Forest Sci., Vol. 27, No. 1, 1981,pp. 49-59
Copyright 1981,by the Society of American Foresters
A Comparisonof Habitat Type and Elevation
]'or Seed-ZoneClassificationof
Douglas-fir in WesternOregon
ROBERT K. CAMPBELL
JERRY F. FRANKLIN
ABSTRACT. Habitat type (identifiedby plant indicators)and elevation were comparedas alternative bases for stratifying forest land into breeding zones or seed zones for reforestation. The
comparisonwas basedon the assumptionthat the geneticvariationin an indigenouspopulation
of Douglas-fir(Pseudotsugamenziesii(Mirb.) Franco) reflectsthe adaptively importantenvironmental variation in the area inhabited by the population. Relative values for habitat type and
elevationwere then estimatedby the amountof geneticvariation explainedby each in classification and regressionmodels. Elevations, habitat types and tree genotypeswere sampledin a
6100-hectarewatershedin the CascadeRange of western Oregon. Parent trees (190) were from
114 locations (sources) in the watershed. Genetic variation was estimated by the performance
(in 15 traits) of 3-year-old seedlingfamilies (190) grown in a common garden. Neither elevation
nor habitat type were completely satisfactoryfor classifyingenvironmentsinto zones, in that
elevation explained only about 56 percent of the source-relatedgenetic variation, habitat type
only about 35 percent. FORESTSCl. 27:49-59.
ADDITIONALKEYWORDS. Pseudotsugamenziesii,adaptation,gertecology,plant indicators.
AN INITIAL PROBLEMin artificial forest regenerationis one of dividing a forest
region into compartmentsof adaptively homogeneousenvironmentsfor use as
seed zones. Developing "breeding zones" for genetic tree improvementis a
closelyrelated task. In each casea classificationbasedon environmentalparameters would be ideal for both uses but is seldom possible. Forest regions, especially mountainousones,are inadequatelysampledby weatherstations;and even
when data are available, the operationalfactors in natural selectionare imperfectly known.
An alternativemethod, pioneeredby Eneroth (1926) and Langlet (1936), is to
describethe source-relatedgeneticvariability in a speciesindigenousto the region
and to use the resultingpattern as an indicatorof environmentalvariability. By
measuringsource-relatedvariability we measurepopulationresponseto natural
selection.By classifyingthis response,we insure that classificationis being applied, indirectly, to the environmentalcomplex that has been active in natural
selection.We further assumethat this complexis the one critical in seedtransfer.
For seed transfer rules, the most useful descriptive model would be the one
which explained most or all of the source-relatedvariation by using easily measuredpredictors.Three generaltypes of modelshave been proposedor used: (1)
the classificationmodel, e.g., by ecotype (Wright and Bull 1963), land productivity classes(Morgenstern1972),elevationalclasses(Langlet 1945),or habitat types
The authors are, respectively, Principal Plant Geneticist and Chief Plant Ecologist, Forestry Sciences Laboratory, Pacific Northwest Forest and Range Experiment Station, USDA Forest Service,
Corvallis, Oregon 97331. Manuscript received 31 January 1980.
VOLUME 27, NUMBER 1, 1981 / 49
TABLE 1. Forest communitiesdescribedby Dyrnessand others(1974) and the
number of trees and locationsfrom which seed was collected within each community.
Number
Loca-
of trees
tions
Psme-Tshe/Coco
Tshe/Cach
4
3
2
2
Tshe/RhmaJGash
21
13
Tshe/RhmaJBene
Tshe/Pomu
69
3
40
2
Tshe/Abam/RhmaJBene
9
6
Tshe-Abam/RhmaJLibo
15
9
Tshe-Abam/Libo
24
13
Abam-Tsme/Xete
5
4
Abam/Vame/Xete
12
6
Abbreviation
Community
Pseudotsuga rnenziesii-Tsuga
heterophylla/Corylus cornuta
Tsuga heterophylla/Castanopsischrysophylla
Tsuga heterophylla/Rhododendron
macrophyllum/Gaultheriashallon
Tsuga heterophylla/Rhododendron
macrophyllum/Berberisnervosa
Tsuga heterophylla/Polystichurnrnuniturn
Tsuga heterophylla-Abies arnabilis/
Rhododendronmacrophyllum/Berberisnervosa
Tsuga heterophylla-Abies amabilis/
Rhododendronmacrophyllum/Linnaea
borealis
Tsuga heterophylla-Abies arnabilis/
Linnaea borealis
Abies arnabilis-Tsuga rnertensiana/
Xerophyllurn tenax
Abies amabilis/Vaccinium
membranaceum/Xerophyllumtenax
Abies amabilis/RhododendronmacrophyllumVaccinium
alaskense/Cornus
canadensis
Abies amabilis/Achlys triphylla
Abies amabilis/Tiarella unifoliata
Abam/Rhma-Vaal/Coca
Abam/Actr
Abam/Tiun
2
2
2
21
2
13
as identifiedby plant indicators(Daubenmire 1976, Rehfeldt 1974);(2) the regression model usingsimpleor complexclines(Schotte 1923, Squillace 1966,Roche
1969, Morgenstern and Roche 1969); or (3) the mixed regression-classification
model by analyzing clines within some subdivisionsof the population, such as
regions (Hattemer and K6nig 1975) or ecotypes(Nienstaedt 1975). Models may
vary in utility dependingon speciesand forest region.
