Predicted Responses of Genetically Improved Populations to Climate Changes Based on Second-Cycle Douglas-Fir Progeny Tests

Abstract

The current planting of economically important timber species, such as Douglas-fir, mainly relies on genetically improved seeds from seed orchards. However, published research on the effects of climate change has largely focused on natural populations. To bridge this gap, data from 80 cooperative second-cycle coastal Douglas-fir progeny tests across eight breeding zones in western Washington and Oregon were analyzed. Climate transfer functions for age-12 growth were derived, showing significant results for the US Pacific Northwest. Region-specific transfer functions (Coast, Inland, and Cascade) displayed stronger correlations. Mean annual temperature and mean coldest month temperature were the most important climatic variables explaining growth. The study found that populations from slightly warmer areas tended to grow better but moving populations from colder to warmer areas by 2 °C (analogous to projected global warming) would result in an 8% genetic loss in age-12 height and a 25% genetic loss in age-12 volume. However, substantial diversity in climatic response was found among full-sib families within large breeding zones, suggesting that breeding and selecting suitable families for future climatic conditions within breeding zones is feasible. The study discusses potential strategies to adapt current breeding programs to address the impacts of future climate change while maintaining high population growth rates in Douglas-fir breeding programs.

1. Introduction

Over the past two decades, western North America has experienced rapid warming and prolonged drought periods [1,2,3,4]. Climate change projections indicate a continuation and intensification of these trends [5,6,7,8], with significant implications for tree growth and forest ecosystem dynamics in the region [9,10,11]. Douglas-fir (Pseudotsuga menziesii var. menziesii), a keystone species of great ecological and economic importance in western North America, is susceptible to these environmental shifts due to its extended lifespan and the protracted timescales required for population-level adaptation to novel climatic conditions [12,13].

The complex interactions between climatic variables and Douglas-fir growth and survival have been studied extensively. Evidence indicates that Douglas-fir growth is significantly influenced by temperature and precipitation patterns, with a positive correlation for precipitation and a negative correlation for temperature during the growing season [14,15,16,17]. Drought has long been recognized as a major limiting factor for the growth and survival of Douglas-fir [18,19]. Additionally, cool fall temperatures are essential to start cold hardening in Douglas-fir, a critical process for winter survival [18,20]. As climate change alters seasonal climatic patterns, concerns about the potential impacts on Douglas-fir growth and overall fitness are increasing. These shifting environmental dynamics have led researchers and forest managers to reassess traditional approaches to seed sources and breeding zones, which have historically been based on static environmental conditions [21,22].

Transfer functions have been widely adopted in forestry research to model the impacts of climate change on forest population growth and survival [23,24,25]. Two main methods are widely employed to investigate and predict climate change effects on forest productivity: long-term provenance tests and short-term common garden nursery trials. Provenance testing, which involves evaluating different geographic populations under varied climatic conditions, has been extensively used to examine population maladaptation resulting from climate change [26,27,28]. Although comprehensive provenance trials for coastal Douglas-fir in the US Pacific Northwest (PNW) are limited, existing research has yielded valuable insights. Aitken et al. [29] proposed that local adaptation is substantial in temperate and boreal tree populations, while Ye and Jayawickrama [30] concluded from an earlier Douglas-fir provenance trial that seed sources could be transferred considerable distances without adverse effects on rotation-age growth. An alternative approach to modeling phenotypic genetic variation is offered by common garden trials, which involve the cultivation of open-pollinated progeny from natural stands [31,32].

The planting of commercially important timber species, including Douglas-fir, is primarily based on genetically improved seeds from seed orchards on public and private forest lands [33,34]. However, most studies on the effects of climate changes have been focused on natural populations, mainly due to the availability of existing large-scale and long-term provenance tests [23,35,36,37]. Family-level studies in common garden trials mainly used open-pollinated seeds from natural stands [27,31,32,38,39,40]. Careful interpretation is apparently required when extrapolating results from natural populations to genetically improved populations, as human selection through tree breeding can significantly alter the natural evolutionary trajectory of tree populations. To our knowledge, there is a notable scarcity of academic research on the climate-driven adaptation of genetically improved populations across extensive geographic regions. This gap in research is likely due to the fact that most existing progeny tests are restricted within a breeding zone with limited geographic and climatic scope.

Douglas-fir tree improvement programs in the PNW have a rich history spanning seven decades [41,42]. The Northwest Tree Improvement Cooperative (NWTIC) plays a crucial role in supporting an extensive network of progeny tests, which includes more than 3000 second-cycle full-sib families on 132 test sites [43,44]. The rigorous management and measurement of these trials, combined with the climatic variability resulting from topographic differences in the PNW, could serve as an invaluable resource for climate transfer studies. This is particularly significant given the relative paucity of comprehensive provenance trials and studies on genetically improved populations of coastal Douglas-fir in the region.

