Dendroclimatic analysis of white pine (Pinus strobus L.) using long-term provenance test sites across eastern North America

If fossil fuel intensive energy generation continues to grow at current rates, atmospheric concentrations of carbon dioxide (CO₂) and other greenhouse gases will increase dramatically. Continued high rates of emissions over the next century could raise mean annual global surface temperature by 2.5 °C to 6.5 °C by 2100 (IPCC 2007). The resulting projected change in spatial and temporal temperature and precipitation regimes, increased frequency of natural disturbance events, and atmospheric CO₂ enrichment is expected to alter the distribution, structure, function, and productivity of northern forests (Mickler et al. 2000; Parmesan and Yohe 2003).

A critical need exists to develop ecologically sustainable forest management approaches that incorporate climate change adaptation to sustain carbon sequestration, fiber production, and other ecological goods and services provided by managed forests (Millar et al. 2007; Alexander and Perschel 2009). The success of such efforts depends in part on knowledge of genetic variation in climatic response of tree species (Rehfeldt et al. 1999; Millar et al. 2007; Wang et al. 2010). Therefore, information generated from long-term provenance tests has substantial potential to improve understanding of tree species responses and adaptability to a rapidly changing climate (Rehfeldt et al. 1999; Wang et al. 2010).

Eastern white pine (Pinus strobus L.) is an ecologically and economically important softwood species that occurs throughout much of eastern North America (Wendel and Smith 1990). White pine was a dominant component of pre-settlement forests of eastern North America that is highly valued for its exceptional wood properties. The effects of large-scale exploitive logging and catastrophic fire in the 18th and 19th centuries, coupled with more recent institutional fire suppression and damage and mortality from white pine blister rust (Cronartium ribicola J.C. Fisch.), white pine weevil (Pissodes strobi (Peck)), and root rot fungi (Armillaria mellea (Vahl) P. Kumm) have dramatically reduced the abundance of white pine on the landscape (Abrams et al. 2000).

Genetic adaptation refers to genetic differences or variation among provenances in the natural range of a tree species that have evolved over several generations and ensure it is in sync with the environmental conditions of its location (White et al. 2007; King et al. 2013; Gray et al. 2016). Examples of genetic adaptation in trees include the link between provenance location and tree growth rate and phenology, i.e., tree provenances from more northern latitudes generally grow more slowly and start growing earlier in spring/stop earlier in fall compared to those from more southern latitudes. White pine typically exhibits significant adaptive genetic variation in response to broad climatic gradients, the product of natural selection and genetic processes (Wright 1970; Genys 1987), and as a result is found as a component of most upland forest types in its range. This variation has resulted in tree species populations that are genetically adapted to local and regional environmental conditions. However, as the climate continues to change, white pine may become increasingly unsynchronized with and maladapted to the prevailing climate, resulting in decreased vigor, productivity, and wood quality. The geographic shifts and reductions in suitable climatic habitat projected by climate change models (McKenney et al. 2007; Iverson et al. 2008; Joyce and Rehfeldt 2013) represent a serious threat to white pine throughout much of eastern North America, especially as changes in climate will also influence other abiotic and biotic stressors (Millar et al. 2007).

A network of provenance test sites established in the early 1960s has provided insight into the adaptive genetic variation of white pine (Wright 1970; Genys 1987; Joyce and Rehfeldt 2013; Zalesny and Headlee 2015). This scientific resource has renewed value as a long-term experiment that can increase understanding of the effect of climate change on white pine tree populations growing outside the climate envelope to which they are naturally adapted (Wang et al. 2006). For example, it is known that southern sources of white pine grow faster but are more susceptible to frost damage when planted farther north (Joyce and Rehfeldt 2013). Combining such knowledge with constantly improving global climate models and climate change scenarios can inform seed deployment decisions to produce nursery stock that is best adapted to a given planting site (Joyce and Rehfeldt 2013; Yang et al. 2015). In addition, such knowledge can guide management of natural regeneration where local provenances are currently well adapted to projected future climate conditions.

Dendrochronological methods can be effectively applied to describe the genetic adaptation of tree provenances to environmental conditions (e.g., Chhin 2010; McClane et al. 2011; Pluess and Weber 2012; King et al. 2013; Eilmann et al. 2014; Chhin 2015) and to examine geographic variation in response to climate (e.g., Chen et al. 2010). In response to seasonal changes in climatic drivers of tree phenology, the annual growth cycle of temperate and boreal tree species in North America alternates between phases of summer growth and winter dormancy (Fritts 1976). Examination of the seasonal periodicity of climatic effects on radial stem growth processes through tree-ring analysis (dendrochronology) provides insight into the seasonal timing of radial growth–climate relationships (Fritts 1976; Chhin et al. 2008) and therefore can be used to quantify genetic adaption of tree provenances to environmental conditions (Chhin 2010; Pluess and Weber 2012; Chhin 2015).