In this paper we compare elevation and habitat type as contrastingbasesfor
classifyingseed zones. Elevation has a long history of such use in the Pacific
Northwest (see Isaac 1949). Vegetation classificationin the PacificNorthwest has
only recently progressed(Dyrness and others 1974)to a stagewhere habitat type
is a feasible alternative.
We also examine the combination
of elevation and hab-
itat, since zones based on both predictors might be better than zones based on
either alone. We report on several model types: (1) classificationmodels for
elevation and habitat type, analyzed separately, (2) an elevational clinal model,
(3) an elevational clinal model including an index for habitat type, and (4) a
classificationmodel including both elevation and habitat type.
Comparisonswere restricted to the coastal Douglas-fir [Pseudotsugamenziesii
vat. menziesii (Mirb.) Franco] populationwithin a 6100-hectarewatershedin the
central Oregon CascadeRange where vegetation of the watershed had been intensively studiedand the populationhad been sampledfor an earlier experiment.
By samplingonly one watershed the comparisonof elevation and habitat was
unencumberedby extraneousgenetic effects associatedwith suchfactors as latitude or distance from the ocean.
50 / FOREST SCIENCE
TABLE 2.
Description of traits measuredin nurserybed.
Trait 1. Seed weight
2. Germination
rate
3. Days to 50 percent germination 4. Cotyledon number
Description
Based on 60-seed averaget
Cumulative germination on
probit scale Interpolated from probit graph
Based on 20-seed average•
5. Budset in 1972
First visible terminal-bud
6. Budburst 1973
First green needles from
terminal
7. Budburst 1973 variability Unit
scales
mg/seed
probits/day days cotyledon/seedling
weeks after 11 Aug
half-weeks
after 24 March bud Variability in (6) among
1ogn•(variance in trait 6) 16-19 seedlings• 8. Budset 1973
First visible terminal-bud
9. Budburst
First green needles from
1974
scales
weeks after 6 July
half-weeks
after 31 March
terminal bud
10. Budburst 1974variability 11. Second flush
12. Height Variability in (9) among
16-19 seedlings• Proportion of 3-year seedlings
with lammas growth
3-year total height
logt0(variance in trait 9) arcsin (percent)
cm
13. Height variability Variability in (12) among
16-19 seedlings
1og•0(variance in trait 12)
14. Diameter
3-year stem diameters,
cotyledon height
mm
15. Dry weight
3-year top-dry-weight
Proportion of 3-year seedlings
survivingdroughttreatment
gm x 10
16. Survival
arcsin (percent)
Base number representsthe number of seed (seedlings)per plot in each of two replications.
PROCEDURES
Data for the comparisonscamefrom a studyreportedpreviously(Campbell1979),
which providedgenotypicvalues(as estimatedby family means)for a sampleof
190 parent-treesin 114 locationsdisperseduniformly (i.e., roughlyproportional
to habitat types and elevations)in the H. J. Andrews Experimental Forest on the
west slope of Oregon's CascadeRange. For each location, elevation above sea
level and habitat type (usingthe classificationof Dyrness and others 1974)were
recorded (Table 1).
The Andrews Forest is an intensive study site for the U.S. International Biological Program's ConiferousForest Biome Project and has been thoroughlydescribed elsewhere (Zobel and others 1976). It includes the watershed of Lookout
Creek in the west central Oregon Cascade Range and is approximately the shape
of a right trianglewith maximumnorth-southand east-westdimensionsof 12 and
18 km, respectively. Elevations within the watershedrange from 500 to 1,600 m.
