D., Kavangh, K., Hedao, P., Hurley, P.T., Carney, K.M., Abell, R. & Walters, D.

(1999). Terrestrial Ecoregions of North America. Island Press, Washington D.C.

Sagarin, R.D. & Gaines, S.D. (2002). The ‘abundant-centre’ distribution: to what extent is it a biogeographical rule? Ecology Letters, 5, 137-147.

Schwarz, P.A., Fahey, T.J. & McCulloch, C.E. (2003). Factors controlling spatial variation of tree species abundance in a forested landscape. Ecology, 84, 1862-1867.

Silvertown, J. (2004). Plant coexistence and the niche. Trends in Ecology and

Evolution, 19, 605-611.

Stevens, G.C. (1989). The latitudinal gradient in geographical range: how so many species coexist in the tropics. American Naturalist, 133, 240-256.

24 BIOSKETCH

Helen Murphy is a doctoral candidate in Biological Sciences at the University of

Windsor, Canada. Her research interests involve the effects of landscape fragmentation and location in the geographic range on plant population biology.

Jeremy VanDerWal is a doctoral student in Biological Sciences at the University of

Windsor. He is interested in statistical modeling applied to macroecology and conservation biology. Current projects involve the development and application of models based upon analysis of spatial patterns of rare-species occurrences to establish priority areas for conservation.

Jon Lovett-Doust is a Professor in Plant Ecology at the University of Windsor with research interests in integration of landscape- and population-level perspectives.

25 Table 1 – Mean proportion of 20 x 20 km cells in each importance value (IV) class within the range of each species, number of species with maximum IV value falling within the class and number of species with at least 1% of total cells within the range in the IV class.

IV Mean SE Number of Number of

proportion species with species with

of cells maximum IV >1% of their

(%) falling within cells in the

the class class

0 59.63 2.03 0 134

1-5 29.39 1.35 4 132

6-10 5.96 0.51 12 102

11-15 2.61 0.33 4 72

16-20 1.53 0.23 5 45

21-25 0.91 0.17 11 22

26-30 0.68 0.17 11 14

31-40 0.72 0.24 20 12

41-60 0.61 0.22 25 7

61-80 0.28 0.11 17 2

81-100 0.16 0.06 25 1

26

Figure 1 – Area for which abundance (importance values) are available (in grey shading), and range centers (●) for each species (n = 134). The black circle contains the range centre for 76 of the 134 species and covers approximately 16% of the study area.

27 40

30

20 Frequency

10

0 0 1,000 2,000 3,000 4,000 5,000 6,000 7,000 Range size

Figure 2 – Frequency distribution of range sizes for 134 eastern North American trees

(km2 x 103)

28 3,500 B ± 1 SE 3,000 3 B 2,500 x 10 2 2,000 A 1,500

1,000 A

500

0 Mean Range Size km

1 2 3 4 Abundance Category

Figure 3 – Range size (mean occupancy ± 1 SE) and abundance category: Category 1

species (‘everywhere sparse’, n = 20); Category 2 species (‘somewhere abundant’, n =

47); Category 3 species (‘somewhere high abundance’, n = 42); and Category 4

species (‘somewhere very high abundance’, n = 25). The same letters above

categories indicate no significant difference in mean occupancy (results of Tukey’s

post hoc test).

29 2.0 (a) (b) 4

1.5 3

1.0 2

Mean IV +- 1 SE 0.5 1 0.0

6 (c) 8 (d)

7 5 6 4 5

3 4 Mean IV +- 1 SE 3 2 2

1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 Distance class

Figure 4 - Mean IV (± 1 SE) in all cells (□) and occupied cells only (o) in each distance class for (a) Category 1 species (‘everywhere sparse’); (b) Category 2 species

(‘somewhere abundant’); (c) Category 3 species (‘somewhere high abundance’); and

Category 4 species (‘somewhere very high abundance’). Note the different y-axis scales.