The objective of this study was threefold: (1) to model and predict the response of Douglas-fir breeding populations to climate variables and climate change through the use of transfer functions at both state and regional levels in the PNW; (2) to investigate the variation among full-sib families in their adaptability to climate changes; and (3) to offer practical guidelines for existing coastal Douglas-fir breeding programs in the region, aiming to balance adaptation and growth potential amid evolving environmental conditions.

2. Materials and Methods

2.1. Test Materials and Experimental Designs

The research analyzed data from 80 second-cycle progeny tests of Douglas-fir established between 1999 and 2011. These trials were selected from a larger set of 132 tests, including only those with appropriately aged data. The study spanned eight breeding zones (second breeding cycle) across western Washington (WA) and western Oregon (OR), as outlined in

Table 1. List of second-cycle Douglas-fir testing programs (or breeding zones).

Breeding Zone	State	Number of Trials	Number of Parents	Number of Full-Sib Families	Spacing (m)	Measurement Age
Washington Coast	WA	5	132	106	2.4	12
Puget Sound	WA	9	331	292	2.4/2.8	12
Washington Cascades	WA	16	312	290	2.4	12
Trask Coast	OR	11	347	295	2.4	11
Trask Inland	OR	10	293	232	2.4	11
NOCTIC	OR	10	442	378	2.4	11/12
Coos Bay CL98	OR	4	194	280	1.8	10
South Central Coast	OR	15	321	342	2.8	10/12

.

The tested full-sib families were mainly produced from parents and forward selections within first-cycle breeding zones, ensuring effective inbreeding management and the preservation of locally adapted gene complexes. Among the 1517 parents tested, 1485 of them were tested in two or more breeding zones. Additional information about the testing program and breeding strategies can be found in Jayawickrama and Ye [42].

Field trials were conducted across a latitudinal gradient of 42.81° N to 48.95° N, a longitudinal gradient of 121.62° W to 124.45° W, and an elevational range of 200 to 1000 m. Each trial was set up in an “alpha” design, wherein full replicates were divided into smaller, more homogeneous, incomplete blocks to decrease the residual variation among plots and enhance the accuracy of genotype mean estimation. All trials used single-tree plots, with the tree spacing ranging from 2.4 × 2.4 m to 2.8 × 2.8 m (apart from CL98, which had a spacing of 1.8 × 1.8 m). Each breeding zone was associated with a single testing program (multi-environmental trial, or MET), which typically involved the testing of 100 to 380 full-sib families across 4 to 16 distinct sites.

From 2010 to 2020, data were collected on the total height (HT), diameter at breast height (DBH), and branching characteristics of all living trees aged 10 to 12 years. The stem volume index (VOL) was approximately calculated as HT × DBH². This study primarily focuses on the height growth of the trees.

2.2. Climate Data

Decadal climate data from 2011 to 2020 for each test site and parent were acquired using ClimateNA software version 7.30 [45], based on their respective latitude, longitude, and elevation information. The collected data consist of the following climate variables: mean annual temperature (MAT), mean coldest month temperature (MCMT), mean annual precipitation (MAP), summer heat/moisture index (SHM), and mean warmest month temperature (MWMT).

In addition, the MAT and MAP data for future periods, including the 2055s (2041–2070) and 2085s (2071–2100), were projected using the 8-GCM ensembles in ClimateNA software.

2.3. Population Climate Transfer Function

2.3.1. Construction of Virtual Populations

Classifying populations based on geographic factors does not effectively define the climatic envelopes of second-cycle parent trees due to the variability within geographic regions. Consequently, we grouped the families according to climatic clusters to enable more meaningful inferences at the population level. The k-means clustering algorithm [46] was used to divide all tested parents into 20 virtual populations based on their standardized MAT and MAP. This algorithm operates by identifying optimal positions for cluster centroids, ensuring that data points within a cluster are closer to their assigned centroid than to any other centroid. These virtual populations were tailored for each area of interest or group of breeding zones. The Python library scikit-learn [47] was used for this process.

2.3.2. Assessment of Population Growth Relative to the Local Source in Each Trial

We took multiple steps to estimate population growth rate relative to the local source.

