3.2.1. Differences between 2005 and 2006

In both the 2005 and 2006 growing seasons, leaves did not reach their maximum temperature when TA was warmest, but rather after several consecutive days without precipitation (4 and 7 days in 2005 and 2006, respectively). No large differences between the days when the maximum TL are reached in each year were found in terms of radiation, wind speed and VPD.

Given the same combination of traits, Anet,cum was systematically higher in 2006 than in 2005 growing season (Figure 2, top), while TL,max was systematically higher in 2005 than in 2006. Indeed, some combinations of traits led to thermal damage during 2005, while this threshold was not reached by any combination of traits during 2006. Conversely, no marked differences in terms of Anet,max and ΔT90 were found (Figure 2).

**Figure 2.** Comparison of model outputs for 2005 and 2006. Value of the four metrics Anet,cum (in μmol m–2), Anet,max (in μmol m–2s–1), TL,max and ΔT90 (in ◦C) obtained for the 100 combination of traits under the growing conditions 2005 and 2006. The dashed line corresponds to the 1:1 line.

#### 3.2.2. Future Growing Conditions

To investigate the effects of likely future conditions, increases of temperature of 0◦ C to 10 ◦C with respect to current conditions were combined with values of soil water potential ranging from −0.5 to −3 MPa. Those changes were applied by considering both a standard day and a warm day of the 2005 growing season as baselines.

Taking the currently normal thermal conditions as baseline, combination 3 (HP-HR) led to the highest assimilation rates under TA lower than 18◦C (Figure 3 bottom row). For TA below 16◦C, TL reached temperatures similar to combinations 2 (HP-LR) and 4 (LP-LR). Combination 1 (LP-HR) led to the highest difference between TA and TL, while combinations 2 (HP-LR), 3 (HP-HR) and 4 (LP-LR) showed similar TL values over a wide range of TA conditions (Figure 3 top row). TL was not sensitive to Ψ<sup>s</sup> in any of the four combinations. The Anet contour curves showed a parabolic shape with a vertex in TA equal to 8◦C, 10◦C, 10◦C and 12◦C for combination 1, 2, 3 and 4, respectively.

**Figure 3.** Leaf temperatures (top row; in ◦C ) and net assimilation rates (bottom row; in μmol m–2 s–1) for different air temperatures and soil water potentials. Four trait combinations are explored (Table 2). The lowest air temperature corresponds to the median day for the growing season 2005. As a term of comparison, the left column shows the frequency of the air temperature at midday reported in the study site during the growing seasons from 1996 to 2016. LP-HR: Low Productivity-High Risk of thermal damage; HP-LR: High Productivity-Low Risk of thermal damage; HP-HR: High Productivity and High Risk of thermal damage; and LP-LR: Low Productivity and Low Risk of thermal damage.

For the warm day baseline conditions, combinations with high lt—i.e., combinations 1 (LP-HR) and 3 (HP-HR)—were more likely to result in thermal damage, with TL exceeding TCRIT even under current thermal conditions (Figure 4 top row). Due to this high TL, the assimilation rates were also reduced to the point that combination 3 (HP-HR) showed lower Anet even though its maximum assimilation rate was higher than the corresponding one in combination 4 (LP-LR) (Figure 4, bottom row). Conversely, combination 2 (HP-LR), and to some extent combination 4 (LP-LR), showed higher thermoregulation capacity (i.e., lower TL) and higher assimilation rates. When compared, combination 2 (HP-LR) appeared to be more beneficial, resulting in lower TL and higher assimilation rates. Note that for all four combinations, an increase in TA with respect to the warm day baseline is detrimental to plant productivity, as Anet is reduced and leaves experience thermal damage already under current

conditions (combination 1 and 3), or for temperature increases of 1 to 2.5 ◦C (combination 4 and 2, respectively). TL was sensitive to Ψ<sup>s</sup> under dry conditions, particularly in combination 2 (HP-LR) (Figure 4 top row). Contrarily, in all combinations, Anet rates decreased as did water availability (Figure 4 bottom row).

**Figure 4.** Leaf temperatures (top row; in ◦C) and net assimilation rates (bottom row; in μmol m–2 s–1) for different air temperatures and soil water potentials. The baseline temperature corresponds to the warmest day of the growing season 2005. All the other parameters are as in Figure 3.

