Transcript Potential impact of climate change on growth and wood quality in

Potential impact of climate change on growth and wood quality in white spruce
Christophe ANDALO1,2, Jean BEAULIEU1 & Jean BOUSQUET2
1
Table 3. Percentage in growth and wood quality deviations from “local” source predicted for 4 different scenarios of
climate change (cf Table 2 for the details of tested scenarios).
Wood quality
For the regression models relating wood quality variables to climatic data, the minimum average temperature was the most important explanatory factor (Fig. 2). Despite a large
proportion of the variation in wood quality explained by this single climate variable (Table 1), the validation procedure did not permit to confirm the predictive power of the models
obtained.
Growth
In estimating regression models, forward selection made it possible to identify maximum temperature as the most explicative climate variable for all the growth traits (Fig. 3). Despite
the validation of their predictive values (p-value < 0.05), the univariate models explained only a very small proportion of the inter-provenance variation in height or diameter (Table
1). Thus, other climate parameters should be taken into account to improve models’ fitting (see below models using canonical scores).
Figure 1
Local adaptation
Regression models did not peak near the origin, which is null climatic distance (Figs. 2 and 3). Consequently, provenances did not appear locally adapted when taking into account
only one climatic variable at a time to describe the local environment. As a general trend, provenance growth seemed to be better at sites cooler than their origin [climate(sourceplanting) > 0].
Figure 2
Figure 3
Height deviation from « local » source (%)
A regional white spruce provenance-progeny test replicated on three sites was analyzed 22
years after planting (Fig. 1). Phenotypic traits studied were growth (height and diameter) and
wood quality (density and proportion of late wood) (see Table 1). All the climatic data in
relation to seed sources and study sites were obtained from simulation models using data
from strategically located meteorological stations as input (Régnière 1996).
The models developed were tested using data collected in a second provenance-progeny test.
To do so, provenances common to both provenance-progeny tests as well as provenances
tested only in the second one were used.
Density deviation from « local » source (%)
Materials
Minimum temperature difference (source - planting)
Provenance-progeny test in Québec
3 sites ( )
2 blocks per site
41 provenances ( )
4 open-pollinated families per provenance
2 trees per family plot
Statistical analyses
Data analysis followed a procedure outlined by Schmidtling (1994). Regression analyses were
used to relate phenotypic observations to climate distance between provenance origin and
planting site. The following steps were carried out:
-Phenotypic data centered on site means were first regressed on climatic distances to
identify the climatic variable explaining the largest proportion of the inter-provenance
variation.
-Local source performances were estimated for each site with a quadratic regression model
using the independent variables identified in the first step (Phenotype = climate + error;
climate = 0 for the local source);
-General regression models were obtained by combining data of the three sites and using as
dependent variable the deviation between the provenance tested and the local source
(phenotype(%) = climate + error);
-Validation was conducted by correlation analysis between observed and predicted
deviations in the second provenance-progeny test.
Table 1. Comparison of different models. Validations were obtained from correlation analysis between deviations from “local” sources predicted by
regression models and observed deviations in the second provenance-progeny test.
Independent variables :
Wood density (juvenile)
0 (transition)
Wood density
0
Prop. of late wood (juvenile)
Prop. of late wood (transition)
Minimum temperature
R2
Validation
0.463 ***
0.148 ns
0.488 ***
0.085 ns
0.361 ***
0.531 ***
Maximum temperature
R2
Validation
Canonical scores
R2
Validation
0.443 ***
0.359 **
0.600 ***
0.309 *
0.187 ns
0.174 ns
22-year height
22-year diameter(DBH)
0.121 ***
0.233 ***
0.271 **
0.359 ***
0.318 ***
0.584 ***
0.394 **
0.379 **
0.323 ***
0.473 ***
0.342 ***
0.380 ***
Models using canonical scores
For wood quality and growth, based on both the proportion of inter-provenance variation explained and a positive validation (significant positive correlation between observed and
predicted values), the models obtained using the canonical scores as independent variables were better than univariate or bivariate models (Table 1). Here again, local provenances did
not seem to be the best adapted within the region sampled, even when taking into account simultaneously temperature and precipitation.
Potential impact of global climate change
Four scenarios were assessed to estimate the potential impact of climate change. In these scenarios (Table 2), we tested different combinations of temperature and precipitation
variables. The range used neither exceeded the range of our actual data set nor the maximum predicted values already published for the region studied (Environment Canada 1997).
Table 2. Delineation of different climatic scenarios based on climatic differences between location of origin and the
planting sites.
Scenarios
In order to take into account more than one climatic variable at once, a canonical correlation
analysis was also used to develop linear combinations of the climatic distances as independent
variables and then the same procedure described above was followed.
