• climate change;
  • drought;
  • Minnesota;
  • mortality;
  • prairie–forest ecotone;
  • Quercus macrocarpa;
  • tree-rings


  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

1.Quercus macrocarpa (bur oak) is the dominant tree species along much of the prairie–forest border in the northern-central United States, and movement of Q. macrocarpa in response to climate change may determine the rate at which the prairie–forest ecotone shifts. To investigate likely controls over Q. macrocarpa performance at the edge of its range, we used tree rings to establish the links between drought, growth-rate and mortality for three sites spanning the prairie–forest border in Minnesota.

2.Quercus macrocarpa growth during the 20th century correlates strongly with the Palmer Drought Severity Index (PDSI) and more weakly with raw temperature and precipitation values for all three sites. However, the sensitivity of annual growth rates to drought has steadily declined over time as evidenced by increasing growth residuals and higher growth rates for a given PDSI value after 1950 compared with the first half of the century. We hypothesize that increased atmospheric carbon dioxide concentration may lead to increased water-use efficiency, although we cannot rule out other environmental factors.

3. Because growth is an excellent predictor of Q. macrocarpa mortality, growth–climate relationships provide information on whether oak forests will contract, because of individual tree death, when climate changes. For Q. macrocarpa, declining sensitivity of growth to drought translates into lower predicted mortality rates at all sites. At one site, declining moisture sensitivity yields a 49% lower predicted mortality from a severe drought (PDSI = −8, on par with the worst 1930s ‘American Dust Bowl’ droughts in our study region).

4. Unless the changing relationship between growth and climate is incorporated into forest simulation models, the predicted rate of established tree dieback in a warmer, drier climate may be exaggerated.

5.Synthesis. Adult Quercus macrocarpa trees appear to be increasingly insensitive to drought-induced mortality. Because the species is dominant at the prairie–forest ecotone in the northern-central United States, movement of the ecotone in response to climate change may be delayed for decades.


  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

Quercus macrocarpa (bur oak) is the dominant species along the prairie–forest border (Fig. 1) in the north-central United States (Burns & Honkala 1990; Prasad et al. 2007), and thus the response of this species to climate change serves as a proxy for the fate of the ecotone. For example, when Q. macrocarpa seedlings establish in grasslands because environmental conditions such as water availability or fire frequency become favourable (Danner & Knapp 2001), forests expand and the prairie–forest border moves. Conversely, as mortality of adult Q. macrocarpa increases in response to environmental change, grasslands may expand and forests retreat.


Figure 1.  Study sites along the prairie–forest border in western-central Minnesota, USA. Map with biome boundaries based on data available from the Minnesota Department of Natural Resources ( MW, Maplewood State Park; GL, Glacial Lakes State Park; SB, Sibley State Park.

Download figure to PowerPoint

With continued increases in global greenhouse gas emissions, climate models predict that, by the end of the 21st century, Minnesota summer temperature will increase by 4–9 °C and summer precipitation will slightly decrease (Kling et al. 2003; Christensen et al. 2007). Some forest models and extrapolations from the paleoecological record suggest that, in response to increased temperature and/or drought, forests may retreat to the extreme north-eastern parts of the state (Pastor & Post 1988; Hamilton & Johnson 2002; Galatowitsch, Frelich & Phillips-Mao 2009). Umbanhowar et al. (2006) documented a 1000-year forest retreat from south-central Minnesota during the mid-Holocene in response to a warming of fewer than 4 °C, with the bulk of forest collapse taking place in the first 200–300 years and without charcoal evidence of catastrophic woodland fire. Breshears et al. (2005, 2009) documented a recent, rapid drought-induced forest retreat in the south-western United States, but uncertainty in 21st century precipitation forecasts make drought-based forest dieback in Minnesota difficult to predict with confidence (Williams, Shuman & Bartlein 2009).

Records indicate that atmospheric CO2 concentrations rose by 10 p.p.m. during the first half of the 20th century and by an additional 57 p.p.m. from 1950 to 2000 (Keeling & Whorf 2005). Since 2000, the rate of increases for atmospheric CO2 has accelerated (Canadell et al. 2007). Numerous theoretical, laboratory and field studies have shown that elevated levels of CO2 have a direct effect on a variety of plants by increasing their water-use efficiency and decreasing their sensitivity to drought (Leavitt et al. 2003; Long et al. 2004; Ainsworth & Long 2005). Thus, the direct effects of increased CO2 levels might at least partially offset the warming-induced increases in evapotranspiration for some species. The balance between these two forces may determine the success of tree species growing at the prairie–forest border.

Several recent tree-ring studies have demonstrated a decline over time in the strength of the correlation between climate and gymnosperm tree-ring width in a variety of forest types, perhaps as a result of CO2-induced decoupling (Briffa et al. 1998a,b; Knapp, Soulé & Grissino-Mayer 2001; Soulé & Knapp 2006; Wang, Chhin & Bauerle 2006; Leal et al. 2008; but see Biondi 2000 and Carrer & Urbinati 2006 for counter examples). To our knowledge, however, no study to date has examined temporal trends in the strength of the ring width–drought link for populations of an angiosperm species or for trees growing in the prairie–forest ecotone.

Annual growth rate serves as an integrating measure of tree health, and there is an established link between tree growth and the likelihood of mortality (Buchman, Pederson & Walters 1983; Kobe et al. 1995; Wyckoff & Clark 2002). Thus, the sensitivity of tree growth to future climates is crucial for determining the rate of forest contraction through death of established trees. Individual-based forest gap models are widely used to simulate forest dynamics across various scales and forest types, both directly and as submodels driving landscape models (Risch, Heiri & Bugmann 2005; Xiaodong & Shugart 2005), and these models incorporate functions linking climate to growth and growth to mortality. One application of these models is to predict impacts of 21st century climate change (Guertin, Easterling & Brandle 1997; Guo et al. 2004;Hanson et al. 2005).

Both the growth–climate and growth–mortality relationships in existing models have drawn extensive criticism (Pacala & Hurtt 1993; Loehle & LeBlanc 1996; Wyckoff & Clark 2002) and new methods for estimating these relationships from field data are beginning to appear in the literature (Wyckoff & Clark 2000; Wunder et al. 2007). One critique arises because individual-based forest simulators typically lump species into coarse categories when assigning climate–growth and growth–morality functions. Model predictions change when species-specific functions are used (Wyckoff & Clark 2002). A further complication arises if species’ functions change temporally (Wunder 2007) or with geography (Kobe 1996).

Here, we offer insight into how established Q. macrocarpa may fare in the current century by: (i) using tree rings to establish the relationship between drought and Q. macrocarpa growth for three sites along Minnesota’s prairie–forest border, (ii) calculating the current relationship between growth and mortality for adult Q. macrocarpa and (iii) using the distributions of current growth rates for Q. macrocarpa to predict the susceptibility of current populations to droughts of varying strength. Furthermore, we look for temporal trends in the correlation between Q. macrocarpa growth and climate, hypothesizing that increases in CO2 may lead to weaker relationships between drought and tree growth over time. We finish by examining the potential impacts of changing tree-growth drought sensitivity on drought-induced mortality and discuss the implications for modelling future forest dynamics.