Regarding the effects of plant traits on soil water depletion, similar Ψ<sup>s</sup> emerged among the four trait combinations after the same number of consecutive days without precipitation. For all four combinations, after intervals shorter than 15 consecutive days without precipitation, soil water potential (i.e., lower Ψs) was only slightly affected by increasing temperatures. The lowest Ψs—around –0.9 MPa after more than 25 consecutive dry days—was reached by combination 2 (HP-LR) and 4 (LP-LR) (Figure 5).

**Figure 5.** Soil water potential (upper panels; MPa) and cumulated net assimilation (lower panels; μmol m–2 s–1) for each focal trait combination (Table 2) for specific TA (y-axis) and after a certain number of consecutive days without precipitation (x-axis). The baseline temperature corresponds to the warmest day in 2005. The rest of the growing conditions correspond to those observed in the study site during the same period within the growing season of 2005.

#### **4. Discussion**

#### *4.1. Role of Traits: Potential Mechanisms Explaining the Dominance of Plant Traits*

The GSA allowed identifying the role of six traits affecting net CO2 assimilation, Anet, leaf temperature and TL, and hence the risk of thermal damage. As expected, both the maximum and cumulated net CO2 assimilation are strongly dependent on the maximum carboxylation rate, VCMAX,25. They are also affected by the effective leaf thickness, lt, although less markedly. In particular, higher lt values were also associated with higher cumulated net CO2 assimilation, Anet,cum, but only up to lt ~0.1m, above which this trend reverts. In general, higher lt leads to higher TL, by reducing the leaf boundary layer conductance gb [60]. Increase in TL may have opposite effects on Anet, depending on the initial temperature and the species: when TL is below the temperature that maximizes net CO2 assimilation, an increase in TL stimulates Anet [11]. In our results relative to current climates, TL was generally below such threshold, so that higher lt values were associated with higher Anet,cum. Nevertheless, as apparent from our results, there is a maximum lt, above which increases in lt are no longer beneficial for CO2 assimilation because of an insufficient leaf cooling. Note that most boreal forests have needles, i.e., low lt. It is important to note that plants grown in high light generally have thick leaves to protect them from high-irradiance damage [61,62]. Conversely, in low light available conditions (frequent in boreal forests), leaf thickness reduces to maximize the light capturing area and reduce self-shading. As such, the low lt typical of many boreal species enhances light use, and hence potentially net CO2 assimilation via light capture as opposed to optimal temperature for photosynthesis. Finally, within the range of values observed in mid-to-high latitudes (shaded area in Figure 1), the two

parameters of the photosynthesis model (g1, β), and both albedos (αPAR and αNIR) did not have any conspicuous influence on the net assimilation rates.

Regarding the thermal metrics, the range of high photosynthesis, ΔT90w, was equally influenced by the traits lt, g1 and VCMAX,25. Conversely, the maximum temperature TL,max, was mostly influenced by lt, while g1 and VCMAX,25 played secondary roles. In particular, lower lt values and higher g1 values at a given VCMAX,25 lead to higher boundary layer and stomatal conductances, respectively, preventing thermal damage and resulting in lower TL. Similarly, higher VCMAX,25 were associated to higher gs, which in turn enhances the cooling effect of transpiration, resulting in lower TL. This mechanism might explain the effect of VCMAX,25 on TL,max. However, with focus on values observed in mid-to-high latitudes (shaded area in Figure 1), the role of VCMAX,25 in regulating maximum TL diminished. Therefore, under warm conditions, the cooling effect of the transpiration process is generally neither benefited from nor hampered by the assimilation capacity of plants, but rather regulated by the traits lt and g1.

There is no evidence in the literature of how ΔT90 changes according to plant traits, to the best of our knowledge. The width of the temperature range that realizes >80% of the maximum photosynthetic rate varies among plant functional types, in particular at low growth temperatures [1,63,64]; however, thus far these differences have not been explained based on specific traits. Our results suggest that VCMAX,25, lt and g1 are the key traits that determine ΔT90, possibly through the regulation of stomatal and boundary layer conductances. The influence of VCMAX,25 might also be due to the dependence of VCMAX on TL per se, as discussed in the Supplementary Materials. As apparent from Figure 1, leaves that are well coupled with the atmosphere (i.e., high gb and gs and, therefore, high g1 and low lt) correspond to wider ΔT90, whereas leaves less coupled with the atmosphere present narrower ΔT90. This might mirror two different plant strategies. Plants that are well coupled with the atmosphere respond more easily to changes on the atmospheric conditions so that TL fluctuates in a wider range than plants that are less coupled with the atmosphere. Therefore, plants that have lower assimilation rates and are well coupled have a wider range of optimal temperatures so that the lower assimilation rates and wider TL fluctuations are compensated by a wider range of semi-optimal conditions. In contrast, the ones with higher assimilation rates and less coupled with the atmosphere present a narrower range of optimal temperatures.