30

26 (a) (b) (c) (d)

l

l

e c

/ 24

s

e

i

c e

p 22

s

r e

b 20

m

u N 18 12345678910 12345678910 12345678910 12345678910 Distance class Figure 5 –Average number of co-occurring species per cell (± 1 SE) at each distance class for (a) Category 1 species (‘everywhere sparse’); (b) Category 2 species

(‘somewhere abundant’); (c) Category 3 species (‘somewhere high abundance’); and

Category 4 species (‘somewhere very high abundance’).

31 (a) 25

20

15

10 Mean species/cell +- 1 SE 5 31 33 35 37 39 27 29 41 43 45 47 49

1.5 (b)

1.0

0.5 Mean IV +- 1 SE

0.0 31 33 35 37 39 27 29 41 43 45 47 49 8.0 (c)

6.0

4.0

Mean IV +- 1 SE 2.0

0.0 35 31 33 37 39 27 29 41 43 45 49 47 Latitude

Figure 6 – (a) Mean number of species per cell (± 1 SE) from southern to northern latitudes, and mean abundance (± 1 SE) from southern to northern latitudes for (b)

Category 1 and 2 species (everywhere sparse to somewhere abundant), and for (c)

Category 3 and 4 species (somewhere high to very high abundance). Note the different y-axis scales.

32 G

X

0.10 b

G I

s W ' X

n b

a 0.00 W WG r W o b W b

X b M b bW b b GX W X b XX W X X W G X G -0.10 G W G G G

12345678910 Spatial lag

Figure 7 – Spatial autocorrelogram showing Moran’s I at increasing distances between cells for Category 1 (●), Category 2 (x), Category 3 (▲) and Category 4 (□) species. One unit on the horizontal axis indicates one interval of the maximum distance between cells, with 1 representing cells in the closest 10 % of the range of distances.

33 Supplementary Information

Supplementary Table 1 – 134 eastern North American tree species used in the analysis, including range size (km2 x 103) and extent of occurrence (number of cells for which an occurrence has been recorded). Taxonomy follows (Kartesz 1994).

U.S.D.A. Extent of Forest Range occurrence Maximum Abundance Service Genus Species Family Order Size (km2 (number of IV category1 Reference x 103) cells) No.

12 Abies balsamea Pinaceae Pinales 3331.73 1197 78 3

43 Chamaecyparis thyoides Cupressaceae Pinales 140.35 366 14 1

68 Juniperus virginiana Cupressaceae Pinales 2807.2 5697 94 4

71 Larix laricina Pinaceae Pinales 5893.15 1813 54 3

94 Picea glauca Pinaceae Pinales 6667.73 837 23 2

95 Picea mariana Pinaceae Pinales 6814.23 1227 50 3

97 Picea rubens Pinaceae Pinales 403.35 8074 37 2

34 U.S.D.A. Extent of Forest Range occurrence Maximum Abundance Service Genus Species Family Order Size (km2 (number of IV category1 Reference x 103) cells) No.

105 Pinus banksiana Pinaceae Pinales 3172.58 656 51 3

107 Pinus clausa Pinaceae Pinales 60.03 146 71 3

110 Pinus echinata Pinaceae Pinales 1112.45 2805 45 3

111 Pinus elliottii Pinaceae Pinales 263.83 653 100 4

115 Pinus glabra Pinaceae Pinales 231.33 595 7 1

121 Pinus palustris Pinaceae Pinales 523.08 1324 50 3

123 Pinus pungens Pinaceae Pinales 93.45 246 8 1

125 Pinus resinosa Pinaceae Pinales 1186.53 1306 31 2

126 Pinus rigida Pinaceae Pinales 480.58 1202 65 3

128 Pinus serotina Pinaceae Pinales 313.93 804 44 3

35 U.S.D.A. Extent of Forest Range occurrence Maximum Abundance Service Genus Species Family Order Size (km2 (number of IV category1 Reference x 103) cells) No.

