This file was created by scanning the printed publication. Errors identified by the software have been corrected; however, some errors may remain.

AmericanJournal of Botany80(3): 330-343. 1993.

GENETIC VARIATION IN THE PONDEROSAE OF THE SOUTHWEST'

GERALD E. REHFELDT IntermountainResearch Station, U.S. Departmentof Agriculture, Forest Service, 1221 S. Main Street,Moscow, Idaho 83843

Ninety-fiveseedling populations of southwesternponderosa ( var. scopulorum) along with single populationsof Pinus engelmannii and Pinusarizonica were compared in fourenvironmentally disparate common gardens. Differentiationamong ponderosa pine populations was detected for a diverseassortment of variables that included patterns of shootelongation, measures of growthpotential, winter and springfreezing damage, and leafcharacteristics. Multiple regressionmodels accounted for as muchas 85% ofthe variance among populations and describedcomplex clines that were dominatedby elevationaland latitudinaleffects. Although P. ponderosa,P. arizonica,and P. engelmanniiwere readily differentiated,theperformance of progeniesfrom one populationsuggested introgression primarily involving P. ponderosa and P. arizonica but also implicatingP. engelmannii.

Microevolutionaryprocesses allow natural systemsof and Blake, 1977; Read, 1980; Rehfeldt,1990, 1991), tol- genetic variabilityto be molded by environmentalhet- eranceto environmentalstress (Read and Sprackling,1981; erogeneityto produce populations geneticallyattuned to Rehfeldt,1986a, b), disease resistance(Hoff, 1988b, 1990, a local environment.As demonstratedrepeatedly by the 1991), and allozymes (Mitton et al., 1977; Mitton,Stur- experimentalapproach ofClausen, Keck, and Hiesey (e.g., geon, and Davis, 1980; Linhart et al., 1981; Hamrick, 1940), culturingplant populationsin common gardensso Blanton,and Hamrick, 1989). frequentlydemonstrates geographic variation that adap- These disparateresults document extensive population tive differentiationof populations commonlyis assumed differentiation,much of which is interpretableas adap- (Mayr, 1970). Species, however,face environmentalhet- tation to heterogeneousenvironments. Nevertheless, to erogeneitywith diverse assortments of genetic variability. P. p. var. ponderosaof the Inland Northwest,the grain As a result,responses to natural selection are varied. In bywhich the environment is perceivedapparently is much the Rocky Mountains, for example, clines in adaptive finerthan forpopulations of P. p. var. scopulorumon the traitssuggest differentiation among populationsseparated Colorado Plateau. Population differentiationis associated in altitudeby 200 m in Pseudotsuga menziesii(Rehfeldt, with habitats that differby at least 35 frost-freedays in 1989), 250 m in Pinus contorta(Rehfeldt, 1988), 350 m the Northwest(Rehfeldt, 1991) but by only 22 frost-free in Pinus ponderosa var. ponderosa (Rehfeldt,1991), and days on the Colorado Plateau (Rehfeldt, 1990). In ad- 450 m in Larix occidentalis(Rehfeldt, 1982). But in Pinus dition,genetic variation within populations tends to occur monticola,clines cannot be demonstratedfor either mor- in patches (Linhart, 1989) and is pronounced for nu- phological traits(Rehfeldt, Hoff, and Steinhoff,1984) or meroustraits reflecting growth (Conkle, 1973; Namkoong allozymes (Steinhoff,Joyce, and Fins, 1983). Yet in the and Conkle, 1976; Rehfeldt,1980), allozymes (Linhartet mountainsof northernIdaho, thesespecies frequentlyco- al., 1981), and adaptation to the biotic (Hoff, 1988a, b, occur (Daubenmire and Daubenmire, 1968) across as 1991) and abiotic environment(Rehfeldt, 1992). much as 1,000 m of elevation,an intervalassociated with The Southwestis a regionin whichthe genetic structure a differenceof about 90 frost-freedays (Baker, 1944). of ponderosa pine (P. p. var. scopulorum)populations is Populations of these wind-pollinatedsympatric poorly understood. While rangewide provenance tests have respondedmuch differently to selectionalong similar unanimously attest to genetic differentiationof south- environmentalgradients. An environmentperceived as westernpopulations from those to the north (Squillace beingcoarse-grained to Pseudotsugamenziesii and P. con- and Silen, 1962; Hanover, 1963; Wells, 1964; Read, 1980), tortais apparentlyfine-grained to P. monticola. patternsof variation among southwesternpopulations are To the widespread ponderosa pine (P. ponderosa), the obscure. This paper reportscommon garden studies and grain of the spatially variable environmentis unques- presents models that describe genetic variation. While tionablycoarse. Geographic races are recognizedwithin differentiationcan be eitherrandom or systematic,it is each of three varieties (Conkle and Critchfield,1988). the systematicpatterns that invariably correspond to en- Genetic variation among populations within races is vironmentalgradients and, therefore,most likely result abundant fora varietyof charactersreflecting growth and from natural selection. Because systematicpatterns are survivalin long-termfield tests (Conkle, 1973; Read, 1983; predictable,functional models can be applied to topics Sheppard and McElderry,1986; Van Haverbeke, 1986), rangingfrom artificial reforestation to gene conservation. growthand development in common gardens (Madsen Two factorscomplicate studies of geneticvariation in southwesternponderosa pine. First,Conkle and Critch- ' Receivedfor publication I June1992; revisionaccepted 12 No- field(1988) separate the Rocky Mountain race fromthe vember1992. southwesternrace of P. p. var. in southern The authorthanks A. K. Arbaband theNavajo ForestryDepartment scopulorum forclose cooperation; R. M. Jeffersand variouspersonnel of theGila Utah and southeasternColorado. The abruptnessof the NationalForest for support; S. P. Wellsfor technical assistance; and D. transitionbetween races, however, is largelyunknown. T. Lesterfor stimulating criticism. Secondly,in southeasternArizona and southwesternNew 330 March 1993] REHFELDT-GENETIC VARIATION IN PONDEROSAE 331

UTAH A/ COLORADO

370

5 fARIZONA 320

tX C-SangredeCristoMtsi. % X% % K )t

Fig. 1. Map of the region of studyshowing the distributionof ponderosa pine (shading,from Little, 1971) and populations sampled. Letters kieygeographic localities referencedin the text.Circle ponderosa pine; triangle= pine; square = pine.

Mexico,ponderosa pine co-occurs with two other , ulationsotherwise sampled the ecologic, geographic, and Arizonapine (P. arizonica)and Apachepine (P. engel- elevationaldistribution of the speciesin the Southwest mannii),the threeof whichare interfertile(Conkle and (Fig. 1). In thispaper, populations reference groups of Critchfield,1988) members of the Ponderosae subsection adaptivelysimilar, interbreeding individuals. The local- of theDiploxylon subgenus of Pinus(Little and Critch- itiesof Fig. 1 are physiographicprovinces that may con- field,1969). Until recently,in fact,Arizona pine was tainnumerous populations. consideredto be a varietyof ponderosapine (cf.,Perry, The 97 populationsincluded single populations of Ar- 1991).Studies of phenotypic variation in terpenesand in izonaand Apache pines (Fig. 1). After the experimentation coneand leaf morphologies have prompted the conclusion wasunder way, it became obvious that a population(Bar- thatnatural hybridization among these species is common foot)in theChiricahua Mountains (locality J, Fig. 1) was (Peloquin,1971, 1984). The extentof introgression, how- eitherintrogressed or containeda mixtureof ponderosa ever,is unknown.While concentrating on geneticvari- and Arizonapines. As a result,the working hypothesis ationin thesouthwestern race of P. p. var. scopulorum, was adoptedthat Barfoot was a mixtureof noninterbreed- thisreport assumes secondary objectives of contributing ingspecies. Under this assumption, eliminating all seed- to an understandingof theinterrelationships among the lingsfrom Barfoot with the leaf characteristics ofArizona southwesternPonderosae. pine (fouror fivenarrow leaves/fascicle) from the Pina- lenoMountains (Fig. 1) allowedBarfoot to be considered MATERIALS AND METHODS as a populationof ponderosapine. Populationdifferentiation was studiedby comparing Fieldtests- Seedlings from each population were grown thegrowth and developmentof seedlingsfrom 97 pop- in plasticcontainers (65 cm3)in a shadehousein Moscow, ulationsin fieldand greenhousetests. Each population Idaho (latitude46.70 N, longitude1170 W), and first-year was representedby a bulkedsample of eight to tenwind- seedlingswere planted in commongardens near Priest pollinatedcones from each of tentrees. To decreasethe River,Idaho; WindowRock, Arizona; and SilverCity, possibilitiesof co-ancestry, sampled were at least30 New . PriestRiver is 190 km northof Moscow; m apartand wereseparated by the crowns of at leasttwo WindowRock and SilverCity are shownin Fig. 1. The interveningtrees. Although no collectionswere made from PriestRiver and WindowRock testswere planted in the thelowest limits of thespecies' distribution where pon- fall,while the SilverCity test was plantedthe following derosapine occurs as scatteredindividuals in woodlands spring.At all sites,ten seedlingsfrom each population dominated by Pinus edulis or Juniperusspp., these pop- wereplanted in rowplots within each of four randomized 332 AMERICAN JOURNAL OF BOTANY [Vol. 80

