Skógræktarritið - 15.05.2001, Blaðsíða 179

Table 1. Mean concentrations (mg/g) of total phenolics, gallotannins, carbohy-
drates and proteins ofthe birch trees included within each class, numberof trees
in brackets. Values correspond to leaves collected in late June 1997. Different
letters indicated significant differences between classes at P < 0.05 (Student-
Newman-Keuls analyses).
Class Total phenolics Total gallotannins Total carbohydrates Total proteins
Gallotannins
High(l) 142.17 a 27.61 a 85.57 a 178.29 a
Medium (3) 117.32 a 11.79 b 113.64 a 135.75 a
Low (3) 93.78 a 2.24 c 114.09 a 128.62 a
Proteins
Hifih (3) 120.96 a 15.04 a 103.17 a 169.43 a
Low (41 103. L5 a ÖJli! 1 14.82 a H5.78h
All pots were placed in 4 rows (=
blocks) in the same greenhouse
bench with a randomized block
design and saplings from each
mother-tree were randomly
located within rows, separated by
10 cm, with doubie distance
between rows.
Overwintering E. autumnata eggs
were taken from seven different
broods and placed at room tem-
perature until they hatched. One
larva per family was placed on
each plant up to a total of seven
larvae per sapling. Larvae were
allowed to freely feed within the
assigned plant. After molting to
the second instar every larva was
individually weighed, marked with
fast drying paint, and reweighed
after marking. Weight at the end
of the instar was also recorded,
and growth during the instar was
thus estimated as the difference
between final weight and weight
after marking.
Data analyses
All statistical analyses were con-
ducted with SAS-package (SAS
Institute 1996).
Differences between mother
trees (trees hereafter) in the pro-
portion of larvae that ballooned or
survived to the end of second
instar was analyzed by fitting a
Generalized Linear Model (GEN-
MOD Procedure, distribution =
binomial, link function = probit;
SAS Institute 1996). Overdisper-
sion problems associated with the
binomial distribution models
were controlled by estimating the
dispersion parameter as Pearson's
chi-square (SAS Institute 1996).
In the case of survival in the
greenhouse experiment pre-
planned contrasts between trees
differing in either concentration
of gallotannins, concentration of
proteins, or both were done to
test whether these factors were
affecting affected larval survivor-
ship. Variation among trees on
growth during second instar was
also studied. Since growth of lar-
vae was normally distributed, dif-
ferences between trees were ana-
lyzed by fitting a General Linear
Model (GLM Procedure). Again
pre-planned contrasts between
trees differing in either concen-
tration of gallotannins, concen-
tration of proteins, or both were
conducted to test for the effects
of these factors on larval growth.
Power of the design to detect dif-
ferences between trees was cal-
culated with GPOWER (Buchner
et al. 1997), note that the power
of pre-planned contrasts is al-
ways higher (SAS Institute 1996).
Larvae that lose their mark or
were weighed after molting to
3rd third instar were excluded
from growth analyses.
70
NUMBER OF BALLOONS (out of 15)
Fig 1. Percentage of birch trees from
which we observed different numbers
of neonate Epirrita autumnata larvae
ballooning.
Results
Ballooning experiment
The experimental procedure was
found to be effective since most
of larvae placed on both the
glass bar and the pine branch
moved away in less than one
hour. However, only 14 out of
459 neonate larvae did balloon
from the stems of mountain
birch branches of the 30 experi-
mental birch trees. Although
these balloons were not uniform-
ly distributed among trees
(Wald’s x2 = 131.6, df= 28, P«
0.0001) in no case we observed
more than 3 larvae ballooning
(Fig. 1) suggesting that variation
is very low and presumably bio-
logically irrelevant.
G reenhouse experimenl
Only two larvae out of the 196
initially placed on plants did not
survive to the second instar.
Mean (± SD) larvae body mass at
the beginning of the second
instar was on average (± SD) 0.73
±0.12 mg, and differences in the
mean body mass of larvae feed-
ing on different trees were not
statistically significant (P > 0.5).
Power of the design at P = 0.05
was higher than 0.90 from an
effect size of 0.31, that represent-
ed a difference of 0.037 mg in
larval weight.
Eighteen percent of the larvae
SKÓGRÆKTARRITIÐ 2001 l.tbl.
177