Materials and methods

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

Study area

At the time of European settlement in the mid-1800s, a narrow prairie–forest ecotone split Minnesota into approximately 1/3 prairie and 2/3 forest ( Since then, changes in climate, fire, land use and herbivore pressure have served to make the prairie–forest transition less distinct (Briggs, Hoch & Johnson 2002). Our three study sites, Maplewood State Park (MW), Glacial Lakes State Park (GL) and Sibley State Park (SB), are located in western-central Minnesota along the prairie–forest border (Fig. 1). All three parks are a mosaic of prairie and forested patches. January and July mean temperatures are −14 and 21 °C at MW, −12 and 22 °C at GL and −13 and 22 °C at SB, respectively. Summer precipitation (June through August) is approximately 15% less at MW (279 mm) and GL (277 mm) then at SB (323 mm) (, while modelled potential evapotranspiration (PET) is similar at all sites (138 mm in July for MW, 140 mm for GL and SB) (Mark White, The Nature Conservancy, Duluth, MN, USA, personal communication), probably resulting in less water stress at the southern SB site. Soils at all three sites are deep and fertile and formed on grey till and glacial outwash. Soils at MW and GL are exclusively mollisols, while soils at SB include a mixture of alfisols and mollisols (Cummins & Grigal 1981).

Historical photographs and park histories indicate expansion of forest cover at all three sites following European settlement and the accompanying reduction in fire frequency and bison grazing pressure. We used data from thirty 20 × 20 m permanent vegetation plots – 10 per site – established in 2002 and 2003 to determine current forest composition in the areas sampled for this study. Forests at each site contain similar tree basal area (MW: 28.3 m2 ha−1; GL: 30.8 m2 ha−1; SB: 28.2 m2 ha−1). Quercus macrocarpa is dominant only along ridgetops at the northern site but dominates everywhere we sampled at the two southern sites (73% of total basal area at GL and 78% at SB). At all sites, tree rings reveal that the oldest Q. macrocarpa individuals date to the mid-1800s, consistent with rapid post-settlement forest expansion (data not shown).

Data collection

Our field efforts included the collection of four data sets: (i) complete increment cores for linking Q. macrocarpa growth and climate; (ii) partial increment cores for linking growth and mortality; (iii) census data for estimating underlying mortality rate and (iv) an additional, independent set of partial Q. macrocarpa cores to quantify the current growth rate distribution for the species.

Data set 1

To establish the relationship between climate and Q. macrocarpa growth, we used increment borers to obtain tree ring samples at each of our three study sites (17 trees at MW, 20 at GL and 19 at SB) during the summer of 2002. Trees were selected by searching near our permanent vegetation plots. Individuals were limited to canopy dominants and co-dominants to increase climate signal. Two cores extending from the bark to the pith were obtained from each tree at approximately 30 cm above-ground level.

Data set 2

In addition to establishing the link between climate and tree growth, we sought to measure the relationship between Q. macrocarpa growth and mortality. We walked transects at GL to locate dead adult Q. macrocarpa during the summer of 2003, obtaining partial increment cores (typically capturing the 20–30 most recent growth rings) from 33 recently dead (average size: 27.8 cm d.b.h.) and 30 nearby living individuals matched to be approximately the same size and canopy position (average size: 30.6 cm d.b.h.). Again, two cores were obtained from each tree approximately 30 cm above-ground level. To separate growth-related mortality from catastrophic mortality, sampled dead individuals were limited to those with intact crowns. ‘Recent death’ was assured by selecting dead trees with bark present that yielded intact and moist increment cores.

Data set 3

Our method for establishing the relationship between tree growth and mortality is based on Wyckoff & Clark (2000), and requires both growth rates for living and recently dead trees and an estimate of background mortality rate. To estimate mortality rate for Q. macrocarpa at our field sites, we followed 219 established individuals in our permanent vegetation plots from plot establishment (2002 and 2003) until 2007. Trees were checked for mortality in 2004, 2005 and 2007. Average size of tracked individuals at the time of plot establishment was 30.4 cm d.b.h. Mortality rate estimates were supplemented by mortality data for USDA Forest Service Forest Inventory and Analysis (FIA) plots in the 10 Minnesota counties along our transect of study sites for 1997–2004.

Data set 4

A more extensive, independent estimate of the Q. macrocarpa growth rate distribution was necessary to allow us to use our estimates of growth–climate and growth–mortality relationships to explore the risk of tree death under various drought conditions. Two partial increment cores were obtained from approximately 30 cm above-ground for 97 additional Q. macrocarpa individuals (later reduced to 93 to better match the size distribution of our growth–mortality data set). Trees were located by walking transects near our permanent vegetation plots at GL in 2003. The average size of the sampled trees was 31.2 cm d.b.h.

The growth–climate relationship

All cores collected to determine the relationship between growth and climate (Data set 1) were dried, mounted, sanded and initially cross-dated using skeleton plots (Stokes & Smiley 1996). Rings were then measured to the nearest 0.01 mm with a Velmex measuring system ( Formal cross-dating was performed and measurements were checked for accuracy using cofecha (Holmes 1999). Cores that could not be accurately cross-dated were excluded from further analysis. Cores were de-trended using arstan (Cook & Holmes 1999), producing a standard chronology for each site, which, after discarding problem cores, was based on 21 radii at MW, 33 at GL and 29 at SB. Average estimated age for individual trees used in chronology construction was 127 for MW, 114 for GL and 130 at SB.

We wished to examine the potential changes in growth–climate relationships over the course of decades, so following Knapp, Soulé & Grissino-Mayer (2001), conservative detrending techniques (i.e. linear and negative exponential) were used to remove the effects of tree age on ring width while preserving as much variance from other sources as possible. Conservative standardization is utilized in many similar recent published studies (Knapp, Soulé & Grissino-Mayer 2004; Soulé, Knapp & Grissino-Mayer 2004; Li et al. 2006, 2007; Soulé & Knapp 2006; Underwood 2007; Leal et al. 2008). Compared with more flexible fits, conservative approaches have the disadvantage of preserving more series autocorrelation, and to examine this trade-off we used the data from our SB site to compare conservative standardization with a variety of smoothing splines, in increasing order of flexibility (all generated using arstan): 126, 100, 80, 60, 40, 20, 10 and 5 years. Speer et al. (2009) suggest that, beyond considerations of the trade-off between preserving decadal trends and eliminating autocorrelation, strength of climate correlation is the preferred metric for deciding among detrending techniques.

Sample depths were sufficient to examine the relationship between growth and climate for the period after 1900 (sample depth in 1900: 13 radii for MW, 11 for GL and 18 for SB). The chronologies we developed were correlated with 20th century monthly temperature, precipitation and levels of the Palmer Drought Severity Index (PDSI) for Minnesota Climate Region 4 (west-central) ( This was done using S-Plus statistical software ( PDSI is commonly used in tree ring studies examining the impacts of drought on growth (Li et al. 2006, 2007; Stahle et al. 2007; Speer et al. 2009) because it captures the joint role temperature and precipitation play in determining moisture availability, a key limiting factor at drought-prone sites such as the prairie–forest border. Kempes et al. (2008) reported that PDSI better predicts growth of Pinus edulis at sites in New Mexico and Arizona than did either raw precipitation or three alternative drought indices.