129 Pinus strobus Pinaceae Pinales 1854.95 2608 46 3

131 Pinus taeda Pinaceae Pinales 871.85 2187 80 3

132 Pinus virginiana Pinaceae Pinales 533.75 1354 36 2

221 Taxodium distichum Taxodiaceae Pinales 889.08 2040 60 3

222 Taxodium distichum var. 100 4 nutans Taxodiaceae Pinales 370.85 892

241 Thuja occidentalis Cupressaceae Pinales 1638.43 1292 59 3

261 Tsuga canadensis Pinaceae Pinales 1214.33 2237 30 2

311 Acer barbatum Aceraceae Sapindales 427.7 1098 8 1

313 Acer negundo Aceraceae Sapindales 4296.5 5953 100 4

36 U.S.D.A. Extent of Forest Range occurrence Maximum Abundance Service Genus Species Family Order Size (km2 (number of IV category1 Reference x 103) cells) No.

314 Acer nigrum Aceraceae Sapindales 996.13 2088 32 2

315 Acer pensylvanicum Aceraceae Sapindales 764.63 1188 30 2

316 Acer rubrum Aceraceae Sapindales 3245.5 6527 100 4

317 Acer saccharinum Aceraceae Sapindales 2529.95 5538 84 4

318 Acer saccharum Aceraceae Sapindales 2286.43 4106 92 4

319 Acer spicatum Aceraceae Sapindales 2069.05 1822 12 1

331 Aesculus glabra Hippocastanaceae Sapindales 937.35 1908 22 2

332 Aesculus octandra Hippocastanaceae Sapindales 261.68 653 9 1

355 Alnus glutinosa Betulaceae Fagales 2772.8 5667 48 3

367 Asimina triloba Annonaceae Magnoliales 1583.1 3837 40 2

37 U.S.D.A. Extent of Forest Range occurrence Maximum Abundance Service Genus Species Family Order Size (km2 (number of IV category1 Reference x 103) cells) No.

371 Betula alleghaniensis Betulaceae Fagales 1521.68 2291 36 2

372 Betula lenta Betulaceae Fagales 504.6 1261 24 2

373 Betula nigra Betulaceae Fagales 1418.58 3563 17 2

375 Betula papyrifera Betulaceae Fagales 6879.98 1653 67 3

379 Betula populifolla Betulaceae Fagales 416.73 633 20 2

381 Bumelia lanuginosa Sapotaceae Ebenales 1071.68 1548 5 1

391 Carpinus caroliniana Betulaceae Fagales 2510.55 5798 25 2

401 Carya aquatica Juglandaceae Fagales 668.98 1666 72 3

402 Carya cordiformis Juglandaceae Fagales 2540.4 5828 67 3

403 Carya glabra Juglandaceae Fagales 1948.55 4744 33 2

38 U.S.D.A. Extent of Forest Range occurrence Maximum Abundance Service Genus Species Family Order Size (km2 (number of IV category1 Reference x 103) cells) No.

404 Carya illinoensis Juglandaceae Fagales 690.05 1214 33 2

405 Carya laciniosa Juglandaceae Fagales 596.38 1450 26 2

407 Carya ovata Juglandaceae Fagales 2045.5 4693 56 3

408 Carya texana Juglandaceae Fagales 443.35 944 33 2

409 Carya tomentosa Juglandaceae Fagales 1920.43 4736 24 2

421 Castanea dentata Fagaceae Fagales 760.7 1878 7 1

452 Catalpa speciosa Bignoniaceae Scrophulariales 44.93 106 6 1

461 Celtis laevigata Ulmaceae Urticales 1620.8 3159 86 4

462 Celtis occidentalis Ulmaceae Urticales 2071.98 3774 100 4

471 Cercis canadensis Fabaceae Fabales 2049.53 4461 54 3

39 U.S.D.A. Extent of Forest Range occurrence Maximum Abundance Service Genus Species Family Order Size (km2 (number of IV category1 Reference x 103) cells) No.