Genotypic values of sample trees in the watershed's primarily 450-year-old
Douglas-firstandwere estimatedby growing seedlingsfrom open-pollinatedseed
collectionsin a nursery bed in Corvallis. The family resultingfrom each collection
was randomly assignedto rows within two replications. Twelve traits of the
internal 16 to 20 seedlingsin each plot were measuredthrough the third growing
season (Table 2).
Trait meansper plot for each trait were analyzedby analysisof variancein a
VOLUME 27, NUMBER 1, 1981 / 51
TABLE 3. Model of analyses showing expected mean squaresfor genetic differences among zones (habitat or elevation) and within zones.
Source variation
Zones
d.f. •
12
(5)
Locations within zones
101
(108)
Parameters estimated by mean squares
tr2 + 2trr• + 3.42•Lz + 25.76•z •
(2)
(4.07)
(60.17)
tr2 + 2trr• + 3.32trL2
(2)
(3.29)
Families in locations
76
(76)
•
Error
190
tr 2
+ 2•rr•
(2)
where: trz• = variance of zone effects
tr•2 = variance of effects due to locations within zones
trr a = variance of effects due to trees within locations
variance of plot effects (replicationswithin families in locationsin zones)
• Degreesof freedom and coefficients(in parentheses)are for zonesbasedon six elevationalclasses
of 175-mwidth; others are for habitat types (Table 1).
classification model for effects due to habitat (or elevational zone), locations
within habitat (or within elevational zone), and trees within locations. The hierarchies in elevation and habitat models were identical, but degreesof freedom in
analyseswere different (Table 3) becauseparent-tree locationswere classifiedin
fewer elevational zones than there were habitat types. Elevational bandsnarrower
than about
175 m have not been used for seed zones in the Pacific
Northwest
and, consequently,were not used for our analysis.
Componentsof variance for each effect for each trait were then estimatedby
equating mean squaresto expectations(Table 3). The sum of the variances attributable to habitat (6-z")and to location within habitat (6-L2) representsthe total
source-related genetic variability (i.e., 6-z" + &L"). The remaining variance includes the average variation among means of families from a location (6-/') and
variation among replicationsplus interaction of families with replications(6-").
The ratio of habitat variance, to source-related variance (6-z/(d-z
2 + 6-Z')) is an
estimateof the proportionof source-relatedvariationdue to habitat(Kempthorne
1957, p. 243). It is therefore a measureof the populationalcomponentof genetic
variance associatedwith differences in habitat type.
By dividing the Experimental Forest into elevationalbandsof different widths,
the same hierarchical analysis (and component analysis) was made for each of
three different elevational classifications: six zones of 175-m width, three zones
of 350-m width, and two zones of 525-m width. For a classification model in-
cluding both elevation and habitat, the analysisusing six elevational zones was
partitionedfurther into habitat types within elevationalzones(habitatsin zones,
19 d.f.; locations in habitats, 89 d.f.).
Finally, for each trait, elevationalclineswere describedby fitting family means
to parent-treeelevation by multiple regression.Predictingvariableswere selected
from the following preliminary model:
Y
+
+ 1,x1 a + 11,m,a
where
Y = mean family responseas determinedin the nursery, X• = parent-tree elevation (m), fi• = coefficientsestimatedfrom the data. 52 / FOREST SCIENCE
From this model, a stepwiseprocedure(Draper and Smith 1966, p. 171)
selectedan equationin which all includedpredictingvariables(X0 contributed
significantly(P < 0.05) to reducingsumsof squaresin the responsevariable
(D.
For comparingthe simplehabitat and clinal modelswith one incorporating
their combinedeffects, an identical stepwiseprocedurewas used to select an
equationfor describing
the combinedeffects.The expandedpreliminarymodel
was
=/3o
+
where
Y and X• are as above,
X., = X-axis coordinate of the vegetative ordination of Dyrness and
others (1974),
/3, = coefficientsestimatedfrom the data.