Significance for correlations between annual growth and monthly metrics from the current and preceding growing season were assessed via bootstrapping (1000 iterations) using Dendroclim2002 (Biondi & Waikul 2004), a program shown to be conservative in estimating significance compared with other commonly used tree-ring analysis methods.

Because Durbin–Watson tests showed retained autocorrelation in our standardized chronologies and (as expected) in PDSI but not in temperature or precipitation records, we elected to adopt a P < 0.01 significance test for regressions relating growth and any climate variable used in the further analyses described below (Knapp, Soulé & Grissino-Mayer 2001).

20th century changes in the growth–climate relationship?

The consistency of the relationship between tree growth and climate was assessed using three approaches. First, following Knapp, Soulé & Grissino-Mayer (2001), we determined the temporal trend in studentized residuals when the chronologies for each of our three sites were regressed against Region 4 PDSI. An increase in residuals with time would indicate that moisture has become a poorer predictor of growth, i.e. Q. macrocarpa became less sensitive to drought. Second, for each site, we compared the regression lines for standardized growth versus PDSI for the periods 1901–1950 and 1951–2000 using an F-test to measure differences between regression coefficients (Sokal & Rohlf 1995). Several prior studies have suggested that pre- and post-1950 tree-ring–climate relationships may be different, possibly because of the large jump in atmospheric CO2 after 1950 (Graumlich 1991; Knapp, Soulé & Grissino-Mayer 2001; Soulé & Knapp 2006). Finally, Dendroclim 2002 (Biondi & Waikul 2004) software was used to detect temporal trends in correlations between the standard chronologies for each of our three sites and Region 4 PDSI. For this analysis, our 20th century data were assessed using moving intervals, with interval length up to 50 years.

Growth–mortality relationship for established Q. macrocarpa

To examine the relationship between growth and mortality for Q. macrocarpa (as in Wyckoff & Clark 2000), we combined three types of data: (i) growth rates of living individuals, (ii) growth rates of recently dead individuals and (iii) estimated annual mortality, θ. All cores collected to determine the distribution of growth rates of living and recently dead individuals (Data set 2) were mounted, sanded and measured to the nearest 0.01 mm as before. We employed informal visual cross-dating, but formal cross-dating with cofecha was not conducted because we only required measurements of recent annual rings for this analysis and because a prior study (Wyckoff & Clark 2002) found that growth rings of dying trees showed no discernable climate signal. A Bayesian approach was used to iteratively combine Forest Service FIA data with the mortality data obtained from our permanent vegetation plots (Data set 3) (eqns 13–18 in Wyckoff & Clark 2000) and to obtain a posterior estimate of θ.

Our growth–mortality function describes how mortality risk increases as growth rate declines. If d is the event that an individual dies in a given time interval and a is the event that it survives, then let the probability of death be represented by p(d). The probability of survival, p(a), is equal to 1 − p(d). For a data set GN containing a sample of N trees, the likelihood that D trees die during this interval is the binomial

  • image( eqn 1a)

where the growth–mortality function p(d|gi;β) is the probability of death conditioned on a risk factor (growth rate) gi and fitted parameters β. If there is no relationship between growth and mortality (i.e. all individuals experience the same risk), then eqn 1a simplifies to the binomial

  • image( eqn 1b)

where, in this case, the parameter θ represents the overall probability of death p(d) and its maximum likelihood (ML) estimate is

  • image( eqn 2)

However, when there is a relationship between growth and mortality, the numbers of live and dead trees must be scaled in accordance with θ for estimation. This scaling is necessary because dead trees are rare. We directly fit the growth–mortality relationship p(d|g), using a Weibull model with fitted parameters b and c. The likelihood from eqn 1a thus becomes:

  • image( eqn 3)

Equation 3 does not explicitly show the mortality rate, θ. However, this likelihood depends implicitly on θ, because θ represents the fraction of dead trees in the sample (eqn 2).

Parameter estimates were determined by a nonparametric bootstrap (Efron & Tibshirani 1993) with 500 resamples. We used the resampling procedure to weight the contributions of living and dead trees. Each resample included: (i) a sample size of D dead trees, drawn with replacement from measured growth rates of recently dead trees, equal to the number of dead trees in the data set; and (ii) a sample of living trees of size:

  • image( eqn 4)

drawn with replacement from measured growth rates of living trees. Each resample drew from a beta distribution of posterior estimates of θ and thus we incorporated error both in our estimates of tree growth and in our estimate of background mortality.

Maximum likelihood parameter estimates were obtained for each resample using eqn. 3. The ML for the model is taken as the mean ML over the bootstrapped sample, because the under-representation of live growth rates means that there is no likelihood for a raw data set. Our method allows for a probability statement regarding how growth rates influence mortality, using a likelihood ratio test and a null model of growth-independent mortality (eqn 2), but it does not allow for the calculation of confidence intervals.

Impact of changing drought tolerance on mortality

To examine the potential impacts of changing drought sensitivity on mortality of adult Q. macrocarpa, we prepared 186 cores from 93 additional adult trees as before, measuring the most recent 15–30 rings in each core (Data set 4). Cores were not formally cross-dated, but the 1988 drought was used as a marker ring. This data set became the starting point for this analysis.

Five-year average growth rates were calculated for each core for 1998–2002 (the most recent growth period). For each study site, growth rates obtained from each core were scaled for each integer value on the PDSI scale from −10 (severe drought) to +5 (quite wet) based on pre-1950 regressions of PDSI versus detrended growth and the ratio of:

  • image

The same step was then repeated using post-1950 regressions of PDSI versus detrended growth rate, but modelled droughts only extended to PDSI = −8 owing to a lack of data from more severe droughts in the post-1950 period. As an example, a core obtained from our MW site with an average annual growth rate of 0.600 mm for 1998–2002 (average PDSI = 0.856) would have a ‘scaled’ growth rate of 0.338 mm for PDSI = −6 based on the 1901–1950 PDSI-detrended growth rate relationship but a higher scaled growth rate of 0.489 mm for the same PDSI based on the 1951–2000 PDSI-detrended growth rate relationship.

To examine the variance inherent in the growth rate data set, the above process was repeated through 500 bootstrap iterations for each study site, with a new tree growth data set constructed for each iteration by sampling (with replacement) from the 186 core-based growth rates. Scaled growth rates for each iteration were then used as input into our growth–mortality function, with another 500 bootstrap iterations performed for each integer unit of PDSI. Output was thus a distribution of mortality risk given growth rate, for varying moisture conditions.


