5-6 Forestry paper with graphs, due 4.15

American Journal of Botany 95(1): 66–76. 2008.

66

Movement of seeds from their collection site to other environ- ments within a species range for reforestation or restoration may increase the risk of maladaptation ( Campbell, 1979 ). Reduced growth or mortality resulting from maladaptation could reduce the success of restoration projects, and gene fl ow from maladapted planted trees into adjacent native populations could negatively affect adaptation to local conditions ( McKay et al., 2005 ). How- ever, the planting of individuals adapted to new environmental conditions, e.g., a warmer climate, could be a method to facilitate migration and provide a source of genotypes well adapted to local populations. Seed transfer should be guided by natural levels of genetic variation and local adaptation in quantitative traits spe- cifi c to the species in question ( Morgenstern, 1996 ; Hufford and Mazer, 2003 ; McKay et al., 2005 ). Understanding genetic struc- ture is also necessary for managing breeding programs, evaluat- ing conservation of genetic resources, and predicting the possible effects of climate change ( St. Clair et al., 2005 ).

The ranges of many trees species are predicted to shift higher in latitude and elevation as a result of climate change ( Davis and Shaw, 2001 ; Hamann and Wang, 2006 ). However, at a local scale, projected vegetation responses include a combination of eleva-

tional, aspect, and microsite adjustments because the location of suitable conditions for each taxon shifts within a region ( Bartlein et al., 1997 ). The potential impacts of predicted warming under- score the importance of understanding genetic structure and adap- tation of populations to their local environment. For species threatened by pests and diseases in addition to climate change, minimizing maladaptation may mean the difference between es- tablishing or maintaining viable populations and local extirpation.

Whitebark pine ( Pinus albicaulis Engelm., Pinaceae) is a high elevation, fi ve-needle pine, and the only North American member of the stone pines ( Pinus subsection Cembrae ) ( Arno and Hoff, 1989 ; Price et al., 1998 ; but see Gernandt et al., 2005 ). Although of little commercial value, it has tremendous ecologi- cal value and is considered a keystone species ( Tomback et al., 2001 ). The large, wingless seeds of whitebark pine are an im- portant food source for the Clark ’ s nutcracker ( Nucifragia co- lumbiana Wilson), which is its primary dispersal agent and mutualist ( Tomback, 1978 ; Hutchins and Lanner, 1982 ; Lanner, 1982 ; Tomback, 1982 ). However, whitebark pine is in decline throughout most of its range from a synergism of natural and human-driven causes. Outbreaks of mountain pine beetle ( Dendroctonus ponderosae Hopkins) and decades of fi re sup- pression have led to mortality and successional replacement by shade-tolerant species. However, the greatest agent driving the current decline is the introduced disease white pine blister rust (caused by the fungus Cronartium ribicola J. C. Fisch. ex Rabh.). Scientists agree that whitebark pine ecosystems require immediate restoration to reduce the effects of fi re exclusion and blister rust ( McCool and Freimund, 2001 ). Silvicultural tech- niques can be used to encourage natural regeneration, but in stands with a compromised seed source or those that need to be regenerated quickly, planting seedlings (if available) is the sug- gested restoration practice ( Hoff et al., 2001 ), using blister-rust- resistant seedlings when they are available. There is a widespread need for restoration and often a limited supply of seed for white- bark pine, thus geographic guidelines on seed transfer are needed for restoration and conservation of this species.

1 Manuscript received 21 September 2006; revision accepted 8 November 2007.

The authors thank the USDA Forest Service regions one, fi ve, and six; the British Columbia Ministry of Forests; E. C. Manning and Tweedsmuir Provincial Parks of British Columbia; and B. Brett of Snowline Ecological Consulting, Whistler, B.C. for seed. Many people provided assistance to this project, including D. Kolotelo, J. Tuytel, C. Chourmouzis, D. Watson, K. Keir, M. Harrison, D. Szohner, P. Smets, J. Krakowski, S, Trehearne, and all of the members of the Aitken laboratory at UBC. Climate data were provided by Drs. T. Wang and G. Rehfeldt. Funding for this study came from the British Columbia Forestry Investment Account through the Forest Genetics Council of B.C. to the Centre for Forest Conservation Genetics at UBC. Thank you to Drs. A. Yanchuk, M. Whitlock, J. Whitton, Y. El- Kassaby, S. Graham, D. Tomback, B. St. Clair, and an anonymous reviewer for their helpful comments on earlier drafts of this manuscript.

2 Author for correspondence (e-mail: [email protected])

ECOLOGICAL GENETICS AND SEED TRANSFER GUIDELINES FOR PINUS ALBICAULIS (PINACEAE) 1

ANDREW D. BOWER 2 AND SALLY N. AITKEN

Centre for Forest Conservation Genetics, Department of Forest Sciences, University of British Columbia, 3401-2424 Main Mall, Vancouver, British Columbia V6T 1Z4 Canada

Whitebark pine ( Pinus albicaulis Engelm.) has greatly declined throughout its range as a result of introduced disease, fi re sup- pression, and other factors, and climate change is predicted to accelerate this decline. Restoration is needed; however, no informa- tion regarding the degree of local adaptation is available to guide these efforts. A seedling common-garden experiment was employed to assess genetic diversity and geographic differentiation ( Q ST ) of whitebark pine for traits involved in growth and ad- aptation to cold and to determine climatic variables revealing local adaptation. Seedlings from 48 populations were grown for two years and measured for height increment, biomass, root to shoot ratio, date of needle fl ush, fall and spring cold injury, and survival. Signifi cant variation was observed among populations for most traits. The Q ST was low (0.07 – 0.14) for growth traits and moderate (0.36 – 0.47) for cold adaptation related traits, but varied by region. Cold adaptation traits were strongly correlated with mean temperature of the coldest month of population origins, while growth traits were generally correlated with growing season length. We recommend that seed transfer for restoration favor seed movement from milder to colder climates to a maximum of 1.9 ° C in mean annual temperature in the northern portion of the species range, and 1.0 ° C in the U. S. Rocky Mountains to avoid maladapta- tion to current conditions yet facilitate adaptation to future climates.

Key words: genetic variation; geographic differentiation; local adaptation; Pinus albicaulis ; quantitative traits; seed transfer; whitebark pine; white pine blister rust.

67January 2008] BOWER AND AITKEN — ECOLOGICAL GENETICS OF WHITEBARK PINE

eight replications had ambient soil temperature (ambient treatment) and the re- maining four replications (cold treatment) had cooled water pumped through hoses buried approximately 25 cm below the surface, which kept soil tempera- ture consistently ~8 ° C cooler during the warmest part of the day. Populations that were represented in fewer than half of the replications (i.e., < 4 in the ambi- ent or < 2 in the cold treatment) because of mortality were excluded from the analysis. The fi nal data set included 40 populations in the ambient treatment and 37 in the cold treatment, with 33 populations common to both treatments ( Table 2 ). The AlphaPlus program ( Mann, 1996 ) was used to design the plant- ing layout and assign seedlings randomly within replications. Seedlings were planted at 9.5 × 10 cm spacing, with one row of buffer trees surrounding each raised nursery bed for which data were not collected. They were kept well wa- tered and were fertilized and weeded as needed to provide conditions optimal for growth for most temperate conifers. Timing of initiation of growth in the spring was observed for 2003 and 2004, and at the end of the 2004 growing season, survival, height growth, aboveground and belowground oven-dry bio- mass of seedlings were measured on all replications. Artifi cial freeze testing was performed on 5-mm needle segments from all seedlings in the ambient treatment in three replications in the fall of 2003 and four replications in the spring of 2004. The electrolyte leakage method was used to quantify cold in- jury. Details of cold hardiness testing are given in Bower and Aitken (2006) .

The fi nal data set contained 10 quantitative variables; data from all trees in both temperature treatments were available for third-year height increment, root biomass, shoot biomass, total biomass, root to shoot ratio, date of needle fl ush in 2003 and 2004 ( Table 3 ). Measurements of fall and spring cold injury were available from the ambient treatment only. In addition, the percentage survival in each soil temperature treatment was tested for treatment effects.