Briefly, lt, VCMAX,25 and, to some extent, g1 are key for both assimilation rates and thermoregulation capacities, mainly via the regulation of stomatal and boundary layer conductances. With focus on mid-to-high latitudes, while VCMAX,25 has no large influence on the maximum temperature reached by plants under warm conditions, it is key on their thermoregulation capacity under normal conditions (Figure 1). Moreover, over this region, g1 has no effect on assimilation rates, which are completely dominated by VCMAX,25. Finally, the traits β, αPAR and αNIR seem to play secondary roles for all four metrics analyzed here and they will therefore not be further discussed.

It is important to acknowledge that the results of this analysis might slightly change if other growing conditions were to be used. However, this potential limitation is mitigated by the method employed to split the 100 combinations of traits between Group 1 and Group 2 (Section 2.3.2), which identifies suitable threshold irrespective of the model outputs. As such, similar conclusions on the relevance of the different traits would be drawn when considering other realistic growing conditions. A further aspect not accounted for here is thermal acclimation. Hence, assimilation rates and thermoregulation capacities and how they are influenced by the analyzed traits might change along with growth temperature, even in existing leaves, at scales of few days to weeks [1,63]. This might represent a limitation when comparing species exposed to contrasting growing conditions and/or with different acclimation capability. However, this analysis still provides key information by identifying the most dominant traits, their inter-relation and the potential mechanisms that explain plant responses.

#### *4.2. Role of Growing Conditions: the Timing of Precipitatio A*ff*ects the Risk of Thermal Damage*

In combination with its inherent traits, plant CO2 assimilation and the risk of thermal damage are also determined by the environmental conditions. The contrasting weather conditions of the growing seasons 2005 and 2006 showed how the same combinations of traits led to different values for the four metrics. While no substantial differences were found in terms of maximum net CO2 assimilation rate, Anet,max, and the range of high photosynthesis, ΔT90, substantial differences emerged in terms of maximum temperature, TL,max and cumulated net CO2 assimilation, Anet,cum (Figure 2). Specifically, despite the lower mean air temperature and higher precipitation of 2005 over 2006, warm periods were more damaging in the growing season 2005 than 2006 regardless of the combination of traits (Figure 2). This difference can be ascribed to the timing of rainfall events: in 2005, the longest dry spell occurred during the warmest days, while this was not the case in 2006. As a result, in 2006, plants with sets of traits that reached thermal damage under growing conditions of 2005 could cope with similar or higher air temperatures and even benefit from these slightly warmer conditions, as photosynthetic capacity was enhanced and water was available for evaporative cooling to stave off the risk of thermal damage.

This result also emerges when exploring in more detail the combined effect of temperature and water availability as part of the future scenario analyses. While leaf temperature, TL, was sensitive to soil water potential, Ψs, under warm conditions (Figure 4, top row), TL did not show any substantial change with Ψ<sup>s</sup> under normal ones (Figure 3, top row). Moreover, higher air temperature enhances soil water depletion but only under warm conditions (Figure 5, top row). These conclusions are not affected by the length of the growing season, which was held constant, since the aim is to identify differences in assimilation rates and leaf temperatures for different trait combinations but given the same abiotic conditions. The results highlight the relevance of the timing of high temperature and lower water availability and the importance of considering these events in conjunction [65–68]. The joint effects of high temperatures and low water availability are expected to become even more important in the future since high temperatures and water deficiency during the growing season are likely to become more frequent in boreal regions [69]. Hence, there will be a likely increase in the probability that boreal forests will need to cope with more severe combination of heat waves and droughts [15].

#### *4.3. Interactions of Traits and Growing Conditions: Most Suitable Traits for Enhanced CO2 Assimilation and Reduced Risk of Thermal Damage Under Current and Future Climates*

Within the general pattern discussed above, specific trait sets can reduce or enhance thermal risks in specific climatic conditions. Therefore, attempts to identify the set of traits that could maximize productivity while preventing thermal damage under current and future conditions necessarily require the joint analysis of traits and specific growing conditions. Thus, we tested four combination of traits that presented contrasting assimilation and thermal responses during the 2005 growing season—a normal season at the reference site—and analyzed their response under multiple scenarios of future climatic conditions. The values for each trait within each combination are restricted to those observed in boreal tree species—in line with our focus on boreal forests—to account for the fact that trait values per se are related to climatic conditions. For example, global VCMAX,25 distribution has been recently proved to be mainly explained by climate [70].