  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

Growth chronologies

Length of chronologies for our three sites ranged from 123 to 180 years (Table 1). Interseries correlation, an indication that trees are responding to a common environmental signal, is strong at each site. Similarly, expressed population signal (EPS) values for each chronology exceed the 0.85 threshold commonly used in judging chronology adequacy (Wigley, Briffa & Jones 1984; Fowler & Boswijk 2003). With conservative standardization, mean sensitivity (the percentage change in ring width from year to year), standard deviation and first-order autocorrelation all indicate chronologies well suited for our subsequent analyses (Table 1). More flexible detrending reduces autocorrelation, but at the cost of decreased correlation with climate and the removal of low-frequency variation. As an example, for our SB site, a 20-year spline yields a resulting series with < 0.2 first order autocorrelation (compared with 0.41 for the conservative detrending) but correlation falls 20% for PDSI and 9% for precipitation with the flexible spline (for months with significant growth–climate correlations), causing us to use only the more conservative approach for subsequent analysis. Our study trees were part of a forest expansion in the late 1800s and probably never experienced the extended periods of suppression followed by periods of more rapid growth that might recommend flexible standardization.

Table 1.   General descriptive statistics for Quercus macrocarpa tree-ring chronologies
Study sitePeriodMean sensitivityStandard deviationFirst-order autocorrelationInterseries correlationEPS
Maplewood State Park1822–20020.260.370.560.590.97
Glacial Lakes State Park1879–20020.290.360.480.590.97
Sibley State Park1873–20020.270.370.410.630.98

Drought– growth relationships for Q. macrocarpa

Detrended Q. macrocarpa ring widths correlate positively with monthly PDSI from current and prior-year growing seasons (Fig. 2), with the months of the current growing season showing the strongest relationship. The trees at the northernmost site (MW) consistently exhibited the strongest sensitivity to drought, while the trees at the southernmost (but wettest) site (SB) tended to show the least sensitivity. Correlations with PDSI were stronger than with temperature or precipitation (Table 2), and current-year July PDSI was the most predictive month at all three sites. Subsequent analyses use only July PDSI, and regressions of July PDSI versus standardized growth were highly significant for all sites (P < 0.0001).


Figure 2.  Prior and current-year growing season monthly correlations between detrended Quercus macrocarpa tree-ring growth indices and the Palmer Drought Severity Index for three study sites. MW, Maplewood State Park; GL, Glacial Lakes State Park; SB, Sibley State Park.

Download figure to PowerPoint

Table 2.   Bootstrapped correlations between climate and tree growth calculated using Dendroclim2002 (Biondi & Waikul 2004). Only relationships significant at the 95% level are shown. The strongest correlation for each site was July Palmer Drought Severity Index (PDSI) (bold)
 Prior MayPrior JunePrior JulyPrior AugustPrior SeptemberPrior OctoberMayJuneJulyAugustSeptemberOctober
Maplewood State Park
 Precipitation  0.23   0.300.320.32   
 Temperature  −0.31−0.16 0.20 −0.38−0.20   
Glacial Lakes State Park
 Precipitation   0.20 0.270.390.280.28   
 Temperature   −0.19   −0.30    
Sibley State Park
 PDSI  0.220.330.330.370.470.520.540.490.490.47
 Precipitation   0.20 0.370.370.180.27   
 Temperature       −0.31    

Q. macrocarpa sensitivity to climate over time

We followed methods developed in Knapp, Soulé & Grissino-Mayer (2001) to determine temporal trends in drought sensitivity at our sites. The studentized residuals from linear regression relating growth and drought increased through the 20th century at all three sites (significant at MW and GL), suggesting a decreased growth response to any particular level of drought in recent years (Fig. 3a–c; Table 3). By 2000, trend lines showed actual growth exceeding predicted growth (e.g. positive residuals in Fig. 3) at all three sites. We next calculated separate linear regressions relating tree growth and drought for the periods 1901–1950 and 1951–2000 for each site. Significant differences in the resulting pre- and post-mid century regression lines occur at the two northern sites (MW and GL –Fig. 4a,b, Table 3). Differences at the third site were not significant. Moving interval analysis using Dendroclim2002 software showed the same pattern, with decreasing drought sensitivity over time at all three sites (not shown).


Figure 3.  Studentized residuals increase with time for (a) Maplewood State Park (MW) (P = 0.03), (b) Glacial Lakes State Park (GL) (P = 0.01) and (c) Sibley State Park (SB) (P = 0.12).

Download figure to PowerPoint

Table 3.   Fitted parameter estimates
 Parameter 1Parameter 2    
  1. PDSI, Palmer Drought Severity Index; MW, Maplewood State Park; GL, Glacial Lakes State Park; SB, Sibley State Park.

 Studentized residuals vs. year (Fig. 3)
a (SE)b (SE)Fd.f.P-valuer2
MW−14.779 (6.646)0.008 (0.003)4.95990.0280.05
GL−16.496 (6.610)0.009 (0.003)6.23990.0140.06
SB−10.514 (6.713)0.005 (0.003)2.45990.120.03
 Detrended growth vs. PDSI (Fig. 4)
a (SE)b (SE)Fd.f.P-valuer2
1901–1950 0.062 (0.007)0.921 (0.0220)68.7448< 0.0010.59
1951–2000 0.030 (0.013)1.089 (0.033)5.59480.0220.10
1901–1950 0.039 (0.007)0.957 (0.021)31.148< 0.0010.39
1951–2000 0.028 (0.008)1.024 (0.021)12.4548< 0.0010.21
 Growth-mortality function (Fig. 5)  
 bcNegative log likelihoodP-value (vs. null model)  
GL 0.0070.300159.87< 0.0001  

Figure 4.  Detrended Quercus macrocarpa tree-ring growth indexes vs. Palmer Drought Severity Index (PDSI) for (a) Maplewood State Park (MW) and (b) Glacial Lakes State Park (GL) sites. On each panel, circles (○) and solid regression lines are for data from 1901 to 1950, and crosses (×) and dashed regression lines correspond to data from 1951 to 2000. Thin solid and dashed lines indicate 1 SE confidence intervals. Panels shown only for sites where pre- and post-1950 regression lines differed significantly at P < 0.05. See Table 3 for regression statistics.

Download figure to PowerPoint

Growth-mortality relationship for Q. macrocarpa

The 30 living and 33 recently dead Q. macrocarpa trees cored for our growth–mortality analysis were of similar average size, but the dead trees grew 40% more slowly on average during the 5 years prior to their death than did the individuals that remained alive (living tree: mean = 0.91 mm year−1, SD = 0.57 mm year−1; dead tree: mean = 0.55 mm year−1, SD = 0.45 mm year−1; two sample t-test comparing means, P = 0.0025). The average dead tree began to grow more slowly than the average live tree up to 20 years prior to coring (Fig. 5a). Visual cross-dating reveals a similar pattern of recent growth deviations for both series (Fig. 5b), suggesting that the typical dead tree had died in the year prior to sampling.


Figure 5.  (a) Trends in average annual radial growth for living and recently dead trees show increasing deviation over a quarter century. (b) Similar recent patterns of deviation from the 25-year trend for each series suggest that both sampled living and recently dead trees responded to a similar environmental signal. Note different axis scales for (a) and (b).

