Increased recovery in coarse‐root secondary growth improves resilience to drought in transition forests

Interaction of global change drivers affects forest resilience. Land‐use changes (land abandonment) and climate change (a higher frequency and intensity of droughts) are interacting in the Mediterranean Region. Components of resilience in secondary stem growth have been widely studied but, despite the importance of root systems in forest functionality and resilience, non‐previous studies have assessed them in coarse roots. In this study, we use Juniperus thurifera tree‐ring chronologies in coarse roots and stems to assess biomass allometry and tree resilience to drought events comparing two stages of a forest expansion gradient (mature forests and transition zone) in Alto Tajo Natural Park. We extracted cores of stems and coarse roots in 48 trees distributed in different developmental stages and calculated cross‐sectional area increments, root‐stem allometric relationship and resilience components for both organs in each individual for two drought events (2005 and 2012). Stem and root growth as well as its allometric exponent were higher in the transition zone than in mature forests. Both organs exhibited a trade‐off between resistance and recovery in mature forests but maintenance of higher values in the transition zone. Resilience did not show differences between organs being higher in the transition zone than in mature forests. However, relative resilience in roots in the transition zone was higher than in mature forests, without differences in stems between stages. Finally, the 2012 drought event showed a higher impact on the components of resilience than the 2005 drought event. Synthesis. This study extends the knowledge of root response to drought events and highlights the potential of land‐use legacies to reduce the negative impact of climate change by promoting increased root recovery after drought events in trees established in past agricultural lands.


| INTRODUC TI ON
Land-use changes and, particularly, land abandonment are the main factors that have influenced the increase of forest areas in Europe. Almost two-thirds of recent European forests occur in former agricultural land, especially in Eastern and Southern Europe (Palmero-Iniesta et al., 2021). Several positive and negative effects on ecosystem services have been described because of forest expansion . The process of expansion and densification of forest area is accompanied by the current increase in environmental adversity due to climate change, especially the growing mean temperatures and in the frequency and intensity of drought periods . Thus, the interaction between land-use changes and more frequent droughts associated with climate change, particularly in the Mediterranean region, affects forest functioning (Guerrieri et al., 2021) and forest resilience.
Forest resilience is the ability of a forest to maintain its state and functions after the impact of a perturbation (Holling, 1973). The study of the impact of drought on forest resilience using tree-ring widths has increased in the last decade (Kannenberg et al., 2020).
Tree ring data provide long-term records of the radial growth of individual stems and can be used to study tree response before, during and after drought events. This approach assumes that tree cores collected at breast height and the resilience indices derived from them, those developed by Lloret et al. (2011), reveal the whole-tree response to drought without considering other processes like carbon allocation to root growth (Doughty et al., 2014).
Roots are estimated to make up 20%-40% of the biomass of trees (Jackson et al., 1997). Root systems of woody species are composed of two types of roots: fine roots (<2 mm in diameter), which function are water and nutrients uptake and coarse roots (>2 mm) which are responsible for anchoring trees to the soil and serve as organs to transport water from deeper soil horizons (Brunner et al., 2015).
Although there is an increasing interest to study drought effects in the structure and growth of both coarse and fine roots (Kozlowski & Pallardy, 2002), coarse roots are less studied due to the difficulty of sample extractions. The analysis of root growth dynamics and the biomass allocation between roots and stems is, therefore, fundamental to a comprehensive understanding of drought and land abandonment effects in whole-tree drought resilience and forest functionality (Brunner & Godbold, 2007).