As discussed in Section 4.2, the combination of low water availability and heat stress causes a disproportionate damage compared to each stress component occurring in isolation [65,66,71–73]. Due to the compound effect of low water availability and high temperatures,the focus of the multi-scenarios assessment was on these abiotic stressors, holding constant (i.e., as observed in 2005) the other climatic variables (chiefly, relative humidity, wind speed and atmospheric CO2 concentration). Nevertheless, reduced wind speed may lead to a further decoupling of leaf and air temperature, with potentially high and damaging temperatures in sunlit leaves. Declines of long-term wind speed (stilling) have been reported in both hemispheres [74,75]. Therefore, all else being equal, even higher leaf temperatures, TL, are to be expected as the result of the reduced cooling due to stilling, in particular for leaves with

large effective leaf thickness lt. Similarly, the stomatal closure potentially caused by enhanced CO2 concentration may further increase TL.

Our model results clearly show that, during the warmest period of the growing season, future warmer and drier conditions may cause reductions in net CO2 assimilation Anet regardless of plant traits (Figure 4 bottom row). Conversely, during periods with lower air temperatures, warming temperatures might be still beneficial (Figure 3 bottom row). However, the analysis of temperature data over the period 1996–2006 (upper row histogram in Figure 3) shows that, for the site of reference, higher temperatures were either not beneficial for productivity or even harmful in 75% to 93% of the days of the growing season depending on the trait combination. Similarly, future warmer and drier conditions may substantially increase the risk of thermal damage regardless of the plant traits. Nonetheless, how damaging these new conditions are depends on the trait set (Figure 4, upper row).

The GSA clearly shows that high photosynthetic capacity (i.e., high maximum carboxylation rate, VCMAX,25) is needed to ensure high Anet; small lt is the key trait to prevent thermal damage. Under the growing conditions of 2005, enhancement of cumulated net CO2 assimilation Anet,cum emerges when VCMAX,25 increases, as apparent when comparing the Anet,cum for combination 1 (Low Productivity-High Risk; LP-HR) and combination 4 (Low Productivity-Low Risk, LP-LR) against combinations 2 (High Productivity-Low Risk; HP-LR) and 3 (High Productivity-High Risk; HP-HR) (Table 2). Indeed, under standard conditions, the combination with the highest VCMAX,25 (combination 3; HP-HR) showed the highest Anet (Figure 3 bottom row). However, this enhancement does not hold under warm conditions: in this case, assimilation rate is also limited by leaf temperature as the optimal temperature for photosynthesis is exceeded (Figure 4 bottom row). Temperatures above this threshold might constrain assimilation rates to the extent that plants with high VCMAX,25 can show lower assimilation rates than plants with lower VCMAX,25, as apparent by comparing combination 4 (LP-LR) versus combination 3 (HP-HR) under the warmest conditions within the growing season. This pattern is also partly due to the sensitivity to temperature of VCMAX increasing with VCMAX,25. Moreover, lower VCMAX,25 was also associated with broader the range of high photosynthesis ΔT90, i.e., a broader range of conditions that allow high assimilation rates. In fact, only those plants with the ability to be well-coupled with the atmosphere (i.e., low lt) can keep high assimilation rates as shown by combination 2 (HP-LR) and, to a lower extent, combination 4 (LP-LR) (Figure 4 bottom row), even though the former has substantially higher assimilation rates and lower TL than the latter.

Nonetheless, combination 3 (HP-HR), which has high lt, was the one with the highest Anet,cum within the growing season 2005. This is because, up to air temperature TA around 18 ◦C, combination 3 (HP-HR) shows the highest Anet (Figure 3 bottom row). Based on the observed TA at midday in the reference site during the growing seasons in 1996-2016, TA was lower than 18 ◦C in the 70.3% of days, making combination 3 (HP-HR) the most appealing one in terms of overall productivity. However, this combination can also lead to thermal damage (i.e., TL exceeding the critical temperature for leaf damage, TCRIT; Section 2.3.2) for TA ≥26 ◦C, particularly under limited water availability (Figure 4 top row). The maximum TA at midday reported in the reference site exceeded this value in half the years between 1996 and 2016 (Figure 4, top left). Therefore, high lt might be beneficial for assimilation over the growing season at the expense of increasing the likelihood of thermal damage, particularly under water stress. Indeed, across all plant types, leaf size is distributed geographically according to a combination of the mean temperature of the warmest month of the year and the mean annual precipitation: higher mean temperature in the warmest month and lower mean annual precipitation correspond to smaller leaf sizes [76]. Thus, the selection of lt is not straightforward since it might represent a trade-off between productivity and prevention of thermal damage. The net result depends largely on the climatic conditions.

