Testing scenarios for assisted migration of forest trees in Europe

Author

• Marta Benito-Garzón, 
• Juan F. Fernández-Manjarrés

Abstract

One approach to compensating for rapid climate change and protecting biodiversity is assisted migration (AM) of key tree species. However, tools for evaluating the sensitivity of target sites and identifying potential sources have not yet been developed. We used the National Forest Inventories of Spain and France to design scenarios for AM between and within both countries. We characterized sensitivity to climate change as the expected changes in volume and mortality of Pinus halepensis Miller and Pinus pinaster Aiton between the present and 2050. Target zones were selected from provenances with high sensitivity and seed zones from provenances with low sensitivity to climate change; the latter can be considered “seed refugia” as the climate changes. Three plausible scenarios for translocation to the target zone were developed on the basis of volume simulations calibrated with different planting strategies: (1) seeds only from foreign provenances; (2) foreign provenances plus local seeds; and (3) only local seeds. The results for both species show that models based on foreign “top-three” provenances always increased the standing volume of the target zone. Models run with only local seeds predicted increased volume for P. halepensis but not for P. pinaster. Our results suggest that volume and mortality trends are not always correlated with seed sources and targets, that projected provenances mortality do not follow always a southern–northern pattern and that seed refugia, if any, may be useful for compensating for the effects of climate change only in a subset of provenances.

Introduction

Human-mediated adaptation of forests to new climate conditions might be necessary to protect biodiversity, maintain functional systems and minimize economic climate-related risk to tree plantations (Schwartz et al. 2012). One adaptation option increasingly discussed by scientists and stakeholders is the intentional translocation of species, populations or genotypes to compensate for observed or future climate changes (Richardson et al. 2009). Assisted migration (AM), managed relocation, translocation, population reinforcement and assisted colonization are all terms used in the literature for this management option. In forestry, AM could offer economic benefits for countries such as the United States and Canada, where the large amount of new forest planted each year suggests a high potential for AM (Pedlar et al. 2012). However, the benefits and drawbacks of AM in forests are largely unknown in Europe where only very few plantations are planted each year, and given the multiple functions provided by most of the forests, creating stable productive forests without compromising local biodiversity and ecosystems services challenges scientists and managers (Fernández-Manjarrés and Tschanz 2010). Drought-induced tree mortality has increased in Europe in recent years, especially in the Mediterranean region (Carnicer et al. 2011), and is expected to continue to increase in the coming years (Benito-Garzón et al. 2013c). However, the assertion that southern populations facing climate-induced risks can be safely translocated to northern regions within the species’ ranges needs to be explored. Similarly, whether core or northern populations would benefit from population reinforcement with seeds from populations at the southern edge of the distribution remains unclear.

The risks and benefits of AM have generated intense debate in the last decade (Richardson et al. 2009; Pedlar et al. 2012; Lunt et al. 2013; Neff and Larson 2014). The main benefit of AM is the potential to protect species and prevent extinctions; the risk of invasion, the difficulty of implementation and poor justification for the translocations are among the drawbacks (Hewitt et al. 2011). In forestry, the main goal of AM is to maintain forest productivity and health under climate change (Pedlar et al. 2012). There are two main hazards of AM. The first is the potential for hybridization with new species. The second is possible maladaptation of southern populations to extreme cold events in the north; this could lead to important economic losses, as has happened in the past (Benito-Garzón et al. 2013a, b).

Given the challenges of AM, building scenarios and associated ecological models before developing programs in the field can improve understanding of the risks associated with lower fitness in translocated populations. Seed transfer functions under climate change scenarios can be calibrated by fitness-related measures surveyed in provenance tests that were originally planned for commercially important trees (O’Neill et al. 2008; O’Neill and Nigh 2011; Oney et al. 2013). Provenance tests provide information on local adaption and phenotypic plasticity of fitness-related traits that can be incorporated into the models (Benito Garzón et al. 2011; O’Neill and Nigh 2011; Oney et al. 2013; Valladares et al. 2014). However, long-term provenance tests suitable for testing the effects of new climates on trees are scarce, and at least in Europe, they do not usually cover the entire range of the species, possibly biasing the calibration of the models (Benito Garzón et al. 2011; Valladares et al. 2014). An alternative to provenance trial data for generating AM scenarios is the use of fitness-related measures surveyed in National Forest Inventories (NFI); these usually cover the entire range of a species and simultaneously provide several demographic variables. Forest inventory data cannot be used directly to calculate seed transfer functions but are extremely useful for capturing statistical relationships between climate and fitness-related traits for thousands of observations that can then be easily projected under expected climate change scenarios. The use of forest inventories restricts the analysis of AM migration to within a species’ distribution because colonization outside the known range cannot be estimated. In Europe, where species’ distributions often cover several countries, the challenge is to synthesize fitness-related measures when NFI methods differ among countries.