Biomass allocation response to a perturbation can be classified following two theories: the optimal partitioning theory (OPT) and the allometric biomass partitioning theory (APT; . According to APT, resource allocation patterns between different organs change with plant size and it is not affected by variation in the local environmental conditions (Enquist & Niklas, 2001;Müller et al., 2000). In contrast, OPT states that a plant always invests in improving the access to the current limiting factor. Thus, in drought events, plants would invest in root growth to uptake water (Bloom et al., 1985). Field and laboratory manipulation experiments, particularly in seedlings, have supported OPT, showing that many tree species respond to dry conditions with an increase in root-shoot ratio (Poorter et al., 2012). However, it has been shown that this increase is influenced by the severity of the stress and that only severe droughts promote C allocation to roots (Poorter et al., 2012). Other studies have analysed biomass allocation by studying the root-stem growth allometry relationship, showing unclear results: some studies have supported APT (Nikolova et al., 2021), OPT  or none of them .
Juniperus is an evergreen gymnosperm genus with broad geographical distribution, occurring in the Arctic, Africa, East Africa, Central Asia, Central America and South Asia, and it has significant ecological and socioeconomic importance in the Northern Hemisphere (Farjon, 2010;Tavankar, 2015). Moreover, adaptative architectural features in the species of this genus, such as extensive lateral root systems, large root/shoot ratios and tracheids with small diameters (Krämer et al., 1996;Martínez-Vilalta et al., 2002), confer them high resistance to drought and enhance their ability to persist in dry environments. In certain areas, juniper encroachment has been considered a negative process because of the invasive role of the genus over other species (Juniperus virginiana L. in USA;Torquato et al., 2020). In other areas, juniper expansion is at the expense of crop abandonment and contributes to increasing ecosystem services (Juniperus thurifera L. in Spain; Martín-Forés et al., 2020). In the last years, the number of works that study the ecological processes underlying the expansion of Juniperus forest in Mediterranean ecosystems has increased , but there are no studies that have analysed how this expansion affects to below-ground organs.
Due to the adaptative characteristics to drought of Juniperus species roots (Krämer et al., 1996) and the expansion that has been shown in previous studies , we use J. thurifera tree-ring chronologies both in coarse roots and stems to assess the biomass allometry and tree resilience to drought events in different stages of forest expansion gradient. Specifically, we analyse the following: (i) variation in secondary growth of stems and roots between two stages of gradient expansion forests, (ii) root-stem allometry between these two stages and (iii) differences in resistance, recovery, resilience and relative resilience between organs, stages of forest expansion gradient and drought events. We hypothesised that stems and roots of J. thurifera individuals show greater annual growth in the transition zone than in mature forests as previous studies of the species showed in stems . Thus, we expect a greater allometric exponent of trees in the transition zone than in mature forests due to the greater annual growth in the two organs in the transition zone. We hypothesize that components of resilience will vary between organs and stages in J. thurifera forest expansion. Drought will affect less the growth rates of trees established in past agricultural lands than in mature forests due to ecological mechanisms associated with past land uses (i.e. location of agricultural lands in flat areas and deeper soils, continuous amelioration of soil structure by agricultural practices and higher soil nutrient concentration) that could buffer new forests from severe drought stress (Vilà-Cabrera et al., 2017). Thus, we propose that higher intensity of drought in mature forests will promote OPT strategy (Poorter et al., 2012), whereas APT will prevail in the transition zone. Finally, adaptive characteristics acquired by trees that colonise open areas together with land-use legacies in the transition zone could promote an increase in resilience in trees established in this stage.