Data analysis — SAS version 8 ( SAS Institute, 1999 ) was used for all statisti- cal analyses. Preliminary analysis showed an increase in variability of residuals with an increase in predicted values, so a natural-log transformation was ap- plied to height increment, root, shoot, and total biomass, and root to shoot ratio for all analyses, which helped to equalize variance. For testing for differences between soil temperature treatments and genotype-by-environment interactions in the quantitative traits, PROC MIXED was used with the following model for populations included in both treatments:

y ijklmn = µ + t i + r ( t ) ij + b ( rt ) ijk + p l + pt il + pr ( t ) ijl + f ( p ) l m + e ijklmn , (Eq. 1)

where y ijklmn is the observed value for tree n in family m w ithin population l in incomplete block k in rep j in soil temperature i , µ is the overall mean, t i is the effect of temperature i, r ( t ) ij is the effect of rep j nested within temperature i , b ( rt ) ijk is the effect of incomplete block k nested within rep j within tem- perature i, p l is the effect of population l , pt il is the interaction of temperature i and population l , pr ( t ) ijl is the interaction of population l and rep j within temperature i , f ( p ) lm is the effect of family m nested within population l , and e ijlkmn Temperature, population, and population-by-temperature interaction were considered fi xed, while all other effects were considered random. The same model was used to analyze each geographic region separately. Population

Genetic variation and population differentiation have been assessed in whitebark pine using molecular markers and mono- terpenes ( Yandell, 1992 ; Jorgensen and Hamrick, 1997 ; Brued- erle et al., 1998 ; Stuart-Smith, 1998 ; Rogers et al., 1999 ; Krakowski et al., 2003 ), and results have indicated average to above average expected heterozygosity compared to other pines (average H e of 0.16 for whitebark pine vs. 0.13 – 0.16 for pines in general [Ledig, 1998]). Population differentiation in white- bark pine was reported to be low to moderate in all studies ( F ST or G ST < 0.09) ( Table 1 ), with signifi cant evidence of inbreed- ing ( F is signifi cantly greater than zero). Populations in the northern (western British Columbia), eastern (Rocky Moun- tains), and southern regions of the species range (California and Oregon) are differentiated for monoterpenes ( Zavarin et al., 1991 ), isozymes ( Yandell, 1992 ) and organelle DNA ( Richard- son et al., 2002b ). However, levels of genetic variation and population differentiation in phenotypic traits potentially in- volved in local adaptation in whitebark pine have not previ- ously been determined.

In this study, we analyze geographic variation and genetic differentiation in phenotypic seedling traits in a common-gar- den experiment in whitebark pine and evaluate degree of local adaptation to climate for the purpose of developing seed trans- fer recommendations and predicting the ability of whitebark pine to adapt to climate change.

MATERIALS AND METHODS

Sample materials — Open-pollinated seeds from 48 populations of white- bark pine from across most of the species range ( Table 2, Fig. 1 ) were germi- nated in 2002 following seed stratifi cation using the protocol described by Burr et al. (2001) . Germinants were sown into individual 10 in 3 (164 cm 3 ) Ray Leach Cone-tainer super cells (Stuewe and Sons, Corvallis, Oregon, USA) for their fi rst growing season. In November 2002, 10-mo-old seedlings were trans- planted into a raised nursery bed common garden in Vancouver, British Colum- bia (49 ° 13 ’ N, 123 ° 6 ’ W) and grown for two growing seasons. Seedlings were planted in an incomplete block alpha design ( Patterson and Williams, 1976 ) with 12 replications, and 10 four-tree by four-tree incomplete blocks within replications. Each replication contained 160 test trees, with populations repre- sented by 1 – 18 families (mean 7.9, SE 0.37), with each family usually repre- sented once per replication. Because temperatures in Vancouver are higher than those in the native environment, two temperature treatments were imposed:

TABLE 1. Reported values of genetic differentiation for whitebark pine ( Pinus albicaulis) and other stone pine ( Pinus subsection Cembrae ) species.

Populations Species Area F ST or G ST Reference

14 P. albicaulis BC, ID, MT, OR 0.075 A. Bower unpublished manuscript 30 P. albicaulis USA rangewide and northern AB 0.034 Jorgensen and Hamrick 1997 14 P. albicaulis USA Great Basin 0.088 Yandell 1992 29 P. albicaulis Canadian Rockies 0.062 Stuart-Smith 1998 17 P. albicaulis British Columbia 0.061 Krakowski et al. 2003 18 P. albicaulis Rangewide 0.046 a Richardson et al. 2002b

8 P. sibirica Russia ~0.042 Goncharenko et al. 1993b 11 P. sibirica Russia 0.025 Krutovskii et al. 1995

P. koraiensis Coastal Russia 0.016 Potenko and Velikov 2001 19 P. koraiensis Russia 0.015 Potenko and Velikov 1998 3 P. koraiensis Russian far east 0.040 Krutovskii et al. 1995 5 P. pumila Russia 0.043 Goncharenko et al. 1993a 3 P. pumila Kamchatka penn., Russia 0.021 Krutovskii et al. 1995 18 P. pumila Japan 0.170 Tani et al. 1996 5 P. cembra Alps and eastern Carpathians, Ukraine 0.040 Belokon et al. 2005

Notes: AB = Alberta, BC = British Columbia, Canada; ID = Idaho, MT = Montana, USA. a Φ ST from cpDNA microsatellite data

68 AMERICAN JOURNAL OF BOTANY [Vol. 95

genetic variance. In this study the variance component for population ( σ 2 p ) was used as the among-population variance, and three times the variance compo- nent for family within-population (3 σ 2 f ( p ) ) was used as the within-population variance. The within-population genetic variation was approximated as three times the family variance instead of four as is used for true half-sibs, because open-pollinated progeny of whitebark pine are more closely related than half- sibs due to moderate inbreeding and correlated paternity ( Squillace, 1974 ; Kra- kowski et al., 2003 ; Bower and Aitken, 2007 ). Values of Q ST were compared to all published estimates for whitebark pine for genetic markers ( F ST or G ST ).

Climatic variables used in the analyses were mean annual tempera- ture, mean temperature of the coldest month, mean temperature of the warmest month, mean annual precipitation, mean summer precipitation, annual heat : moisture index, summer heat : moisture index, and frost-free period. Cli- matic variables for populations north of 48 ° N were obtained from PRISM cli- matic data corrected for local elevation using the Climate BC model described by Wang et al. (2006a) . For populations south of 48 ° N, climatic data were

means were used to test for differences between treatments for survival percent- age using the above model with only the temperature and population effects, and their interaction.

To test differences among populations within each soil temperature, PROC MIXED was used with the REML variance component estimator and the fol- lowing model:

y ijklm = µ + r i + b ( r ) ij + p k + rp ik + f ( p ) kl + e ijklm , (Eq. 2)

where terms for each effect are the same as in Eq. 1 without the effect of soil temperature. All terms were considered random except for population, which was fi xed. To obtain estimates of variance components, the analysis was re- peated with all effects considered random.

Genetic differentiation among populations was estimated for all quantitative traits by calculation of Q ST ( Spitze, 1993 ): Q ST = σ 2 b / ( σ 2 b + 2 σ 2 w ), where σ 2 b is the among-population variance and σ 2 w is the within-population additive

TABLE 2. Pinus albicaulis populations, number of seedlings tested, geographic and climatic data.

Site no. Region Name

No. trees

Lat. o N Long. o W Elev. (m) MAT ( ° C) MTWM ( ° C) MTCM ( ° C) FFP (d) SH:MAmbient Cold