In Western Europe and the Mediterranean area in particular, AM would necessarily require the sharing of seeds among countries. Some species that are economically important for timber production, such as maritime pine (Pinus pinaster Aiton), have a recent history of failed provenance translocation from southern to northern countries, leading to the only case where the European Union has banned the introduction of foreign seeds (Benito-Garzón et al. 2013a). Other species are central to the AM debate because they have high drought tolerance; for example, some populations of Aleppo pine (P. halepensis Miller) can survive in a semi-arid climate (Atzmon et al. 2004; Klein et al. 2011). Climate change will likely prompt increased aridity in some parts of the Mediterranean (Gao and Giorgi 2008), and having a pool of tree populations or species with high tolerance to hydric stress would protect some Mediterranean regions from desertification in the near future.

Two contrasting approaches can be taken to the design of AM programs. The classical approach asks which seed sources are best adapted to withstand climate change at a target site. The second approach identifies the areas within a species’ distribution that seem less vulnerable (i.e., less sensitive) to climate change and uses these areas as sources for seed transfers. This second approach assumes there are “seed refugia” not needing AM, but serving as long-term sources for AM programs. We predict that the approaches would be combined in practice.

Here, we propose AM scenarios for P. halepensis and P. pinaster. The scenarios are based on a range-wide sensitivity analysis to climate change to identify potential candidate zones for AM, appropriate seed sources for the candidate zones and eventual “seed refugia” and their utility for AM. Sensitivity to climate change is traditionally estimated by measuring the ecophysiological responses of a given species when exposed to changes in the environment (Williams et al. 2008; Moritz and Agudo 2013). In our context, species sensitivity to climate change is estimated by models of volume and mortality based on the French and Spanish NFIs for current and 2050 (A1B scenario) climates, averaged by provenance regions. After the sensitivity of the provenance regions has been estimated, seed sources and target zones for AM are identified. Finally, we simulate the performance of plantations from seed sources in the target zones by 2050 and discuss how these translocations can help decision-making concerning assisted migration of species within their distribution.

Materials and Methods

National Forest Inventories

We harmonized plot information recorded by the Spanish and French National Forest Inventories (NFI) to calibrate models with information on juvenile trees from two countries that will likely share seeds in the near future. The forest inventories share some species that are considered important for AM programs; for example, Pinus pinaster has been extensively planted within its range for timber production, and P. halepensis will be a likely target for AM because of its high resistance to drought.

The Spanish National Forest Inventory (hereafter SI), conducted from 1997 to 2007, provides measures of tree diameter at breast height, total tree height and mortality per plot, measured in permanent plots from the third forest inventory. The French National Forest Inventory (hereafter FI), conducted from 2005 to 2012, provides non-repetitive measures of circumference at breast height, total height per tree and mortality per plot. Plots are circular with a 25 m radius for both NFIs. Mortality, calculated by both NFIs as the number of dead trees per plot, was used to calibrate mortality models. The SI provides measures of diameter of the trees at breast height and total height, whereas the FI provides the circumference at breast height and the total height of the trees. Therefore, standing volume per tree was calculated for both inventories using a simple equation for trunk volume:

Volume=Π⋅r2⋅htot$$\text{<span style="font-family: -webkit-standard;"><span style="background-color: rgba(255, 255, 255, 0);">Volume</span></span>}\mathrm{<span\; style="font-family:\; -webkit-standard;"><span\; style="background-color:\; rgba(255,\; 255,\; 255,\; 0);">\Pi </span></span>}{\mathrm{<span\; style="font-family:\; -webkit-standard;"><span\; style="background-color:\; rgba(255,\; 255,\; 255,\; 0);">r</span></span>}}^{\mathrm{<span\; style="font-family:\; -webkit-standard;\; font-size:\; small;"><span\; style="background-color:\; rgba(255,\; 255,\; 255,\; 0);">2</span></span>}}\mathrm{<span\; style="font-family:\; -webkit-standard;"><span\; style="background-color:\; rgba(255,\; 255,\; 255,\; 0);">h</span></span>}\mathrm{<span\; style="font-family:\; -webkit-standard;"><span\; style="background-color:\; rgba(255,\; 255,\; 255,\; 0);">t</span></span>}\mathrm{<span\; style="font-family:\; -webkit-standard;"><span\; style="background-color:\; rgba(255,\; 255,\; 255,\; 0);">o</span></span>}\mathrm{<span\; style="font-family:\; -webkit-standard;"><span\; style="background-color:\; rgba(255,\; 255,\; 255,\; 0);">t</span></span>}$$

where r is the radius of the trunk at breast height (1.30 m), htot is the total height measured for the tree, and the constant Π is 3.1416. The radius was calculated from the diameter for the SI (r = d/2, where d is diameter at breast height and r is the radius) and from the circumference for the trees measured in the FI (r = c/2 · Π where r is the radius and c the circumference at breast height and Π = 3.1416).