Download figure to PowerPoint

To calculate a growth–mortality relationship, we needed to weight the living and dead tree samples based on the relative number of living and dead trees in the population. The required ratio was obtained by combining our permanent plot-based data with regional FIA data and yielded a posterior Bayesian estimate of annual mortality (θ) for Q. macrocarpa of 0.017 (95% CI: 0.009–0.027). Trees ranging from 15 to 65 cm d.b.h. contributed to the estimate of θ, with an average d.b.h. of 25 cm for the dead trees. When growth data and θ were combined, the resulting growth–mortality function was highly significant compared with the null model of no relationship between growth and mortality (P < 0.0001). Our calculated function predicts negligible mortality risk for individual trees exhibiting 5-year average radial growth rates greater than 1.5 mm year−1, but risk increases as trees grow more slowly than that threshold (Fig. 6; Table 3).


Figure 6.  Fit Weibull growth–mortality function (solid line) shows that as growth rate increases, annual probability of mortality decreases. Average growth rate for the five most-recent years of growth are shown for living (×) and recently dead (◊) individual trees.

Download figure to PowerPoint

Potential impact of a changing growth–climate relationship on Q. macrocarpa mortality

When combined, (i) the pre- and post-1950 growth–drought sensitivity regressions for each site, (ii) the calculated growth–mortality function (not site specific), and (iii) our independent set of core-based tree-ring growth suggest that declining drought sensitivity may translate into substantially reduced mortality risk for a given level of drought at the MW site (Fig. 7a). At a PDSI of −8, a drought on par with the worst of the American Dust Bowl drought of the 1930s, predicted mortality would be 49% less now than it was then. Predicted mortality risk is more modestly reduced at GL (21%) for a PDSI = −8 drought (Fig. 7b), and only mildly decreased at SB (not shown).


Figure 7.  Predicted annual Quercus macrocarpa mortality for various levels of drought as measured by Palmer Drought Severity Index (PDSI) for (a) Maplewood State Park (MW) and (b) Glacial Lakes State Park (GL) sites. Solid lines are based on pre-1950 drought sensitivity and include mean and bootstrapped 95% CI response. Dashed lines are based on lessened post-1950 drought sensitivity.

Download figure to PowerPoint


  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

We find that Q. macrocarpa growth is strongly related to drought, implying that expected climate change will have a substantial negative impact on the performance of this dominant species in the prairie–forest ecotone. However, individual Q. macrocarpa trees are less sensitive to drought stress now than they were a century ago, suggesting that elevated levels of CO2 may be mitigating the impact of drought. Additionally, given the strong relationship between growth and mortality, our combined results suggest both that the pace of forest retreat may be difficult to predict from historic patterns and that many current models may overstate the expected pace of change in the current century.

The strong relationship between drought and ring width for Q. macrocarpa reported here (Fig. 2, Table 2) is consistent in magnitude with results reported by Dean et al. (1996) for sites in North Dakota, South Dakota and Iowa, USA. Our results are also similar to those reported by Tardif & Conciatori (2006) for Q. alba and Q. rubra growing in south-western Quebec. New here is the documentation of changes in the strength of the relationship between drought and growth across the 20th century (Figs 3 and 4).

Although we do not provide direct evidence for a causal relationship between rising levels of atmospheric carbon dioxide and the declining drought sensitivity seen at our study sites, an accumulating body of theoretical and circumstantial evidence suggests a link. Free-air carbon enrichment (FACE) studies report generally decreased transpiration and increased water-use efficiency for trees grown under elevated levels of carbon dioxide (Long et al. 2004; Ainsworth & Long 2005), although recent studies suggest that the effect is modest for European oaks (Ferrio et al. 2003; Leuzinger & Korner 2007). Feng (1999) used measurements of 13C discrimination in the rings of a single oak tree to document increased water-use efficiency since 1950. Recent isotopic studies expand this analysis (Reynolds-Henne et al. 2007; Roden & Ehleringer 2007; Peñuelas et al. 2008), although there are substantial limitations to the technique (Cooper & Solis 2003).

Recent non-isotopic studies infer a decline in drought sensitivity for adult soft-wood trees growing in Oregon and California, USA (Knapp, Soulé & Grissino-Mayer 2001; Soulé & Knapp 2006), Manitoba, Canada (Wang, Chhin & Bauerle 2006), and Austria (Leal et al. 2008). Davis et al. (2007) document evidence of decreased drought stress and higher soil moisture for Q. ellipsoidalis seedlings growing under elevated CO2 concentrations at the Cedar Creek FACE site in Minnesota. Gedney et al. (2006) suggest that CO2-induced decreases in transpiration are partly to blame for increased flooding in the Mississippi River basin – the location of our field sites – although maize (Zea mays) and soybeans (Glycine max) are undoubtedly more important contributors to this effect than our target species.

There may be alternative explanations for the declining sensitivity to drought with time for Q. macrocarpa at our study sites. Other environmental factors may play a role. Heitholt, Johnson & Ferris (1991) suggest that increased nitrogen fertilization can reduce drought sensitivity in wheat, although Wu et al. (2008) report that increased nitrogen increased drought sensitivity for a shrub, possibly through a shift away from allocation to roots. Another obvious candidate is a change in sensitivity with tree age. Voelker et al. 2006 suggest that the fertilization effect of CO2 on growth of Q. velutina and Q. coccinea are most pronounced for younger trees, although they did not consider age-related trends in climate sensitivity. The individual trees in our growth chronologies were on average nearly 30 years old in 1900 (or the year they entered the calculations if after 1900). Although we used detrending techniques to take out low-frequency effects of age on ring widths, any high-frequency age-related variation was retained.

It is possible that the decline in sensitivity to year-to-year variations in climate is seen only because the trees were getting older. However, we think this unlikely, because there is no significant trend in growth residuals after factoring in moisture for any of our sites when the period 1901–1950 is considered alone (not shown). By contrast, two sites (MW and SB) show significant trends in studentized residuals with time when the period considered is 1951–2000 (not shown). For the latter period, individual trees were older and presumably any age, size and status effects would have dissipated. In our permanent plots, Q. macrocarpa growth rates (no data available on drought sensitivity) increased with tree size, but the slope became flat when subcanopy individuals (< 35 cm d.b.h.) were removed. No growth decline from senescence was seen in our relatively young study forests (data not shown). Most of the individuals used in constructing our growth chronologies would have been in the canopy by 1950 and would be decidedly ‘middle aged’ for the remaining duration of the study.

The oldest Q. macrocarpa trees were found at our northernmost site (MW – see Fig. 1). Trees at this site showed the greatest overall drought sensitivity (Fig. 2) along with the greatest decline in sensitivity with time (Fig. 3a). Our northern MW site got 15% less summer precipitation than our southern SB site, while modelled PET was essentially unchanged across our three sites (a variation beneath the resolution of our regional PDSI data). Therefore, contrary to the expected pattern, our southernmost site was the least droughty. It may be that the trees growing at the northern MW site are both the most water stressed and the most susceptible to relief of such stress by rising atmospheric CO2 concentrations (Huang et al. 2007). If our interpretation is correct, this would further support our hypothesis that increased CO2 concentration is the cause of the declining drought sensitivity.