For each plant community (Table 1), a value for X,, (the xeric index) was
obtainedby averagingx-coordinatevalues obtainedby Dyrness and others(1974)
from two-dimensionalgradient analysesof 300 reference plots within or near the
H. J. Andrews ExperimentalForest. Gradient analysesfor plots from high- and
low-elevation zones involved different sets of vegetative descriptorsand might
not measureidentical moisture gradients. In our analysis, however, we used X,,
values derived from both zones because: 1) Dyrness and others (1974) hypothesized that x-axes measuredin the two zones covered similar moisture gradients,
and 2) x-values from the two gradient analyses have been shown to be strongly
correlated with maximum summer moisture stressin reference plots (r"'s ranged
from 0.79 to 0.95 dependingon year and zone; Zobel and others 1976). We did
not use y-axis gradient values as a variable in the model. The y dimensionhas
been hypothesizedby Dyrness and others (1974) to reflect populationresponse
to temperature;it therefore may measurethe samecomplextemperaturegradient
that elevation is commonly assumedto measure.
After an equation had been selectedfrom above models for each trait, the
equationwas testedfor lack of fit to the data (Draper and Smith 1966, p. 63). The
two or more trees sampledat some locations were repeat observationsof genotypic value at a location and mean squaresfor trees within locations could thus
be used as an estimateof "pure error" for testinglack of fit. Significantlack of
fit indicated that genotypic values of some trees deviated significantlyfrom the
cline as describedby the regressionequation. When this occurred,source-related
variation was not completelyaccountedfor by the equation, usually becausenot
all pertinent descriptivevariables had been included in the model.
RESULTS
Averaged over all traits except seed weight, 6.2 percent of total variation among
plot means was associatedwith habitat type in our sample (Table 4, col. 2(i)).
Seed weight was excludedbecauseit may be unduly influencedby nongenetic
maternal effects. A further 8.7 percent of variation was connectedwith location
of parent trees within habitat type (col. 3(i)). Thus about 14.9 percent of variation
could be attributed to parent-tree location within the watershed. The remainder
representedgenetic variation among trees within a location (23.1 percent, col.
4(i)) and error (62.0 percent, col. 50)).
Thirty-five percentof the geneticvariationthat was associatedwith parent-tree
location was also associatedwith habitat type (Table 4, col. l(i)), when averaged
VOLUME 27, NUMBER 1, 1981 / 53
TABLE 4. Partitioning of variation in plot means by genetic categories when
families are classifiedaccordingto origin: (i) habitat type---13zones, (ii) 175-m
elevational bands•6 zones, (iii) 350-m elevational bands--3 zones.
Zone
related •
variation
among
Location s
Trees
within a
location
within
within
Trait and
sources
Zone •
zone
zones
Error •
(origin class)
(1)
(2)
(3)
(4)
(5)
a. Seed weight
(i)
(ii)
(iii)
b. Germination
4.8
4.4
0.0
14.7
19.1
19.1
76.6**
77.0**
77.0**
3.9
3.9
3.7
0.0
33.1
14.7
0.0
3.9**
1.7'
11.7
7.9
9.9
14.4'
14.6'
14.6'
74.0
73.6
73.8
1.5
0.0
0.0
15.0
16.8'
16.3'
0.1
1.7
1.7
83.0
81.6
82.1
rate
(i)
(ii)
(iii)
c. Germination
19.5
22.4
0.0
date
(i)
(ii)
(iii)
9.1
0.0
0.0
d. Cotyledon number
(i)
(ii)
(iii)
21.9
73.4
44.5
2.5
9.4**
5.7**
8.9
3.4
7.1
44.7**
43.9**
43.9**
43.8
43.3
43.3
100.0
100.0
100.0
14.8'*
19.4'*
23.3**
0.0
0.0
0.0
49.7**
47.9**
45.9**
35.5
32.7
31.3
0.0
100.0
100.0
0.0
3.1'
3.4'*
4.3
0.0
0.0
47.4**
48.0**
47.8**
48.3
48.9
48.8
0.0
0.0
0.0
3.3
3.4
2.9
9.4
9.4
9.4
87.4
87.3
87.7
70.5
98.5
100.0
4.3
6.6*
8.7**
1.8
0.1
0.0
50.7**
50.4**
49.3**
43.2
42.9
42.0
16.5
27.9
30.1
4.0
6.9**
7.7**
20.2**
17.8'
17.9'
19.7'*
19.8'*
19.6'*
56.1
55.5
54.8
61.4
63.4
62.9
19.0'*
19.9'*
21.2**
11.9*
11.8*
12.5'
15.3**
15.1'*
14.6'*
53.8
53.5
51.7
e. Budset in 1972
(i)
(ii)
(iii)
f. Budburst
in 1973
(i)
(ii)
(iii)
g. Budset in 1973
(i)
(ii)
(iii)
0.0
0.0
0.0
h. Budburst in 1974
(i)
(ii)
(iii)
i. Second flush
(i)
(ii)
(iii)
j. Height
(i)
(ii)
(iii)
54 / FOREST SCIENCE
TABLE
4.