Only trees with a diameter of less than 200 mm and more than 75 mm were considered because growth is expected to be strongly correlated with climate in the early-middle age of trees (Weiner 2004), but we avoid regeneration (trees below 75 mm of diameter) because they would have a more sensitive response to climate than young adult trees. For P. pinaster, we used 308,868 trees (285,917 and 22,951 from the Spanish and French NFIs, respectively) recorded from 14,402 plots (10,962 and 3440 from the Spanish and French NFIs, respectively; Supplementary Figure S1). For P. halepensis, we used a total of 135,421 trees (130,704 and 4,717 from the Spanish and French NFIs, respectively) distributed in 10,416 plots (9,529 and 887 from the Spanish and French NFIs, respectively; Supplementary Figure S1).

Environmental drivers

We used monthly climate data from WorldClim (Hijmans et al. 2005) for current conditions and future climate scenarios from the CGIAR climate change program (http://www.ccafs-climate.org/data/). Future scenarios were built by averaging four IPCC A1B storyline scenarios (CCSM3, CSIRO, ECHAM3, IPSL) projected for 2050. Ten environmental variables were used in the models: maximum temperature of the warmest month, minimum temperature of the coldest month, temperature seasonality, precipitation of the warmest quarter, precipitation of the coldest quarter, precipitation seasonality, altitude, two proxies of climate extremes for maximum temperature (dtmax) and minimum temperature (dtmin) and the background mortality per plot. The distances to the maximum (dtmax) and minimum (dtmin) temperatures were calculated as the differences between the maximum and minimum temperatures of the corresponding period and the maximum or minimum temperatures from 1901 to 2005, respectively, recorded by the CRU (http://www.cru.uea.ac.uk/data). In addition, we calculated the background mortality per plot as the total number of dead trees by plot.

Provenance regions

The units developed for commercial regeneration of trees in Europe are provenance regions (= seed collection zones) that are designed independently for each country. These regions are based on environmental characteristics (climate, altitude and soils) and in the physiognomy of the trees. We used the classification of French (http://www.irstea.fr/) and Spanish (http://wwwsp.inia.es/Investigacion/centros/CIFOR/redes/Genfored/Paginas/Regiones%20Procedencia.aspx) provenance regions for both species to regionalize our volume and mortality results and to design the AM sources and target zones. Twenty provenance regions were designated for P. halepensis (1 in France and 19 in Spain) and 33 for P. pinaster (5 in France and 28 in Spain).

Modeling set up: sensitivity analysis

We designed a modeling framework to simulate forest AM in three main steps. First, we assessed the sensitivity of trees by calculating volume and mortality models for present and future conditions. Non parametric models on standing volume were run by species and by individual. Mortality models were calibrated using the average mortality by plot and by species. To estimate the sensitivity of each species to climate change, volume and mortality models were bootstrapped with 10 iterations using the environmental variables described above. The models were calibrated on the current climate with the random forest algorithm (R library randomForests (Liaw and Wiener 2002)) using 2/3 of the total data and then projected for current and 2050 climatic conditions for the entire range of the species. The remaining 1/3 of the data were used for model evaluation. The random forests algorithm builds multiple individual random trees for regression (Breiman 2001). One regression tree is grown for each of the groups made with multiple individual trees (n = 500) bootstrapped from the original data. The final prediction is the average prediction of all the regression trees grown without pruning. The random forests algorithm estimates the percentage of the variance explained by the model drivers and the capacity of generalization of the models is measured by the R2 coefficient. Final predictions of volume and mortality were conducted for current climate and for future scenarios at 1 km resolution. Then, the predicted differences in standing volume and mortality for the 2050-present period were averaged by provenance regions. This allowed us to classify the global sensitivity of the trees to future climate change for each provenance region and to define the target regions for assisted migration. This simulation was also used as a baseline for AM because it represented a plantation with an admixture of seeds from all the provenance regions.

Modeling set up: assisted migration simulations

Second, we defined the target zone as a provenance region where trees showed no increase in volume by 2050 and/or the predicted mortality of trees by 2050 was high in the sensitivity analysis (baseline simulation). We propose these selected target regions as scenarios for experimenting with AM prior to any large scale implementation to stimulate the debate about AM in Europe. The target choice for each species was based on the sensitivity analysis described above performed with mortality and standing volume.

Third, in addition to the baseline simulation (seed admixture, Table 1), three simulations of standing volume were run to cover three additional scenarios of AM. After the target zones for assisted migration were selected, we developed three seed source combinations involving seed translocations (Table 1). We could not use mortality models to calibrate our simulations for AM scenarios because the models’ explanatory value and generalization power was too low for predictive modeling when they were calibrated independently for each provenance region. Therefore, the evaluation of seed sources for target areas was based only on the estimates of future standing volume.