491 Cornus florida Cornaceae Cornales 1968.75 4793 27 2

521 Diospyros virginiana Ebenaceae Ebenales 1927.73 4570 36 2

531 Fagus grandifolia Fagaceae Fagales 2288.1 4749 66 3

541 Fraxinus americana Oleaceae Scrophulariales 2858.2 5995 58 3

543 Fraxinus nigra Oleaceae Scrophulariales 2201.18 2557 77 3

544 Fraxinus pennsylvanica Oleaceae Scrophulariales 4912.27 7346 100 4

546 Fraxinus quadrangulata Oleaceae Scrophulariales 346.03 809 9 1

551 Gleditsia aquatica Fabaceae Fabales 291.8 715 27 2

552 Gleditsia triacanthos Fabaceae Fabales 1733.78 3759 100 4

555 Gordonia lasianthus Theaceae Theales 234.15 591 21 2

40 U.S.D.A. Extent of Forest Range occurrence Maximum Abundance Service Genus Species Family Order Size (km2 (number of IV category1 Reference x 103) cells) No.

571 Gymnocladus dioicus Fabaceae Fabales 719.53 1607 21 2

580 Halesia spp. Styracaceae Ebenales 221.18 565 1 1

591 Ilex opaca Aquifoliaceae Aquifoliales 1238.5 3080 22 2

601 Juglans cinerea Juglandaceae Fagales 1463.53 3205 29 2

602 Juglans nigra Juglandaceae Fagales 2306.25 4994 100 4

611 Liquidambar styraciflua Hamamelidaceae Hamamelidales 1420.9 3524 61 3

621 Liriodendron tulipifera Magnoliaceae Magnoliales 1585.28 3901 35 2

641 Maclura pomifera Moraceae Urticales 120.25 106 39 2

651 Magnolia acuminata Magnoliaceae Magnoliales 505.55 1293 7 1

652 Magnolia grandiflora Magnoliaceae Magnoliales 395.3 1008 9 1

41 U.S.D.A. Extent of Forest Range occurrence Maximum Abundance Service Genus Species Family Order Size (km2 (number of IV category1 Reference x 103) cells) No.

653 Magnolia virginiana Magnoliaceae Magnoliales 705.23 1757 80 3

654 Magnolia macrophylla Magnoliaceae Magnoliales 157.57 416 8 1

682 Morus rubra Moraceae Urticales 2807.68 5938 84 4

691 Nyssa aquatica Nyssaceae Cornales 469.85 1158 54 3

692 Nyssa ogechee Nyssaceae Cornales 80.3 203 7 1

693 Nyssa sylvatica Nyssaceae Cornales 1980.95 4856 26 2

694 Nyssa sylvatica var. biflora Nyssaceae Cornales 80.08 199 37 2

701 Ostrya virginiana Betulaceae Fagales 3054.55 6281 100 4

711 Oxydendrum arboreum Ericaceae Ericales 722.83 1836 19 2

721 Persea borbonia Lauraceae Laurales 481.33 1144 32 2

42 U.S.D.A. Extent of Forest Range occurrence Maximum Abundance Service Genus Species Family Order Size (km2 (number of IV category1 Reference x 103) cells) No.

731 Platanus occidentalis Platanaceae Hamamelidales 2579.48 5799 46 3

741 Populus balsamifera Salicaceae Salicales 5892.18 1247 43 3

742 Populus deltoides Salicaceae Salicales 3824.68 5274 100 4

743 Populus grandidentata Salicaceae Salicales 2041.63 3057 28 2

746 Populus tremuloides Salicaceae Salicales 7482.63 2743 100 4

761 Prunus pensylvanica Rosaceae Rosales 4563.75 2347 26 2

762 Prunus serotina Rosaceae Rosales 3536.75 6775 65 3

763 Prunus virginiana Rosaceae Rosales 5993.95 3702 59 3

766 Prunus americana Rosaceae Rosales 2468.78 4444 36 2

802 Quercus alba Fagaceae Fagales 2508.7 5938 100 4

43 U.S.D.A. Extent of Forest Range occurrence Maximum Abundance Service Genus Species Family Order Size (km2 (number of IV category1 Reference x 103) cells) No.