TABLE 1. Physicalcharacteristics, mean performance, and generalcli- TABLE 2. Descriptionof the variables analyzed mateat thefield test sites Variable, Description Test site PRHT4 Age 4 height Characteristic PriestRiver Silver City Window Rock PRDIA Age 3 diameter Latitude (0N) 48.5 32.8 35.9 PRDEV Deviationfrom regression of 4-yr height on 2-yr Elevation (m) 750 1,900 2,400 height Frost-free PRCL Scoresof the color (green = 0 or blue = 1) ofthe suc- period (d) 90 210 110 culentshoot Dry season Summer Winter-Spring Winter-Spring PRRTO Ratioof the 3-yr height to the3-yr diameter Survival (%) 99.5 97.6 89.8 PRLL Leaflength from a fasciclenear the center of the 3-yr 4-yrheight (cm) 80.8 57.1 26.1 terminalshoot PRLW Leafwidth from a fasciclenear the center of the 3-yr terminalshoot PRLNM Averagenumber of leaves in tenfascicles distributed completeblocks. Rows wereseparated by 0.6 m, while throughoutthe 3-yr terminal shoot 0.3 m separatedseedlings within rows. All siteswere tilled SCHT4 Age4 height and fencedbefore planting and were irrigated and weeded SCSF Scores(I to 4) ofdamage to elongatingshoots from a periodicallyfor three growing seasons after planting. Test- springfrost in year3 SCDEV3 Deviationfrom regression of 3-yrheight on 2-yr ingwas completedafter the trees reached age 4. height Because environmentaleffects at theseplanting sites SCDEV4 Deviationfrom regression of 4-yr height on 3-yr differentiallyaffected the growth and development of trees height (Table 1), a differentset and numberof variableswas WRHT4 Age4 height necessaryfor describing growth and developmentat each WRDEV Deviationfrom regression of 4-yr height on 2-yr site(Table 2). Fieldtests thus contributed a diverse array height WRWI Death offoliage from winter desiccation during years of 17 variablesthat included morphometric traits, spring 2 and 3 and winterfreezing damage, leaf morphology, and sur- WRLL Leaflength from a fasciclenear the center of the 3-yr vival. Of thesevariables, note that the deviations from shoot regressionaccount for the autocorrelation of theannual WRDEAD Scoreof mortality at anyage shootgrowth of treesand are, thereby,relatively inde- GHEL Lengthof the terminal shoot produced in year2 pendentof prioreffects. These values thus can reflect GHS2 Initiationof elongation: the day by which the 2-yr ter- to a environmentin a minalshoot had elongated2 mm adaptation particular shortperiod GHS8 Startof elongation: the day by which the 2-yr termi- oftime. The variablesPRDEV and WRDEV reflect4-yr nal shoothad elongated8 mm heightas if all individualshad been the same heightin GHEN Cessationof elongation: the day by which all but2 yr2. Becauseof springfrost damage at thebeginning of mmof elongation had occurred thethird growing season at SilverCity, SCDEV3 reflects GHDR Durationof elongation: the number of days between theeffects of springfrost injury on thethird-year growth theinitiation and cessation of from at SCDEV4 GHRT Elongationper day during the period for which 20% trees a commonheight yr 2; and to 80% ofthe shoot elongated reflectsthe growth that occurred in yr4 independentlyof GHHT2 Age2 height thefrost damage that occurred at thebeginning of yr 3. GHDIA Age 2 diameterat thesoil surface GHLL Averagelength of leaves from ten fascicles distributed Greenhousetests -Seedlings from each population were throughthe 2-year terminal shoot grownfor 6 mo in plasticcontainers (740 cm3)in a shade- GHLNM Averagenumber of leaves in 10 fasciclesdistributed throughthe 2-yr terminal shoot house at Moscow,Idaho. The experimentaldesign con- GHRTO Ratio ofthe to sistedof nine seedlingsgrowing in row plotsin each of 2-yrheight the2-yr diameter threeblocks. Trays of containerscontaining three plots a Firsttwo letters of variable code the test site: PR = PriestRiver, SC weretransferred into an unheatedgreenhouse for the win- -Silver City, WR = Window Rock, and GH = Greenhouse. termonths and, in earlyMarch of the secondgrowing season,were exposed to a daytimetemperature of about individualtrees (Table 2). Additionalmeasurements pro- 25 C, whichwas allowedto cool to a minimumof 13 C duced fivemore variables that reflected growth and de- at night.All seedlingswere measured three times each velopmentin thegreenhouse environment. weekuntil elongation of the terminal shoot was well under way.Thereafter, each seedlingwas measuredtwice each Patterns of variation-Population differentiationwas weekuntil elongation of thepreformed shoot was com- assessedfrom analyses of variance(SAS Institute,1985, plete. using Type III estimablefunctions), which were per- Periodicmeasurements allowed shoot elongation of in- formedaccording to thefollowing model of randomef- dividualtrees to be modeledwith a logisticfunction with fects: a hyperbolictime term (Rehfeldt and Wykoff,1981): Yijk = + Pi + Bj + Ei. + Wijk Y = (1 + be{-rx + (c/X)})-l whereYijk is an observationon seedlingk in blockj from whereY is theproportion of total increment attained by populationi; ,uis the mean; Pi and Bj are theeffects of day X; b, r,and c are regressioncoefficients; and e is the populationsand blocks,respectively; Eii is the experi- base ofnatural logarithms. Regression statistics produced mentalerror, the interaction of blocks with populations; bythis function allowed calculation of six variablesthat and Wijkis thesampling error. Under the assumption that wereused to describethe pattern of shootelongation of blocksand populationsare randomvariates, the experi- March 1993] REHFELDT-GENETIC VARIATION IN PONDEROSAE 333 mentalerror becomes the variance appropriatefor testing Using lsd to assess rates of differentiationintuitively differencesamong populations. For these analyses, the suggeststhat interpretations are dependenton sample sizes harmonicmean ofobservations per plotwas 9.92 at Priest and experimentalerrors (uniformityof cultural condi- River, 9.70 at Silver City, 8.62 at Window Rock, and tions). However, because the cones from ten trees had 8.84 in the greenhouse. been bulked, variances within plots are composed not An attemptwas made to reduce the numberof dimen- only of environmentaleffects at the plantingsite, but also sions about which differentiationwas being expressedby of genetic variances within populations. Consequently, usingprincipal component analyses (SAS Institute,1985) lsd would still reflectthe geneticvariances withinpopu- on the data fromeach test site. However, differentiation lations even if samples were large and the controlof mi- of populationswas so pronouncedthat the principal com- croenvironmentaleffects was complete. ponentanalysis allowed the numberof variatesfor which populations differedsignificantly to be reduced by only Interrelationamong species-Canonical discriminant 8. Because ofthis,because theuse ofprincipalcomponents analyses (SAS Institute,1985) were used to assess mul- resultsin a loss of information(Johnson and Wichern, tivariaterelationships among species.These analyseswere 1982), and because principal components oftenare im- performedseparately on data fromeach test site. They practicalin the generalapplication of regressionmodels, used observations on individual seedlingsfrom each of subsequent analyses involved only the originalvariates. the populations of Apache and Arizona pines, fivepop- Multiple regressiontechniques were used to develop a ulations of ponderosa pine, and the Barfootpopulation. generalmodel ofgeneticvariation according to procedures The fiveponderosa pine populationswere geographically detailed earlier(Rehfeldt, 1989): proximal and elevationallysimilar to the Arizona pine, 1) Deriving independentvariables fromlatitude (LT), Apache pine, and Barfootpopulation. Of the five,three longitude(LN), and elevation that could serve as surro- came fromthe Mogollon Rim and one each came from gates for the complex three-dimensionalenvironmental theTularosa and Pinaleno Mountains(Fig. 1). These anal- gradientsthat have operated in naturalselection. For the yses also used all trees fromBarfoot, regardless of leaf presentanalyses, independent variables included the first morphology.Barfoot was of particularinterest because and second powers of elevation and the first,second, and the performanceof its progenies suggestedthat the pa- thirdpowers of LT, LN, LT x LN and LT . LN. The rentalpopulation was eithera hybridswarm or a mixture lattertwo of these variables produced a gridfrom north- of Arizona and ponderosa pines. west to southeast and fromnortheast to southwest,re- spectively; 2) screeningthe independentvariables by stepwisere- RESULTS gression,the best model of which was judged relativeto statisticalsignificance, the Mallows statistic,and patterns The resultsconsider first, variation among populations displayed by the residuals (Draper and Smith, 1981); of ponderosa pine, and second, biosystematicimplica- 3) refininga stepwise model with multipleregression tions. to develop the most parsimonious model; 4) and finally,plotting elevational and geographicpat- Population differentiation-Differences among popu- ternsof variationto assure thatthe models were sensible lations were detected(P < 0.05) forall but one variable, biologically. withthe effectsof populationsaccounting for at least 40% A fifthstep, verificationof the model, could not be of the total variance for ten of the variables (Table 3). attemptedbecause independentdata were not available. Because of the experimentaldesign that was used, weak Models were developed forponderosa pine by exclud- effectsfor blocks (Table 3) meant thatthe intraclasscor- ing populations of Apache and Arizona pine. This left95 relation for the effectsof populations approximates the populations forthe regressionanalyses. Excluding seed- ratio of the total geneticvariance to the phenotypicvari- lingsfrom Barfoot that had leafcharacteristics of Arizona ance. The size of these intraclasscorrelations, therefore, pine removednine treesfrom greenhouse tests, three from atteststo pronounced differentiationof populations. PriestRiver tests,three from Silver City,and none from The strongesteffects of populations were associated Window Rock where winterinjuries and mortalityhad with the cessation of shoot elongation (GHEN), the du- decimated the Barfootpopulation. ration of elongation(GHDR), and number of leaves per Rates of differentiationalong geographicor elevational fascicle(PRLNM and GHLNM), variablesfor which pop- clines were interpretedrelative to the least significant ulation effectsaccounted foras much as 72% of the total difference(Steel and Torrie, 1960) among populations at variance. For most of the variables, however, much of the 20% significancelevel (lsd 0.2). Values of lsd were. the total variance was associated with sampling errors used because stepwisemodels developed fromnumerous that are composed of microenvironmentaleffects at the independentvariables are subjectto overfittingand over- test site and geneticvariances withinpopulations. Since parameterizing(Draper and Smith, 1981). The use of lsd geneticvariances formany of these traitstend to be pro- guardedagainst accepting fallacious results. The 20% sig- nounced (Rehfeldt,1992), large samplingerrors also can nificancelevel was used to guard against accepting no be expected.The deviation fromregression of 4-yrheight differencesamong populations when differencesactually on 2-yrheight at Window Rock (WRDEV) was the only exist (type II errors);such errorsprovide the greateptpo- variable for which no differenceswere detected among tentialfor faulty interpretations when models are applied. populations,a resultsuggesting that growth potential (in- Values of lsd werecalculated from the interaction of blocks nate ability to produce and assimilate photosynthatein and populations in the analysis of variance. the absence of environmentaleffects that mask the ge- 334 AMERICAN JOURNAL OF BOTANY [Vol. 80