Table 1

Theoretical scenarios of assisted migration selected, the seed zones were defining by calibrating local models on specific provenance regions and target zones were defined spatially-defined as the region in which the model was projected

AM scenario	Model calibration = seed zones	Model target
1. Seed admixture	All provenance regions	All provenance regions
2. Only foreign provenances	The three provenances with the highest volume projected by 2050 in the general model	Target provenance
3. Foreign provenances + local	The three provenances with the highest volume projected by 2050 in the general model + target provenance	Target provenance
4. Only local provenances	Target provenance	Target provenance

The case (1) is the general model where the calibration and the projection of the model are done for the entire distribution area that considers all the provenance regions. The (2), (3) and (4) cases are designs were calibration has been restricted to one (4) or several provenance regions (2 and 3) and only projected in the target provenance

These scenarios include (a) plantations in the target zone that only consider foreign seeds, (b) foreign and local seeds, and (c) only local seeds to calibrate the models (Table 1). In parallel with these simulations, we looked for provenance zones that might be favored by climate change and that may constitute “seed refugia” for AM programs. When evaluating seed sources, models of standing volume were first calibrated using only the three provenance regions where the volume prediction for 2050 was the highest in the global sensitivity models (“seed refugia”, Fig. 4). The second calibration added the local provenance to the seed refugia; the third scenario represented the case where no AM is implemented and only local seeds are used. In all three cases, the predictions were only for the target zone, to simulate the variations in standing volume depending on the seed source used by an AM program (Table 1).

Results

Global sensitivity to climate change: volume and mortality models

Both standing volume and mortality were highly correlated with current environmental variables, as indicated by the percentage of the variance explained by the drivers (PEV) and the generalization power indicated by the R2 coefficient (Table 2). PEV ranged from 42.73 for the Pinus halepensis volume model to 85.63 for the P. pinaster mortality model (Table 2). The generalization power of the models ranged from 0.71 for the P. halepensis volume predictions to 0.97 for the P. pinaster mortality model. The background mortality was the most important variable in the mortality models for both species (Supplementary Figure S1). P. pinaster volume models were strongly dependent on the precipitation of the coldest quarter (BIO19) and the background mortality; the most important predictor variable for P. halepensis was the background mortality (Supplementary Figure S1).

Table 2

Species, provenance regions used to calibrate the models, projection area of the models, model type (volume or mortality model), total number of trees (NT) used for the model (for volume models) or total number of plots (NP) used for the model (for mortality models), percentage of the variance explained by the model (PEV) and R2

Species	Model calibration	Model projection	Model type	NT/NP	PEV	R2
Phalepensis	All provenance regions (1)	All provenance regions	Volume	51,411	42.73	0.71
Phalepensis	All provenance regions	All provenance regions	Mortality	10,253	68.64	0.95
Phalepensis	3 seed foreign sources + target region (2)	Target region	Volume	18,623	48.09	0.17
Phalepensis	3 seed foreign sources (3)	Target region	Volume	15,729	41.34	0.12
Phalepensis	Target region (4)	Target region	Volume	2894	37.41	0.65
Ppinaster	All provenance regions (1)	All provenance regions	Volume	95,894	54.15	0.74
Ppinaster	All provenance regions	All provenance regions	Mortality	13,897	85.63	0.97
Ppinaster	3 seed foreign sources + target region (2)	Target region	Volume	6784	76.72	0.63
Ppinaster	3 seed sources (3)	Target region	Volume	6295	77.83	0.60
Ppinaster	Target region (4)	Target region	Volume	489	54.8	0.14

The results were averaged for 10 model bootstrap runs. Numbers on model calibration 1–4 follow the names of the Table 1

The random mixture of seeds from all the provenances produced contrasting results for the species in the general models, i.e., models calibrated and tested with all the data. The P. halepensis general model predicted increased volume by 2050 in the provenance regions at the core of the species’ distribution. In contrast, marginal regions showed a slight or no volume increase by 2050 (Figs. 1a, 2). For this species, there was a clear pattern of increased mortality in the south, decreasing to the north (Figs. 1b, 2). The P. pinaster general model produced a contrasting pattern; the largest increase in volume occurred in provenance regions at the northern part of the species’ distribution in France, and the lowest expected volume for 2050 occurred in provenance regions in the western/central areas of Spain (Figs. 1c, 3). The mortality rates for P. pinaster are expected to increase in all regions of Spain by 2050, especially in one provenance region in central-western Spain (Fig. 1d). However, the among-population variation in the prediction of standing volume and mortality rates was always smaller for P. pinaster (Fig. 3) than for P. halepensis (Fig. 2).