• Predicted parental and population performance: Parental breeding values were predicted using a mixed-model approach (family model), as outlined by Isik et al. [48]. The analyses were conducted with the ASReml software version 4.2 [49]. These breeding values were subsequently used to evaluate the growth performance of the parents. The population performance of growth ($Y$) was determined by averaging the breeding values of all parents within the virtual population.
• Climate transfer distance: Climate transfer distance (CTD) was calculated for each virtual population as: CTD = (site climate) − (population mean climate). Thus, local populations are the population with zero CTDs, and negative CTD values represent the transfer of a population from a warmer or wetter climate to one cooler or drier.
• Individual transfer function and relative growth performance: The individual transfer function is a mathematical function frequently used to describe the relationship between population growth and climate transfer distance [25,26,37,50,51]. In this study, we opted to use a stepwise quadratic regression approach to fit a transfer function, i.e., population growth = f (CTD), where f is a quadratic function that was constructed step by step through interactive means, and only independent variables that displayed statistical significance were retained. Finally, we computed the relative growth performance of each virtual population as $Y_{rel}=Y/Y_{local}$, where $Y_{local}$ denotes the fitted value of the population at zero CTD from the transfer function. The process of model fitting was carried out using the Python library statsmodels [52].

2.3.3. Pooled Transfer Function across Trials

Several scholars have cautioned that individual transfer functions should be used judiciously, as they may only be applicable to specific geographic locations and cannot be easily generalized to other regions [23,50,51]. As a more robust alternative to individual transfer functions, many researchers have recommended using a “pooled” transfer function [25,28,50,53]. However, if pooling is to be implemented, it has been suggested that the standardization of population responses is essential to account for productivity differences among test sites [25,53].

We followed the method of O’Neill and Nigh [53], which involves creating transfer functions using CTD as the predictor variable and $Y_{rel}$ as the response variable. The population climate was initially incorporated into the models to account for potential variations in responses across different CTD ranges; however, it was subsequently excluded due to a lack of statistical significance.

By definition, the expected $Y_{rel}$ value is 1 at zero CTD. However, when the sample size is not large enough, the fitted curve may not align with this expectation perfectly. To address this issue, constrained quadratic regressions were used instead of regular ones, ensuring that each regression curve passed through point (0, 1). The pooled model was fitted using the curve_fit function from the Python library scipy [54]. To guarantee that the fitted curves always passed through point (0, 1), we added this point to the list of points to fit and assigned it a significantly higher weighting compared to the other points. The resulting R² values were very similar between the models with and without constraints.

2.4. Family Adaptation to Climate Changes

A subset of 31 progeny tests was selected from the initial 80 tests. This selection encompassed three second-cycle breeding zones in WA: the Coast, Puget Sound, and Cascades zones. The subset included more than 460 full-sib families that were evaluated. Many of these full-sib families were assessed across more than one breeding zone. The age-12 height measurement data was subjected to the following multivariate family model analysis.

$$\mathbf{y}=\mathbf{X}\mathbf{\tau }+\mathbf{Z}\mathbf{u}+\mathbf{e}$$

(1)

The fixed-effects parameters ($\mathbf{\tau }$) include the overall mean and site effect. The vector of random-effects parameters (u) includes parental GCA, full-sib family SCA, replicate (REP), and block within replicate (BLK) effects. The variances for effects u and e are as follows:

$$\mathbf{V}\mathbf{a}\mathbf{r}\left[\begin{array}{c}\mathbf{u}\\ \mathbf{e}\end{array}\right]=\left[\begin{array}{cc}\mathbf{G}& \mathbf{0}\\ \mathbf{0}& \mathbf{R}\end{array}\right]=\left[\begin{array}{ccccc}{\mathbf{G}}_{rep}& \mathbf{0}& \mathbf{0}& \mathbf{0}& \mathbf{0}\\ \mathbf{0}& {\mathbf{G}}_{blk}& \mathbf{0}& \mathbf{0}& \mathbf{0}\\ \mathbf{0}& \mathbf{0}& {\mathbf{G}}_{gca}& \mathbf{0}& \mathbf{0}\\ \mathbf{0}& \mathbf{0}& \mathbf{0}& {\mathbf{G}}_{sca}& \mathbf{0}\\ \mathbf{0}& \mathbf{0}& \mathbf{0}& \mathbf{0}& \mathbf{R}\end{array}\right]$$

(2)

where 0s are null matrices; ${\mathbf{G}}_{rep}$, ${\mathbf{G}}_{blk}$, ${\mathbf{G}}_{gca}$, and ${\mathbf{G}}_{sca}$ are covariance matrices corresponding to the above random effects, respectively. R is the covariance matrix for residuals. The covariance matrices ${\mathbf{G}}_{rep}$, ${\mathbf{G}}_{blk}$, and R were assumed to have the simple variance component structure with a block-diagonal format for sites: ${\mathbf{G}}_{rep}={⨁}_{i=1}^s{\sigma }_{{rep}_i}^2{\mathbf{I}}_{r_i}$, ${\mathbf{G}}_{blk}={⨁}_{i=1}^s{\sigma }_{{blk}_i}^2{\mathbf{I}}_{b_i}$, and $\mathbf{R}={⨁}_{i=1}^s{\sigma }_{e_i}^2{\mathbf{I}}_{n_i}$ where Is are identical matrices; ${\sigma }^2$s are site-specific variance components; s is the number of sites; r is the number of replicates; b is the number of blocks; and n is the number of trees.