Maximum temperature difference (source - planting)
A (T , P)
B (T , P)
C (T , P)
D (T , P)
(Annual mean
temperature)
(coldest
annual mean
temperature)
(warmest
annual mean
temperature)
(number of
days without
frost)
(frost free
period)
(annual
precipitation)
(summer
precipitation)
-4
-4
-1
-1
-4
-4
-1
-1
-4
-4
-1
-1
-35
-35
-14
-14
-40
-40
-20
-20
-100
100
-100
100
-50
50
-50
50
Some dependent variables were transformed in order to obtain normal distributions.
For wood quality, increase in temperature did not have a major effect. Furthermore, this effect did not seem to be dependent of changes in precipitation (Table 3, Fig. 4).
For growth and especially for 22-year height, in the climate warming scenarios tested (+4C or +1C), the significant predictions were always negative and in the range of those
previously published for this species (Table 3, Carter 1996). However, these predictions were strongly influenced by the level of change in precipitation (Fig. 5). With a decrease of
precipitation (-100 mm), the predictions were modified significantly (Table 3).
Scenario A
Scenario B
Scenario C
Scenario D
(T / +4C , P / +100mm)
(T / +4C , P / -100mm)
(T / +1C , P / +100mm)
(T / +1C , P / -100mm)
0.140 (0.009 , 0.272)
0.236 (0.105 , 0.366)
-0.180 (-0.297 , -0.064)
-0.076 (-0.196 , 0.044)
0.195 (0.061 , 0.329)
0.329 (0.196 , 0.462)
-0.241 (-0.360 , -0.122)
-0.101 (-0.223 , 0.021)
Prop. of late wood (juvenile)
0.306 (-0.158 , 0.770)
0.561 (0.099 , 1.022)
-0.268 (-0.681 ,0.144)
-0.125 (-0.550 , 0.300)
Prop. of late wood (transition)
0.553 (0.154 , 0.953)
0.943 (0.545 , 1.340)
-0.639 (-0.994 , -0.283)
-0.268 (-0.633 , 0.098)
22-year height
-4.171 (-5.045 , -3.297)
-2.042 (-2.876 , -1.208)
-1.558 (-2.364 , -0.753)
0.251 (-0.569 , 1.071)
22-year diameter (DBH)
-2.645 (-3.050 , -2.240)
-1.423 (-1.809 , -1.037)
-1.106 (-1.479 , -0.732)
0.257 (-0.123 , 0.639)
Wood density (juvenile)
0 (transition)
Wood density
0
95% confidence interval is given in parentheses.
Figure 4
Density deviation from « local » source (%)
Results and discussion
Height deviation from « local » source (%)
General circulation models (GCM) predict a rapid global change of climate. The mean
annual temperature in the northern hemisphere is expected to rise and the pattern of
precipitation could be modified profoundly. However, GCM’s spatial resolution is still poor
and the prediction of regional climate change remains limited, especially for precipitation.
In this context, our objective was to integrate several different climate variables (see Table 2)
related to temperature and precipitation in order to develop models of tree response to major
environmental disturbances
Natural Resources Canada, Canadian Forest Service, Laurentian Forestry Centre, 1055 du P.E.P.S., P.O. Box 3800, Sainte-Foy,
Quebec, Canada G1V 4C7
2 CRBF, Université Laval, Sainte-Foy, Québec, Canada G1K 7P4
Figure 5
Conclusion
It appears that wood quality of white spruce should not be affected to a large extent within the
range of climate changes tested in this study. However, with increasing temperature, a growth
loss relative to growth expected from a genetically adapted source was observed. With respect
to its migration capacity, white spruce will probably not be able to exploit the more favourable
growth conditions in the short term. We also showed that regression models, taking
simultaneously into account various climatic parameters, allow for a better understanding of
their interactions. Hence, we showed that the effects of precipitation variables should not be
neglected. Given the uncertainty about the precipitation changes in the future, caution must be
exercised in interpreting the predictions of increasing adaptation lag in global warming
scenarios based solely on temperature change.
References
Carter, K.K. 1996. Provenance tests as indicators of growth response to climate change in 10
north temperate tree species. Can. J. For. Res. 26: 1089-1095.
Environment Canada. 1997. The Canada country study: Climate impacts and adaptation.
http://www.ec.gc.ca/climate/ccs/.
Regnière, J. 1996. Generalized approach to landscape-wide seasonal forcasting with
temperature-driven simulation models. Environ. Entomol. 25: 869-881.
Schmidtling, R.C. 1994. Use of provenance tests to predict response to climatic change:
loblolly pine and Norway spruce. Tree Physiol. 14: 805-817.