TABLE3. Mean ofall ponderosapine populations, range of mean diferences among populations, and resultsof analysis of variance. Results of the anlaysesof varianceare presentedas intraclasscorrelations, the ratio of thevariance component for the indicated effects to thesum of all components

Source of variance Range Experimental Variable Units Mean Maximum Minimum Blocks Populations error Sampling error PRHT4 cm 81 97 58 0.07** 0.40** 0.02** 0.51 PRDIA mm 17 21 14 0.07** 0.41 ** 0.06** 0.47 PRDEV cm 0.0 11 -11 0.01** 0. 14** 0.06** 0.80 PRCL Score 0.8 1.0 0.4 0.00 0.10** 0.00 0.89 PRRTO cm/mm 2.9 3.4 2.5 0.02** 0.22** 0.09** 0.67 PRLL cm 17 19 14 0.03** 0.26** 0.04** 0.67 PRLW mm 1.7 1.8 1.6 0.04** 0.17** 0.08** 0.71 PRLNM Count 3.1 3.2 2.0 0.00 0.72** 0.02** 0.26 SCHT4 cm 57 85 28 0.03** 0.41** 0.11** 0.46 SCSF Score 2.2 2.8 1.5 0.01** 0.13** 0.04** 0.82 SCDEV3 cm 0 7 -4 0.01 0.13** 0.11** 0.76 SCDEV4 cm 0 10 -10 0.07** 0.15** 0.19** 0.60 WRHT4 cm 26 33 19 0.01 ** 0.22** 0.16** 0.61 WRDEV cm 0 5 -4 0.06** 0.03 0.25** 0.66 WRWI % 27 92 0 0.05** 0.19** 0.06** 0.70 WRLL cm 7 8 5 0.16** 0.03* 0.23** 0.58 WRDEAD % 11 35 0 0.04** 0.09** 0.16** 0.71 GHEL mm 134 175 71 0.00 0.40** 0.05** 0.55 GHS2 d 4 6 3 0.01* 0.03* 0.15** 0.82 GHS8 d 8 11 7 0.00 0.07* 0.19** 0.74 GHEN d 38 50 29 0.01** 0.50** 0.02 0.48 GHDR d 34 46 25 0.01 ** 0.46** 0.02 0.51 GHRT mm/d 6 8 4 0.00 0.23** 0.04* 0.73 GHHT2 mm 240 319 130 0.00 0.40** 0.07** 0.54 GHDIA mm 6 9 4 0.01 ** 0.41 ** 0.06** 0.53 GHLL cm 14 167 106 0.00 0.20** 0.04* 0.76 GHLNM Count 3.1 3.4 3.0 0.00 0.61** 0.02* 0.37 GHRTO mm/mm 38 44 30 0.01* 0.15** 0.19** 0.65 *Significance of F-value at 0.05 > PF> 0.01. * Significanceof F-value at P < 0.0 1. notype) had been masked in the rigorousenvironment significant(P < 0.01) proportionsof the variance among (Table 1). populations forall variables (Table 4). Values of R2 were The degree of differentiationamong populations is il- as high as 0.85, averaging 0.58. For only eight of the lustratedreadily by the variables describingshoot elon- variables did regressionmodels account forless than half gation. In the greenhouse,shoot elongationof individual of the variance among populations. These results thus treeswas completedbetween 17 and 62 d afterthe green- demonstratethat much ofthe variation follows systematic house was warmed. This meant that seven to 19 obser- vations were available for the logistic regressionsthat 200 describedshoot elongationof individual treesnearly per- fectly:values of R2 ranged from0.93 to essentially1.0, averaging0.99. E Because shoot elongation is one of a sequence of de- ~150-150 velopmental events that must be completed within the ~~~~~~/----H frost-freeseason, differentpatterns of shoot growthillus- F trateadaptation to heterogeneousenvironments. As shown I in Fig. 2, differencesin the startof elongationwere small, z but differencesin therate, duration, cessation, and amount ..D100I-~~~~~~~~~~~~~/ A of elongation were pronounced. In this figure,the high growthpotential and long duration of shoot growthof z '/ treesfrom the Bradshaw Mountains (localityE) describes 0 o 1 -50 / a populationfrom the lowest elevation (1,700 m) sampled w in the study. The remaininggraphs are for populations ~~~~~~~~~~~/P fromabout the same elevation (2,300-2,700 m) but dif- ferentgeographic localities. In contrastto southernpop- ulations,those fromthe north(e.g., San Juan Mountains 0- [localityB] or MarkaguntPlateau [localityA]) combined 0 15 30 45 60 an early cessation of elongationwith low growthpoten- tials. Fig. 2. Mean cumulative shoot elongationof seedlingsfrom six lo- Multiple regressionmodels accounted for statistically calities identifiedin Fig. 1. March 1993] REHFELDT-GENETIC VARIATION IN PONDEROSAE 335