1 N Serb Creek 5 10 54.71 127.57 1385 0.7 11.4 − 10.7 52 32.7 2 N Hunters Basin 13 29 54.53 127.18 1446 0.3 11.1 − 11 48 34.8 3 N Morice Lake 2 — 54.04 127.48 1231 0.6 11.3 − 11.1 31 44.9 4 N Kimsquit river 1 — 53.19 127.18 900 3 12.5 − 7.2 83 32.9 5 N Heckman Pass 6 — 52.52 125.82 1526 0 9.9 − 10.6 58 46.8 6 N Perkins Peak 3 1 51.83 125.05 1916 − 1.8 7.9 − 11.5 35 31.9 7 N Jesamond 42 19 51.27 121.87 1846 0.9 11.4 − 9.1 45 45 8 N Lime Mtn. 19 13 51.10 121.67 1900 0.5 11.1 − 9.4 39 52.9 9 N Darcy 46 9 50.53 122.58 1800 0.5 11.7 − 9.8 46.1 45.3 10 N Blackcomb 57 40 50.10 122.90 1908 0.6 10 − 7.5 43.7 22.4 11 N Thynne Mtn. 20 1 49.71 120.92 1785 1.9 12.7 − 7.7 48.8 27.4 12 N Manning Park 93 40 49.10 120.67 2000 0.3 10.8 − 8.6 43.8 48.1 13 N Baldy 14 — 49.17 119.25 2154 1.2 12.1 − 8.6 39.6 37.5 14 R Copper Butte 11 9 48.70 118.46 2185 − 0.5 10.4 − 10.2 48.2 30.7 15 R Colville 6 4 48.66 118.46 2154 − 0.1 10.7 − 10.1 49 32.5 16 R Snow Peak — 7 48.58 118.48 2185 0.5 11.2 − 9.7 48.7 35.6 17 R Salmo Mtn. 13 11 48.97 117.10 2092 0 10.7 − 9.3 61.2 21.2 18 R Hooknose Mtn. — 2 48.94 117.43 2215 0.5 11.5 − 8.9 54 28.8 19 R Farnham Ridge 31 26 48.84 116.51 1846 1.5 12.4 − 8.3 70.6 35.4 20 R North Baldy — 7 48.55 117.16 1877 2.6 13.9 − 7 89 41.6 21 R Lunch Peak 37 14 48.38 116.19 1846 2.1 12.4 − 6.6 84 23.9 22 R Our Lake 6 13 47.84 112.81 2277 0.2 11.4 − 9.6 31.3 35.1 23 R Sheep Shed 22 19 47.52 112.80 2154 1 12.3 − 8.9 33.4 43.5 24 R Granite Butte 19 8 46.87 112.47 2338 0.5 12 − 9.1 32.8 47.1 25 R Blacklead Mtn. 23 41 46.64 114.86 2062 1.2 12.3 − 8.1 32.1 26.2 26 R Gospel Peak 22 16 45.63 115.95 2154 1 11.9 − 8.3 26.4 33.9 27 R Heavens Gate 10 3 45.38 116.51 2154 1.2 12.3 − 8.3 40.6 49.2 28 R Mudd Ridge 23 15 45.90 113.45 2400 0.2 11.8 − 9.5 17.5 29 29 R Quartz Hill 27 44 45.71 112.93 2646 − 0.8 10.9 − 10.2 15.8 37 30 R Little Bear 22 26 45.40 111.28 2154 2.1 14.4 − 8.9 40.5 43.4 31 R Picket Pin 4 1 45.44 110.05 2892 − 1.8 9.9 − 11 20.6 24.9 32 R Hellroaring II 20 24 45.04 109.45 2892 − 1.4 10.3 − 10.4 21.7 36.5 33 R Sawtel Peak 26 18 44.54 111.41 2400 − 0.1 12.9 − 11.9 25.9 45.8 34 R Vinegar Hill 14 26 44.72 118.57 2338 0.5 11.4 − 8.6 39.8 54.1 35 S Mt. Hood 20 22 45.39 121.66 1969 1.7 10.8 − 4.9 48.4 15 36 S Newberry Crater 30 13 43.72 121.23 2100 2.9 12.9 − 4.7 45.5 58.1 37 S Paulina Peak 18 11 43.69 121.25 2250 2 11.9 − 5.3 42.3 59.6 38 S Batchelor Butte 16 1 43.26 122.68 2200 1.9 10.7 − 3.9 44.1 40 39 S Tipsoo Peak — 13 43.22 122.04 2462 0.7 10 − 5.4 34.4 35.3 40 S Moon Mtn. 6 4 43.20 122.65 2201 1.9 10.8 − 3.9 43.9 45.2 41 S Pelican Butte 40 17 42.51 122.15 2462 1.1 10.4 − 5 35.2 50.6 42 S Ball Mtn. 22 15 41.80 122.16 2363 2.2 11.4 − 4.4 39.1 107.1 43 S Goosenest Summit 6 6 41.72 122.23 2506 1.5 10.6 − 4.6 35.3 99.3 44 S Drakes Peak 26 16 42.30 120.15 2462 2.5 13.1 − 5.3 48.1 80.3 45 S Crane Mountain 19 22 42.07 120.24 2538 2.2 12.7 − 5.6 46.3 72.4 46 S Mt. Rose 11 — 39.30 119.90 2754 2.4 12.6 − 4.8 61 88.4 47 S Stevens Peak 7 — 38.70 119.98 2923 1.6 11 − 5.2 46.6 61.6 48 S Ebbetts Pass 2 — 38.50 119.80 2769 2.5 12 − 4.4 46.9 72

Notes: Region: N = northern, R = Rocky Mountain, S = southern; Lat. = latitude, Long. = longitude, Elev. = elevation, See Table 3 for abbreviations and explanation of variables.

69January 2008] BOWER AND AITKEN — ECOLOGICAL GENETICS OF WHITEBARK PINE

treatment had fewer replications, lack of cold injury testing, and the absence of a few key populations at the northern and southern ends of the range compared to the ambient treatment, thus only data from the ambient treatment were used in the canonical correlation analysis. Climatic data and least-squares population means for each seedling phenotypic trait demonstrating signifi cant ( P ≤ 0.05) population differentiation were included in this analysis. Canonical redundancy analysis was used to determine the proportion of variation in phenotypic traits accounted for by canonical correlations with the climatic or geographic data sets. To assess potential differences in relationships between seedling pheno- typic traits and climatic variables between the two soil temperature treatments, canonical correlation analysis was repeated for the two treatments separately using only the populations common to both.

To develop predictive equations for the construction of seed transfer guide- lines, values of signifi cant canonical variables for the seedling phenotypic traits were regressed on the standardized climatic variable with the highest loading for that canonical variable. The slope of this regression estimates the rate of change in the phenotypic canonical variable relative to the selected climatic variable. Rates of differentiation along climatic gradients were interpreted rela- tive to the least signifi cant difference among populations at the 20% level (least signifi cant difference: LSD 0.2). This conservatively reduces type II error (ac- cepting no differences among populations when differences actually exist) and minimizes maladaptation risk accordingly ( Rehfeldt, 1991 ). Values of LSD for the phenotypic canonical variables were obtained from a Duncan ’ s multiple range test in PROC GLM using the model for testing variation among popula- tions described. The fl oating seed transfer model developed by Rehfeldt (1991 , 1994 ) was used to determine seed transfer guidelines for restoration programs of whitebark pine. The maximum recommended environmental transfer dis- tance between seed collection population and planting site was calculated as the difference in the standardized climate variable associated with the LSD ( P = 0.20) value of the phenotypic canonical variable multiplied by the standard deviation of the climate variable. Univariate regressions of climate variables on latitude, longitude, and elevation were used to determine the geographic dis- tances associated with the rates of differentiation in climate variables to make simple seed transfer recommendations.

RESULTS

Soil temperature effects — Height increment and survival were signifi cantly greater, on average, in the cold treatment than in the ambient treatment (least squares mean = 6.7 and 8.9 mm, P = 0.04 for height increment and 66.9 and 82.3%, P < 0.001 for survival, in the ambient and cold treatment respectively). Means for biomass traits were also greater in the cold treatment, and the temperature treatment difference greater for root mass than shoot mass, although the difference between treatments for these traits was not signifi cant. The date of needle fl ush did not differ signifi cantly between treatments. Population-by-treat- ment interaction was not signifi cant for any of the traits. The foliage of seedlings in the cold temperature treatments gener- ally appeared darker green and healthier than those in the ambi- ent treatment. No treatment-specifi c geographic trends were evident, and separate canonical correlation analyses of individ- ual treatments yielded the same results.

Geographic patterns across the species range — In general, seedlings from populations originating from colder climates had less overall growth, earlier needle fl ush in spring, and less cold injury in fall than seedlings originating from milder cli- mates when grown in the common garden. Populations differed signifi cantly in the ambient soil temperature treatment for all variables except root : shoot ratio and spring cold injury ( Table 5 ). Despite a lack of signifi cant differences among populations in the ANOVA, root : shoot ratio differed signifi cantly among populations in a Duncan ’ s multiple range test.

Growth-related traits generally had low levels of population differentiation (0 ≤ Q ST ≤ 0.14), while the cold-adaptation re- lated traits (date of needle fl ush and fall cold injury) showed

obtained from a model using the thin plate splines of Hutchinson (2000) as il- lustrated for North America by McKinney et al. (2001) . Clines in quantitative traits can be obscured when there are correlations among traits or if geographi- cal structure is complex. In these cases, canonical correlation analysis is more effi cient than regressing each trait on environmental variables separately ( West- fall, 1992 ). Several of the seedling phenotypic traits and climatic or geographic variables were strongly intercorrelated ( Table 4 ), so canonical correlation anal- ysis was used to examine the relationships among these variables. The cold

Fig. 1. Distribution of Pinus albicaulis and locations of populations tested in common-garden experiment. Dashed lines separate the southern, Rocky Mountain and northern regions.

TABLE 3. Description of (A) quantitative and (B) climatic variables.

A) Quantitative trait Abbreviation Unit

3rd year height increment HTINC millimeters Total dry biomass TDM grams Root dry biomass RM grams Shoot dry biomass SM grams Root : shoot ratio RSR unitless 2003 Date of needle fl ush FL03 days from Jan. 1 2004 Date of needle fl ush FL04 days from Jan. 1 Fall cold injury FCI index of injury (%) Spring cold injury SCI index of injury (%)

B) Climatic variable

Mean annual temperature MAT ° C Mean temperature, warmest month MTWM ° C Mean temperature, coldest month MTCM ° C Mean annual precipitation MAP millimeters Mean summer precipitation MSP millimeters Annual heat : moisture index AH:M [(MAT + 10)/(MAP/1000)] Summer heat : moisture index SH:M [(MWMT/(MSP/1000)] Frost-free period FFP days

70 AMERICAN JOURNAL OF BOTANY [Vol. 95

nifi cant ( P = 0.006) and accounted for an additional 15% of the variation. The second pair of variables demonstrates the posi- tive relationship between the length of the frost-free period and growth, both height and biomass ( Table 6 ). The regression of the second phenotypic canonical score on frost-free period was also signifi cant ( P = 0.001) but weak ( r 2 = 0.24). Canonical re- dundancy analysis showed that the fi rst two climatic canonical variables account for 24 and 17% (41% total) of the variation in population phenotypic trait means, indicating substantial ge- netic structure along climatic gradients.