| Study area and species
This study focused on J. thurifera that is a dominant species forming low-density forests that can establish on poor, shallow, rocky soils and tolerates broad temperature range (high temperatures in summer and low ones and frost in winter) typical of Mediterranean continental climate regions (Gauquelin et al., 1999;Montesinos et al., 2009). Its distribution is restricted in the Western Mediterranean basin.
We conducted this study in Alto Tajo Natural Park where J. thurifera forests are expanding towards abandoned lands (Gimeno, Pías, et al., 2012 (Villellas et al., 2020). These studies classified the gradient in three stages: expanding front (areas of recently abandoned lands), mature forests (well-conserved forest) and transition zone between the other stages with intermediate characteristics related to tree density, tree size and age (for further information see Villellas et al., 2020;Acuña-Míguez et al., 2020). Due to the reduced tree size in the expanding front, we conducted this study in mature forests and transition zone (two plots per site). Previous studies showed differences between these two stages at the forest expansion gradient in tree size, age and tree density (Acuña-Míguez et al., 2020).

| Sample collection and tree-ring data
In each plot, we selected four trees based on the frequency of diameter class within each plot. We extracted one core from the stem of each tree at breast height according to the existing methodology (Nikolova et al., 2021;. Then, we excavated and selected two coarse roots of each tree considering the growth variability due to soil characteristics (i.e. stoniness or compaction) and extracted two cores of each coarse root in perpendicular orientation at 30-50 cm distance from the trunk using a Haglöf increment borer ( Figure S1).
We measured the diameter at breast height (DBH) of each tree and the quadratic mean diameter (calculated as the square root of the sum of square diameter at breast height of each stem of a tree; Stewart & Salazar, 1992) of all neighbouring trees in a radius of 6 meters for calculating the local basal area as a measure of competition of each tree. Moreover, we measured the diameter of the cored roots at the position of coring (Table 1; Table S1).
We longitudinally cut each core using a microtome and scanned them at 2400 dpi. We measured ring widths to an accuracy of 0.01 mm using the CooRecorder v9.3 software (Cybis Elektronik 2018). The cross-dating of individual series was checked using the CDendro v9.1 software and COFECHA program (Holmes, 1983).
We calculated the mean growth of the two radii of each core. Then, we calculated cross-sectional area increment (CSAI, cm 2 ) for each core using bai.out function of dplR package (Bunn, 2010). For the following analysis, we used the temporal series from 2000 to 2018 to keep most of the trees in the study. We calculated backwards diameter time series for each root and stem using cross-sectional area increments of each one. With these time series, we applied the basic allometric equation (Huxley & Teissier, 1936), which describes how the studied plant organs main root and stem change with plant size: where 0 is the allometric factor, 1 corresponds with the allometric exponent and dr i and ds i are the root and stem diameter for year i. 1 is the biologically relevant term, which covers both radial increment and plant proportions in the long term. The allometric exponent is 1 when plant growth is in a steady state (Nikolova et al., 2021;Poorter et al., 2012).

| Drought events and resilience indices
We calculated the 6-month SPEI (standardised precipitationevaporation index) from 1901 to 2018 and used 6-month SPEI from (1) ln dr i = 0 + 1 ln ds i , TA B L E 1 Mean ± SD of diameter at breast height (DBH), DR1 (root diameter 1), DR2 (root diameter 2) and local basal area (Local BA) of trees selected in each stage of forest expansion gradient and in each site  [Beguería et al., 2014]). To choose drought events that could impact J. thurifera growth, we focused on the most negative values in spring. Secondary growth initiates in April-May in this species (Camarero et al., 2010) and it has been shown that spring SPEI has an impact in this species ( For roots, we considered components of resilience value as the mean between the indices measured in the two roots of each tree.

| Statistical analyses
We used linear mixed models (LMMs) for each organ to analyse how cross-sectional area increment varied among stages. As fixed effect, we introduced stages of forest expansion gradient, years, the basal area and the interaction among stage and year. For the stem, the random effect was the tree nested in site (Equation 6) and for the root, the random effect was the root nested in tree and nested in site (Equation 7). Cross-sectional area increments and basal area were log-transformed to conform model assumptions.
where y ijl is cross-sectional area increments for stem, for the tree j in the where y rijkl is cross-sectional area increments for root, for the tree root ( We used an LMM to analyse the differences in root-stem allometry between two stages of forest expansion gradient. We introduced root diameter as response variable root and stem diameter, forest expansion stage and its interaction as fixed factors. We introduced root nested in tree as random intercept and stem diameter as random slope (Equation 7). Root and stem diameter were log-transformed to conform model assumptions.
where Droot ijk is root diameter, for the tree root i of tree j in the year k, Dstem ijk is the stem diameter for the tree root i of the tree j in the year k and Stage j is the stage of forest expansion of the tree j.
We conducted four LMMs (one per component of resilience) to analyse how components of resilience varied between organs, drought events and stages of forest expansion gradient. We set as fixed effects the interaction among stages of forest expansion gradient, organs and drought events and a plot and tree nested in plot as random effects.
We analysed if there were differences between two models that differed in random effects, as there were no differences, we selected the simplest one (it means plot as random effect; Equation 8). All components of resilience were log-transformed to conform to normality. In the case of relative resilience, it was transformed with natural logarithm +1 to remove negative values. In case of interactions were significant, we conducted post-hoc Tukey analyses.
where y ij is the log-transformed response variable (Rt, Rc, Rs, RRs) for the tree i in the plot j, Organ j (is either root or stem), Stage of forest expansion of the plot j and Drought is the drought event (2005 or 2012).
We reduced all models to those with the lowest Akaike Information Criterion AIC (the best or most parsimonious models) using the dredge function from the MuMIn R package (Barton, 2020). The method was set to maximum likelihood (ML) during the fixed-effect model selection phase, although the final models are presented using restricted maximum likelihood (REML; Kuznetsova et al., 2017). Models fit was visually checked to ensure model assumptions. We calculated marginal (i.e. the proportion of variance explained by fixed effects) and conditional (i.e. the proportion of variance explained by fixed and random effects) r 2 with the sjplot R package (Lüdecke et al., 2021). All statistical analyses were performed using R version 3.5.1 (R Core Team, 2021).