TABLE4. Resultsof multiple regression analyses. Geographic patterns of variation are keyedto Fig. 4. All regressionswere statistically significant at probabilitiesless than 0.01

Patternof geneticvariation Elevation cline Geographic cline Independent Dependent variable R2 variables Sign Slope Shape Slope Direction PRHT4 0.85 6 Negative Steep Nonlinear Steep South to North PRDIA 0.82 5 a Steep Nonlinear Steep South to North PRDEV 0.59 3 Positive Steep Nonlinear Steep South to North PRCL 0.49 8 Negative Shallow Linear Shallow Southwestto Northeast PRRTO 0.52 3 Negative Steep Linear Shallow North to South PRLL 0.75 6 Negative Steep Nonlinear Moderate South to North PRLW 0.23 3 b Shallow Northeastto Southwest PRLNM 0.64 7 Negative Shallow Nonlinear Shallow South to North SCHT4 0.85 3 Negative Steep Linear Moderate South to North SCSF 0.54 8 Positive Moderate Linear Moderate Northeastto Southwest SCDEV3 0.63 4 Negative Steep Linear Moderate Southwestto Northeast SCDEV4 0.50 2 b Moderate South to North WRHT4 0.71 3 Negative Shallow Linear Moderate South to North WRDEV 0.20 3 b Shallow Southeast to Northwest WRWI 0.79 3 b Steep South to North WRLL 0.40 2 Negative Shallow Linear Shallow Southeast to Northwest WRDEAD 0.26 3 Positive Shallow Linear Shallow Southwestto Northeast GHEL 0.83 4 Negative Moderate Linear Steep South to North GHS2 0.16 5 a Shallow Nonlinear Shallow Southeast to Northwest GHS8 0.27 6 a Shallow Nonlinear Moderate Southwestto Northeast GHEN 0.83 5 Negative Moderate Linear Steep South to North GHDR 0.84 5 Negative Moderate Linear Steep South to North GHRT 0.66 4 b Moderate South to North GHHT2 0.83 5 Negative Moderate Nonlinear Moderate South to North GHDIA 0.83 5 Negative Moderate Nonlinear Steep South to North GHLL 0.60 7 Negative Moderate Linear Moderate South to North GHLNM 0.55 5 a Shallow Nonlinear Moderate South to North GHRTO 0.44 5 Negative Shallow Linear Shallow North to South a No generalsign. b No elevational cline. patternsthat regressionmodels are remarkablycapable Nonlinear elevational clines typifiedthe general re- ofdescribing. Regression equations are available fromthe sponse of manyvariables measuredat PriestRiver, a site author. to whichsome populationswere transferred north as much The models described genetic variation as occurring as 170 of latitude. While linear clines commonly related along both elevational and geographicclines (Table 4). growthpotential to elevation (see GHHT2 and GHLL, Elevational clines (Fig. 3) were detected for all but five Fig. 3), populations fromthe mildestenvironments (low- of the variables and exhibiteda varietyof shapes, slopes, est elevations)evidently were incapable offully expressing and signs. Ten variables exhibitednonlinear elevational theirgrowth potential at the northernsite. Yet, popula- clines, all of which were of shape similar to those of tions fromthe middle and high elevations seemed to be PRHT4 and PRDEV in Fig. 3. While the sign of the unaffected.As a result,the populations from high ele- elevational cline is directlyinterpretable for linear clines, vations had the most growthfrom a common 2-yrheight onlya generalcant describesthe slope of nonlinearclines; (PRDEV). The nonlinearcline for4-yr height (PRHT4), thus,the generalrelationship for PRHT4 is negativeand therefore,resulted first from a differencein growthpo- that for PRDEV is positive (Fig. 3). The strengthof a tentialamong populations and second froma difference cline was interpretedrelative to lsd 0.2: forclines of steep in the degreethat the potentialwas masked. slope, differencesequal to lsd 0.2 are expected to occur Rates of differentiationalong linear clines are readily betweenpopulations withinthe same geographiclocality interpretedin relationto lsd 0.2. For the 4-yrheight of that are separated by less than 300 m of elevation; for treesat SilverCity (SCHT4), thevariable with the steepest those of moderate slope, 300 to 500 m; and forthose of linear cline, populations in the same locality that were shallow slope, more than 500 m. separated in elevation by about 300 m tended to be ge- Together,the clines illustrate declining growth potential neticallydifferent. This variable, incidentally,integrated as the elevation of the population increases. Thus, the genetic differencesin growthpotential with those con- durationof developmental events, height-diameter ratios, trollingtolerance to springfrosts. Frost injuryto popu- leaf lengths,and heightswere negativelyrelated to lations fromhigh elevations accentuatedthe differences elevation. In addition, as elevation increased, the color in heightthat accrue according to thenegative relationship of succulent shoots tended to change from blue-green between growthpotential and elevation. towardgreen, the number of leaves per fascicledecreased, Rates ofdifferentiation along thenonlinear clines, how- damage fromthe springfrost at Silver Cityincreased, and ever, depend on the elevations at which comparisonsare mortalityat Window Rock increased. made. The resultsfor 4-yr height at PriestRiver (PRHT4), 336 AMERICAN JOURNAL OF BOTANY [Vol. 80

325- GHHT2 170 GHLL

E ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ -

o < = , e ~~~~~~~~~~~~~~~~~~~~G s <

125 g- - C]) 0- __

H-~~~~~~~~~~~~~- r - -T

160 2300 3000 1600 2300 30 E LEVATI ON ( M) E LEVATI ON (M )

100L PRHT4 15 PIRDEV

:H

-I C . I, . -.