The study was conducted at 7 field test sites belonging to a range-wide white pine provenance trial established in the early 1960s in the eastern United States and Canada (Wright et al. 1970) (Fig. 1; Table 1). Thirteen white pine provenances were evaluated at Wabeno, Wisconsin, USA; Manistique, Michigan, USA; and Pine River, Michigan, USA (Table 2). The Penobscot, Maine, USA provenance was absent at the Newaygo, Michigan, USA test site. The Newaygo provenance was not tested at Turkey Point, Ontario, Canada; Ganaraska, Ontario, Canada; and Orono, Maine, USA test sites. The field sampling design varied at each test site, with a uniform spacing of 2.13 m × 2.13 m. The field design at the Wabeno, Manistique, and Pine River test sites were planted in 24 blocks with one tree per provenance per block. At the Newaygo and Ganaraska test sites, white pine was planted in 4 blocks comprising 81 trees per provenance (arranged in 9 rows of 9 trees) per block. The Turkey Point test site consisted of 4 blocks, each block containing 11 rows of 11 trees per provenance. At the Orono test site, white pine was planted in 12 blocks and each block contained each provenance planted in 4-tree row plots (King and Nienstaedt 1969; Zalesny and Headlee 2015).

Location	Longitude (°W)	Latitude (°N)	Elevation (m)	Range in number of trees sampleda
Wabeno, Wisconsin	−88.50	45.40	438	2–10
Manistique, Michigan	−86.40	46.00	232	5–14
Pine River, Michigan	−85.80	44.20	268	4–18
Newaygo, Michigan	−85.70	43.50	273	27–28
Turkey Point, Ontario	−80.45	42.70	214	27–30
Ganaraska, Ontario	−78.73	44.95	307	26–30
Orono, Maine	−68.65	44.88	34	9–13

^aRange in number of trees sampled is across all provenances at each test site

Provenance origin	Longitude (°W)	Latitude (°N)	Elevation (m)	Canada seed source #	Canada population #	United States seed source #
Union, Georgia	−84.05	34.77	876	1	272	1633
Greene, Tennessee	−82.80	36.00	625	2	273	1634
Monroe, Pennsylvania	−75.42	41.08	585	3	274	1640
Franklin, New York	−74.30	44.40	702	4	275	1639
Penobscot, Maine	−68.60	44.90	33	5	276	1638
Ashland, Ohio	−82.30	40.80	394	6	277	1632
Allamakee, Iowa	−91.50	43.30	350	7	278	1624
Cass, Minnesota	−94.50	47.40	393	8	279	1622
Forest, Wisconsin	−88.90	45.80	500	9	280	1623
Lunenburg, Nova Scotia	−64.60	44.40	95	10	281	1637
Pontiac, Quebec	−77.00	47.40	366	11	282	1635
Algoma, Ontario	−82.60	46.40	365	12	283	1636
Newaygo, Michigan	−85.70	43.50	273	13	345	1670

Two increment cores were collected at breast height (i.e., 1.37 m) from a subset of trees that had a diameter at breast height (dbh) at or near the median value of all living trees measured in a plot. All increment cores used in this study were taken at dbh in the east cardinal direction using a 5-mm diameter borer. The number of living trees per provenance available for sampling (Table 1) differed due to variable tree mortality among test sites. The cores were permanently mounted and sanded using standard dendrochronology procedures to prepare them for radial growth trend analysis (Stokes and Smiley 1996). Scanned images of individual cores were acquired with a flatbed scanner at a resolution of 720 dpi. Cores and scanned images were visually cross-dated. Ring-widths were measured from the scanned images using an image analysis system (WinDENDRO, Regent Instruments, Québec, Canada). Statistical quality control was conducted using COFECHA software to ensure that missing and false rings were accounted for (Holmes 1983; Grissino-Mayer 2001). To remove non-climatic effects, cross-dated ring-width measurements were de-trended using a conservative de-trending approach that included fitting a negative exponential curve, simple linear regression, and a line through the mean with ARSTAN software (Cook 1985). The standardized version of the chronology was retained for each individual core sample to prevent the loss of any radial growth signals associated with provenances at different test sites. The standardized radial growth indices were then averaged for individual trees of each provenance at each test site. Chronology statistics, including mean sensitivity and the inter-series correlation coefficient, were determined for each provenance location at each test site.