Regional patterns of variation — When populations were analyzed separately by region, some broad-scale geographic differences in patterns of population differentiation emerged ( Table 7 ). In the ambient soil temperature treatment, in the northern region, signifi cant differences were detected among populations for all three biomass traits. In the Rocky Mountain region, only date of needle fl ush in 2004 varied signifi cantly among populations, while in the southern region, only date of

moderate to strong differentiation among populations regard- less of treatment (0.36 ≤ Q ST ≤ 0.65). A comparison of Q ST values with previously published values of F ST for whitebark pine ( Table 1 ) shows that the phenotypic traits with the weakest differentiation are similar to the highest estimates of differen- tiation in presumably neutral molecular markers from rangewide studies ( Jorgensen and Hamrick, 1997 ; Richardson et al., 2002b ), and the quantitative traits with the strongest differentia- tion have substantially higher Q ST estimates.

In the canonical correlation analysis of population means for seedling phenotypic traits and climatic variables for population origins, the fi rst canonical correlation was signifi cantly differ- ent from zero ( P < 0.0001) and explained 72% of the variance in the data. The fi rst pair of canonical variables summarizes relationships between cold-related phenotypic traits and mean temperature of the coldest month ( Table 6 ). Mean temperature of the coldest month explained a substantial proportion of the variance in the fi rst phenotypic canonical score ( r 2 = 0.79, P < 0.0001, Fig. 2 ). The second canonical correlation was also sig-

TABLE 4. Correlations among population means for quantitative, climatic, and geographic variables. See Table 3 for explanation of variables.

A) Northern region

Variable HTINC a TDM a RM a SM a FL03 FL04 FCI SCI Survival Lat. Long. Elev.

MAT − 0.15 − 0.40 − 0.49 − 0.31 − 0.06 − 0.50 − 0.16 0.31 0.26 − 0.04 0.07 − 0.42 MTWM 0.03 − 0.16 − 0.26 − 0.07 0.14 − 0.38 − 0.21 0.30 0.38 − 0.07 0.19 − 0.27 MTCM − 0.37 − 0.34 − 0.48 − 0.21 − 0.25 − 0.04 0.20 0.45 0.13 − 0.59* 0.52 0.12 MAP − 0.18 − 0.37 − 0.41 − 0.32 0.02 − 0.05 0.35 0.37 − 0.03 0.03 − 0.19 − 0.30 MSP − 0.15 − 0.29 − 0.33 − 0.24 0.20 0.03 0.20 0.50 0.02 − 0.09 0.02 − 0.14 AH:M 0.33 0.45 0.47 0.43 − 0.09 − 0.22 − 0.55 − 0.24 0.17 − 0.10 0.31 0.26 SH:M 0.25 0.33 0.31 0.32 − 0.17 − 0.20 − 0.28 − 0.41 0.25 − 0.04 0.16 0.10 FFP − 0.23 − 0.48 − 0.52 − 0.42 − 0.57* − 0.62* − 0.01 − 0.42 0.05 0.25 − 0.34 − 0.63 Lat. 0.41 − 0.01 0.14 − 0.11 0.10 − 0.63* − 0.42 − 0.24 − 0.08 Long. − 0.24 0.19 0.04 0.29 − 0.07 0.55* 0.19 0.29 0.14 − 0.95* Elev. − 0.12 0.39 0.30 0.44 0.01 0.76* 0.21 0.30 − 0.04 − 0.81* 0.86*

B) Rocky Mountain region

Variable HTINC a TDM a RM a SM a FL03 FL04 FCI SCI Survival Lat. Long. Elev.

MAT 0.61* 0.60* 0.53* 0.63* 0.04 0.06 0.48 0.53* 0.24 0.20 − 0.33 − 0.85* MTWM 0.51* 0.58* 0.54* 0.59* 0.08 − 0.14 0.13 0.40 0.13 − 0.18 0.11 − 0.53* MTCM 0.44 0.51* 0.46* 0.51* − 0.12 0.23 0.69 0.55* 0.17 0.29 − 0.45 − 0.71* MAP 0.21 0.08 0.03 0.11 − 0.21 0.01 0.48 0.24 0.18 0.54* − 0.43 − 0.53* MSP − 0.15 − 0.15 − 0.15 − 0.14 − 0.43 − 0.19 − 0.12 − 0.11 0.03 0.35 0.22 0.00 AH:M − 0.14 0.00 − 0.02 0.03 0.06 0.05 − 0.45 − 0.12 − 0.09 − 0.49* 0.40 0.39 SH:M − 0.11 0.08 0.02 0.12 0.34 0.36 0.00 0.17 − 0.17 − 0.52* 0.08 0.09 FFP 0.41 0.15 0.06 0.22 − 0.13 0.00 0.47 0.30 0.33 0.70* − 0.60* − 0.77* Lat. 0.39 0.01 − 0.06 0.07 − 0.07 − 0.09 0.07 − 0.11 0.47* Long. − 0.32 − 0.07 0.02 − 0.12 − 0.14 − 0.27 − 0.66* − 0.27 − 0.22 − 0.51* Elev. − 0.68* − 0.44 − 0.34 − 0.49* − 0.08 − 0.03 − 0.50* − 0.39 − 0.38 − 0.61* 0.66*

C) Southern region

Variable HTINC a TDM a RM a SM a FL03 FL04 FCI SCI Survival Lat. Long. Elev.

MAT 0.12 0.24 0.22 0.25 0.05 − 0.14 0.22 0.35 − 0.32 − 0.12 0.47* − 0.03 MTWM − 0.16 0.00 0.00 0.00 0.18 − 0.02 0.13 0.51 − 0.43 − 0.16 0.70* 0.15 MTCM 0.43 0.43 0.42 0.43 − 0.26 − 0.13 0.17 − 0.35 0.24 0.05 − 0.60* − 0.28 MAP 0.40 0.20 0.22 0.19 − 0.47 − 0.20 0.03 − 0.36 0.51 0.37 − 0.68* − 0.43 MSP 0.46 0.24 0.27 0.22 − 0.50 − 0.21 − 0.05 − 0.29 0.55* 0.55* − 0.67 − 0.57* AH:M − 0.17 − 0.08 − 0.06 − 0.11 0.63* 0.55* − 0.07 0.11 − 0.43 0.02 0.03 − 0.01 SH:M − 0.47 − 0.21 − 0.23 − 0.21 0.59* 0.51 0.00 − 0.10 − 0.61* − 0.55* 0.26 0.54 FFP − 0.23 − 0.15 − 0.19 − 0.13 − 0.08 − 0.21 0.38 0.34 − 0.25 − 0.31 0.65* 0.24 Lat. 0.62* 0.38 0.46 0.31 − 0.25 − 0.06 − 0.23 0.19 0.56* Long. − 0.49 − 0.31 − 0.37 − 0.27 0.11 − 0.16 0.10 0.26 − 0.50 − 0.64* Elev. − 0.76* − 0.52 − 0.58* − 0.47 0.31 0.16 0.10 − 0.15 − 0.60* − 0.96* 0.67*

a Natural log transformed * Signifi cant at α = 0.05

71January 2008] BOWER AND AITKEN — ECOLOGICAL GENETICS OF WHITEBARK PINE

Scores for the fi rst pair of canonical variables from the south- ern (Oregon and California) populations were clearly separated from the Rocky Mountain and Canadian populations ( Fig. 3 ). The southern populations had a large infl uence on the relation- ship between the fi rst phenotypic canonical score and mean temperature of the coldest month ( Fig. 2 ). Regressions con- ducted within each region separately revealed a signifi cant rela- tionship between the fi rst phenotypic canonical score and mean temperature of the coldest month in both the northern region ( r 2 = 0.41, P = 0.05) and the Rocky Mountain region ( r 2 = 0.32, P = 0.01), but not in the southern region ( r 2 = 0.08, P = 0.38). The relationship between the second phenotypic canonical score (largely representing growth traits) and frost-free period was only signifi cant in the Rocky Mountain region ( P = 0.003).

Limits to seed transfer — The interval in mean temperature of the coldest month associated with a signifi cant difference in the fi rst quantitative canonical variable (which largely refl ects date of needle fl ush and fall cold injury) over all regions was esti- mated as 1.1 ° C, which translates to a geographic distance of approximately 2.8 ° latitude or 300 kilometers ( r 2 = 0.49,

needle fl ush (both years) varied signifi cantly among popula- tions. In the cold treatment, height increment differed signifi - cantly among populations in the northern region; all traits except root : shoot ratio differed in the Rocky Mountain region; and date of needle fl ush differed among populations in the southern region. Estimates of population differentiation ( Q ST ) were lower, on average, within regions than those for all popu- lations pooled across region ( Table 7 ). In the ambient treatment, the northern region had higher estimates of Q ST, on average, than other regions, while in the cold treatment, estimates of population differentiation were highest for the Rocky Mountain region.

Correlations among quantitative traits and climatic variables also varied by region ( Table 4 ). In the northern region, only the date of needle fl ush had clinal variation that was positively as- sociated with frost-free period. In the Rocky Mountain region, growth traits (height growth and biomass) and spring cold in- jury were positively associated with temperature variables. In the southern region, survival was correlated positively with summer precipitation and negatively with summer aridity in- dex, and date of needle fl ush was positively correlated with aridity.