The covariance matrix ${\mathbf{G}}_{gca}$, which combines the main GCA effect and GCA × site interaction effect, was fitted using a factor analytic covariance structure of order 2: ${\mathbf{G}}_{gca}=[{\mathbf{\Lambda }}_1{\mathbf{\Lambda }}_1^{\mathrm{T}}+{\mathbf{\Psi }}_1]\otimes \mathbf{A}$, where ${\mathbf{\Lambda }}_1$ is an s × 2 matrix of site loadings for the GCA effect, ${\mathbf{\Psi }}_1$ is an s × s diagonal matrix with elements referred to as specific variances for the GCA effect, and A is the numerator relationship matrix derived from the pedigree of parents. Similarly, the covariance matrix ${\mathbf{G}}_{sca}$, which combines the main SCA effect and SCA × site interaction effect, was fitted as ${\mathbf{G}}_{sca}=[{\mathbf{\Lambda }}_2{\mathbf{\Lambda }}_2^{\mathrm{T}}+{\mathbf{\Psi }}_2]\otimes \mathbf{I}$, where ${\mathbf{\Lambda }}_2$ is an s × 2 matrix of site loadings for the SCA effect, and ${\mathbf{\Psi }}_2$ is an s × s diagonal matrix with elements referred to as specific variances for the SCA effect. These covariance structures imply that both GCA × site and SCA × site effects are correlated between sites.

The linear mixed model was fitted using ASReml software version 4.2 [49]. Genetic gain, expressed as twice the mid-parent GCA as a percentage of the population mean, was predicted for each full-sib family at each of the 31 test sites. Contour plots illustrating the pattern of predicted gain for each family were generated across a grid of site-specific MAT and MAP.

3. Results

The test sites of the eight second-cycle Douglas-fir breeding zones used in this study displayed a wide variation in climatic conditions (Figure 1), which were not distinctly separated by program (Figure 2a). We interpret this to mean that the breeding zones (generally established according to geographic footprint) would not have constituted distinct climate envelopes within the studied area. However, the climates of test sites in the Coast and Cascade regions were evidently dissimilar, with the Coast sites experiencing higher temperature and more precipitation whereas most of the Cascade sites experienced lower temperature and less precipitation (Figure 2b). The climates of the Inland sites were intermediate between those in the other two regions but more akin to the Cascade region.

Significant transfer functions were obtained for western WA, western OR, and the PNW region with a statistical significance of p ≤ 0.05 (Figure 3). However, the values of goodness of fit (R²) were generally low (≤0.2). Among the five climate variables studied, the two temperature-related variables, MAT and MCMT, exhibited relatively higher R² values in explaining height growth at age 12 compared to the other variables. In WA, for example, populations with a −2.8 °C < CTD for MAT < 0 °C performed better than the local population, with a maximum age-12 height gain of 1.0% at CTD = −1.5 °C. This suggests that moving parents or populations from slightly warmer to cooler areas could result in up to 1% gain in age-12 height growth. Similar patterns were observed in OR and the PNW, where moving parents or populations from areas with ≤3 °C warmer temperatures tended to have up to 1% gain in height. Conversely, the three drought-related climate variables (MAP, SHM, WMWT) in WA seemed to follow the expected “local-is-best” patterns for growth. For example, CTD for MAP ranged from −3100~2900 mm, and little differences in growth were observed between populations from areas that were 1000 mm wetter or drier than the local area. In OR and the PNW, however, moving populations from a slightly wetter or lower summer temperature tended to perform slightly better, although such patterns were very weak (R² ≤ 0.04).

Apart from analyzing the impact of climate on height growth at the state level, we also delved into the matter by region (i.e., Coast, Inland, and Cascade, Figure 4). The transfer function patterns for MAT and MCMT were akin to those observed at the state level, but they exhibited greater strength, as indicated by the higher R² values, with the highest being 0.448. In the Cascade region, populations migrating from areas with up to 3.5 °C higher MAT or 4 °C higher MCMT grew at least as rapidly as the local population. The predicted maximum height gain was approximately 2.3% at a CTD of −1.7 °C for MAT or MCMT, with an R² value greater than 0.4. Climatic gradients in temperature and precipitation are steep in the Cascades, both in western WA and western OR, primarily associated with steep gradients in elevation. In the Coast region, maximum height gains (up to 2%~5%) were predicted at a CTD = −1.8 °C for MAT and −4.5 °C for MCMT, with R² between 0.3 and 0.4. In the Inland region, however, only minor height gains (≤0.3%) could be achieved when moving populations from slightly warmer areas. Regarding the MAP, SHM, and MWMT variables, populations from areas in the Cascade region experiencing moderate summer drought demonstrated better growth, with R² values ranging from 0.2 to 0.3 and maximum height growth gains of 1%~2%. In other regions, transfer functions for these variables were either weak (R² ≤ 0.03) or showed <0.3% maximum height growth gain.