2 5 - - 230 30 0 1600 2300 3000 1600 230 ELEVATION (M) ELEVATION (M)

7.5- SCIOEV3 0- H-- H-1 SCHT4

75 - , , ,

1600 2300 3000 1600 2300 3000 ELEVATION (M) ELEVATION (M) Fig. 3. Mean performanceof ponderosa pine populations forsix variables plotted accordingto the elevation of theseed source. Variables are referencedin Table 2. Each regressionline representsa geographiclocality identified in Fig. 1,>and the bracketnear theorigin quantifies lsd 0.2. for instance, imply differentiationof populations from in Fig. 3 by regressionlines of differentintercept. This highelevations (>2,400 m) when separatedby about 220 component is shown in detail for six variables in Fig. 4 m of elevation, but suggestlittle differentiationamong wheregenetic variation among populations predictedfor populations fromlower elevations. Similarly,analyses of an elevation of 2,400 m is representedby isopleths. In growth from a common 2-yr height at Priest River this figure,the interval between isopleths equals 1/2 lsd (PRDEV) suggestdifferentiation among populationsfrom 0.2. This means thatpopulations separated by a geographic low elevation (<2,300 m) if separated by at least 220 m distance equaling two intervalsare expected to differat of elevation, but imply little differentiationamong the about the 20% probabilitylevel. Geographicpatterns for populations at higherelevations. othervariables are documentedin Table 4 wherethe slope The models thus suggestthat elevation and, therefore, ofthecline is describedas steepwhen the distance between lengthof thefrost-free period are closelyrelated to genetic isopleths averages less than 50 km, moderate when av- differentiation.This means that at localities such as the eragingbetween 50 and 100 kkm,and shallow when av- Tularosa Mountains (locality G, Fig. 1) where the ele- eragingmore than 100 km. Here also, in describingthe vational distributionof ponderosa pine is broad (Fig. 3), directionof a cline, the geographicdirection listed firstis geneticdifferentiation across thelandscape is pronounced. that toward which highestvalues are predicted. A geographiccomponent to geneticvariation was de- Althoughgeographic patterns were detected for all vari- tected in all of the variables (Table 3), and is illustrated ables, the model for the growthfrom a common 2-yr March 1993] REHFELDT-GENETIC VARIATION IN PONDEROSAE 337

GHDR WRWI

GHDIA GHRT

GHS8 SCSF

Fig. 4. Geographic patternsof vaniationfor six variables predictedby regressionmodels for populations at a common elevation (2,400 in). Isopleths connect populations of similar performance,and the intervalbetween isopleths equals 1/2 lsd 0.2. The mean isopleth (X) along with the positive (+) and negative(-) deviations fromthe mean are marked. Variables are referencedin Table 2. 338 AMERICAN JOURNAL OF BOTANY [Vol. 80

TABLE5. Simplecorrelations ofselected variables with all variables. 200- Onlythose coefficients with an absolutevalue greater than 0.30 are presented.All coefficients are statistically significant atprobabilities lessthan 0.001 - AZ F GHDR WRWI SCDEV3 PRLL PRCL GHLNM 150 - I PRHT4 0.85 0.71 0.48 0.81 0.47 0.42 G PRDIA 0.83 0.70 0.30 0.76 0.38 0.58 PRDEV 0.57 0.59 0.51 0.56 z AP PRCL 0.46 0.48 0.50 0.43 1.00 ? 100/ PRRTO 0.35 0.69 0.33 0.46 -0.31 PRLL 0.70 0.62 0.39 1.00 0.42 0.30 PRLW PRLNM 0.49 0.53 0.32 0.81 SCHT4 0.81 0.64 0.75 0.75 0.56 - 50- SCSF -0.58 -0.42 SCDEV3 0.44 0.35 1.00 0.39 0.50 SCDEV4 0.63 0.53 0.57 0.49 WRHT4 0.71 0.48 0.59 0.68 0.40 WRDEV WRWI 0.75 1.00 0.35 0.62 0.48 0.53 0 15 30 45 60 WRLL 0.47 0.39 0.40 0.58 D AY WRDEAD 0.38 0.46 0.30 GHEL 0.91 0.70 0.46 0.74 0.48 0.40 Fig. 5. Mean cumulativeshoot elongation of seedlingsof Arizona GHS2 pine(AZ), Apachepine (AP), and threepopulations of ponderosa pine, GHS8 0.30 keyedto Fig. 1. The populationsof ponderosa pine were geographically GHEN 0.99 0.75 0.46 0.71 0.45 0.49 proximaland elevationallysimilar to thoseof the other species. GHDR 1.00 0.75 0.45 0.70 0.46 0.49 GHRT 0.65 0.50 0.41 0.65 0.39 GHHT2 0.82 0.61 0.58 0.72 0.55 inent in regressionmodels for 16 of the variables. This GHDIA 0.88 0.76 0.44 0.72 0.47 0.53 implies that many of the variables were stronglyinter- GHLL 0.59 0.49 0.36 0.72 0.40 correlated(Table 5), a resultcommon in studies of pop- GHLNM 0.49 0.54 0.30 1.00 ulationdifferentiation in ponderosapine (Rehfeldt,1 986a, GHRTO -0.43 b, 1990).

Biosystematicimplications-Some ofthe effects of pop- heightat Window Rock (WRDEV) failed to predictdif- ulations detected by the analysis of variance (Table 3) ferencesthat exceeded lsd 0.2. This variable, moreover, resultfrom differences between ponderosa, Apache, and was the only variable forwhich the effectsof populations Arizona pines. Differencesamong these species are readily lacked statisticalsignificance. Of the 27 remainingvari- illustratedby patternsof shootelongation (Fig. 5). Apache ables, 18 exhibited geographic patterns that were de- pines tended to startelongation the latest, elongated at scribedby latitudinalclines or variationsthereon (Table the slowest rate, and elongated the least, despite having 3); fourof these latitudinalclines are presentedin Fig. 4. a longduration of shootgrowth. Seedlings of Arizona pine All but two of the latitudinalclines were inclinedtoward coupled a high rate of elongationwith a long durationto thesouth; for only the ratios of height to diameter(PRRTO achieve the highestgrowth potential. Seedlings of pon- and GHRTO) were the largestvalues (least stockytrees) derosa pine grew at a rapid rate but ceased elongating found to the north.Together, these clines illustratethat early. when populationsfrom the same elevationare compared, A comparisonof mean values of seedlingsrepresenting growthpotential decreases as latitude increases and the the three species and the Barfoot population (Table 6) lengthof the frost-freeperiod decreases. As a result,vari- shows,first, the numeroustraits that differentiate Apache ables as differentas the duration of shoot elongationin and ponderosa pines. Apache pine's lower growthpoten- the greenhouse(GHDR) and winterinjuries at Window tial, longerleaves, slower rates of shoot elongation,and Rock (WRWI) can exhibitpatterns that are nearlyiden- stockierform are prominent.Differences between the pop- tical (Fig. 4). ulationof Arizona pine and thoseofponderosa pine center While latitudinalclines were the strongestof the geo- on Arizona pine's higherleaf counts, narrower leaves, and graphic patterns,two secondary patternswere also evi- slightlyhigher growth potential. It follows,therefore, that dent. The strongestof these secondarypatterns (Fig. 4) seedlingsof Arizona pine differedfrom those of Apache occurredacross an axis fromnortheast to southwestand pine in nearly all charactersmeasured. The trees from showed thatpopulations from the Rocky Mountains were Barfoot,however, minus those trees with leaf morphol- the earliest to begin shoot elongationin the greenhouse ogies of Arizona pine, differedfrom ponderosa pine for (GHS8) and were the most susceptibleto damage from about one-halfof the charactersthat separated Arizona the earlyspring frost at Silver City (SCSF and SCDEV3). pine and ponderosa pine. These same trees, moreover, Another weak pattern (not presented) occurred from differedfrom ponderosa pine forthree of the characters southeastto northwestand also tended to separate pop- that separated Apache pine fromponderosa pine. ulations from the Rocky Mountains from those of the The canonical discriminantanalyses produced two ei- Southwest. genvalues thataccounted forat least 90% of the variance Clines involvingboth elevationand latitudewere prom- among the fourgroups discussed above. In fact,for all March 1993] REHFELDT-GENETIC VARIATION IN PONDEROSAE 339