Continued.
Zone
Trees
related •
variation
Location s
within a
location
within
within
among
Trait
and
sources
Zone s
zone
(1)
(2)
(3)
49.4
54.4
58.7
8.3**
9.2**
10.8'*
56.8
62.3
66.8
39.6
60.5
47.1
(origin class)
zones
Error •
(4)
(5)
8.5
7.7
7.6
11.1
11.6*
11.4'
72.1
71.5
70.2
13.8'*
15.4'*
18.1'*
10.5'
9.3
9.0
10.3
10.2
9.9
65.4
65.1
63.0
5.7*
8.9**
7.2**
8.7
5.8
8.1
4.3
5.0
5.0
81.4
80.3
80.3
k. Diameter
(i)
(ii)
(iii)
1. Dry weight
(i)
(ii)
(iii)
m. Survival
(i)
(ii)
(iii)
Average•excluding seed weight
(i)
35.4
6.2
8.7
23.1
62.0
(ii)
(iii)
56.1
52.1
8.6
9.0
7.0
7.6
23.1
22.7
61.4
60.8
Estimated as 100 &zV(&z2 + &L2).Symbolsas in Table 3.
Partitioned effects of parent tree locations: Zone effects = 100 &z2/crr
• where drr2= &za + OL• +
+ •. Within-zone effects = 100
Estimated as 100
Estimated as 100
*,** F-ratios significantat P < 0.05 and P < 0.01, respectively.
over 12 traits. This habitat-related variation, as a percentage of source-related
variation, differed greatly among traits, averaging 10 percent for seed traits (lines
b-d, Table 4), 37 percent for developmentalcycle traits (lines e-i), and 56 percent
for growth traits (linesj-l).
When family means were classifiedinto six elevational zones of 175 m, genetic
variation among zones was larger than variation within zones (P < 0.05) for 10
of 13 traits. In similar analysesof habitat type, only 5 of the 13 comparisonswere
significant(Table 4, col. 2). Averaged over all traits except seedweight, variation
among elevational zones accountedfor 56.1 percent of the source-relatedgenetic
variation, as compared to the 35.4 percent accounted for by habitat type (Table
4, col. 1). Thus, classificationof the population by elevation explained more of
the source-related genetic variability than did classificationby habitat type, by
20.7 percentagepoints with 95-percentconfidencelimits of 1.4 to 39.9. Classification by elevation explained more of the total experimental variation, also, by
2.4 percentagepoints (8.6-6.2; Table 4, col. 2) with 95-percentconfidencelimits
1.0 to 3.8. The ordering of families by three elevational classesrather than six
produced approximately the same results (Table 4, col. l(iii) and 2(iii)).
When analyzed by regression, variability among seedlingfamilies was signifi-
VOLUME 27, NUMBE}• 1, 1981 / 55
TABLE 5. Relation of family performance(Y) to parent-treeelevation(X).
Percentage of
sums
squares
explained by-RegresF for
"lack
sion
on ele-
Habitat
regression of fit"
vation
type
F for
Trait
Regressionequation form
Seed weight
Y = bo - b•X + b2X2
- b.•X•
Germinationrate
Y = bo - b•X + b2X•
Mean germination date None significant
Cotyledon number
Y = bo - b•X + b2X2
- baX 3
Budset 1972
Y = b o - b•X 2
Budburst1973
Y = bo + b•X • - b2X3
Budset 1973
None significant
Budburst 1974
Y = bo + b•X
Secondflush
Y = b o - b•X
Height
Y = bo - b•X •
Diameter
Y = bo - b•X
Dry weight
Y = bo - b•X
Survival
Y = bo - b•X
Average•excluding seed weight
d.f.