As depicted in the figures, most of the transfer function curves had relatively flat profiles at their peaks. This indicates that a slight deviation of the climate transfer distance from its optimal value is unlikely to produce a substantial reduction in relative growth. When local populations were not the best performers, however, the observed pattern was asymmetrical relative to the local populations, depending on whether the climate transfer distance was positive or negative. For instance, in both the Coast and the Cascade regions, the projections suggest that the relative height growth would not be expected to decrease if the CTD of the MAT were to increase from −3.5 °C to 0 °C (i.e., moving from warmer areas). In contrast, when the CTD of the MAT increased from 0 °C to 2 °C (i.e., moving from colder areas), the relative height growth is anticipated to drop by 5% to 7%.

Climate change projections indicate that the average annual temperature is likely to rise by approximately 0.8 °C to 1.9 °C by 2055, and 1.5 °C to 2.5 °C by 2085, compared to the 2011–2020 decade (as depicted in Figure 5). This anticipated temperature increase over the next few decades will not be uniformly distributed throughout the PNW. The most significant rise, ranging from 1.8 °C to 2.5 °C, is predicted for the WA Cascade and Puget Sound areas. This will be followed by the WA Coast, which is expected to see a temperature increase of 1.5 °C to 2.1 °C. The Southern Coast, Cascade, and Inland regions of OR are forecasted to experience a temperature rise of 1.4 °C to 1.9 °C. The smallest increase, between 1 °C and 1.7 °C, is projected for the northern coast of OR and the state’s southernmost region. The derived regional transfer functions were used to estimate the maximum potential reductions in age-12 height (Figure 4) and stem volume (see Supplementary Figure S1) as a consequence of a warming climate (

Table 2. Projected rises in MAT and the corresponding declines in age-12 growth over the next 30 to 60 years.

Region	Projected Maximum Increase in MAT		Estimated Maximum Gain Loss in Age-12 Height		Estimated Maximum Gain Loss in Age-12 Volume
	2055	2085	2055	2085	2055	2085
WA Cascade, Puget Sound	1.8 °C	2.5 °C	5%	10%	20%	30%
WA Coast	1.5 °C	2.1 °C	4%	5%	15%	22%
OR Southern Coast, Cascade, Inland	1.4 °C	1.9 °C	3.5%	4.6%	11%	18%
OR Northern Coast, Southernmost Region	1.2 °C	1.6 °C	2.5%	3.7%	7%	12%

). The findings indicated that the maximum loss in height could reach up to 5% by 2055 and 10% by 2085, while the maximum reduction in volume could be as much as 20% by 2055 and 30% by 2085. Regarding precipitation (Figure 6), a decrease is anticipated in the WA Cascade and Puget Sound regions, with reductions ranging from 50 to 300 mm, and in southern OR, where it could decrease by 50 to 400 mm. Conversely, other parts of the PNW are projected to see an increase in rainfall. However, this does not necessarily mean that precipitation will rise during the hot, dry summer months when it would have the greatest impact on tree growth and survival.

We selected 50 full-sib families from each of the top, middle, and bottom groups, based on predicted genetic gains for height, to examine responses to climate change at the family level. The climate response patterns of these families were categorized into six groups, labeled A through F (Figure 7). These findings suggest that the full-sib families within the breeding populations may have distinct preferences for climatic conditions that optimize their growth. As summarized in

Table 3. The distribution of climate-response patterns among family groups. The “Top 50”, “Middle 50”, and “Bottom 50” categories represent the top, middle, and bottom 50 full-sib families, respectively, based on the predicted genetic gains in height.

Pattern	Top 50	Middle 50	Bottom 50	All
A	45 (90%)	14 (28%)		59 (39%)
B		14 (28%)	27 (54%)	41 (28%)
C	3 (6%)	6 (12%)	3 (6%)	12 (8%)
D	2 (4%)	6 (12%)	20 (40%)	28 (19%)
E		5 (10%)		5 (3%)
F		5 (10%)		5 (3%)

, most top-performing families (90%) exhibited pattern A, indicating that they are expected to perform strongly at an MAP of approximately 2000 mm across a wide range of MAT. In contrast, most bottom-ranking families (94%) displayed pattern B or D, with optimal performance at a low MAT (<7.5 °C) and an MAP between 2500 and 3500 mm. Middle-tier families were distributed across all patterns, with 56% exhibiting either pattern A or B. Patterns E and F were rare.