TABLE 5. Signifi cance level of population and family within-population effect in ANOVA, over all populations and by region, in two soil temperature treatments.

Variable a Population F Family in-population Z †

Population F by region

Northern Rocky Mtn. Southern

A-HTINC b 1.84** 1.90* 1.59 1.71 1.66 A-TDM b 1.69** 2.63** 2.23* 1.20 1.73 A-RM b 1.65* 2.12* 2.07* 1.17 1.57 A-SM b 1.73** 2.72** 2.47* 1.21 1.78 A-RSR b 1.09 1.02 1.34 0.82 1.05 A-FL03 6.51** 2.52** 0.76 0.98 3.25* A-FL04 5.49** 1.93* 1.70 1.83* 2.12* A-FCI 2.59** 1.02 1.85 1.17 2.04 A-SCI 1.11 1.25 0.40 0.91 1.76

C-HTINC b 1.57* 1.73* 3.52* 2.55** 0.45 C-TDM b 1.24 2.06* 1.28 2.56** 0.78 C-RM b 1.22 1.86* 1.21 2.31** 0.73 C-SM b 1.23 2.14* 1.15 2.30** 0.85 C-RSR b 1.09 1.50* 0.30 0.99 1.82 C-FL03 3.29** 1.23 0.47 2.30** 2.01 C-FL04 6.36** 0.48 1.99 3.89** 3.31**

a A = ambient soil temperature treatment, C = cold soil temperature treatment, see Table 3 for explanation of variables b Natural log transformed * Signifi cant at α = 0.05 ** Signifi cant at α = 0.01 † Wald test of covariance parameter = estimate/approximate standard error ( SAS Institute, 1999 )

TABLE 6. Correlations between quantitative canonical variables (Quant1 and Quant2) and the quantitative variables, and between climate canonical variables (Clim1 and Clim2) and both climate and quantitative variables.

Variable a Quant1 Quant2 Variable Clim1 Clim2 Variable Clim1 Clim2

HTINC b 0.07 0.84 MAT 0.75 0.42 HTINC 0.07 0.66 TDM b − 0.05 0.58 MTWM 0.22 0.07 TDM − 0.05 0.46 RM b − 0.13 0.55 MTCM 0.95 0.27 RM − 0.12 0.43 SM b − 0.01 0.59 MAP 0.21 0.62 SM − 0.01 0.46 FL03 0.91 − 0.27 MSP − 0.25 0.24 FL03 0.85 − 0.21 FL04 0.90 − 0.31 AH_M 0.20 − 0.31 FL04 0.85 − 0.24 FCI 0.66 0.28 SH_M 0.74 − 0.35 FCI 0.62 0.22

FFP 0.14 0.61

a See Table 3 for explanation of variables. b Natural log transformed

72 AMERICAN JOURNAL OF BOTANY [Vol. 95

able seed prevented the replication of this experiment in differ- ent environments, so the two temperature treatments were intended to assess potential levels of genotype-by-environment interaction. The common-garden environment (at Vancouver, British Columbia; elevation ~100 m, MAT = 10 ° C, MTWM = 17.3 ° C, MTCM = 3.2 ° C) is considerably warmer than white- bark pine ’ s native habitat, where frosts can occur during any month of the year ( Arno and Hoff, 1989) ( Table 2 ). Although milder air temperatures and warmer soil would enhance growth for most tree species, for whitebark pine, the ambient soil tem- perature appeared more stressful than the cold soil treatment, even with the soil kept moist. The darker color and superior health of the seedlings in the cold treatment indicated that higher soil temperature was a stress that appeared to be cumula- tive over the two growing seasons. However, it appears that all populations suffered equally in the warm environment because there was no evidence of genotype-by-environment interaction between soil temperature treatments for any of the traits as- sessed. Although this experiment was outside of the natural range of the species, phenotypic differences in a common gar- den among populations demonstrate genetic differences and provide a better basis for estimating limits to seed transfer and likelihood of maladaptation than do estimates from molecular marker studies.

Genetic variation and population differentiation — We ob- served signifi cant differences among population means in most quantitative traits ( Table 5 ), similar to many other widespread North American conifers ( Morgenstern, 1996 ). In the subalpine environments where whitebark pine grows, traits affecting tol- erance of abiotic stresses most likely play a larger role in deter- mining fi tness than do growth traits related to competitive ability.

The average level of population differentiation for quantita- tive traits studied ( Q ST ) was within the range of previous esti- mates for other conifers ( Merila and Crnokrak, 2001 ; McKay and Latta, 2002 ; Howe et al., 2003 ; Savolainen et al., 2004 ; St. Clair et al., 2005; St. Clair, 2006 ; Mimura and Aitken, 2007 ). Population differentiation ( Q ST ) was strongest across all popu- lations for traits related to phenology and cold injury. However, patterns of population differentiation varied among regions, with populations in the northern region differentiated most

P < 0.0001 for regression of mean temperature of the coldest month on latitude). The difference for the northern region was 1.9 ° C and for the Rocky Mountain region was 1.0 ° C. Mean temperature of the coldest month was most closely associated with latitude in the northern region and with elevation in the Rocky Mountain region ( Table 4 ). The interval in frost-free period associated with the second canonical variable (which primarily comprises growth traits) was 15 d over all regions, which translated to a difference of 1010 m in elevation or 12.2 ° longi- tude ( r 2 = 0.23, P = 0.001 for regression of frost-free period on elevation) and 27.5 d in the Rocky Mountain region.

DISCUSSION

Effects of common-garden environments — Common-gar- den experiments are often replicated in different environments within a species ’ range to detect genotype-by-environment in- teractions as well as population differentiation. This experiment was outside of the natural range of whitebark pine; therefore results may have differed from what would have been observed in the natural habitat of this species. A limited supply of avail-

Fig. 2. Regression of fi rst quantitative canonical score (QS1) on stan- dardized mean temperature of the coldest month (MTCM) for 41 popula- tions of Pinus albicaulis in three geographic regions. Y-axis scale is standard deviation, and bracket indicates value of LSD 0.20.

TABLE 7. Quantitative trait Q ST values over all populations and by region for two temperature treatments.

Variable a All populations Northern Rocky Mtn. Southern

A-HTINC 0.14 0.17 0.13 0.10 A-TDM 0.07 0.24 0.00 0.07 A-RM 0.08 0.37 0.00 0.05 A-SM 0.07 0.30 0.00 0.08 A-RSR 0.00 0.04 0.00 0.00 A-FL03 0.47 0.03 0.00 0.24 A-FL04 0.47 0.08 0.19 0.15 A-FCI 0.36 1.00 0.05 0.25 A-SCI 0.12 0.00 0.05 0.11

C-HTINC 0.13 1.00 1.00 0.00 C-TDM 0.09 0.05 0.89 0.03 C-RM 0.08 0.04 0.48 0.02 C-SM 0.08 0.04 1.00 0.03 C-RSR 0.01 0.00 0.00 1.00 C-FL03 0.43 0.00 0.41 0.12 C-FL04 0.65 1.00 0.47 0.18

a See Table 3 for explanation of variables.

Fig. 3. Scatterplot of fi rst two quantitative canonical scores (QC1 and QC2) based on eight quantitative traits for 41 populations of Pinus albi- caulis . Axis scales are standardized values. Symbols refer to geographic regions shown in Table 2 .

73January 2008] BOWER AND AITKEN — ECOLOGICAL GENETICS OF WHITEBARK PINE

North American conifers, including subalpine fi r [ Abies lasio- carpa (Hook) Nutt.] ( Peterson et al., 2002 ), mountain hemlock [ Tsuga mertensiana (Bong.) Carr.] ( Peterson and Peterson, 2001 ), and Douglas fi r ( St. Clair et al., 2005 ).

Whitebark pine has a high level of cold hardiness compared to other conifers, but signifi cant differences exist among geo- graphic regions ( Bower and Aitken, 2006 ). In the common- garden environment, populations from colder (higher latitude) populations (with lower mean temperature of the coldest month) fl ushed earlier in the spring and suffered greater spring cold injury. Earlier fl ushing of these cold-adapted populations may be due to lower chilling requirements in winter, lower heat sum requirements in spring, or both ( Howe et al., 2003 ). In their natural environments, these populations are likely to experience shorter growing seasons than populations further south, so de- spite a higher risk of spring cold injury, earlier fl ushing in the spring may be a mechanism to allow trees at these higher lati- tudes to extend their growing season relative to trees from lower latitudes ( Sagnard et al., 2002 ). In species that can tolerate frost damage well or that have high recovery potential the length of the growing season may be a more important driving force in adaptation than the avoidance of damage ( Leinonen and Hanninen, 2002 ). Worrall (1983) reported differences in both threshold temperature and heat-sum needed for fl ushing in am- abilis fi r [ Abies amabilis (Dougl.) Forbes] and subalpine fi r and also found that populations from higher elevations fl ushed ear- lier than warmer, lower elevation populations in a common gar- den. A faster response to warming spring temperatures of higher elevation or more northerly sources has also been reported for a number of other conifer species (see references in Campbell and Sugano, 1979 ; Morgenstern, 1996 ). For coastal Douglas fi r ( P. menziesii var. menziesii ), however, where growth may be more limited by drought than cold in some environments, pat- terns were opposite ( Campbell and Sugano, 1979 ).