3;186
2.9*
1.4
4.5
10.6
2;187
-3;186
5.5**
-2.9*
2.4**
-1.8'*
5.5
0
4.4
4.0
5.5
7.3
1;188
3;186
47.8**
2.9
-19.9'*
21.8'*
65.1'*
29.5**
52.0**
25.1'*
1.2
1.6'
-1.4'
2.5**
2.2**
2.4**
2.3**
2.6**
20.3
3.0
0
9.6
10.4
25.7
13.6
21.7
11.8
17.5
2.4
3.0
8.4
8.5
20.2
10.8
15.5
8.5
10.5
9.3
-1;188
1;188
1;188
1;188
1;188
1;188
* Significant at probability P < 0.05. ** Significant at probability P < 0.01. cantly related to elevation for 10 of the 13 traits (Table 5). In general, family
mean for height, diameter, and dry weight decreasedlinearly with elevation of
parent origin. For seed traits and developmental cycle traits, equationsindicated
several types of nonlinear clines with elevation.
Equationsderived by regressionanalysesof the provisionalmodelswhich included the xeric index were identical to those selected from models involving
elevation alone. Thus genetic variability among families was not associatedwith
the part of the environmentalcomplex measuredby the xeric index after correlations of habitat type with elevationwere discounted.Other indexesof habitat
type probably would not have served any better. In an analysisof variance to
check this possibility, elevationalzoneswere partitionedinto habitat types within
zones. In this analysis,habitat type accountedfor source-relatedvariation in only
three traits, seed weight (2.0, 6.7, and 11.9 percent for elevational zone, habitats
in zones, and locations in habitats, respectively), germination date (0, 6.7, and
12.2 percent) and diameter (8.1, 3.4, and 6.2 percent). In all other traits, the
estimated component of variance for habitats within elevational zones was zero.
Therefore, for most traits, the variability among families that could not be accountedfor by elevation also could not be accountedfor by habitat type, whether
analyzedby regressionor analysisof variance.
DISCUSSION
Daubenmire(1976) has proposedthat natural vegetation,in its compositionand
other attributes, integratesall of the environmentalfactors important to plants.
Furthermore, he proposesthat vegetal indicators are superior to human judg56 / FOREST SCIENCE
ments due to the difficulty of quantitatively relating aspectsof climate, topography, and soils. Vegetation or "habitat type" is, therefore, consideredthe best
indicator of environmentalconditions.Consequently,it seemsreasonableto expect that the adaptive genetic variability in Douglas-fir should be more closely
associatedwith habitat type than with any other singledescriptivevariable. Instead, in this study a classificationof tree locations by elevation explained substantially more of the source-relatedvariation than did habitat type. Furthermore,
the elevational regressionmodels, which included a xeric index for habitat based
on community gradient analysis, were no better than those for elevation alone.
Apparently, in our sample and for the traits we measured, the environmental
diversity relevant to adaption in Douglas-firis more closely associatedwith elevation than with habitat type.
There are several possiblereasonswhy habitat types were not more successful
in explaininggeneticvariation in Douglas-fir.First, the field samplingwas concentrated on a few, very widespreadhabitat types (Table 1)•Tshe/Rhma/Bene,
Tshe-Abam/Rhma/Libo, Tshe-Abam/Libo, Tshe/Rhma/Gash, and Tshe-Abam/
Rhma/Bene. These habitat types are indicative of relatively moderate environmental conditionsand cover broad elevational spanswhich are, of course, temperature and moisture gradients. Very few sampleswere obtainedfrom habitat
types indicative of extreme environmentalconditions,such as very dry, wet, or
cold and snowy sites; the hottest and driest forested habitat, Pseudotsugamenziesii/Holodiscusdiscolor (Dyrness and others 1974) was not sampledat all. We
could have sampled specificallyby habitat types or along the major moisture,
temperature, and snowpackgradients so as to include a greater number of extreme sites. This type of sampling,as opposedto our proportional sampleof the
entire landscape, might have resulted in habitat types accountingfor a higher
percentageof the Douglas-firgenetic variability encountered.
Other possible explanationsfor the failure of habitat types have to do with
genetic variability in the indicator species. First, the environmental complex
which influencedadaptationin Douglas-firmay not be identical to the complex
which influenceddistributionsof the indicator species.Second, members of an
indicator speciesat one samplelocation may differ geneticallyfrom membersat
another location. Consequently, the same community in two different areas could
indicate two somewhat different
environments.
