Regression models described population differentiation 

 according to the elevation and geographic location of the 

 seed source. As elevation increases, the length of the 

 frost-free period rapidly decreases. Baker (1944) has 

 shown that the frost-free period in the Inland Northwest 

 varies by about 80 days across an elevational interval of 

 1,000 m. Consequently, populations from low elevation 

 generally exhibit a high growth potential associated with a 

 long duration and late cessation of elongation. Populations 

 from high elevation express a low growth potential largely 

 because developmental events must be completed within a 

 short frost-free period. These general trends are remark- 

 ably similar to those described by the height of 2-year-old 

 trees grovdng in nursery beds (Madsen and Blake 1977) 

 and also are similar to the patterns exhibited by popula- 

 tions from central Idaho (Rehfeldt 1986). For both central 

 Idaho and the Inland Northwest, an elevational interval of 

 about 400 m is sufficient for detecting (80 percent level of 

 probability) population differentiation. 



Geographic patterns of genetic variation at either a 

 constant or the base of elevation (figs. 3 and 4) were pro- 

 nounced, consistent, and uniform. Moreover, these pat- 

 terns roughly correspond to geographic patterns in 

 precipitation and in the frost-free period (fig. 5). Popula- 

 tions of highest growth potential tend to occur in the 

 west-central region where frost-free periods are long and 

 precipitation is relatively high. From this area, growth 

 potential of populations decreases in all directions in a pat- 

 tern that is similar to that of the frost-free period. Toward 

 the northwest, however, the rapid decline in growth poten- 

 tial is associated with only a small decline in the frost-free 

 period but with a huge reduction in precipitation. Thus, 

 the northwestern populations, which may originate from 

 the lowest elevations (400 m), display a low growth poten- 

 tial. In addition, the eastern populations that exhibited the 

 lowest growth potentials come from areas characterized by 

 both low precipitation and a short frost-free period. 



The elevational cline is much steeper than the geogra- 

 phic. In figure 3, the distance between any two isopleths 

 represents half the distance required to detect (80 percent 

 level of probability) genetic differentiation at a constant 

 elevation. Because differentiation can be detected between 

 populations separated by about 400 m in elevation, the 

 distance between geographic isopleths represents the 

 amount of genetic variation that occurs across 200 m of 

 elevation at a single locality. Thus, the total amount of 

 geographic variation that occurs within the region at a 

 constant elevation can also occur across 1,000 m at a 

 single locality. 



Regardless, adaptive differentiation occurs along two 

 interdependent clines. This means that populations of 

 similar adaptive norms recur at different elevations across 

 the landscape. As shown in figure 2, populations with a 

 relatively long duration of shoot elongation, 38 days, occur 

 at about 700 m at locality A (fig. 1), 750 m at E, 1,000 m 

 at D, and 1,300 m at both A and B. Thus, similar geno- 

 types are distributed across the landscape in a pattern 

 that is oblique to elevation. 



Because patterns of genetic variation are systematically 

 related to environmental variation, one logically accepts 

 the patterns as reflecting adaptive differentiation. As 

 such, the patterns have direct application to artificial re- 

 forestation. To maximize productivity, planted trees must 

 be adapted to the planting site. Adaptation is secured by 

 limiting the distance that seeds are transferred from their 

 origin. Consequently, limits to seed transfer must reflect 

 geographic and elevational patterns of variation. 



One estimate of an appropriate limit to seed transfer is 

 the smallest geographic or elevational interval across 

 which differentiation is detected (Rehfeldt 1979a). Dif- 

 ferentiation along the steepest elevational clines (fig. 2) 

 suggests that a seed zone for ponderosa pine should en- 

 compass not more than 400 m of elevation. This means 

 that seed from a single source should not be transferred 

 more than +200 m. In addition, the geographic clines of 

 figure 3 describe seed zones that should not encompass 

 more than two geographic bands between isopleths. 

 Approximately three geographic zones would be suitable 

 for the region, and seeds from a single source should not 

 be transferred a distance equivalent to more than + 1 

 band. 



Although seed zoning is an administratively simple pro- 

 cedure for limiting seed transfer, the procedure tends to 

 be inflexible, inefficient, and uneconomical. This is because 

 seed zones compartmentalize continuous genetic variation 

 and fail to take advantage of the recurrence of similar 

 genotypes at different elevations in geographically 

 separated localities. 



By contrast, floating transfer guidelines are based on 

 the recurrence of similar environments in a pattern that is 

 oblique to elevation. According to floating guidelines, seed 

 can be transferred across isopleths (fig. 3), but each time 

 seed is transferred across a geographic interval equaling 

 the interval between isopleths, the elevations at which the 

 seed is to be used should be adjusted. When transferring 

 across isopleths of high to low value, the interval should 

 be adjusted downward by 200 m. When transferring from 

 lower, to higher value, the interval should be adjusted up- 

 ward by 200 m. 



For example, assume that seed originates from 1,000 m 

 on the isopleth representing the mean of all populations 

 (fig. 3). This seed should be used between 800 and 1,200 m 

 in lands adjacent to the isopleth. In transferring the seed 

 across one isopleth of larger value, the seed should be 

 used between 1,000 and 1,400 m. Transfers of two con- 

 tours should be used between 1,200 and 1,600 m. And in 

 transferring the seed across an isopleth of lesser value, 

 the seed should be used between 600 and 1,000 m. In this 

 way, seed from a single source or seed orchard can serve 

 a much broader geographic area than under the concept of 

 discrete seed zones. 



These recommendations for limiting seed transfer 

 evolved from statistical models based on the performance 

 of young trees under controlled conditions. The models 

 need practical verification. On the one hand, the environ- 

 mental events responsible for the systematic patterns may 



r 



8 