Seed transfer guidelines and climate change — While white- bark pine is distributed over a large latitudinal range, the clinal variation observed indicates that trees from a particular popula- tion are expected to be optimally adapted for only a portion of the environmental conditions experienced across the species range. We have used the fl oating seed transfer model developed by Rehfeldt (1991 , 1994 ) to determine seed transfer guidelines for restoration in current climates of whitebark pine. Of the traits we assessed, the date of needle fl ush gives the strongest signature for local adaptation because it has the highest Q ST es- timates (0.43 – 0.63), thus it was considered fi rst for developing regional estimates of maximum potential seed transfer distances for restoration without substantial risk of maladaptation. In the northern region, it should be possible to move seed from cone collection sites to planting sites that differ by up to 1.9 ° C in mean temperature of the coldest month and maintain growth phenology suitable for the current local climate with acceptable risk of fall cold injury. In the Rocky Mountain region, the cli- matic transfer maximum is reduced to 1.0 ° C in current climates. Restoration ecologists, park managers, and foresters can more easily use seed transfer guidelines based on geographic dis- tances than climatic differences and these differences in mean temperature of the coldest month translate to approximately 4.6 ° latitude, or 505 km, for the northern region, and 320 m in elevation in the Rocky Mountain region, based on signifi cant correlations between climatic and geographic variables. In the southern region, the lack of correspondence between the fi rst seedling phenotypic canonical variable and mean temperature

strongly for growth traits (height growth and biomass), popula- tions in the southern region differentiated by cold adaptation traits (date of needle fl ush and cold injury), while populations in the Rocky Mountain region were differentiated by both height growth and date of needle fl ush ( Table 7 ). The Q ST values for these traits in these regions were greater than estimates of dif- ferentiation in neutral molecular markers ( F ST and G ST ) for whitebark pine ( Table 1 ). Levels of population differentiation ( F ST ) previously reported in whitebark pine range from 0.034 to 0.088 and average 0.058, which is slightly higher than most values reported for other stone pine species ( Pinus subsection Cembrae ) ( Table 1 ). This value indicates that the vast majority of selectively neutral genetic variation in whitebark pine is among individuals within populations, as for most conifers. Our results show that differentiation is stronger for about half of the quantitative traits we studied than for the upper estimate of differentiation for neutral genetic markers in whitebark pine, indicating a moderate degree of local adaptation. Greater dif- ferentiation based quantitative traits than on neutral markers suggests a prominent role of natural selection driving local ad- aptation of populations for many of these polygenic phenotypic traits ( Lynch et al., 1999 ; Whitlock, 1999 ).

There is a general concordance of the patterns of variation we detected for seedling phenotypic traits with those reported by both Zavarin et al. (1991) using monoterpenes and Richard- son et al. (2002b) using mtDNA. Both of these studies found differentiation among the Rocky Mountain, northern, and southern regions of the species range, which may indicate some historical effects such as isolation, genetic drift, and migration on quantitative genetic structure in addition to the effects of selection for local adaptation to climate. Richardson et al. (2002b) also suggested that genetic evidence supports two Pleistocene refugia for whitebark pine, one in the Yellowstone region and one in the southern Oregon Cascades, with subse- quent northward postglacial colonization patterns that have re- sulted in a secondary contact zone in the southern Washington Cascades. An assessment of quantitative traits across the con- tact zone suggested by Richardson et al. (2002a , b ) might help support the molecular results, but we are unable to test for this pattern due to the lack of representation of Washington Cascade populations in our study.

Environmental gradients associated with phenotypic traits — Populations differed signifi cantly for nearly all traits, and at a broad geographic scale, those from higher latitude environ- ments with lower winter temperatures fl ushed earlier in the spring, suffered less cold injury in the fall, and allocated more biomass to shoots in the common-garden environment than those from milder environments. However, clinal variation pat- terns that corresponded to climatic gradients varied by region ( Table 4 ), indicating local adaptation is driven by selection pressures from different environmental factors in different re- gions. In the northern region, local adaptation to available growing season length appears to be important because the date of needle fl ush has a clinal variation associated with frost-free period. In the Rocky Mountain region, annual and seasonal mean temperatures appear to be driving local adaptation, with height growth increasing with temperature at population ori- gins. In the southern region, where survival and date of needle fl ush were both associated with rainfall patterns, water avail- ability appears to be the factor associated with population dif- ferentiation. Regional differences in relationships between phenotypes and source climates have been reported for other

74 AMERICAN JOURNAL OF BOTANY [Vol. 95

One strategy for selecting seed sources that may result in successful establishment of seedlings under current climates, at least some of which can tolerate warming over the next several decades, is to transfer seed unidirectionally to the maximum extent allowable from mild to cold climates on the basis of the estimates from the fl oating seed transfer model. Most popula- tions of temperate conifers have a reasonably broad tempera- ture tolerance, although populations vary in breadth of reaction norm (e.g., Wang et al., 2006b ). To balance the risks of malad- aptation in both current and future environments, the challenge is to plant seedlings into environments near their lower-temper- ature limits to ensure adequate survival and growth, yet have at least some of those trees as adults remain within their tempera- ture tolerances on the warmer side of the reaction norm.

The warming predicted by global circulation models will likely cause phenotypes to shift northward and upward to track conditions to which they are locally adapted ( Davis and Shaw, 2001 ), but climatic differences among regions should maintain clinal variation. The guidelines we propose are established based on a 20% risk of assuming no difference among popula- tions when a difference actually exists. This threshold may be too conservative, given the risks of climate change, and it may become optimal to transfer seed greater distances than esti- mated and accept a greater risk of type II error (assuming popu- lations are the same when they are different). However, exceeding these distances increases the chance of maladapta- tion under current conditions and should only be done after weighing this risk against the need for restoration. Mixing seeds from different populations within the acceptable transfer range may also be a reasonable strategy for mitigating uncertain fu- ture climates because it may increase the probably that at least some trees survive. Outbreeding depression from the mixing of populations is unlikely to be a problem for whitebark pine; there are no published examples of this phenomenon in conifers that we are aware of, and it is unlikely to evolve in species with high levels of gene fl ow (e.g., via wind-dispersed pollen).

Successful restoration plantings with preferential movement of seeds from south to north or to higher elevations, both within and exceeding the current species range, could facilitate popu- lation migration, particularly if planted genotypes have some resistance to white pine blister rust. Without such human inter- vention, whitebark pine may continue to decline, and while selection may result in adaptation by favoring survival and reproduction of those local genotypes adapted to new condi- tions, this may be insuffi cient demographically to maintain vi- able populations. A comparison of predictions of the current range of whitebark pine in British Columbia and the current predicted range based on bioclimatic modeling indicates a considerably poorer fi t of predicted to actual distribution than is typical of most of the 49 forest tree species in the province ( Hamann and Wang, 2006 ). This lack of fi t between actual and current predicted range of whitebark pine may refl ect a migra- tional lag from the dependence on seed dispersal by the Clark ’ s nutcracker and may be indicative of a slow potential rate of migra- tion in response to climate change. This unoccupied potential habitat is also predicted to be one of the few areas that have climates that could support whitebark pine both currently and in seven or eight decades. Facilitated migration via restoration plantings that move blister-rust-resistant genotypes along envi- ronmental gradients and into areas of new potential habitat may be the only way that populations of this species can maintain viability. Experimental fi eld plantings will be needed to test this concept operationally.

of the coldest month suggests that seed can be freely moved within this region. However, in the absence of further data and until populations from the Washington Cascades can be studied and compared with other southern populations, we suggest that transfer between mountain ranges (e.g., Sierra Nevadas and Cascades) should be avoided. Height growth and biomass in whitebark pine populations are signifi cantly correlated with the length of the growing season (frost-free period); however, these clines are gentle, and estimates of maximum seed transfer dis- tances based on these traits are also too large to be of practical use in a conservation or restoration program.