Principal components analysis (PCA) was used to examine the main modes of variation in the standardized radial growth indices of individual provenances at each test site (Graumlich 1993; Legendre and Legendre 1998). The PCA was carried out using the Factor Analysis function in Systat (ver. 13.0, released 2009), which also included selection of a Varimax rotation to improve interpretation of the principal component axes. Selected principal component axes explained more variation than expected, based on a broken stick null model (Legendre and Legendre 1998).

Year scores from 1976 to 2009 were used in dendroclimatic analyses to examine shared provenance radial growth responses to test site climate. For test sites in the United States, primary climate data were obtained from PRISM (Daly 2008) and for test sites in Canada were obtained from ClimateNA (Hamann et al. 2013). Climate data included minimum, maximum, and mean monthly temperature and total monthly precipitation. These primary variables were used as the basis to produce a synthetic monthly climatic moisture index (CMI) variable calculated as total monthly precipitation minus a measure of monthly potential evapotranspiration (Hogg 1997). Since tree radial growth may respond more strongly when climatic variables are expressed at a seasonal compared to a monthly scale, monthly climatic variables were converted to seasonal 3-month periods (i.e., quarters) by averaging across months for temperature-related variables and summing across months for precipitation and moisture index variables. Because climatic conditions in the previous year can influence radial growth in the following growing season, the range of monthly and seasonal periods was examined from April of the prior year (t-1) to October of the current year (t) of ring formation.

Dendroclimatic analysis was conducted to examine the relationship between year scores representing the main modes of variation (i.e., the first (PC1) and second (PC2) principal component axes) shared across provenances at each test site with respect to the monthly and seasonal climate at each test site. Dendroclimatic analysis was conducted using a step-wise multiple regression conducted with R statistical analysis software for the period 1976 to 2009 (Venables and Ripley 2002; Chhin et al. 2008; R Core Team 2014). Standardized partial regression coefficients were also calculated to rank the relative importance of the predictor variables in each regression model (Zar 1999).

Natural populations of white pine have been examined in prior dendroclimatic and dendroecological studies in eastern North America (Graumlich 1993; Abrams et al. 2000; Black and Abrams 2005; Kilgore and Telewski 2004; Kipfmueller et al. 2010, Chhin et al. 2013). This study represents the first dendroclimatic assessment of a long-term range-wide provenance study of white pine, which consisted of 7 test sites in the United States and Canada and 13 provenances obtained from across the species’ natural range. The results revealed complex radial growth responses to climatic and geographical variables among and within provenances of white pine.

The future regional distribution, abundance, and productivity of white pine and white pine forest types under a changing climate have been projected using a variety of modelling approaches and climate change scenarios (McKenney et al. 2007; Joyce and Rehfeldt 2013; Peters et al. 2013; Handler et al. 2014; Iverson et al. 2008, 2017), each having a different structure, limitations, assumptions, and levels of uncertainty (Iverson et al. 2017). Bioclimatic, or species distribution models, have typically shown the area of suitable climatic habitat for white pine to decrease and/or move northward by the end of the century, with larger changes under higher greenhouse gas emissions scenarios, but the degree of change varies with region (McKenney et al. 2007; Joyce and Rehfledt et al. 2013; Iverson et al. 2008, 2017). For example, the future abundance of white pine in three forest regions of the United States was unchanged, or experienced small to large reductions by the end of the century, depending on the forest region and the model used, with projected changes being larger in scenarios where future greenhouse gas emissions remain comparatively high and the climate changes more rapidly (Iverson et al. 2017). In the northern portion of its natural range, white pine may experience modest increases in growth during this century under a slightly warmer climate (Peters et al. 2013; Handler et al. 2014; Yang et al. 2015). Conversely, white pine growing near its warmer range limit, may be poorly adapted to increasing temperatures and suffer significant reductions in growth and survival under climate change (Joyce and Rehfeldt 2013; Yang et al. 2015).

Future climate change presents both challenges and opportunities for management of white pine. The current study indicates potential vulnerabilities of white pine radial growth in the northern portion of its range to higher temperatures, summer moisture stress, and crown damage by high winds and heavy snowfall. Projected future increases in temperature, summer drought stress, and higher frequencies of storm activity (IPCC 2007) in this region will require efforts to both adapt management practices to the future climate and enhance conservation of genetic resources of white pine. On the other hand, the current study suggests that white pine radial growth at some test site locations has been historically limited by cold spring and fall conditions. Since future climatic warming will likely increase the length of the growing season, increased growing season temperature could potentially offset these growth limitations in colder regions of its natural range. These results are consistent with those of Yang et al. (2015) that reported optimal height growth of young white pine occurs at an AMT of about 11 °C, and that white pine in relatively colder locations may benefit, at least initially, from modest climate warming.