Both PSMC and WACTIC breeding zones displayed substantial diversity in climatic response patterns, while most WA Coast families aligned with pattern A (the WA Coast breeding zone is much smaller than the PSMC and WACTIC zones) (

Table 4. Distribution patterns of climate response among the 150 sampled families across the three WA breeding zones.

Program	Pattern
	A	B	C	D	E	F
WA Puget Sound	46	24	8	12	3	2
WA Coast	21	2	1	3	4	2
WA Cascade	28	32	10	26	3	1

). Given the substantial diversity in climate adaptation among families within relatively large breeding zones, it seems feasible to make within-zone selections of suitable families for future climatic conditions, particularly in large breeding zones.

4. Discussion

4.1. Virtual Populations

Unlike provenance trials, where each provenance represents a geographically distinct source population, progeny trials test families derived from hundreds of parents selected across different geographic areas from previous breeding cycles. This difference creates a challenge: classifying populations based on geographic factors fails to accurately reflect the climatic conditions to which these families are adapted, due to the climate variability within geographic regions. To address this issue and enable meaningful population-level analysis of climate adaptation, we aggregated families into virtual populations based on their climatic similarity. These virtual populations function as family pools, representing sets of families potentially adapted to specific site conditions due to comparable climatic origins. This approach aligns with the “species pool” concept in multispecies seed-sourcing strategies [55,56] and the “wide-sense population” concept in local adaptation studies [21,57,58]. Consequently, the term “local populations” we used in the discussion referred to populations consisting of parents originating from areas with climate variables similar to those of the local testing location. While these virtual populations are theoretical constructs, we propose that their use is appropriate for examining the climate adaptation of improved genetic materials at the population level. By averaging variation among families with similar source climates, this method allows for a more focused analysis of population-level responses, effectively reducing noise from individual family differences.

4.2. Survival vs. Growth

The survival rate could be an important factor to consider in assisted migration studies. For instance, Berlin et al. [59] discovered that relocating Scots pine (Pinus sylvestris L.) populations to the south in the boreal regions of Sweden and Finland generally increased the survival rate. In contrast, moving them northward resulted in faster growth but higher mortality rates. Similar findings were observed in Silver Birch (Betula pendula Roth) in Finland [60]. These results underscore the potential conflicting patterns when determining the optimal transfer distance for balancing growth and survival rates.

Climatic transfer functions for survival rate were not reported in this study due to our preliminary investigation yielding notably low R-squared values, indicating a weak explanatory power of climate variables alone. This may be attributed to the various causes of mortality in young Douglas-fir progeny trials, which are influenced by a complex interplay of factors. In addition to climate mismatch [8], mortality also involves competition for resources [61], pathogen infestations [62,63,64], insect damage [64], and extreme weather events [64], among other factors. Given these complexities, we focused on modeling the growth rate in this study, as it reflects both the tree’s genetic potential and its ability to cope with environmental stress. Trees must survive and remain healthy to achieve significant growth, and, therefore, growth rate has been widely used as an indicator of juvenile fitness [20].

4.3. Population Performance under Present Climate Conditions

The forest science and reforestation communities have long operated under the assumption that locally sourced seed outperforms non-local seed, a principle that has guided seed transfer practices for commercially important tree species [20,21,41,58]. This “local is best” paradigm was the foundation for most existing seed transfer guidelines until extensive provenance and progeny trials were established [65,66]. In regions or for species where local adaptation remains understudied, conservative approaches have been adopted, including limiting seed transfer distances for reforestation efforts or sourcing seeds from ecologically similar regions [67,68]. These practices aim to minimize the risk of maladaptation and maintain local genetic integrity [69].

However, recent research has begun to challenge the local is best assumption, particularly in light of rapid climate change. The evidence for local adaptation for coastal Douglas-fir is mixed. For instance, White and Ching [70] observed that local provenances significantly improved their growth rankings from age 9 to 25, indicating potential advantages of local seed sources over time. However, Krakowski and Stoehr [35] found that provenances from relatively distant origins often performed as well as or better than local sources at age 45, challenging the strict interpretation of “local is best”. Ye and Jayawickrama [30] revealed that while local Douglas-fir provenances did not always exhibit superior performance at rotation age, geographically or climatically similar provenances tended to show comparable growth. This suggests that seed transfer guidelines could potentially be broadened to include climatically matched sources from a wider geographic range.

This study found only modest evidence supporting the superiority of local population growth. In fact, populations originating from slightly warmer regions exhibited up to 2% increased height and 8% more stem volume at the age of 12 compared to local ones. Similar results were reported by Silen and Mandel [71]. In the PNW, the practice of transferring improved seed westward and to higher elevations is already implemented by foresters willing to accept minimal risk, particularly when faced by shortages of ideal seedlots. However, transfers from high to low elevations remain infrequent. Our findings generally align with those reported by St. Clair et al. [8] in their Douglas-fir provenance trial study. They concluded that Douglas-fir seed sources could be moved at least 2 °C cooler or warmer while maintaining good long-term productivity.