Althoughdirect evidenceis lacking, the secondhypothesisseemsparticularly
reasonable.Daubenmire (1976) cautionedthat the same speciesin different geographic areas is usually representedby different ecotypes,which by their nature
have different indicator significance.It is likely that source-relatedgenetic variation in some indicator speciesmay be even larger than in Douglas-fir due to
patchinessin distribution and consequentrestrictionsin gene flow. Thus, population differentiation can occur within much smaller regionsthan is connotedby
Daubenmire'sphrase, "different geographicareas." It is also likely that differentiation of indicator species will be clinal as well as ecotypic--in the H. J.
Andrews, it may have paralleleddifferentiationin Douglas-fir,that is, partially
along elevational gradients. If so, the niche inhabited by a community at its
highestelevation may be quite different environmentallyfrom the niche at lower
elevations. This may be particularly true in the widespread habitats where our
samplingwas concentrated.Elevational rangeswere 396 to 807 m in Tshe/Rhma/
Gash, 457 to 932 m in Tshe/Rham/Bene, and 807 to 1,219 m in Tshe-Abam/Libo;
environmental measurementsclearly indicate major variability within these habitat types over this elevational range (Zobel and others 1976).
It is unlikely that any of the models we tried are completely satisfactoryfor
classifyingenvironmentsinto breeding zones or seed-transferzones within the
watershed. If source-relatedgenetic variation can be equated, at least partially,
VOLUME 27, NUMBER 1, 1981 / 57
with environmentalvariation within the Experimental Forest, much of the environmental variability has remained unexplained. Even in the best case, classification by elevationinto six zones, an estimated44 percent of the source-related
genetic variation was associatedwith differencesamong locations within zones.
These differencescannot be attributed to chancegroupingof geneticallyvariable
trees. Previous work within the watershed indicated that virtually all of the
source-relatedvariation in many traits was patterned in three-dimensionalgradients according to elevation and N-S and E-W coordinatesof parent trees
(Campbell 1979). Such complex patterns could not be explained except as resuitingfrom natural selection.Therefore, the variation amonglocationswithin
zones described herein apparently representsadaptation to environmental diversity not associatedwith elevation. The diversity so measuredis hypothesizedto
be in terms of environmental
factors which have contributed
to natural selection.
Consequently, elevational bands, even as narrow as 175 m, may not be adequate
as seed zones, because seedlingsresulting from seed transfer between some locations within zones may be poorly adapted to the new site.
Classificationof source-relatedvariation probably would not have been improved by using narrower elevational zones. Elevation used as a continuous
variable in regressionaccountedfor only slightly more of the total sumsof squares
than did habitat type in the classificationmodel (Table 5). Also, for 8 of the 10
traits in which an elevational cline was demonstrated,the responsesof trees from
some locations did not fit the general elevational trend (significant lack of fit,
Table 5). These deviations are analogous to variation among locations within
elevational zones and indicate unclassifiedenvironmentalheterogeneity. Classification was not improved by including habitat type in conjunctionwith elevational zone either in the regressionor classificationmodels.
Habitat type may be a more satisfactory tool for classifying environments in
a large region. Judgingfrom data presentedby Rehfeldt (1974), habitat type and
elevation were about equally effective for classifyingsource-relatedgenetic variation in Rocky Mountain Douglas-fir(about 50 percent in either case), in an area
encompassingeastern Washington, northern Idaho, and northwestern Montana.
Habitat types may also be more useful in locales where the landscapeis composed
of a mosaic of several, sharply contrasting environments or habitat types rather
than a few, environmentallymoderatehabitat types. In suchlandscapes,gradients
will probably be abrupt rather than extendedand gradual, and contrastsbetween
adjacent habitat types sharp.
In conclusion,in landscapeswhere broad, environmentallymoderatehabitat
types dominate, better indicesto genetic variability appear necessary.Elevation
was a superior alternative in our study area; but it, too, was inadequate for
indexing Douglas-firvariability.
LITERATURE
CITED
CAMPBELL,R. K. 1979. Genecology of Douglas-fir in a watershed in the Oregon Cascades.
Ecology 60:1036-1050.
DAUBENMIRE,R. 1976. The use of vegetationin assessingthe productivity of forest lands. Bot Rev
42(2): 115-143.
DR•'ER, N. R., and H. SMITH. 1966. Applied regressionanalysis.John Wiley and Sons, New York.
407 p.
DYRNESS,C. T., J. F. FRANKLIN,and W. H. MOlR. 1974. A preliminary classificationof forest
communitiesin the central portion of the western Cascadesin Oregon. U.S. IBP (Int Biol Prog),
ConiferousForest Biome Bull 4, 123p. Univ Wash, Seattle.