The current century is predicted to bring not only further increases in carbon dioxide and other greenhouse gas concentrations, but also an accompanying rapid rise in temperature for our study region (Christensen et al. 2007). This increased temperature may lead to more severe regional droughts as PET increases. Many forest models have assumed that temperature increases alone will be sufficient to lead to dieback of forest species, mediated through declined growth rate and a sensitive relationship between tree growth and mortality (see results for the LINKAGES model in Pastor & Post (1988) and Post & Pastor (1996)). In response to criticism that model predictions might be overstating the risk of future forest tree dieback (Loehle & LeBlanc 1996; Loehle 2000), modified models with altered growth–mortality functions have appeared more recently (Wullschleger et al. 2003; Hanson et al. 2005), but growth–climate and growth–mortality relationships generally remain simplistic.

Our results suggest that sophisticated model functions will be required to accurately predict the response of tree species to changing climate. Although establishment of new seedlings may be threatened, established Q. marcocarpa individuals are long-lived. Since climate change pressure will favour a forest retreat, the life stage we study here is crucial for predicting expected pace of range contraction–recruitment–climate relationships only set the pace of a range expansion. The decreasing drought sensitivity of established trees (Fig. 7) may act as a buffer and delay the movement of the prairie–forest ecotone for many decades even in the face of climate change.

Our data have limitations. Although we develop growth–climate regressions for each of three of our study sites, our data only allow for a pooled growth–mortality function (Figs 5 and 6). Kobe (1996) showed that such functions can vary with site for a given species. Perhaps more importantly, it is possible that, like drought sensitivity, the growth–mortality relationship varies with time. We do not have the data required to explore temporal trends in that relationship. Our results should thus be seen as an exploration of the potential impact that changing drought sensitivity may have on the mortality of trees, not an explicit prediction for any given site.

We limited our drought simulations to PDSI ≥ −10 (≥ −8 when the post-1950 drought sensitivity was assumed) because, among other reasons, more severe drought would have implied 5-year average growth rates for some individual trees lower than the lowest growth rate observed in our ‘recently dead’ data set. Quercus macrocarpa is a ring porous species, and the large vessels that mark the beginning of each ring were not observed here to have diameters less than 0.08 mm. Thus, we would not expect the smallest discernable ring to be much less than 0.08 mm. It is not clear how to interpret the output of growth–mortality functions if the input growth rates drop below a biologically plausible level (or even merely below levels actually observed), and this should be taken into account when using forest simulation models to predict the dynamics of Q. macrocarpa and similar species. More generally, we should also acknowledge that we are stretching somewhat when using models to predict forest dynamics in the 21st century that have been parameterized with data from the 20th century. For our study region, July PDSI was −9.65 in 1934 and −7.45 in 1988, but we have no data for even more severe droughts. The American Dust Bowl drought (and other 20th century droughts) may have been too mild to provide guidance for the current century.

One final caveat: the paleorecord suggests at least a gradual oak forest retreat as recruitment fails in a warmer world (Umbanhowar et al. 2006), but current woody encroachment into prairie (Danner & Knapp 2001) indicates the subtlety and complexity required to accurately predict the near future. Our data suggest that drought alone may be less of a threat than it once was to established Q. macrocarpa in western-central Minnesota, but other factors, such as disease or fire, could certainly act to remove stands growing at the prairie–forest border. If a drier climate means a substantial increase in incidence for either threat, then the retreat of the Minnesota oak forest could be rapid.


  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

This work was supported by a University of Minnesota Grant-in-Aid of Research to P. W. and various Undergraduate Research Opportunity Program fellowships. We thank students R. Jansen, A. Oredson, J. Gustaffson, K. Juhlin, G. Meyer, J. Nagel, B. Burns and M. Bombyk for help with field work and data collection. F. Biondi provided access to Dendroclim2002. H. Grissino-Mayer, P. Soulé and E. Sunger provided statistical advice. We also thank C. Cole, J. HilleRisLambers, A. Hodgson, T. J. O Wyckoff and three anonymous referees for comments which greatly helped to improve this manuscript.