As explained in the Methods, the trait combinations were generated randomly because this study aimed to identify the role of each trait when combined with other traits and under specific growing conditions. Thus, the trait combinations do not to represent specific species. However, the combination of traits that leads to the highest thermal damage risk and/or is more likely to experience sub-optimal

conditions for assimilation can be used to speculate which species are likely to be more vulnerable to future growing conditions. For example, *Pinus sylvestris* L. and *Picea abies* L.—the two dominant species in Northern Europe - have small and rounded needle leaves (i.e., low effective leaf thickness lt) in common but *Pinus sylvestris* L. usually exhibits higher photosynthetic capacity (i.e., higher VCMAX,25) than *Picea abies* L. e.g., [77]. Although our results are not aimed to be representative of specific boreal species, *Pinus sylvestris* L. can be considered similar to combinations 2 (HP-LR) and *Picea abies* L. to combination 4 (LP-LR). As such, based on our conclusions, one could expect *Picea abies* L. to be more vulnerable to warming climates than *Pinus sylvestris* L., even suffering thermal damage under current conditions. Indeed, negative effects of warming on *Picea abies* L. species have been already reported [78], while *Pinus sylvestris* L. appears less sensitive to warming [79,80].

#### **5. Conclusions**

Understanding how specific plant traits, and combination of traits, prevent or enhance heat stress and C uptake is vital to predict how projected increases in frequency of heat waves and droughts in future climate may affect boreal forests. To disentangle the role of plant traits on thermoregulation and C uptake, we focused on six traits: the maximum carboxylation rate, two parameters that regulate the stomatal conductance and its sensibility to water stress, the effective leaf thickness and the PAR and NIR albedos. Four performance metrics related to leaf temperature and assimilation were evaluated, with a focus on the growing season.

Among the analyzed traits, photosynthetic capacity (as represented by maximum carboxylation rate at 25 °C, VCMAX,25) and the effective leaf thickness, lt, were the dominant ones regarding both thermoregulation and assimilation. Higher values of VCMAX,25 are needed to enhance assimilation under current and future conditions. To prevent thermal damage, high VCMAX,25 should be combined with low lt. However, the selection between low or high lt is not straightforward since lt seems to represent a trade-off between thermal damage prevention and productivity.

Moreover, the climate change scenario analyses highlighted that the projected joint changes in temperature and water availability needs to be considered in our prognosis of future boreal forest wellbeing, because combination of traits that prevent thermal damage under current growing conditions will not be able to limit the occurrence of thermal damage under warmer and/or drier conditions. Likewise, substantial differences were observed when considering currently normal versus warm conditions within the same growing season. This suggests that trait selection should not only rely on the overall productivity over the whole growing season but should also consider specifically the warmest period.

Further analyses exploiting databases of traits and how they are distributed geographically and regional projections of growing conditions, combined with the understanding of the role played by plant traits provided by our results, can support the identification of species and regions most vulnerable to climate change, where appropriate forest management should be focused.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/1999-4907/10/5/398/s1, Supplementary Materials document.

**Author Contributions:** This article is the result of the joint effort between G.R.-P., G.V. and S.L. The tasks were divided as follows: conceptualization, G.R.-P., G.V. and S.L.; methodology, G.R.-P. and G.V.; software, S.L.; validation, G.R.-P. and G.V.; formal analysis, G.R.-P.; investigation, G.R.-P. and G.V.; resources, G.V. and S.L.; writing—original draft preparation, G.R.-P.; writing—review and editing, G.R.-P., G.V. and S.L. supervision, G.V. and S.L.; and funding acquisition, G.V.

**Funding:** This research was supported by the Swedish Research Council Formas through grant 2018-01820 and by the Swedish government through the project Trees and Crops for the Future (TC4F). S.L. was also supported by the Academy of Finland, through the Academy Research Fellow project CLIMOSS (No. 296116 and 307192).

**Conflicts of Interest:** The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.