TABLE 6. Mean valuesforseedlingsfromfivepopulations ofponderosa ponderosa pine, althoughthe rangeand densityof plotted pine as comparedto thoseof theArizona pine, Apache pine, and values are accuratelydepicted. Discriminant scores for Barfootpopulations. For populationsother than ponderosa pine, data fromSilver City and Window Rock are not presented onlythose means are presented that deviatefrom ponderosa pine by an amountgreater than lsd 0.01. The populationsof ponderosa because theresults were of lesserresolution. While it may pinewere geographically proximal and elevationally similar to those seem anomalous forthe two least naturalof the testsites ofthe other groups to provide the results of greatestresolution, one must recallthat 1) Arizona and Apache pines performedpoorly Variable Units Apache pine Arizona pine Ponderosa pine Barfoot in the harsh environmentat Window Rock, and 2) the PRHT4 cm 68.1 105.5 90.8 leafcharacters and ratiosof heightto diameterthat prom- PRDIA mm 25.0 18.9 inentlydistinguished the species (Table 6) were not mea- PRDEV cm 7.9 0.9 9.8 sured at Silver City. PRCL Score 0.66 1.00 0.89 1.00 For both the PriestRiver and greenhousedata sets,the PRRTO cm/mm 1.7 3.0 2.6 discriminantfunction nicely separated the threespecies. PRLL cm 21.3 18.5 PRLW mm 1.1 1.7 1.5 On the one hand, the greenhousedata suggestedthat the PRLNM Count 4.57 3.04 3.30 Barfootpopulation is a mixtureof ponderosa and Arizona SCHT4 cm 56.2 68.7 pines, the two of which could be separated by leaf mor- SCSF Score 1.3 2.1 phology. Thus, the three individuals fromBarfoot that SCDEV3 cm 0.7 1.7 had leaf morphologies of Arizona pine fell within the SCDEV4 cm 12.0 13.5 -0.9 clusterof Arizona pines (Fig. 6). But on the otherhand, WRHT4 cm 17.1 28.8 WRDEV cm -1.1 the distributionof Barfoot trees for Priest River data WRWI % 95.0 95.2 56.4 92.5 implied introgressionthat may even implicate Apache WRLL cm 7.0 pine. Noteworthyis one seedling that displayed a leaf WR morphologyof Arizona pine but otherwise resembled DEAD % 92.5 21.0 ponderosa pine. GHEL mm 111.5 153.6 GHS2 d 6.9 4.0 GHS8 d 13.6 8.6 GHEN d 51.4 52.3 42.9 48.8 DISCUSSION GHDR d 44.5 47.8 38.9 44.8 When grown in environmentallydisparate common GHRT mm/d 4.2 7.0 GHHT2 mm 190.7 276.8 gardens,seedling populations representingthe Pondero- GHDIA mm 9.4 7.5 9.3 sae of the Southwestexhibited genetic differences for 27 GHLL cm 15.7 of the 28 morphologicaland developmental traitsana- GHLNM Count 3.29 4.86 3.09 3.67 lyzed.While some ofthisvariation was due to-interspecific GHRTO mm/mm 20.7 37.2 29.8 differences,most reflectedgenetic differentiation among the 95 populations of ponderosa pine that were studied. but the Window Rock data set, the firsttwo eigenvalues accountedfor at least 96% ofthe variance between groups. Populationdifferentiation -Mathematical models were In Fig. 6, scores foreach seedlingare plottedfor the first remarkablysuccessful in describingpatterns of genetic two canonical variables forPriest River and greenhouse variation across the landscape. Because intercorrelations data. Only about one-halfof the plots are presentedfor among traits were strong,similar patternsof variation

6-E 10- EE E E

E E E C28 E x ~~~~~~~~~E E E EE E E E E E E E E E

< 3 E E EEE

- o A A A p E E * * EA A -1 A A < tpFp

-3 - 1 35 A -10-26 -2 - 0 2

FIRST CANONICAL A X IS FI R ST C A NONI C AL A XI S Fig. 6. Resultsof canonical discriminant analyses for greenhouse studies (left) and PriestRiver tests (right). P = ponderosapine; A = Arizona pine;E = Apache pine;asterisk = Barfoottrees with leafmorphologies of?P. arizonica; circle = otherBarfoot trees. 340 AMERICAN JOURNAL OF BOTANY [Vol. 80 were evident. The most prominentpatterns implicated resents1/2 lsd 0.2, a value that equals the amount of dif- elevation and latitude, two variables whose relation to ferentiationexpected between populations separated by environmentalfactors is well known. 110 m (10 frost-freed) along theelevational cline. If,then, Correlatedsets of traitsapparently have resultedfrom thelatitudinal clines reflect differentiation associated with parallel selectionto produce coherent(sensu Clausen and thefrost-free period, there should be about seven isopleths Hiesey, 1960) geneticsystems involving the components separatingthe San Juan and Tularosa Mountains. Of the of an annual sequence of developmentalevents. This se- latitudinalclines in Fig. 4, nine isopleths separate these quence begins with dehardeningin the spring;includes localitiesfor GHDR, seven forWRWI, eightfor GHDIA, shoot elongation,leaf expansion, bud development,di- and six forGHRT. Clearly,much of the variation asso- ametergrowth, and lignification;and concludes withcold ciated withboth latitudeand elevationreflects adaptation acclimation. As described by Dietrichson(1964), the se- to the lengthof the frost-freeperiod. quence has been molded to fitwithin a growingseason Secondary geographicclines were much weaker than offinite length, and, therefore,the duration of events tends the latitudinalclines and tended to separate populations to be intercorrelated.As a result,a diverse assortmentof fromthe Rocky Mountains fromthose of the Southwest variablesbecomes correlatedeven thoughthe traits might (see GHS8 and WRWI, Fig. 4). This separationwas also be measured in verydifferent environments. In thiscase, apparentfor populations from the Colorado Plateau (Reh- the duration of shoot growth(GHDR), the number of feldt, 1990) and seemed to reflectthe transitionfrom a leaves per fascicle(GHLNM and PRLNM), winterinjury climate dominated by winter-springdroughts in the at WindowRock (WRWI), leaflengths(PRLL and GHLL), Southwestto that characterizedby summer droughtsin stem color (PRCL), spring frostdamage at Silver City the Rocky Mountains. Populations fromthe continental (SCDEV3 and SCSF), and various expressionsof growth climate of the Rocky Mountains initiateshoot elongation potentialare stronglyintercorrelated. Because shootgrowth rapidlywhile those fromthe Southwestinitiate shoot ac- in ponderosa pine is predetermined(Sacher, 1954), the tivitymore slowly. Because of this, southernColorado pattern of shoot elongation can act as a surrogatefor populations were the most susceptibleto damage froma understandingthe adaptation of the entiresequence to a late springfrost at Silver City (SCSF, Fig. 4). heterogeneousenvironment. Together,the elevational and geographicclines describe Geneticvariation that is systematicallydistributed along complex patternsacross the landscape. Populations in- environmentalgradients undoubtedly arises fromnatural habitingthe same elevation in differentmountain ranges selection. Selection apparentlyhas molded a systemof tend to be differentgenetically (Fig. 3). This also means loosely intercorrelatedtraits that jointly adapt popula- thatpopulations capable of similarresponses are expected tions to the Southwest'sspatially heterogeneous environ- to recurat differentelevations in differentmountain rang- ments,a conclusion compatible withresults of studiesof es. For instance,populations capable of developing rel- the same species from the Colorado Plateau (Rehfeldt, atively long leaves (e.g., GHLL = 150 mm, Fig. 2) are 1990), the Inland Northwest(Madsen and Blake, 1977; expectedat 1,750 m in the Bradshaw Mountains (locality Rehfeldt,1991), the Sierra Nevada (Callaham and Lid- E), 2,000 m in the Sacramento Mountains (locality H), dicoet, 1961; Conkle, 1973), and the easternslopes of the 2,300 m in the Pinaleno Mountains (localityI), and 2,800 Rocky Mountains (Read, 1980, 1983). m in the Tularosa Mountains (locality G). Elevational clines have direct microevolutionaryin- Examining the frequencyby which populations that terpretations.As elevation increases, temperaturesde- exhibit similar responses recur with respect to several crease, with a reductionin mean annual temperatureof variablesis facilitatedby usingthe regression model. Each 1 C leaving 12 fewerfrost-free days (Baker, 1944). Con- regressionequation can be used to generatea data base sequently,seedlings frompopulations distributedalong containingpredicted values forthe entiregeographic and an elevational gradientdisplay adaptations to growing elevational distributionof ponderosa pine withinthe re- seasons of differentlength. When compared in a common gion of study.Then, by surroundingeach observationin environment,populations fromlow elevations expressa the data base with a confidenceinterval of ? 1/2 lsd 0.2, high growthpotential, grow fora relativelylong period, populations readilycan be grouped accordingto similar and become large; populations adapted to shortgrowing responses. seasons cease developmentearly and tend to be small. In The expectedrecurrence of populationscapable of sim- the Southwest,an elevational intervalof 1,000 m tends ilar responsesis illustratedin Fig. 7 forsix targetedpop- to be associated witha changeof 90 frost-freedays (Baker, ulations(pyramids). For populationsin the centerof large 1944). As shown in this study,populations separated in continuous distributions(Tularosa Mountains [locality elevation by about 220 m tend to be differentgenetically. G] and Sangre de Cristo Mountains [localityC]) recur- This suggeststhat populations occupyingenvironments rence is widespread. But recurrenceis considerablyre- thatdiffer by 20 frost-freedays tend to differgenetically. stricted for populations in isolated mountain ranges This conclusion is remarkablysimilar to thatreached for whetherthe localityis on theperiphery (Bradshaw Moun- the same varietyon the Colorado Plateau (22 d) but is tains [localityE] and SacramentoMountains [localityH]) much less than that obtained forP. p. var. ponderosa in or in the center (Defiance Plateau [locality D]) of the the Inland Northwest(35 d). species distribution.For the isolated ranges in south- The latitudinalclines also seem to reflectadaptation to eastern Arizona (Pinaleno Mountains [localityl]),how- frost-freeperiods of variable length.Baker (1944) shows ever, recurrenceis extremelylimited (Fig. 7). that fora constantelevation, the frost-freeperiod differs Practical uses of models of genetic variation are nu- by about 70 d between the San Juan Mountains and the merous and diverse (see Rehfeldt, 1991). For artificial Tularosa Mountains (Fig. 1). In Fig. 4, each isoplethrep- reforestation,one can assume thatthe targetedlocations March 1993] REHFELDT-GENETIC VARIATION IN PONDEROSAE 341