Transfer functions for several drought-related climate variables indicated slight growth advantages for local Douglas-fir populations; however, their overall model fits were generally weak. The relatively flat peaks of the transfer function curves suggest that genetically improved Douglas-fir populations may possess greater climate change resilience than previously assumed. This finding could potentially lead to more flexible seed transfer guidelines.

4.4. Impacts of Climate Change on Population Growth and Implications for Assisted Migration

The projected rising temperature and shifts in precipitation patterns due to anthropogenic climate change are expected to have significant negative impacts on the productivity and overall fitness of tree populations [14,72,73]. Consequently, it is crucial for tree breeders and forest managers to quantify the potential loss of genetic gains due to climate change impacts at the end of tree rotation periods [36,74,75]. Our results on climate transfer functions illustrated the potential future losses in growth for existing Douglas-fir genetically superior materials without assisted migration or breeding efforts. The anticipated future temperature rise would be equivalent to moving the current populations from cooler to warmer environments, which may lead to a decrease in tree height by up to 10% and a reduction in stem volume by up to 30% at 12 years of age. The losses in absolute growth could be significantly greater when projected to rotation age. Such reductions in tree growth would have significant impacts on the future of tree breeding programs and reforestation itself, as selection for populations with faster growth and higher productivity is and will still be the main goal for timber species like Douglas-fir [76].

Our results suggested that properly implementing assisted migration efforts using genetically improved families from southern regions or lower elevation areas could effectively mitigate the impact of future temperature increases by approximately 0.8 °C by 2055 and 1 °C by 2085. This strategy would limit the reductions in height and stem volume to 2%–5% and 10%–15%, respectively, in the WA Cascade and Puget Sound regions, and 2%–3% and 12%–18%, respectively, in the OR Coast region. However, other strategies besides northward migration are also viable and will be discussed in the next section. These projections emphasize the urgency of incorporating assisted migration and other appropriate climate change mitigation efforts into Douglas-fir breeding programs in the region.

4.5. Implications for the Current Douglas-Fir Breeding Programs in the PNW

Since the early 1950s, Douglas-fir operational breeding programs in the PNW have evaluated more than 24,000 first-cycle parents and 2900 second-cycle full-sib families through progeny tests at approximately 1100 sites [42]. Most seedlings now used in reforestation are grown from seeds collected from seed orchards established using genetically selected trees within breeding zones [34,44]. Cooperative members have invested heavily in these programs, and third-cycle breeding is currently in progress. With anticipated climate change, there is growing concern that current plantations may not be well-adapted to future climate conditions. Consequently, Douglas-fir breeders in the PNW aim to adapt current breeding strategies to address the impacts of future climate change while preserving high population growth rates, instead of launching new breeding programs.

Our results emphasize that selecting and deploying fast-growing trees or families from warmer regions could effectively enhance the resilience of existing breeding populations. This strategy could help mitigate the impacts of future global warming, particularly in the northern Cascades area. The approach involves transferring seeds from southern to northern locations and from lower to higher elevations. While small-scale seed transfers between breeding zones are now fairly common in the PNW, larger-scale transfers remain infrequent. Since each breeding zone in the PNW involves distinct cooperative memberships, the existing meta-cooperative ownership agreements have imposed restrictions on the transfer of germplasm. These agreements should be reviewed.

While assisted migration could be a valuable strategy, long-distance seed transfer is not free of risk. When moving tree populations from south to north, even if climatic conditions are similar, the longer photoperiod during the growing season can potentially increase the risk of frost damage in both spring and fall. This risk stems from the complex interactions between genetic adaptations, temperature, and photoperiod cues that regulate tree phenology [20,77,78]. These may underscore the importance of considering photoperiod effects in assisted migration strategies, as simply matching climatic conditions may not account for the complex phenological responses of trees to changing day length cues. On the other hand, seed transfer from lower to higher elevations appears to be safer. This is primarily due to the consistency of photoperiodic cues across elevational gradients. As noted by Hopkins [79], there is an inverse relationship between elevation and temperature, with a decrease of approximately 1.4 °C observed for every 305 m gained in altitude within the United States. In addition, occasional incursions of very cold weather systems can still pose risks to southern sources [80]. The assisted migration strategy appears particularly appealing for breeding zones in northern regions and/or at higher elevations, such as the WA Cascades, WA Puget Sound, and OR Cascades.

