which ancestral populations grew, and therefore coadap- 

 tive complexes of traits can readily be perpetuated once 

 such complexes arise. 



Thus, a broad ecological distribution, a unique cycle of 

 population establishment, and occupancy of an extremely 

 heterogeneous environment make P. contorta an ideal 

 species for studying adaptedness, the degree by which in- 

 dividuals are physiologically attuned to their environment. 



MATERIALS AND METHODS 



Genetic variation was studied in seedlings from 83 pop- 

 ulations (fig. 1) that represented the geographic and 

 ecological distribution of P. contorta in the Northern 

 Rocky Mountains. Populations ranged in elevation from 

 640 to 2,700 m and represented habitat types as diverse 

 as the Tsuga heterophylla/Pachistima myrsinites on moist 

 sites and the Abies lasiocarpa/Vaccinium scoparium in 

 subalpine communities. 



Because of pronounced genetic heterogeneity within 

 populations (Rehfeldt 1985b; Ying and others 1985), cone 

 collections were conducted in a manner to assure that 

 subsequent seedling populations would represent a large 

 number of parental trees. Thus, about 250 wind-pollinated 

 cones, each containing about 25 viable seeds, were 

 selected from several squirrel caches in each population. 

 Genetic diversity was further assured by selecting a varie- 

 ty of cone morphologies, sizes, and colors from each cache. 



Seedlings were grown for 6 months in plastic containers 

 (65 cm^) in a shadehouse at Moscow, ID (lat. 48.5° N., 

 long. 116.7° W.). In the fall, seedlings were planted at 

 three locations (fig. 1): at 640 m and 1,500 m elevation on 

 the Priest River Experimental Forest (PREF), and at 

 1,500 m in Lost Valley. The experimental design consisted 

 of a random allocation of linear seedling plots within rows 

 of a rectangular planting. A plot comprised eight seedlings 

 from each population. Nine plots represented each popula- 

 tion at PREF sites, and six plots represented each popula- 

 tion at Lost Valley. Thus, 72 seedlings represented each 

 population at both PREF sites while 48 represented each 

 population at Lost Valley. The spacing of seedlings within 

 and between rows was: 0.5 m and 1 m, respectively, at 

 PREF 640; 0.5 m and 0.5 m at PREF 1,500; and 1 m and 

 1 m at Lost Valley. 



Intensive culture at PREF included control of competing 

 vegetation, gophers, and white grubs. The site at 640 m 

 was irrigated once during years 2 and 3. No control of ex- 

 traneous environmental effects was provided at Lost 

 Valley. Variable cultural regimes and spacing of trees in 

 different physical environments sample a range of condi- 

 tions under which populations exist naturally. Survival was 

 76 percent at PREF 640, 83 percent at PREF 1,500, and 

 65 percent at Lost Valley. 



Periodic measurements or scores provided the following 

 variables: 



1. Mean height for each plot after 3 years. 



2. Height of individual trees after 7 years. 



3. Adjusted height: the 7-year height of individual trees 

 adjusted by regression on 5-year height. 



4. Late growth of trees at PREF: the amount of the 

 7-year, predetermined shoot that elongated after the date 



when the uppermost leaves of the shortest trees were ap- 

 proximately 2 cm long (June 4 at 640 m; July 3 at 

 1,500 m). 



5. Needle cast: the proportion of 6-year leaves of in- 

 dividual trees at PREF 640 that were infected with 

 Lophodermella concolor (Dearn.) Darker, scored in July of 

 year 7 with values of 1 to 5, which coded <5 percent, 25 

 percent, 50 percent, 75 percent, or >95 percent infected 

 leaves. 



6. Mites: the presence or absence of Trisetacus camp- 

 nodus Keifer (Hunt 1981) on 7-year shoots of individual 

 trees at PREF 640. 



7. Shoot borers: the presence or absence of Eucosma 

 sonomana Kearfott on the 7-year shoots of individual trees 

 at PREF 640. 



8. Frost injury: presence or absence of spring frost in- 

 jury to the developing 7-year shoot at PREF 640. 



9. Snow damage: scored in the spring of year 7 as the 

 presence or absence of basal injuries of individual trees at 

 PREF 1,500 sufficient to expose the inner bark. 



Population differentiation was assessed with data from 

 individual trees for all variables except snow damage, 

 frost injury, mites, and shoot borers, for which the propor- 

 tion of injured trees in each plot was analyzed. Allometric 

 traits were transformed to logarithms because variances 

 were proportional to the square of mean values. Scores of 

 needle cast were transformed to \fX to normalize distribu- 

 tions. Three-year height was not subjected to rigorous 

 analysis but was used only for correlation. 



Statistical analyses followed a general model of 

 unweighted means (Steel and Torrie 1960): 



Yijki = M + s, Pj + sp,j + dk^,j) + e 



where Y,ji;i = the performance of the Ith tree of the A;th 

 plot of the jth population at the ■ith planting site, ^ = the 

 overall mean, s = the effect of the planting site, p = the 

 effect of the population, sp = the interaction of sites and 

 populations, d = the effect of plots within populations and 

 sites, and e = the residual. 



Multiple regression models were used to relate genetic 

 variation to the elevation and geographic location of the 

 seed source. Independent variables included elevation, 

 latitude, longitude, northwest departure, southwest depar- 

 ture, and their squares. Northwest and southwest depar- 

 tures were derived by rotating the grid of latitude and 

 longitude by 45°. Geographic variables were nested within 

 four geographic regions that had proven useful in previous 

 analyses (Rehfeldt 1980; Rehfeldt and Wykoff 1981): Idaho 

 north or south of the Salmon River, and Montana east or 

 west of the Continental Divide (fig. 1). However, the geo- 

 graphic regions were not represented by dummy variables 

 because such variables force continuous genetic variation 

 to be described discontinuously. Interactions of elevation 

 and geographic variables were not considered because 

 preliminary analyses indicated that the effects of elevation 

 could be described by similar regression coefficients in all 

 geographic regions. Thus, 34 independent variables were 

 screened by a stepwise regression model for maximizing 

 (SAS 1982) according to the general model: 



j,k J,k 



3 