The problem with this approach for estimating symmetrical, static seed transfer maxima based on climate is that it assumes local genotypes are optimally adapted to current climates and that these climates will remain constant. The evidence that we are in a period of record rates of warming is mounting ( IPCC, 2001 ). While populations of most temperate and boreal tree species have high levels of genetic variation that could enable adaptation to new environments ( Hamrick, 2004 ), long genera- tion lengths will greatly constrain their ability to adapt to rapid climate change ( Burger and Lynch, 1995 ), and seed dispersal is unlikely to be suffi ciently rapid to allow populations to migrate and track climates spatially ( Davis and Shaw, 2001 ). One re- sponse to modest climate change may be for trees to migrate within local areas among microsites or aspects. However, whitebark pine inhabits a relatively small range of aspects, slopes, and microsites, so this is not a likely mechanism for maintaining high levels of adaptive diversity. Whitebark pine has reasonably high levels of variation; however, it requires several to many decades to reach sexual maturity, its habitats are discontinuous, largely consisting of high-elevation “ islands ” separated by lower-elevation valleys, and its migration is de- pendent on the Clark ’ s nutcracker ( Arno and Hoff, 1989 ). All reduce the likelihood that this species can adapt or migrate suf- fi ciently rapidly to avoid the collapse of many populations. Cli- mate models predict a dramatic decrease in the range of habitat suitable for whitebark pine with increases in temperature and CO 2 ( Romme and Turner, 1991 ; Bartlein et al., 1997 ; Hamann and Wang, 2006 ). While increasing temperature may result in new habitat available north of its current range, it is also likely to lead to an upward shift of the timberline and the range of whitebark pine, resulting in a smaller potential area for it to oc- cupy. Many populations currently have negative growth rates due to white pine blister rust infection; fi re exclusion and re- sulting competition from shade-tolerant, fi re-intolerant, faster- growing conifer species; and mountain pine beetle (Kendall and Keane, 2001). Interactions among this introduced disease, the current mountain pine beetle epidemic, and climate change could drive populations toward extirpation and the species to- ward extinction ( Hamrick, 2004 ).

New approaches to developing seed transfer recommenda- tions should balance anticipated future climates with the need to restrict seed movement to environmental distances that can lead to successful seedling establishment under current, albeit transient, conditions. If seed is transferred an excessive distance from warmer to colder climates in anticipation of future warm- ing, cold injury, or mortality may result during establishment; yet if predictions of future climate are ignored, local seed sources that are fi t in current local environments may result in restoration plantings of trees that are not adequately competi- tive with other species or will never achieve reproductive matu- rity as a result of slow growth rates under the conditions of the decades to come.

75January 2008] BOWER AND AITKEN — ECOLOGICAL GENETICS OF WHITEBARK PINE

IPCC . 2001 . Third Assessment Report of the Intergovernmental Panel on Climate Change IPCC (WG I & II). Cambridge University Press, Cambridge, UK.

JORGENSEN , S . M . , AND J . L . HAMRICK . 1997 . Biogeography and popula- tion genetics of whitebark pine, Pinus albicaulis. Canadian Journal of Forest Research 27 : 1574 – 1585 .

KENDALL , K . C . , AND R . E . KEANE . 2001 . Whitebark pine decline: infection, mortality, and population trends. In D. F. Tomback, S. F. Arno, and R. E. Keane [eds.], Whitebark pine communities: Ecology and restora- tion, 221 – 242. Island Press, Washington, D.C., USA.

KRAKOWSKI , J . , S . N . AITKEN , AND Y . A . EL - KASSABY . 2003 . Inbreeding and conservation genetics in whitebark pine. Conservation Genetics 4 : 581 – 593 .

KRUTOVSKII , K . V . , D . V . POLITOV , AND Y . P . ALTUKHOV . 1995 . Isozyme study of population genetic structure, mating system and phylogenetic relationships of the fi ve stone pine species (subsection Cembrae , section Strobi , subgenus Strobus ). In P. Baradat, W. T. Adams, and G. Muller- Starck [eds.], Population genetics and genetic conservation of forest trees, 279 – 304. SPB Academic Publishing, Amsterdam, Netherlands.

LANNER , R . M . 1982 . Adaptations of whitebark pine for seed dispersal by Clark ’ s Nutcracker. Canadian Journal of Forest Research 12 : 391 – 402 .

LEINONEN , I . , AND H . HANNINEN . 2002 . Adaptation of the timing of bud burst of Norway spruce to temperature and boreal climate. Silva Fennica 36 : 695 – 701 .

LITTLE , E . L . JR . 1971 . Atlas of United States trees, vol. 1, Conifers and im- portant hardwoods. USDA Forest Service Miscellaneous Publication 1146. Washington, D.C., USA.

LYNCH , M . , M . PFRENDER , K . SPITZE , N . LEHMAN , J . HICKS , D . ALLEN , L . LATTA , M . OTTENE , F . BOGUE , AND J . COLBOURNE . 1999 . The quan- titative and molecular genetic architecture of a subdivided spe- cies. Evolution; International Journal of Organic Evolution 53 : 100 – 110 .

Mann, A. D. 1996 . Alpha+: Experimental designs for variety trials and many-treatment experiments. Biometrics and Statistics Scotland, Edinburgh, Scotland.

MCCOOL , S . F . , AND W . A . FREIMUND . 2001 . Threatened landscapes and fragile experiences: Confl ict in whitebark pine restoration. In D. F. Tomback, S. F. Arno, and R. E. Keane [eds.], Whitebark pine com- munities: ecology and restoration, 263 – 284. Island Press, Washington, D.C., USA.

MCKAY , J . K . , AND R . G . LATTA . 2002 . Adaptive population divergence: markers, QTL and traits. Trends in Ecology & Evolution 17 : 285 – 291 .

MCKAY , J . K . , C . E . CHRISTIAN , S . HARRISON , AND K . J . RICE . 2005 . “ How local is local? ” A review of practical and conceptual issues in the ge- netics of restoration. Restoration Ecology 13 : 432 – 440 .

MCKINNEY , D . W . , M . F . HUTCHINSON , J . L . KESTEVEN , AND L . A . VENIER . 2001 . Canada ’ s plant hardiness zones revisited using modern cli- mate interpolation techniques. Canadian Journal of Plant Science 81 : 129 – 143 .

MERILA , J . , AND P . CRNOKRAK . 2001 . Comparison of genetic differentiation at marker loci and quantitative traits. Journal of Evolutionary Biology 14 : 892 – 903 .

MIMURA , M . , AND S . N . AITKEN . 2007 . Adaptive gradients and isolation by distance with postglacial migration in Picea sitchensis. Heredity 99 : 224 – 232 .

MORGENSTERN , K . E . 1996 . Geographic variation in forest trees. UBC Press, Vancouver, British Columbia, Canada.

PATTERSON , H . D . , AND E . R . WILLIAMS . 1976 . A new class of resolvable incomplete block designs. Biometrika 63 : 83 – 92 .

PETERSON , D . W . , AND D . L . PETERSON . 2001 . Mountain hemlock growth responds to climatic variability at annual and decadal time scales. Ecology 82 : 3330 – 3345 .

PETERSON , D . L . , D . W . PETERSON , AND G . J . ETTL . 2002 . Growth responses of subalpine fi r to climatic variability in the Pacifi c Northwest. Canadian Journal of Forest Research 32 : 1503 – 1517 .

POTENKO , V . V . , AND A . V . VELIKOV . 1998 . Genetic diversity and differen- tiation of natural populations of Pinus koraiensis (Sieb. et Zucc.) in Russia. Silvae Genetica 47 : 202 – 208 .

LITERATURE CITED

ARNO , S . F . , AND R . J . HOFF . 1989 . Silvics of whitebark pine ( Pinus albi- caulis ), General Technical Report. INT-GTR-253. U.S. Department of Agriculture, Forest Service, Intermountain Research Station, Ogden, Utah, USA.

BARTLEIN , P . J . , C . WHITLOCK , AND S . L . SHAFER . 1997 . Future climate in the Yellowstone National Park region and its potential impact on vegeta- tion. Conservation Biology 11 : 782 – 792 .

BELOKON , M . M . , Y . S . BELOKON , D . V . POLITOV , AND Y . P . ALTUKHOV . 2005 . Allozyme polymorphism of Swiss stone pine Pinus cembra L. in mountain populations of the Alps and eastern Carpathians. Russian Journal of Genetics 41 : 1268 – 1280 .

BOWER , A . D . , AND S . N . AITKEN . 2006 . Geographic and seasonal variation in cold hardiness of whitebark pine ( Pinus albicaulis Engelm.). Canadian Journal of Forest Research 36 : 1842 – 1850 .

Bower , A. D. , and S. N. Aitken . 2007 . Mating system and inbreeding de- pression in whitebark pine ( Pinus albicauls Engelm.). Tree Genetics and Genomes 3: 379 – 388. [Online] DOI

BRUEDERLE , L . P . , D . F . TOMBACK , K . K . KELLY , AND R . C . HARDWICK . 1998 . Population genetic structure in a bird-dispersed pine, Pinus albi- caulis (Pinaceae). Canadian Journal of Botany 76 : 83 – 90 .

BURGER , R . , AND M . LYNCH . 1995 . Evolution and extinction in a changing environment: a quantitative genetic analysis. Evolution; International Journal of Organic Evolution 49 : 151 – 163 .

BURR , K . E . , A . ERAMIAN , AND K . EGGLESTON . 2001 . Growing whitebark pine seedlings for restoration. In D. F. Tomback, S. F. Arno, and R. E. Keane [eds.], Whitebark pine communities: Ecology and restoration, 325 – 345. Island Press, Washington, D.C., USA.

CAMPBELL , R . K . 1979 . Genecology of Douglas-fi r in a watershed in the Oregon cascades. Ecology 60 : 1036 – 1050 .