In response to projected climate change scenarios, it is reasonable to anticipate that future forest ecosystems will be subject to an array of climate-induced stressors. These may include, but are not limited to, more frequent and severe winter cold snaps and extended summer droughts [29,81]. Thus, it is imperative that future seed transfer and progeny testing protocols incorporate a focus on these stressors. Previous research on Douglas-fir has identified trade-offs between drought tolerance and growth [40,82], as well as between cold hardiness and growth [83]. Notably, Bansal et al. [75] discovered a correlation between winter cold hardiness and summer drought tolerance in coastal Douglas-fir, with trees from regions experiencing colder winters demonstrating enhanced cold and drought tolerance. Similar results were found in other studies as well [34,82,84,85]. Within the cooperative Douglas-fir breeding programs, artificial freezing tests for fall cold hardiness have been conducted in the northern breeding zones. Additionally, a limited number of small-scale nursery and field tests have been carried out in the region. However, it is worth noting that most seed orchard families have yet to undergo evaluation for cold and/or drought resistance.

In the realm of assisted migration and seed deployment strategies, researchers have suggested using simple tools like SeedWhere [86] and Seedlot Selection Tool [87]. These tools mainly use climate matching techniques to identify the best seedlots for deployment, which can be particularly useful when selecting provenances or seed sources in assisted migration contexts. However, it is important to recognize certain limitations when applying this methodology to seed orchards and breeding populations. The models behind these tools were not developed using data from genetically improved families, which could lead to results that differ significantly from those based on provenance. For instance, our research indicates that some transfer functions exhibited relatively low R-squared values, implying that climate variables explain only a small portion of the variations seen among genetically improved populations. Moreover, this approach does not explicitly account for the genetic values of the predicted trees, which could significantly influence the overall effectiveness of reforestation efforts. To improve the success of assisted migration strategies in Douglas-fir breeding programs, it would be prudent to incorporate genetic testing results into the decision-making process.

Assisted migration is not the sole strategy used to alleviate the consequences of climate change. It is important to note that ample genetic diversity exists within breeding populations. The results of this study demonstrate considerable differences in climate adaptation patterns among full-sib families, particularly within larger breeding zones. Significant family-by-site interactions found in other progeny tests also imply the existence of family-by-climate interactions [88,89,90]. These indicate that selecting fast-growing families better adapted to future climate conditions is a viable strategy for enhancing Douglas-fir stand resilience and productivity in the face of climate change, even without long-distance assisted migration.

The forest lands in the PNW are highly diverse, with significant variations in elevation, notable differences between south-facing (warm and dry) and north-facing (cold and snowy) slopes, and a variety of soil types. For future Douglas-fir progeny trials, it is recommended to include locations that represent more extreme environmental conditions. Greater emphasis should be placed on breeding and selecting families that exhibit robust growth in well-drained soils within warmer and drier environments. For the existing large-scale progeny testing network, sharing measurement data among breeding zones is highly recommended. Using appropriate statistical methods, such as across-zone multivariate analysis, allows tree breeders to model and predict the genetic gains of tested families under projected climate conditions. This approach facilitates the evaluation of the adaptation range of genetically superior families and their deployment for maximum benefit.

Long-term model-based forecasts are typically associated with inherent unpredictability and are subject to potential future revisions. As a result, maintaining genetic diversity is considered the most effective strategy to counter environmental changes and mitigate uncertainty [91]. Cooperative breeding programs in the region have strived to maintain high levels of genetic diversity through clone banks and seed pools within breeding zones at each breeding cycle. Given the anticipated climate change, we should expand these pools to include seed sources beyond current breeding zone boundaries.

These results are based on data from 12-year-old trees. Given the strong belief that the adverse impacts of climate on temperate conifers intensify with age [8,92], it would be valuable to investigate these trends in older progeny tests or in these same trials at a later date.

5. Conclusions

Our comprehensive analysis of climatic transfer functions, leveraging data from extensive Douglas-fir breeding programs in the Pacific Northwest, offers valuable insights for future tree breeding strategies. By examining growth data from 80 second-cycle progeny trials across western Washington and Oregon, we have uncovered critical patterns in climate adaptation among genetically improved Douglas-fir populations.

Key findings include the following:

• Genetically improved populations from slightly warmer areas showed growth rates comparable to or better than local populations, particularly in the northern Cascade region.
• Populations transferred from colder to warmer climates exhibited significant growth reductions, indicating an asymmetrical transfer function pattern.
• Substantial variation in climate adaptation was observed among full-sib families within larger breeding zones.
• Within-zone selection for families suited to future climates emerged as a promising strategy to enhance Douglas-fir stand resilience.
• The study proposes methods to adapt current breeding strategies to address climate change impacts while maintaining high growth rates in regional Douglas-fir breeding programs.

These findings underscore the importance of considering climate adaptation in tree breeding efforts and offer practical approaches for mitigating the effects of climate change on Douglas-fir populations in the Pacific Northwest.