| Cross-sectional area increment
Linear mixed-effects model (according to AIC), which better explained variation in stem and root CSAI, was that with year, stages of forest expansion gradient and basal area as fixed effects. We found that basal area had a significant positive effect on stem and root Ig (Tables S2 and S3, respectively). Stem CSAI (Figure 2; marginal r 2 = 0.736; conditional r 2 = 0.932) and root CSAI (Figure 2; marginal r 2 = 0.760; conditional r 2 = 0.941) were higher in the transition zone than in mature forests.

| Root-stem allometry
We found that root-stem allometry varied between stages of forest expansion gradient (marginal r 2 = 0.476; conditional r 2 = 0.995; Figure 3). Allometric exponent was higher in the transition zone than in mature forest (1.619 and 1.306, respectively; Table 2).

| Components of resilience
Best linear-mixed effects model (according to AIC) that explained variation in resistance included year and the interaction between stages of forest expansion and organs (Table S4). We found that resistance was significantly lower in 2012 than in 2005 (Figure 4a). In mature forests resistance was higher in roots than in stems, but no differences between organs were found in the transition zone (Figure 4b). Trees in the transition zone showed higher stem resistance than in mature forests (Figure 4b; marginal r 2 = 0.225; conditional r 2 = 0.367). Best linear model for recovery included the interaction between years and organs and the interaction between stages and organs (Table S4) Best linear effects models for relative resilience included interactions between organs and years and between organs and stages of forest expansion (Table S4). Relative resilience measured in stems was higher in 2005 than in 2012 (Figure 6a). There were no differences between stems of trees established in different stages (Figure 6b

| DISCUSS ION
The stage of forest expansion gradient affects secondary growth of roots and stems as well as biomass allocation to roots, being (8) higher in trees established in the transition zone in both cases. The strategies of the trees to face drought events were different between the stages. In mature forests, we found a trade-off between resistance and recovery (higher resistance of roots and higher recovery of stems) while roots and stems of trees established in the transition zone maintained high values of both indices. We did not find differences between organs in resilience, but tree resilience was higher in the transition zone than in mature forests. However, relative resilience measured in stems was not different between stages, we found that relative resilience of roots was higher in the transition zone suggesting a greater relative resilience of trees in this stage.

F I G U R E 2
Mean ± SE for cross sectional area increment of stem and roots in mature forest (red points and bar) and transition zone (blue points and bar) along the studied temporal series. Values are log-transformed. Arrows indicate both drought events occurrence.

F I G U R E 3
Relationship between root diameter and stem diameter in mature forest (blue) and transition zone (red). Coefficients of model predictions were transformed to arithmetic scale.

| Cross-sectional area increments and root-stem allometry between stages of forest expansion gradient
Lower stem growth in J. thurifera trees established in mature forests has been previously observed in this system . Some studies have associated this decrease to competition Gimeno, Camarero, et al., 2012) while others have not found this relationship (Granda et al., 2013). In this study, we show a decrease in root and stem growth in mature forests without an effect of local basal area around each individual in the growth of both organs (Tables S5 and S6).
Forests that colonise abandoned fields can benefit from previous land uses (i.e. agricultural uses increase soil nitrogen levels; Nadal-Romero et al., 2018). Thus, we proposed that land-use legacies could reduce nutrient limitation and increase growth of roots and stems.
Biomass allocation to roots with size has been related previously to a decrease in competition (Nikolova et al., 2021). In our case, this pattern was not related to local basal area around each tree (Table S7). We found that the increase in biomass allocation to roots was higher in the transition zone than in mature forests.
The establishment in open areas in Mediterranean climate implies greater water stress during summer than other regions, triggered by high irradiance due to lower canopy cover. In water-limited environments, the allocation to fine roots tends to be strongly related to aboveground size (Magnani et al., 2002). In contrast, conifers increase stem growth at the expense of root growth under favourable soil moisture conditions (Thurm et al., 2016).
Moreover, previous studies in young trees have shown that biomass allocation to roots increases to capture water and nutrients to TA B L E 2 Estimated fixed effects for allometry model (Equation 8) for root diameter (ln dr) in relation to stem diameter (ln ds) and stage of forest expansion, being ds root and stem diameter, respectively

| Resistance and recovery: Different strategies of trees between stages of forest expansion
Our results support the idea that trees of J. thurifera established in different stages of forest expansion gradient displays different strategies to cope with drought. Although, this species is among the most drought-resistant conifers known (Olano et al., 2017), the difference F I G U R E 5 Model prediction of resilience for (a) the effect of the interaction between year and organs, (b) the effect of the stage of forest expansion gradient. In panel a, different lowercase letters show differences between organs for each year (p < 0.05), and different capital letters show differences between year for each organ (p < 0.05). In panel b, different letters indicate significant differences (p < 0.05).