  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  • Ainsworth, E. & Long, S. (2005) What have we learned from 15 years of free-air CO2 enrichment (FACE)? A meta-analytic review of the responses of photosynthesis, canopy properties and plant production to rising CO2. New Phytologist, 165, 351372.
  • Biondi, F. (2000) Are climate-tree growth relationships changing in north-central Idaho, USA? Arctic, Antarctic and Alpine Research, 32, 111116.
  • Biondi, F. & Waikul, K. (2004) Dendroclim2002: a C++ program for statistical calibration of climate signals in tree-ring chronologies. Computers & Geosciences, 30, 303311.
  • Breshears, D.D., Cobb, N.S., Rich, P.M., Price, K.P., Allen, C.D., Balice, R.G. et al. (2005) Regional vegetation die-off in response to global change-type drought. Proceedings of the National Academy of Sciences of the United States of America, 102, 1514415148.
  • Breshears, D.D., Myers, O.B., Meyer, C.W., Barnes, F.J., Zou, C.B., Allen, C.D., McDowell, N.G. & Pockman, W.T. (2009) Tree die-off in response to global change-type drought: mortality insights from a decade of plant water potential measurements. Frontiers in Ecology and the Environment, 7, 185189.
  • Briffa, K.R., Schweingruber, F.H., Jones, P.D., Osborn, T.J., Harris, I.C., Shiyatov, S.G., Vaganov, E.A. & Grudd, H. (1998a) Trees tell of past climates: but are they speaking less clearly today? Philosophical Transactions of the Royal Society of London B, 353, 6573.
  • Briffa, K.R., Schweingruber, F.H., Jones, P.D., Osborn, T.J., Shiyatov, S.G. & Vaganov, E.A. (1998b) Reduced sensitivity of recent tree-growth to temperature at high northern latitudes. Nature, 391, 678682.
  • Briggs, J., Hoch, G. & Johnson, L. (2002) Assessing the rate, mechanisms, and consequences of the conversion of tallgrass prairie to Juniperus virginiana forest. Ecosystems, 5, 578586.
  • Buchman, R.G., Pederson, S.P. & Walters, N.R. (1983) A tree survival model with application to species of the Great Lakes region. Canadian Journal of Forest Research, 13, 601608.
  • Burns, R.M. & Honkala, B.H., tech. coords. (1990) Silvics of North America: 2. Hardwoods (Vol 2), Agriculture Handbook 654. U.S. Department of Agriculture, Forest Service, Washington, DC.
  • Canadell, J., Le Quéré, C.L., Raupach, M., Field, C., Buitenhuis, E., Ciais, P., Conway, T., Gillett, N.P., Houghton, R.A. & Marland, G. (2007) Contribution to accelerating atmospheric CO2 growth from economic activity, carbon intensity, and efficiency of natural sinks. Proceedings of the National Academy of Sciences, 104, 1886618870.
  • Carrer, M. & Urbinati, C. (2006) Long-term change in the sensitivity of tree-ring growth to climate forcing in Larix decidua. New Phytologist, 170, 861872.
  • Christensen, J.H., Hewitson, B., Busuioc, A., Chen, A., Gao, X., Held, I. et al. 2007: Regional climate projections. Climate change 2007: the physical science basis. Contribution of Working Group I to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change (eds S.Solomon, D.Qin, M.Manning, Z.Chen, M.Marquis, K.B.Avery, M.Tignor & H.L.Miller), pp. 847940. Cambridge University Press, Cambridge, UK/New York, NY, USA.
  • Cook, E. & Holmes, R. (1999). User’s Manual for Program arstan. University of Arizona, Tuscon, AZ, USA.
  • Cooper, L. & Solis, C. (2003) O and 13C in leaf litter versus tree ring cellulose as proxy isotopic indicators of climate change. North American Temperate Deciduous Forest Response to Changing Precipitation Regimes (eds P.Hanson & S.Wullschleger), pp. 140157. Springer Verlag, New York.
  • Cummins, J.F. & Grigal, D.F. (1981). Legend to Map: Soils and Land Surfaces of Minnesota 1980, Rep. No. Soil Series No. 110, Miscellaneous Publication 11. Department of Soils Science, University of Minnesota, St Paul, MN.
  • Danner, B.T. & Knapp, A.K. (2001) Growth dynamics of oak seedlings (Quercus macrocarpa Michx. and Quercus muhlenbergii Engelm.) from gallery forests: implications for forest expansion into grasslands. Trees, 15, 271277.
  • Davis, M., Reich, P., Knoll, M., Dooley, L., Hundtoft, M. & Attleson, I. (2007) Elevated atmospheric CO2: a nurse plant substitute for oak seedlings establishing in old fields. Global Change Biology, 13, 23082316.
  • Dean, W.E., Ahlbrandt, T.S., Anderson, R.Y. & Bradbury, J.P. (1996) Regional aridity in North America during the middle Holocene. The Holocene, 6, 145155.
  • Efron, B. & Tibshirani, R. (1993) An Introduction to the Boot-Strap. Chapman & Hall, New York.
  • Feng, X. (1999) Trends in intrinsic water-use efficiency of natural trees for the past 100–200 years: a response to atmospheric CO2 concentration. Geochimica et Cosmochimica Acta, 63, 18911903.
  • Ferrio, J.P., Florit, A., Vega, A., Serrano, L. & Voltas, J. (2003) Δ13C and tree-ring width reflect different drought responses in Quercus ilex and Pinus halepensis. Oecologia, 137, 512518.
  • Fowler, A. & Boswijk, G. (2003) Chronology stripping as a tool for enhancing the statistical quality of tree ring chronologies. Tree-Ring Research, 59, 5362.
  • Galatowitsch, S., Frelich, L. & Phillips-Mao, L. (2009) Regional climate change adaptation strategies for biodiversity conservation in a midcontinental region of North America. Biological Conservation, 142, 20122022.
  • Gedney, N., Cox, P.M., Betts, R.A., Boucher, O., Huntingford, C. & Stott, P.A. (2006) Detection of a direct carbon dioxide effect in continental river runoff records. Nature, 439, 835838.
  • Graumlich, L.J. (1991) Subalpine tree growth, climate and increasing CO2: an assessment of recent growth trends. Ecology, 72, 111.
  • Guertin, D.S., Easterling, W.E. & Brandle, J. (1997) Climate change and forests in the Great Plains: issues in modeling fragmented woodlands in intensively managed landscapes. BioScience, 47, 287295.
  • Guo, Q., Brandle, J., Schoeneberger, M. & Buettner, D. (2004) Simulating the dynamics of linear forests in Great Plains agroecosystems under changing climates. Canadian Journal of Forest Research, 34, 25642572.
  • Hamilton, J.D. & Johnson, S. (2002) Playing with Fire: Climate Change in Minnesota, 3rd edn. Minnesotans for and Energy-Efficient Economy, St Paul, MN, USA.
  • Hanson, P., Wullschleger, S.D., Norby, R., Tschaplinski, T.J. & Gunderson, C. (2005) Importance of changing CO2, temperature, precipitation, and ozone on carbon and water cycles of an upland-oak forest: incorporating experimental results into model simulations. Global Change Biology, 11, 14021423.
  • Heitholt, J.J., Johnson, R.C. & Ferris, D.M. (1991) Stomatal limitation to carbon dioxide assimilation in nitrogen and drought-stressed Wheat. Crop Science, 31, 135139.
  • Holmes, R. (1999) User’s Manual for Program cofecha. University of Arizona, Tuscon, AZ, USA.
  • Huang, J.-G., Bergeron, Y., Denneler, B., Berninger, F. & Tardif, J. (2007) Response of forest trees to increased atmospheric CO2. Critical Reviews in Plant Sciences, 26, 265283.
  • Keeling, C.D. & Whorf, T.P. (2005) Atmospheric CO2 Records from Sites in the SIO Air Sampling Network. Trends: A Compendium of Data on Global Change. Oak Ridge National Laboratory, Oak Ridge, TN.
  • Kempes, C.P., Myers, O.B., Breshears, D.D. & Ebersole, J.J. (2008) Comparing response of Pinus edulis tree-ring growth to five alternate moisture indices using historic meteorological data. Journal of Arid Environments, 72, 350357.
  • Kling, G.W., Hayhoe, K., Johnson, L.B., Magnuson, J.J., Polasky, S., Robinson, S.K. et al. (2003) Confronting Climate Change in the Great Lakes Region: Impacts on our Communities and Ecosystems. Union of Concerned Scientists and Ecological Society of America, Washington, DC.
  • Knapp, P., Soulé, P. & Grissino-Mayer, H. (2001) Detecting potential regional effects of increased atmospheric CO2 on growth rates of western juniper. Global Change Biology, 7, 903917.