Fig. 1

Projected differences in volume (a, c) and mortality (b, d) between the 2050 A1B scenario and present conditions for Pinus halepensis (on the left) and Pinus pinaster (on the right). These general models were calibrated using the entire database of the French and Spanish National Forest Inventories and projected for the all the territory so they represent the random use of all the seed sources in all areas. The final maps of volume and mortality are averaged by provenance region. Predicted differences in volume between 2050 and the present are shown in m−3 and mortality is given as the probability of dead trees between 0 and 1

Fig. 2

Standing volume (m−3) and mortality (probability) predicted for the models by provenance region for Pinus halepensis for present conditions (light green and orange, respectively) and for the A1B 2050 scenario (dark green and red, respectively). Provenance regions code is shown in the x-label and located in a map. (Color figure online)

Fig. 3

Standing volume (m−3) and mortality (probability) predicted for the models by provenance region for Pinus pinaster for present conditions (light green and orange, respectively) and for the A1B 2050 scenario (dark green and red, respectively). Provenance regions code is shown in the x-label and located in a map. (Color figure online)

Assisted migration scenarios

Consistent with the criteria of choosing AM targets in areas with high sensitivity to climate change, i.e., no increase in volume by 2050 and moderate increased mortality in the general sensitivity models, the northernmost provenance region in Mediterranean France was selected as the P. halepensis target region. Neighboring provenances in Catalonia exhibited even more mortality in the general model, but we used the northern French Mediterranean provenance for the exercise of moving provenances from south to north. Three southern seed source regions (Provenances number 5, 9 and 18; Fig. 2) were selected among the regions with the highest volume expected by 2050 (Fig. 4). The target zone of P. halepensisrepresents a classic case of south-to-north AM translocation, where northern populations are not likely to increase their standing volume even if the increase in mortality in the future is not high. The rationale behind is to bolster populations at the leading edge by translocating reproductive material from southern populations that will adapt more quickly to warmer climates.

Fig. 4

Seed sources (black) and target zones (grey) selected for modelling AM based on the expected sensitivity (volume and mortality, see Fig. 1) to climate change of Pinus halepensis (a) and P. pinaster (b) by 2050 (see Fig. 1). Geographic areas correspond to current provenance zones in both countries. See text for the criteria used in selecting target and seed sources zones

For P. pinaster, one target zone that matched the criteria of high sensitivity to climate change in the Southern Central System was selected; it showed no increase in volume by 2050 and the highest expected mortality rate among the provenance regions in the general model. In this case, AM would be intended not for the most northerly provenance but for a very sensitive provenance. Again, seed sources were chosen from the three seed regions that showed the highest predicted volume in the general sensitivity model (Fig. 4; Provenances 7, 31 and 36 Fig. 3). Models for sensitivity analysis for maritime pine do not show a clear geographical trend in volume or mortality, i.e., no “northern” provenance appeared more sensitive than a southern or central one. For this species, the selected target zone is therefore a medium latitude provenance where sensitivity patterns show an increase in mortality combined with no future increase in standing volume. The rationale for studying this target zone is to understand what happens when a marginal population that is not at the leading edge needs to be reinforced with other provenances to increase its standing volume under climate change.

Overall, the seed-source AM models calibrated with the three provenance regions with the highest predicted volume by 2050 explained a high percentage of the variance, ranging from 41.34 for P. halepensis to 77.83 for P. pinaster (Table 2). The R2 values were lower than for the general sensitivity models in some cases (Table 2). The R2 ranged from 0.12 for P. halepensis calibrated with the three provenance regions with the highest predicted volume by 2050 in the general sensitivity model to 0.65 for P. pinaster calibrated only with the local (target) provenance.

The resulting maps of estimated volume showed that the highest increase in volume by 2050 was obtained for both species when only the three provenances with the highest volume were used to calibrate the model (Fig. 5a, d). A smaller volume increase was obtained for both species when the models were calibrated with the local (target) provenance in addition to the three provenances with the highest volume (Fig. 5b, e). However, patterns differed between the species when the models were calibrated with trees coming only from the target region (local seed). P. halepensis had a large volume increase by 2050; the volume of P. pinaster did not increase by 2050 (Fig. 5c, f).

Fig. 5

Projected differences in volume between 2050 and present conditions for the target zones selected (see main text and Fig. 2) for Pinus halepensis (on the left part of the panel) and P. pinaster (on the right part of the panel). a, d The results of calibrating the models with the three provenance regions with a greater increase in volume estimated by 2050. b, e The results of calibrating the models with the three provenances with higher increase in volume plus the trees coming from the target provenance for the AM. c, f The results of calibrating the models only with the target region

For Aleppo pine, the standing volume predicted by the baseline model (sensitivity) that considers all provenances in the calibration was higher than the observed values estimated from the NFIs for the target zone (Fig. 6). Moreover, these results are higher than the use of only local seeds or any of the AM options. On the contrary, estimated standing volume with the baseline model that considers all provenances in the calibration with for maritime pine were essentially similar to either the use of only local trees, only imported provenances or a mixture of both.