ENEROTH,O. 1926. Studier 6vet risken vid anvSndandeav tallfr6 av frSmmandeproveniens.Medd
Statens Skogsf6rs6ksanst23:1-62.
58 / FOREST SCIENCE
HATTEMER,
H. H., and A. KrNIG. 1975. Geographic
variationof earlygrowthandfrostresistance
in Douglas-fir.Silvae Genet 24:97-106.
ISAAC,L. A. 1949. BetterDouglas-fir
forestsfrom betterseed.Univ WashPress,Seattle.64 p.
KEMPTHORNE,
O. 1957. An introduction
to geneticstatistics,
JohnWileyand Sons,New York.
545 p.
LANGLEY,
O. 1936.StudletiSvertallensfysiologiska
variabilitetochdesssamband
med
klimatet.Ett
bidragtill kiinnedomen
omtallensikotyper.MeddelSkogsfiSrsiSksanst
29:219-470.
Stockholm.
LANGLEY,O. 1945. Om miSjligheterna
att skogsodla
medgran-ochtallfri5av ortsframmande
provenlens. Svenska SkogsforsAnst43:68-79.
MORGENSTERN,
E. K. 1972.Progenytestsof blackspruce(Picearnariana(Mill.) B. S. P.) in boreal
environments.Proc Meet Work Party on ProgenyTesting,p 48-55. GeorgiaForestRes Counc,
Macon.
MORGENSTERN,
E. K., and L. ROCHE. 1969. Using conceptsof selectionto delimit seed zones.
SecondWorld Consult on Forest Tree Breeding, FAO-FO-FTB-69-2/16.
NIENSTAEDT,
H. 1975. Adaptivevariationsmanifestations
in tree speciesandusesin forestmanagementandtree improvement.
ProcFifteenthMeet of CanTree ImproyAssoc,p 11-23.
REHFELDT,
G. E. 1974.Geneticvariationof Douglas-fir
in the northernRockyMountains.USDA
Forest Serv Res Note INT-184, 6 p.
ROCHE,
L. 1969.A genecological
studyof thegenusPiceain BritishColumbia.New Phytol68:505554.
SCHOTTE,
G. 1923.Talletsproveniens-Norrlands
viktigasteskogsodlingsfrfiga.
MeddrStatensSkogf'6rsiSksanst 20:305-400.
SQUILLACE,
A. E. 1966. Geographicvariationin slashpine. ForestSci Monogr 10, 56 p.
WRIGHT,J. W., andW. K. BULL. 1963.Geographic
variationin Scotchpine.SilvaeGenet12:1-25.
ZOnEL,D. B., A. McKEE,G. M. HAWK,andC. Y. DYRNESS.
1976.Relationships
of environment
to composition,
structure,anddiversityof forestcommunities
of the centralwesternCascades
of Oregon. Ecol Mongr 46:135-156. Forest Sci., vol. 27, No. 1, 1981, pp. 59•01
Copyright 1981, by the Society of American Foresters
Rapid Determination of Southern Pine Specific
Gravity With a Pilodyn Tester
Fred W. Taylor
ABSTRACT. A Pilodyn Wood Tester was used to estimate the specific gravity of 21-year-old,
plantation-grown,1oblollypine. The Pilodynvalues(depthof penetrationof a spring-powered
pin)
were correlatedwith wood specificgravity. The procedurehas potentialfor rapid determination
of specificgravity of forest trees. FORESTSC!. 27:59-61.
ADDITIONAL
KEY WORDS.
Pinus laeda.
THE PILODYN WOOD TESTER,1 developed in Switzerland to measure soft rot in wooden
poles, has potential for rapidly indicating specificgravity. Operation involves injecting a
tension-springpowered pin into wood and measuringthe depth of penetration. The utility
of the instrument to estimate specific gravity of standing trees was recognized by Hoff-
l Information about the instrument and its operation may be obtained from Mr. Carl Bechgaard,
Kai R. Spangenberg,Hovedgaden 26 Postziro 83669, 2970 Horsholm, Denmark.
The author is Assistant Director, Forest Products Laboratory, Mississippi State University, P.O.
Drawer FP, MississippiState, MS 39762. Manuscript received 18 January 1980.
VOLUME 27, NUMBER 1, 1981 / 59