Appendix A. Additional details on data analysis

*Analysis
of canopy stature variables*

Canopy dimensions were reduced to three variables to minimize autocorrelation. For overall stature, the first axis of a PCA for height, crown area, crown volume, maximum stem diameter, mean stem diameter and basal area (all log transformed) explained 94.1% of all variance; the species scores on the first eigenvector were used as the variable STATURE (height is reported in Table 1 for comparative purposes). For canopy spread, analysis of covariance (log(crown diameter) with species as a discrete factor and log(height) as a covariate) indicated no differences in the slopes of crown diameter vs. height among species. Therefore, a standardized measure of crown spread was calculated as the expected value for each species in this ancova, with no interaction term (i.e., assuming constant slopes across species). This was positively correlated with a direct calculation of the crown diameter/height ratio, but corrected for allometric shifts associated with absolute size (a slight trend towards higher diameter/height ratios in smaller species). This standardized measure of crown diameter was included as the variable SPREAD. Finally, log(no. stems) was included as a measure of shrub vs. tree growth form.

*Complete methods for phenology data collection*

In December 1998,
two shoots per plant on the ten primary plants of each species were marked to
examine the phenology of vegetative growth and reproduction. Shoots selected
for study were in relatively exposed positions in the crown. Shoots were censused
once per month (at approximately mid-month) until December 1999. At the first
census, the base of the shoot was permanently marked as a reference point for
length measurement. For deciduous species the following traits were measured
at each census: shoot length (from base to apex); number of leaves produced
in the current month; and number of leaves surviving from each previous month's
cohort. For evergreen species, the transition points between cohorts of the
previous year(s) were marked based on morphological characteristics (e.g., bud
scale scars) and preliminary measurements made during the previous growing season.
At each census during 1999, shoot length, the total number of leaves produced
in 1999, and the number of leaves surviving from each previous years' cohort
were recorded. The growth form of *Adenostoma* made our phenological methods
prohibitively time consuming so no data are included in this analysis (leaf
life span was obtained from Jow et al. 1980).

Vegetative phenology data were analyzed to obtain a maximum of 18 summary variables for each shoot, followed by calculation of mean and standard deviation for each species. Six variables each were calculated for leaf production, leaf abscission and twig extension: month of first activity (e.g., month of first new leaf production), month of last activity, month of peak activity, duration of activity (last first month), peak level of activity (e.g., greatest number of leaves gained or lost, or maximum twig extension, in one month), and total activity for the year. Total leaf production and twig extension were included in the leaf/twig functional morphology data set (see above).

Preliminary analysis indicated that many of the variables above were tightly correlated, and the following three parameterss were selected to reflect different aspects of vegetative phenology: (1) Vegetative onset (Onset) was determined as the PCA first axis score (explaining 90.5% of the variation) based on species means for first leaf production, peak leaf production, and first twig extension. (2) Vegetative duration (Dur) was based on the PCA first axis score (explaining 93.3% of the variation) for the duration of leaf production and the duration of twig extension. (3) First leaf drop was included as a measure of the onset of abscission, of particular importance in relation to summer drought.

Leaf life span: For deciduous species, leaf life span was calculated as the average for all leaves produced in 1999, based on the repeated censuses of the number of leaves remaining in each month's cohort. For evergreen species, leaf life span was calculated as the average for all leaves which were lost during 1999, based on the month in which they were dropped and the annual cohort to which they belonged, and using the month of peak leaf production as the average month of birth for all leaves.

There were fairly
high levels of shoot loss due to deer browsing, physical damage, or other sources
of mortality, and levels of reproduction were low for some species in the year
of study. Data for a species were dropped from analysis of variance if *N*
< 3, resulting in elimination of the following variables: *Arbutus*:
fruiting; *Arctostaphylos*: leaf production; *Baccharis*: all vegetative
traits (leaf life span was obtained from K.A. Preston, *unpublished data*);
*Dirca*: fruiting; *Holodiscus*: flowering and fruiting. Data obtained
for *N* = 1 or 2 were included in the correlation and multivariate analyses,
despite their high uncertainty, to maintain data matrix completeness (especially
important for Principal Components Analyses, in which missing data are otherwise
replaced by the grand mean for all species).

Literature cited

Jow, W. M., S. H.
Bullock, and J. Kummerow. 1980. Leaf turnover rates of *Adenostoma fasciculatum*
(Rosaceae). American Journal of Botany **67**:256–261.