Fig. 6

Standing volume projected (m−3) for the A1B 2050 scenario (dark green) and present conditions (light green) for the target zones selected (see Fig. 2) for Pinus halepensis (a) and P. pinaster (b). The x-axis represents the four cases of AM described in Table 1: (1) Seed admixture; (2) only foreign provenances; (3) foreign provenances + local; and (4) only local provenances. The grey linerepresents the average standing volume surveyed by the NFI for the target zones. (Color figure online)

Discussion

Despite the increasing number of articles discussing the benefits and drawbacks of assisted migration of species (Richardson et al. 2009; Fernández-Manjarrés and Tschanz 2010; Aubin et al. 2011; Hewitt et al. 2011; Pedlar et al. 2012; Schwartz et al. 2012; Lunt et al. 2013; Neff and Larson 2014), this is, to the best of our knowledge, the first attempt to create scenarios of assisted migration based on field data. Representations of assisted migration such as the ones presented here can help diminish the costs and failures of assisted migration programs in the field and provide guidance for the current debate by, for example, guiding future provenance tests.

Sensitivity to climate change based on volume and mortality models

The necessary first step for implementing AM in forestry is to evaluate the sensitivity to climate change of a species throughout its range. Certain regions would be seed sources, even if threatened by climate change, and other regions would be targets of seed translocation. Moreover, some areas could persist as “seed refugia” if conditions are stable or improve in a region for a given species. Changes in standing biomass and mortality in trees during recent decades are often attributed to changes in climate (Allen et al. 2010; Carnicer et al. 2011; Pretzsch et al. 2014). Using our approach, volume is projected to increase by 2050 (A1B scenario) for the provenances at the core of the distribution of Aleppo pine. Northern provenances of maritime pine are predicted to increase their volume in the future, suggesting that plantations would perform better at the leading edge of the distribution. The likely increase in mortality projected in our models for both species also agrees with the increasing mortality rates associated with climate change-induced drought (Fig. 1; Allen et al. 2010; Carnicer et al. 2011).

A novel aspect of our approach is that it includes simulations of standing volume and mortality for most of the range of the maritime and Aleppo pines; to our knowledge, this is the first time that the Spanish and French National Forest Inventories have been harmonized. Compared with volume-based models that include only the French or Spanish forest inventories in their analysis (Benito-Garzón et al. 2013c), this enhancement of the data significantly improved the percentage of the variance explained and the capacity of generalization of the models, making the global results more reliable.

Choice of target zones

Our analysis suggests that some zones may not have alternative seed sources to compensate for climate change. The analysis of global sensitivity based on the differences in mortality and standing volume between 2050 and current conditions allowed us to set one target region (according to our arbitrary criteria) per species, as well as the source seed zones used to simulate AM (Table 1). Although we acknowledge that many other target zones could have been selected for testing assisted migration scenarios, we limited our results to four hypothetical seed source scenarios in one target zone for each species (Table 1).

The global sensitivity of Pinus halepensis presents a typical scenario of what is expected under drier and warmer conditions for many Mediterranean species, an increase in the mortality rate from southern to northern provenance regions that will put southern populations at risk (Fig. 1). In general, demographic events differ at the core and margins of the species’ distribution (Purves 2009), which can influence standing volume and mortality projections. Our simulations for 2050 seem to follow this core-to-edges geographical pattern, with provenance regions where Aleppo pine is expected to increase its volume at the core (and at altitude) but not at the margins of the range (Fig. 1). Under these predictions, our AM scenario designed to increase the volume of the trees at the leading edge, where mortality is predicted to be low (Fig. 2a), seems complementary to the protection of seed sources in provenances that may become key players in the long term.

Pinus pinaster presents a different trend in global sensitivity that does not differentiate between the geographical core and the margin. Mortality is almost uniform and low across the region. The volume projected by 2050 is very high for the southwest of France, which could be explained by the extensive plantations that have been established for centuries (Benito-Garzón et al. 2013a). In this case, we chose a target zone with the highest mortality rate and zero increase in volume expected by 2050 in a marginal population in central Spain (Fig. 2b). In other words, we propose an AM action that is similar to a “classic” idea of ecosystem reinforcement and not necessary in the sense that AM is generally used, i.e., to find a place to ‘save’ the climatically endangered population. These results highlight that climate change may favor provenances that are not necessarily a high priority today and that AM programs may be more complicated than simply moving species northward or higher in altitude.

A remarkable difference between the species is the level of variation among provenances, particularly for the standing volume projections (Figs. 2, 3). AM programs would benefit from the higher variation in standing volume of Aleppo pine, whereas solutions in addition to AM would be necessary to manage adaptation to climate change in maritime pine. In other words, genetic reinforcement seems a promising alternative for P. halepensis, while onsite management, such as reduced competition for water, may be more relevant for P. pinaster.

P. halepensis volume simulations suggest that the increase in volume expected in many of the provenances is linked to altitude, suggesting that this species could compensate for climate change without latitudinal migration in (Supplementary Figure 3). In contrast, simulations of standing volume in P. pinaster do not show this altitudinal compensation in the future, suggesting the changing climate poses a higher risk to P. pinaster than to P. halepensis (Supplementary Figure S4).