ELEVATION) G ELEVAON C

2900 - 2900 -

2500- 2500-

2100 -2100-

114 38.5 114 38.o

108 LONGfWDE ic8 ZU LATMDELATITUDE LAI~~~~~~~~~~~~~~~~~~~~~~~ATITUDE D .0 105 105

ELEVATION(M) E ELEVATONO) H

12002900 290001700

114 38.5 114 38.

LONGITUDEioe LOITUDE 108 LATITUDE LATfflJDE 105t.0 105

ELEVATION(M) ELEVATION(M)

2900 2900-

2500 -200

2100 2100

1700 -4 P-- ~~~~~~~~~~~~~~~~~~1700

LONGITUDE 108 . LATiTUDE t

Fig. 7. Using models of geneticvariation to locate populations (balloons) expected to exhibitresponses in common gardensthat are similarto those of a targetedpopulation (pyramid).Letters key the generalgeographic locality (Fig. 1) of the targetpopulation. 342 AMERICAN JOURNAL OF BOTANY [Vol. 80 representeither planting sites or seed productionareas. ables could have been interpretedas introgressionwith The model can thenbe used to 1) locate sources of seeds Apache pine, while those forthe othersix variables could that should be geneticallycompatible with the environ- have been interpretedin termsof introgressionwith Ar- ment at the target,or 2) select plantingsites for which izona pine. Althoughlogical, such interpretationsmust seeds gatheredfrom the targetedpopulation should be be tempered by the fact that only two ponderosa pine adapted. The model mightalso be used to locate disjunct populations were sampled from the isolated ranges in populations so unique geneticallythat gene conservation southern Arizona, and the southernmostof these was programsmay be desirableas, forexample, in the Pinale- Barfoot.Since Barfootseems to be introgressed,it is likely no Mountains (localityI). Additional uses mightinclude thatthe changein the slope of geographicclines in south- assessingthe impact ofclimate change on theadaptedness ern Arizona was due to the fittingof the regressionto the of populations,understanding phenotypic variation, and performanceof Barfoot progenies ratherthan to wide- delimitingseed orchardsand breedingzones. spread introgression. The usefulnessof a model, however, depends on its Nevertheless,these results along withthose of Peloquin credibility,and a firststep in acquiringcredibility involves supportintrogression, primarily involving ponderosa and verification.Unfortunately, independent data currently Arizona pines but also implicatingApache pine. Yet the are not available forverifying the presentmodel. There- degreeof hybridization,the amount and directionof gene fore,the model should be used with discretion.In addi- flow,and the ecological genetics of species of the Pon- tion, genetic variation mightbe occurringalong clines derosae in theSouthwest deserve a thoroughexamination. independentof those already detected.This would mean thatthe recurrence of similarpopulations across theland- scape is more restrictedthan the resultsimply. For this LITERATURE CITED reason, functionalmodels require periodic updating as additional mathematical descriptorsbecome available. BAKER,F. S. 1944. Mountains climates of the westernUnited States. Finally, credibilityrequires that the appropriatespecies EcologicalMonograph 14: 223-254. CALLAHAM,R. Z., AND A. R. LIDDICOET. 1961. Altitudinalvariation are ecologicallysuited to the elevations and localities for at 20 years in ponderosa and Jeffreypines. Journalof Forestry 59: which predictionsare made. This means that data bases 814-820. must be firmlycoordinated with the ecological distribu- CLAUSEN,J., AND W. M. HIESEY. 1960. The balance betweencoherence tion of the species; untilphysiographic predictors are re- and variation in evolution. Proceedingsof the National Academy placed by environmentalvariables, models should not be of Sciences, USA 46: 494-506. used forextrapolation. , D. D. KECK, AND W. M. HIESEY. 1940. Experimentalstudies on the nature of species. I. The effectsof varied environmentson Nevertheless,the resultsattest to pronouncedlevels of westernAmerican . Carnegie Instituteof Washington,Pub- geneticvariation among populations of southwestern pon- lication 520. Stanford,CA. derosa pine. It seems reasonable to conclude that much CONKLE,M. T. 1973. Growthdata from29 yearsfrom the California of the variation has been molded by heterogeneousen- elevational studyof ponderosa pine. ForestScience 19: 31-39. vironments. Microevolutionaryprocesses undoubtedly , AND W. B. CRITCHFIELD. 1988. Genetic variationand hybrid- have been furtheredby ponderosa pine's discontinuous ization of ponderosa pine. In D. M. Baumgartnerand J. E. Lotan [eds.], Ponderosa pine: the species and its management,27-43. distributionin the Southwest. Such distributionslimit WashingtonState University,Pullman, WA. gene flowand therebypromote selective differentiation. DAUBENMIRE,R., AND J.B. DAUBENMIRE.1968. Forest vegetationof easternWashington and northernIdaho. WashingtonAgricultural Biosystematic implications- Conkle and Critchfield ExperimentStation Technical Bulletin 60. WashingtonState Uni- (1988) separate the southwesternrace of P. p. var. sco- versity,Pullman, WA. pulorumfrom the Rocky Mountain race in southernUtah DIETRICHSON,J. 1964. The selection problem and growthrhythm. Silvae Genetica13: 178-184. and southwesternColorado. Yet the resultsof this study, DRAPER, N. R., AND H. SMITH. 1981. Applied regressionanalysis. like those fromthe Colorado Plateau (Rehfeldt, 1990), Wiley, New York, NY. demonstratecontinuous genetic variation along geograph- HAMRICK, J. L., H. M. BLANTON, AND K. J. HAMRICK. 1989. Genetic ic and elevationalgradients. To be sure,populations from structureof geographicallymarginal populations of ponderosa pine. southernArizona and differtremendously AmericanJournal of Botany 76: 1559-1568. fromthose in Utah and Colorado. While thesedifferences HANOVER, J. R. 1963. Geographicvariation in ponderosa pine leader growth.Forest Science 9: 86-95. may be sufficientto justifythe racial classifications,the HoFF, R. J. 1988a. Resistance of ponderosa pine to the goutypitch transitionbetween races is unquestionablybroad. midge (Cecodpmyiapiniinopsis). Research Paper INT-387. U.S. The resultsalso supportPeloquin's (1971) contention Departmentof Agriculture, Forest Service, Intermountain Research that introgressionamong species of Ponderosae is com- Station. Ogden, UT. mon in southeasternArizona. While Peloquin reached 1988b. Susceptibilityof ponderosa pine to the needle cast his conclusions fromphenotypic observations in natural fungusLophodermium baculiferum. Research Paper INT-386. U.S. Departmentof Agriculture, Forest Service, Intermountain Research populations,the present results suggest introgression from Station. Ogden, UT. the performanceof progeniesfrom Barfoot. One might, 1990. Susceptibilityof ponderosa pine to westerngall rust moreover,interpret several of the patternsof geographic withinthe middle Columbia River System. Research Paper INT- variation in ponderosa pine according to introgressive 416. U.S. Departmentof Agriculture, Forest Service, Intermountain hybridization.As illustratedby theduration of shoot elon- Research Station. Ogden, UT. gation in the greenhouse(GHDR) and winterinjuries at 1991. Resistance to westerngall rustin artificiallyinoculated ponderosa pine. CanadianJournal of Forest Research 21: 1316- Window Rock (WRWI) in Fig. 4, the slope of geographic 1320. cline for nine variables became' the steepest in south- JOHNSON, R. A., AND D. W. WICHERN. 1982. Applied multivariate easternArizona. Of thesenine, the patterns for three vari- statistics.Prentice-Hall, Englewood Cliffs,NJ. March 1993] REHFELDT-GENETIC VARIATION IN PONDEROSAE 343