804 Quercus bicolor Fagaceae Fagales 870.85 2016 22 2

806 Quercus coccinea Fagaceae Fagales 934.55 2355 46 3

808 Quercus durandii Fagaceae Fagales 179.88 194 5 1

809 Quercus ellipsoidalis Fagaceae Fagales 285.52 631 30 2

812 Quercus falcata var. falcata Fagaceae Fagales 1339.6 3320 22 2

813 Quercus falcata var. 28 2 pagodifolia Fagaceae Fagales 722.28 1750

816 Quercus ilicifolia Fagaceae Fagales 207.88 535 82 4

817 Quercus imbricaria Fagaceae Fagales 659.7 1578 36 2

819 Quercus laevis Fagaceae Fagales 374.3 943 36 2

44 U.S.D.A. Extent of Forest Range occurrence Maximum Abundance Service Genus Species Family Order Size (km2 (number of IV category1 Reference x 103) cells) No.

820 Quercus laurifolia Fagaceae Fagales 515.88 1276 50 3

822 Quercus lyrata Fagaceae Fagales 912.93 2243 85 4

823 Quercus macrocarpa Fagaceae Fagales 2470.15 3495 100 4

824 Quercus marilandica Fagaceae Fagales 1840.63 3959 52 3

825 Quercus michauxii Fagaceae Fagales 957.83 2390 12 1

826 Quercus muehlenbergii Fagaceae Fagales 1741.9 3734 33 2

827 Quercus nigra Fagaceae Fagales 1050.33 2515 51 3

828 Quercus nuttallii Fagaceae Fagales 270.8 661 36 2

830 Quercus palustris Fagaceae Fagales 910.4 2155 48 3

831 Quercus phellos Fagaceae Fagales 1005.33 2481 41 3

45 U.S.D.A. Extent of Forest Range occurrence Maximum Abundance Service Genus Species Family Order Size (km2 (number of IV category1 Reference x 103) cells) No.

832 Quercus prinus Fagaceae Fagales 798.3 1965 39 2

833 Quercus rubra Fagaceae Fagales 2888 5738 65 3

834 Quercus shumardii Fagaceae Fagales 1481.05 3235 14 1

835 Quercus stellata Fagaceae Fagales 2132.48 4424 70 3

837 Quercus velutina Fagaceae Fagales 2299.4 5439 68 3

838 Quercus virginiana Fagaceae Fagales 516.17 598 100 4

840 Quercus stellata var. 16 2 margaretta Fagaceae Fagales 394.6 991

901 Robinia pseudoacacia Fabaceae Fabales 440.18 1112 24 2

921 Salix amygdaloides Salicaceae Salicales 3382.13 1990 100 4

46 U.S.D.A. Extent of Forest Range occurrence Maximum Abundance Service Genus Species Family Order Size (km2 (number of IV category1 Reference x 103) cells) No.

922 Salix nigra Salicaceae Salicales 3159.18 6387 100 4

931 Sassafras albidum Lauraceae Laurales 2173.45 5294 41 3

935 Sorbus americana Rosaceae Rosales 1888.08 1017 3 1

951 Tilia americana Tiliaceae Malvales 2223.32 4355 69 3

952 Tilia heterophylla Tiliaceae Malvales 568.03 1433 18 2

971 Ulmus alata Ulmaceae Urticales 1217.18 2890 100 4

972 Ulmus americana Ulmaceae Urticales 5216.83 8107 100 4

973 Ulmus crassifolia Ulmaceae Urticales 434.63 471 55 3

975 Ulmus rubra Ulmaceae Urticales 2873.6 5855 75 3

977 Ulmus thomasii Ulmaceae Urticales 804.2 1601 10 1

47

48 1. Kartesz J.T. (1994). A synonymized checklist of the vascular flora of the United States,

Canada, and Greenland. Timber Press, Portland.

49