Assisted migration scenarios: local versus foreign resources

The response of trees to changing climates can vary among populations, as shown by niche models that incorporate local adaptation and phenotypic plasticity (O’Neill et al. 2008; Benito Garzón et al. 2011; Oney et al. 2013; Valladares et al. 2014). This complicates interpretations of AM as a simple south–north or lowland-highland translocation procedure. We used provenance regions that were originally designed to account for the phenotypic variation among populations (forest ecoregions). We can therefore assume that trees coming from the same provenance region will likely respond to climate change similarly.

Probably the most popular advice for increasing the adaptive capacity of forests is planting trees from a random admixture of seeds collected across the range of a species (Millar et al. 2007). Interestingly, this option did not produce the desired increased volume between current and expected future conditions (Fig. 1a, c); for P. halepensis, however, it is the option that will likely maintain the highest standing volume until 2050 (Fig. 6a).

AM in forestry has the main objective of increasing forest production while maintaining forest health (Pedlar et al. 2012). Translocations with foreign seeds from highly productive locations, alone or in addition to local seeds, must be tested (strategies 2 and 3 in Table 1, respectively). However, this criterion means that not all sites will necessarily have a source with “optimal” seeds under climate change. Nevertheless, for the studied species, the results suggest that these two solutions (planting foreign resources alone or in combination with local seeds) always increase the standing volume at the target site. These two cases represent what we would expect if climate change is not hampering tree growth in the target zone: the seed sources coming from provenances with the highest volume in the southern populations are those that perform better in the target provenance as well. In other words, it appears that AM would be beneficial for production in areas that are not so exposed to climate stress.

The question remains as to whether AM is truly necessary for a target zone. We addressed this question by using models calibrated on only local seeds. This experiment resulted in large differences in predicted volume for both pines (strategy 4 Table 1; Fig. 3). The maritime pine AM model did not show increases in volume, as in the baseline model (Figs. 1c, 5f); In contrast, Aleppo pine volume increased by 2050 (Fig. 1). This suggests that collecting seed mixes from a species’ entire range might not always be the best method for establishing new plantations and that the local option needs to be seriously considered before importing seeds to increase the local genetic pool. In addition, these results indicate the necessity of calibrating models at different scales when estimating the sensitivity of forests to climate change. For example, our results from models calibrated only with local seeds (strategy 4, Table 1) agree with the results of the general models that were calibrated only for France. For this general models calibrated only for France an increase in volume is predicted for Mediterranean France along with a slight increase in mortality whereas no increase in volume is predicted for P. pinaster.

Limitations and further work

We acknowledge that many other simulations could have been implemented to test all the possible scenarios of AM. For example, testing AM of a species outside its geographical range would help with species selection in places where local species might be at risk of extinction, but we cannot do this with inventory data because there is no baseline for comparison. This reasoning could also be extended to the translocation of completely exogenous species. In the European context, this could be the case with the southernmost populations of many Mediterranean species, such as Aleppo pine, where mortality is expected to increase in the near future and no resources from the same species are available in nearby countries. This project, however, was not intended to review all AM possibilities, but to demonstrate how to evaluate AM scenarios before implementing programs in the field. Ideally, these would start with short-term provenance tests of several populations, from which the most promising ones would be selected.