LINHART,Y. B. 1989. Ecological andevolutionarystudies ofponderosa . 1986a. Adaptive variation in Pinus ponderosa from Inter- pine in the Rocky Mountains. In D. M. Baumgartnerand J. E. mountainregions. I. Snake and Salmon River basins. ForestScience Lotan [eds.], Ponderosa pine: the species and its management,77- 32: 79-92. 89. WashingtonState University,Pullman, WA. . 1986b. Adaptive variation in Pinus ponderosa from Inter- ,J.B. MITON, K. B. STURGEON,ANDM. L. DAvis. 1981. Genetic mountain regions. II. Middle Columbia River System. Research variation in space and time in a population of ponderosa pine. Paper INT-373. U.S. Department of Agriculture,Forest Service, Heredity46: 407-426. IntermountainResearch Station. Ogden, UT. LITTLE, E. J. 1971. Atlas of United States trees,vol. 1, Conifersand 1988. Ecological geneticsof Pinus contortafrom the Rocky importanthardwoods. Miscellaneous Publication,U.S. Department Mountains (USA): a synthesis.Silvae Genetica 37: 131-135. of AgricultureNo. 1146. Washington,DC. * 1989. Ecological adaptationsin Douglas-fir(Pseudotsuga men- ,IAND W. B. CRITCHFIELD. 1969. Subdivisionsofthe genus Pinus ziesiivar. glauca): a synthesis.Forest Ecology and Management (pines). Miscellaneous Publication,U.S. Departmentof Agriculture, 28: 203-215. Forest Service, No. 1144. Washington,DC. 1990. Geneticdifferentiation among populations ofPinuspon- MADSEN, J. L., AND G. M. BLAKE. 1977. Ecological geneticsof pon- derosa from the upper Colorado River Basin. Botanical Gazette derosa pine in the northernRocky Mountains. SilvaeGenetica 26: 151: 125-137. 1-8. 1991. Models of geneticvariation for Pinus ponderosa in the MAYR, E. 1970. Populations, species and evolution. Harvard Uni- Inland Northwest(U.S.A.). Canadian Journal of Forest Research versityPress, Cambridge,MA. 21: 1491-1500. MITTON,J. B., Y. B. LINHART,J. L. HAMRICK,AND J. BECKMAN.1977. * 1992. Early selection in Pinus ponderosa: compromises be- Observationson geneticstructure and matingsystems in ponderosa tweengrowth potential and growthrhythm in developingbreeding pine ofthe Colorado FrontRange. TheoreticalandApplied Genetics strategies.Forest Science 38: 661-677. 51: 5-13. , R. J. HoFF, AND R. J. STEINHOFF. 1984. Geographic patterns , K. B. STURGEON, AND M. L. DAVIS. 1980. Genetic differen- ofgenetic variation in Pinus monticola.Botanical Gazette 145: 229- tiationin ponderosa pine along a steep elevational transect.Silvae 239. Genetica29: 100-103. , AND W. R. WYKoFF. 1981. Periodicityof shoot elongation NAMKOONG,G., AND M. T. CONKLE. 1976. Trend analysis in genetic among populations of Pinus contortafrom the northernRocky controlof heightgrowth in ponderosa pine. ForestScience 22: 2- Mountains. Annals of Botany 48: 371-377. 12. SACHER, J. A. 1954. Structureand seasonal activityof the shoot apices PELOQUIN,R. L. 1971. Variation and hybridizationpatterns in Pinus of Pinus lambertianaand Pinusponderosa. American Journal of ponderosaand Pinusengelmannii. Ph.D. dissertation,University Botany42: 82-91. of California.Santa Barbara, CA. SAS INSTITUTE. 1985. SAS user's guide: statistics,version 5, ed., vol. 1984. The identificationof three-specieshybrids in the pon- 2. SAS Institute,Inc., Cary, NC. derosa pine complex. SouthwesternNaturalist 29: 115-122. SHEPPARD, W. D., ANDS. E. McELDERRY. 1986. Ten-yearresults of a PERRY, J. P., JR. 1991. The pines of Central America and Mexico. ponderosa pine progenytest in the Black Hills. WesternJournal of Timber Press, Portland,OR. AppliedForestry 1: 79-83. READ, R. A. 1980. Genetic variationin seedlingprogeny of ponderosa SQUILLACE,A. E., ANDR. R. SILEN. 1962. Racial variationin ponderosa pine provenances.Forest Science Monograph 23: 1-59. pine.Forest Science Monograph 2: 1-26. 1983. Ten-yearperformance of ponderosa pine provenances STEEL,R. G. D., AND J. H. TORRIE. 1960. Principlesand procedures in the Great Plains of North America. Research Paper RM-250. of statistics.McGraw-Hill, New York, NY. U.S. Departmentof Agriculture,Forest Service, Rocky Mountain STEINHOFF,R. J.,D. G. JOYCE,AND L. FINs. 1983. Isozyme variation Forest and Range ExperimentStation. Fort Collins, CO. inPinusmonticola. Canadian Journal ofForestResearch 113: 1122- , AND J. A. SPRACKLING. 1981. Hail damage variationby seed 1131. source in a ponderosa pine plantation. Research Note RM-4 10. VAN HAVERBEKE, D. F. 1986. Genetic variation in ponderosa pine: a U.S. Departmentof Agriculture,Forest Service, Rocky Mountain 15-yeartest of provenances in the Great Plains. Research Paper Forest and Range ExperimentStation. Fort Collins, CO. RM-265. U.S. Departmentof Agriculture, Forest Service. Fort Col- REHFELDT, G. E. 1980. Genetic gains fromtree improvement of pon- lins, CO. derosa pine in southernIdaho. Research Paper INT-263. U.S. De- WELLS, 0. 0. 1964. Geographic variation in ponderosa pine. I. The partmentof Agriculture,Forest Service, IntermountainForest and ecotypesand theirdistribution. Silvae Genetica 13: 89-103. Range ExperimentStation. Ogden, UT. 1982. Differentiationof Larix occidentalispopulations from the northernRocky Mountains. SilvaeGenetica 31: 13-19.