  • Knapp, P.A., Soulé, P.T. & Grissino-Mayer, H.D. (2004) Occurrence of sustained droughts in the interior Pacific Northwest (A.D. 1733-1980) inferred from tree-ring data. Journal of Climate, 17, 140150.
  • Kobe, R.K. (1996) Intraspecific variation in sapling mortality and growth predicts geographic variation in forest composition. Ecological Monographs, 66, 181201.
  • Kobe, R.K., Pacala, S.W., Silander, J.A. & Canham, C.D. (1995) Juvenile tree survivorship as a component of shade tolerance. Ecological Applications, 5, 517532.
  • Leal, S., Eamus, D., Grabner, M., Wimmer, R. & Cherubini, P. (2008) Tree rings of Pinus nigra from the Vienna basin region (Austria) show evidence of change in climatic sensitivity in the late 20th century. Canadian Journal of Forest Research, 38, 744759.
  • Leavitt, S.W., Idso, S.B., Kimball, B.A., Burns, J.M., Sinha, A. & Stott, L. (2003) The effect of long-term atmospheric CO2 enrichment on the intrinsic water-use efficiency of sour orange trees. Chemosphere, 30, 217222.
  • Leuzinger, S. & Korner, C. (2007) Water savings in mature deciduous forest trees under elevated CO2. Global Change Biology, 13, 24982508.
  • Li, J., Gou, X., Cook, E.R. & Chen, F. (2006) Tree-ring based drought reconstruction for the central Tien Shan area in northwest China. Geophysical Research Letters, 33, L07715.1L07715.5.
  • Li, J., Chen, F., Cook, E.R., Gou, X. & Zhang, Y. (2007) Drought reconstruction for North Central China from tree rings: the value of the Palmer drought severity index. International Journal of Climatology, 27, 903909.
  • Loehle, C. (2000) Forest ecotone response to climate change: sensitivity to temperature response functional forms. Canadian Journal of Forest Research, 30, 16321645.
  • Loehle, C. & LeBlanc, D. (1996) Model-based assessments of climate change effects on forests: a critical review. Ecological Modeling, 90, 131.
  • Long, S., Ainsworth, E., Rogers, A. & Ort, D. (2004) Rising atmospheric carbon dioxide: plants FACE the future. Annual Review of Plant Biology, 55, 591628.
  • Pacala, S.W. & Hurtt, G.C. (1993) Terrestrial vegetation and climate change: integrating models and experiments. Biotic Interactions and Global Change (eds J.Kingslover, P.Kareiva & R.Huey), pp. 5774. Sinaur Associates Inc, Sunderland, MA.
  • Pastor, J. & Post, W.M. (1988) Response of northern forests to CO2-induced climate change. Nature, 334, 5558.
  • Peñuelas, J., Hunt, J.M., Ogaya, R. & Jump, A.S. (2008) Twentieth century changes of tree-ring δ13C at the southern range-edge of Fagus sylvatica: increasing water-use efficiency does not avoid the growth decline induced by warming at low altitudes. Global Change Biology, 14, 10761088.
  • Post, W.M. & Pastor, J. (1996) Linkages – an individual-based forest ecosystem model. Climatic Change, 34, 253261.
  • Prasad, A.M., Iverson, L.R., Matthews, S. & Peters, M. (2007-ongoing) A Climate Change Atlas for 134 Forest Tree Species of the Eastern United States [database]. Northern Research Station, USDA Forest Service, Delaware, OH. Available at:
  • Reynolds-Henne, C.E., Siegwolf, R.T.W., Treydte, K.S., Esper, J., Henne, S. & Saurer, M. (2007) Temporal stability of climate-isotope relationships in tree rings of oak and pine (Ticino, Switzerland). Global Biogeochemical Cycles, 21, GB4009.
  • Risch, A., Heiri, C. & Bugmann, H. (2005) Simulating structural forest patterns with a forest gap model: a model evaluation. Ecological Modelling, 181, 161172.
  • Roden, J. & Ehleringer, J. (2007) Summer precipitation influences the stable oxygen and carbon isotopic composition of tree-ring cellulose in Pinus ponderosa. Tree Physiology, 27, 491501.
  • Sokal, R. & Rohlf, F.J. (1995) Biometry, 3rd ednn W. H. Freeman and Company, New York. 887 p.
  • Soulé, P. & Knapp, P. (2006) Radial growth rate increases in naturally occuring ponderosa pine tree: a late-20th century CO2 fertilization effect? New Phytologist, 171, 379390.
  • Soulé, P., Knapp, P. & Grissino-Mayer, H.D. (2004) Human agency, environmental drivers, and western juniper establishment during the late holocene. Ecological Applications, 14, 96112.
  • Speer, J.H., Grissino-Mayer, H.D., Orvis, K.H. & Greenberg, C.H. (2009) Climate response of five oak species in the eastern deciduous forest of the southern Appalachian Mountains, USA. Canadian Journal of Forest Research, 39, 507518.
  • Stahle, D., Fye, F., Cook, E. & Griffin, R. (2007) Tree-ring reconstructed megadroughts over North America since A.D. 1300. Climatic Change, 83, 133149.
  • Stokes, M. & Smiley, T. (1996) An Introduction to Tree-Ring Dating. The University of Arizona Press, Tuscon, AZ, USA.
  • Tardif, J.C. & Conciatori, F. (2006) Influence of climate on tree rings and vessel features in red oak and white oak growing near their northern distribution limit, southwestern Quebec, Canada. Canadian Journal of Forest Research, 36, 23172330.
  • Umbanhowar C.E., Jr, Camill, P., Geiss, C.E. & Teed, R. (2006) Asymmetric vegetation responses to mid-Holocene aridity at the prairie–forest ecotone in south-central Minnesota. Quaternary Research, 66, 5366.
  • Underwood, C.A. (2007) Century-scale trends in climatic variability for the Pacific Northwest from western juniper (Juniperus occidentalis Hook. ssp. occidentalis) tree-ring data. MS Thesis, The University of Tennessee, Knoxville, TN.
  • Voelker, S., Muzika, R.-M., Guyette, R. & Stambaugh, M. (2006) Historical CO2 growth enhancement declines with age in Quercus and Pinus. Ecological Monographs, 76, 549564.
  • Wang, G.G., Chhin, S. & Bauerle, W. (2006) Effect of natural atmospheric CO2 fertilization suggested by open-grown white spruce in a dry environment. Global Change Biology, 12, 601610.
  • Wigley, T.M.L., Briffa, K.R. & Jones, P.D. (1984) On the average value of correlated time series, with applications in dendroclimatology and hydrometeorology. Journal of Applied Meteorology, 23, 201213.
  • Williams, J.W., Shuman, B. & Bartlein, P.J. (2009) Rapid responses of the prairie–forest ecotone to early Holocene aridity in mid-continental North America. Global and Planetary Change, 66, 195207.
  • Wu, F., Bao, W., Li, F. & Wu, N. (2008) Effects of drought stress and N supply on the growth, biomass partitioning and water-use efficiency of Sophora davidii seedlings. Environmental and Experimental Botany, 63, 248255.
  • Wullschleger, S.D., Gunderson, C.A., Tharp, M.L., West, D.C. & Post, W.M. (2003) Simulated patterns of forest succession and productivity as a consequence of altered precipitation. North American Temperate Deciduous Forest Responses to Changing Precipitation Regimes (eds P.Hanson & S.Wullschleger), pp. 433446. Springer-Verlag, New York.
  • Wunder, J. (2007) Conceptual advancement and ecological applications of tree mortality models based on tree-ring and forest inventory data. PhD Thesis, Swiss Federal Institute of Technology, Zurich.
  • Wunder, J., Reineking, B., Matter, J.-F., Bigler, C. & Bugmann, H. (2007) Predicting tree death for Fagus sylvatica and Abies alba using permanent plot data. Journal of Vegetation Science, 18, 525534.
  • Wyckoff, P.H. & Clark, J.S. (2000) Predicting tree mortality from diameter growth: a comparison of maximum likelihood and Bayesian approaches. Canadian Journal of Forest Research, 30, 156167.
  • Wyckoff, P.H. & Clark, J.S. (2002) The relationship between growth and mortality for seven co-occuring tree species in the southern Appalachian Mountains. Journal of Ecology, 90, 604615.
  • Xiaodong, Y. & Shugart, H.H. (2005) FAREAST: a forest gap model to simulate dynamics and patterns of eastern Eurasian forests. Journal of Biogeography, 32, 16411658.