References

1. Allen CD, Macalady AK, Chenchouni H et al (2010) A global overview of drought and heat-induced tree mortality reveals emerging climate change risks for forests. For Ecol Manage 259:660–684. doi:10.1016/j.foreco.2009.09.001CrossRef
2. Atzmon N, Moshe Y, Schiller G (2004) Ecophysiological response to severe drought in Pinus halepensis Mill. trees of two provenances. Plant Ecol 171:15–22CrossRef
3. Aubin I, Garbe C, Colombo S et al (2011) Why we disagree about assisted migration1: ethical implications of a key debate regarding the future of Canada’s forests. For Chron 87:755–765CrossRef
4. Benito Garzón M, Alía R, Robson TM, Zavala MA (2011) Intra-specific variability and plasticity influence potential tree species distributions under climate change. Glob Ecol Biogeogr 20:766–778. doi:10.1111/j.1466-8238.2010.00646.xCrossRef
5. Benito-Garzón M, Ha-Duong M, Frascaria-Lacoste N, Fernández-Manjarrés J (2013a) Habitat restoration and climate change: dealing with climate variability, incomplete data, and management decisions with tree translocations. Restor Ecol 21:530–536CrossRef
6. Benito-Garzón M, Ha-Duong M, Frascaria-Lacoste N, Fernandez-Manjarrés JF (2013b) Extreme climate variability should be considered in forestry-assisted migration. Bioscience. doi:10.1525/bio.2013.63.5.20
7. Benito-Garzón M, Ruiz-Benito P, Zavala MA (2013c) Inter-specific differences in tree growth and mortality responses to environmental drivers determine potential species distribution limits in Iberian forests. Glob Ecol Biogeogr 22:1141–1151CrossRef
8. Breiman L (2001) Random forests. Mach Learn 45:5–32. doi:10.1023/A:1010933404324CrossRef
9. Carnicer J, Coll M, Ninyerola M et al (2011) Widespread crown condition decline, food web disruption, and amplified tree mortality with increased climate change-type drought. Proc Natl Acad Sci USA 108:1474–1478. doi:10.1073/pnas.1010070108PubMedCentralCrossRefPubMed
10. Fernández-Manjarrés JF, Tschanz L (2010) Assisted colonization: protect managed forests. Science 330(80):1319. doi:10.1126/science.330.6009.1319-aCrossRefPubMed
11. Gao X, Giorgi F (2008) Increased aridity in the Mediterranean region under greenhouse gas forcing estimated from high resolution simulations with a regional climate model. Glob Planet Change 62:195–209CrossRef
12. Hewitt N, Klenk N, Smith AL et al (2011) Taking stock of the assisted migration debate. Biol Conserv 144:2560–2572. doi:10.1016/j.biocon.2011.04.031CrossRef
13. Hijmans R, Cameron S, Parra J et al (2005) Very high resolution interpolated climate surfaces for global land areas. Int J Climatol 25:1965–1978CrossRef
14. Klein T, Cohen S, Yakir D (2011) Hydraulic adjustments underlying drought resistance of Pinus halepensis. Tree Physiol 31:637–648. doi:10.1093/treephys/tpr047CrossRefPubMed
15. Liaw A, Wiener M (2002) Classification and Regression by randomForest. http://CRAN.R-project.org/doc/Rnews/. R News 2:18–22
16. Lunt ID, Byrne M, Hellmann JJ et al (2013) Using assisted colonisation to conserve biodiversity and restore ecosystem function under climate change. Biol Conserv 157:172–177. doi:10.1016/j.biocon.2012.08.034CrossRef
17. Millar C, Stephenson N, Stephens S (2007) Climate change and forests of the future: managing in the face of uncertainty. Ecol Appl 17:2145–2151CrossRefPubMed
18. Moritz C, Agudo R (2013) The future of species under climate change: resilience or decline? Science 341(80):504–508. doi:10.1126/science.1237190CrossRefPubMed
19. Neff MW, Larson BMH (2014) Scientists, managers, and assisted colonization: four contrasting perspectives entangle science and policy. Biol Conserv 172:1–7. doi:10.1016/j.biocon.2014.02.001CrossRef
20. O’Neill GA, Nigh G (2011) Linking population genetics and tree height growth models to predict impacts of climate change on forest production. Glob Chang Biol 17:3208–3217. doi:10.1111/j.1365-2486.2011.02467.xCrossRef
21. O’Neill GA, Hamann A, Wang T (2008) Accounting for population variation improves estimates of the impact of climate change on species growth and distribution. J Appl Ecol 45:1040–1049. doi:10.1111/j.1365-2664.2008.01472.xCrossRef
22. Oney B, Reineking B, O’Neill G, Kreyling J (2013) Intraspecific variation buffers projected climate change impacts on Pinus contorta. Ecol Evol 3:437–449. doi:10.1002/ece3.426PubMedCentralCrossRefPubMed
23. Pedlar JH, McKenney DW, Aubin I et al (2012) Placing forestry in the assisted migration debate. Bioscience 62:835–842CrossRef
24. Pretzsch H, Biber P, Schütze G et al (2014) Forest stand growth dynamics in Central Europe have accelerated since 1870. Nat Commun 5:4967. doi:10.1038/ncomms5967PubMedCentralCrossRefPubMed
25. Purves DW (2009) The demography of range boundaries versus range cores in eastern US tree species. Proc R Soc B 276:1477–1484. doi:10.1098/rspb.2008.1241PubMedCentralCrossRefPubMed
26. Richardson DM, Hellmann JJ, McLachlan JS et al (2009) Multidimensional evaluation of managed relocation. Proc Natl Acad Sci USA 106:9721–9724. doi:10.1073/pnas.0902327106PubMedCentralCrossRefPubMed
27. Schwartz MW, Hellmann JJ, Mclachlan JM et al (2012) Managed relocation: integrating the scientific, regulatory, and ethical challenges. Bioscience 62:732–743. doi:10.1525/bio.2012.62.8.6CrossRef
28. Valladares F, Matesanz S, Araujo MB et al (2014) The effects of phenotypic plasticity and local adaptation on forecasts of species range shifts under climate change. Ecol Lett 17:1351–1364CrossRefPubMed
29. Weiner J (2004) Allocation, plasticity and allometry in plants. Perspect Plant Ecol Evol Syst 6:207–215. doi:10.1078/1433-8319-00083CrossRef
30. Williams SE, Shoo LP, Isaac JL et al (2008) Towards an integrated framework for assessing the vulnerability of species to climate change. PLoS Biol 6:2621–2626. doi:10.1371/journal.pbio.0060325CrossRefPubMed

For further details log on website :
http://link.springer.com/article/10.1007/s11056-015-9481-9