CAMPBELL , R . K . , AND A . I . SUGANO . 1979 . Genecology of bud-burst phe- nology in Douglas-fi r: response to fl ushing temperature and chilling. Botanical Gazette (Chicago, Ill.) 140 : 223 – 231 .

DAVIS , M . B . , AND R . G . SHAW . 2001 . Range shifts and adaptive responses to Quaternary climate change. Science 292 : 673 – 679 .

GERNANDT , D . S . , G . GEADA L Ó PEZ , S . ORTIZ GARC Í A , AND A . LISTON . 2005 . Phylogeny and classifi cation of Pinus. Taxon 54 : 29 – 42 .

GONCHARENKO , G . G . , V . E . PADUTOV , AND A . E . SILIN . 1993a . Allozyme variation in natural populations of Eurasian pines. 1. Population struc- ture, genetic variation, and differentiation in Pinus pumila (Pall) Regel from Chukotsk and Sakhalin. Silvae Genetica 42 : 237 – 246 .

GONCHARENKO , G . G . , V . E . PADUTOV , AND A . E . SILIN . 1993b . Allozyme vari- ation in natural populations of Eurasian pines. 2. Genetic variation, di- versity, differentiation, and gene fl ow in Pinus sibirica DuTour in some lowland and mountain populations. Silvae Genetica 42 : 246 – 253 .

HAMANN , A . , AND T . WANG . 2006 . Potential effects of climate change on ecosystem and tree species distribution in British Columbia. Ecology 87 : 2773 – 2786 .

HAMRICK , J . L . 2004 . Response of forest trees to global environmental changes. Forest Ecology and Management 197 : 323 – 335 .

HOFF , R . J . , D . E . FERGUSON , G . I . MCDONALD , AND R . E . KEANE . 2001 . Strategies for managing whitebark pine in the presence of white pine blister rust. In D. F. Tomback, S. F. Arno, and R. E. Keane [eds.], Whitebark pine communities: Ecology and restoration, 346 – 366. Island Press, Washington, D.C., USA.

HOWE , G . T . , S . N . AITKEN , D . B . NEALE , K . D . JERMSTAD , N . C . WHEELER , AND T . H . H . CHEN . 2003 . From genotype to phenotype: unraveling the complexities of cold adaptation in forest trees. Canadian Journal of Botany 81 : 1247 – 1266 .

HUFFORD , K . M . , AND S . J . MAZER . 2003 . Plant ecotypes: Genetic differentia- tion in the age of ecological restoration. Trends in Ecology & Evolution 18 : 147 – 155 .

HUTCHINS , H . E . , AND R . M . LANNER . 1982 . The central role of Clark ’ s Nutcracker in the dispersal and establishment of whitebark pine. Oecologia 55 : 192 – 201 .

HUTCHINSON , M . F . 2000 . ANUSPLIN version 4.1 user ’ s guide. Australian National University, Centre for Resource and Environmental Studies, Canberra, Australia.

76 AMERICAN JOURNAL OF BOTANY [Vol. 95

POTENKO , V . V . , AND A . V . VELIKOV . 2001 . Allozyme variation and mat- ing system of coastal populations of Pinus koraiensis Sieb. et Zucc in Russia. Silvae Genetica 50 : 117 – 122 .

PRICE , R . A . , A . LISTON , AND S . H . STRAUSS . 1998 . Phylogeny and classifi - cation of Pinus . In D. M. Richardson [ed.], Ecology and biogeography of Pinus , 49 – 91. Cambridge University Press, Cambridge, UK.

REHFELDT , G . E . 1991 . A model of genetic variation for Pinus ponderosa in the Inland Northwest (U.S.A.): applications in gene resource man- agement. Canadian Journal of Forest Research 21 : 1491 – 1500 .

REHFELDT , G . E . 1994 . Adaptation of Picea engelmannii populations to the heterogeneous environments of the Intermountain West. Canadian Journal of Botany 72 : 1197 – 1208 .

RICHARDSON , B . A . , S . J . BRUNSFELD , AND N . B . KLOPFENSTEIN . 2002a . Assessing Clark ’ s nutcracker seed-caching fl ights using ma- ternally inherited mitochondrial DNA of whitebark pine. Canadian Journal of Forest Research 32 : 1103 – 1107 .

RICHARDSON , B . A . , S . J . BRUNSFELD , AND N . B . KLOPFENSTEIN . 2002b . DNA from bird-dispersed seed and wind-disseminated pollen provides in- sights into postglacial colonization and population genetic structure of whitebark pine ( Pinus albicaulis ). Molecular Ecology 11 : 215 – 227 .

ROGERS , D . L . , C . I . MILLAR , AND R . D . WESTFALL . 1999 . Fine-scale genetic structure of whitebark pine ( Pinus albicaulis ): Associations with wa- tershed and growth form. Evolution; International Journal of Organic Evolution 53 : 74 – 90 .

ROMME , W . H . , AND M . G . TURNER . 1991 . Implications of global climate change for biogeographic patterns in the Greater Yellowstone ecosys- tem. Conservation Biology 5 : 373 – 386 .

SAGNARD , F . , C . BARBEROT , AND B . FADY . 2002 . Structure of genetic diver- sity in Abies alba Mill. from southwestern Alps: multivariate analysis of adaptive and non-adaptive traits for conservation in France. Forest Ecology and Management 157 : 175 – 189 .

SAS Institute . 1999 . The SAS system for Windows, version 8.0. SAS Institute, Cary, North Carolina, USA.

SAVOLAINEN , O . , F . BOKMA , R . GARC Í A - GIL , P . KOMULAINEN , AND T . REPO . 2004 . Genetic variation in cessation of growth and frost hardiness and consequences for adaptation of Pinus sylvestris to climatic changes. Forest Ecology and Management 197 : 79 – 89 .

SPITZE , K . 1993 . Population structure in Daphnia obtusa : Quantitative ge- netic and allozymic variation. Genetics 135 : 367 – 374 .

SQUILLACE , A . E . 1974 . Average genetic correlations among offspring from open-pollinated forest trees. Silvae Genetica 23 : 149 – 156 .

ST . CLAIR , J . B . 2006 . Genetic variation in fall cold hardiness in coastal Douglas-fi r in western Oregon and Washington. Canadian Journal of Botany 84 : 1110 – 1121 .

ST . CLAIR , J . B . , N . L . MANDEL , AND K . W . VANCE - BORLAND . 2005 . Genecology of Douglas fi r in western Oregon and Washington. Annals of Botany 96 : 1199 – 1214 .

STUART - SMITH , G . J . 1998 . Conservation of whitebark pine in the Canadian Rockies: Blister rust and population genetics. M.Sc. thesis, University of Alberta, Edmonton, Alberta, Canada.

TANI , N . , N . TOMARU , M . ARAKI , AND K . OHBA . 1996 . Genetic diversity and differentiation in populations of Japanese stone pine ( Pinus pumila ) in Japan. Canadian Journal of Forest Research 26 : 1454 – 1462 .

TOMBACK , D . F . 1978 . Foraging strategies of Clark ’ s Nutcracker. Living Bird 16 : 123 – 161 .

TOMBACK , D . F . 1982 . Dispersal of whitebark pine seeds by Clark ’ s Nutcracker: A mutualism hypothesis. Journal of Animal Ecology 51 : 451 – 467 .

TOMBACK , D . F . , S . F . ARNO , AND R . E . KEANE . 2001 . The compelling case for management intervention. In D. F. Tomback, S. F. Arno, and R. E. Keane [eds.], Whitebark pine communities: Ecology and restoration, 3 – 25. Island Press, Washington, D.C., USA.

WANG , T . , A . HAMANN , D . L . SPITTLEHOUSE , AND S . N . AITKEN . 2006a . Development of scale-free climate data for western Canada for use in resource management. International Journal of Climatology 26 : 383 – 397 .

WANG , T . L . , A . HAMANN , A . YANCHUK , G . A . O ’ NEILL , AND S . N . AITKEN . 2006b . Use of response functions in selecting lodgepole pine popula- tions for future climates. Global Change Biology 12 : 2404 – 2416 .

WESTFALL , R . D . 1992 . Developing seed transfer zones. In L. Fins, S. T. Friedman, and J. V. Brotschol [eds.], Handbook of quantitative forest genetics, 313 – 398. Kluwer, Dordrecht, Netherlands.

WHITLOCK , M . C . 1999 . Neutral additive genetic variance in a metapopula- tion. Genetical Research 74 : 215 – 221 .

WORRALL , J . 1983 . Temperature-bud burst relationships in Amabalis and subalpine fi r provenance tests replicated at different elevations. Silvae Genetica 32 : 203 – 209 .

YANDELL , U . G . 1992 . An allozyme analysis of whitebark pine ( Pinus albi- caulis Engl.). M.Sc. thesis, University of Nevada, Reno, Nevada, USA.

ZAVARIN , E . , Z . RAFII , L . G . COOL , AND K . SNAJBERK . 1991 . Geographic monoterpene variability of Pinus albicaulis. Biochemical Systematics and Ecology 19 : 147 – 156 .