F I G U R E 6
Model prediction of relative resilience for (a) the effect of the interaction between year and organs, (b) the effect of the interaction between the stage of forest expansion gradient and organs. In panel a, different lowercase letters show differences between organs for each year (p < 0.05), and different capital letters show differences between year for each organ (p < 0.05). In panel b, different lowercase letters show differences between organs for each stage (p < 0.05), and different capital letters show differences between stage for each organ (p < 0.05).
between root and stems indices in each stage indicates that the resistance and recovery to drought is organ and stage dependent. Gazol et al. (2017) showed that there is a global trade-off between resistance and recovery being the most resistant trees those that established in the driest areas. In the present study, we have shown that the trade-off between these two components of resilience can be found between organs in the same individual, at least in mature forests. Roots of trees in mature forests showed higher resistance, but lower recovery than stems. During drought events, trees in mature forests allocated more resources to roots (higher resistance). After drought events, mature forests trees showed higher stem growth than before the drought event (values of recovery >1). Thus, we suggest that OPT is occurring in mature forests with a greater investment in roots than in stems during drought that enables the opposite investment after drought due to a more developed below-ground hydraulic system that improves water acquisition. In the transition zone, we found that both organs maintained high rates of resistance and recovery without differences between organs, suggesting that APT was acting during and after drought.
We suggest that allometric partitioning in trees established in the transition zone is a plastic response of tree growth acquired through the process of colonisation of past agricultural lands. J.
thurifera juvenile trees of this species show a more efficient phenotype with greater water-use efficiency and greater growth rates in new forests than in mature forests (Benavides et al., 2022) and these differences are maintained in adult trees Alfaro-Sánchez et al., 2021). Higher efficiency in trees established in past agricultural lands could be related to amelioration of drought effects due to land-use legacies (Vilà-Cabrera et al., 2017), as previously shown for the same ecosystem, region and species (Gimeno, Camarero, et al., 2012). Thus, we propose that, although colonisation process of past agricultural lands could be limited by the higher irradiation due to less tree cover, land-use legacies (i.e. higher nutrient concentration and deeper soils) allow that trees established in the transition zone invest more in both organs to increase the likelihood to overcome the first summer droughts by improving the acquisition and use of the most limiting factor (water). On the other hand, trees established in mature forest would benefit from a greater tree cover, which reduces excessive irradiation. However, negative effects of tree cover during dry years could be happening in mature forests (Valladares et al., 2008) promoting OPT strategy in trees of this stage due to a greater impact of drought events (Poorter et al., 2012).
The different strategies followed by trees at each stage (APT in the transition zone and OPT in mature forests) caused changes in resilience among stages. In mature forests, the trade-off between resistance and recovery and the higher investment in roots than in stems during drought decrease the resilience of trees established in this stage without differences between organs. Higher recovery found in stems in mature forests might reflect a compensation strategy, that is, preferential carbon allocation to stem after drought to rebuild damaged xylem. However, OPT strategy of trees in this stage, and particularly, stems growth suppression during drought events, indicates a greater drought growth sensitivity that might led to chronic stress in the long-term (Zoblin, 2022). Thus, this strategy might cause higher sensitivity to drought in trees established in mature forests under a scenario of climate change that increase frequency and intensity of drought events (Gessler et al., 2020). Moreover, roots in the transition zone showed higher recovery than roots in mature forests. Hagedorn et al. (2016) showed that trees prioritise the investment in roots after drought to restore trees' capability to acquire water and nutrients. Thus, it seems that the optimal strategy to cope with drought is to maintain high rates of resistance and recovery in both organs, in other words, APT strategy.
Moreover, resilience components were dependent of drought events identity (2012 and 2005). The main difference between these two drought events was the duration (Figure 1). 2012 drought was longer, with autumn and winter of 2011 showing low SPEI values, depleting soil water reserves before the start of the growing season.
Previous studies of J. thurifera in this area has shown that tree-ring formation was strong dependent on previous autumn environmental conditions (Granda et al., 2013) what was associated to the capacity of conifers to store carbohydrates that can be used to grow the

| The importance of measuring both organs in relative resilience
Although components of resilience described by Lloret et al. (2011) have been widely used, most studies have not shown their results of relative resilience. Relative resilience is the component of resilience that considers the damage during the drought event (Equation 4).
When the value of this index is lower than 1 the drought effect persists after the disturbance (Lloret et al., 2011).
We found differences in relative resilience between the two stages of forest expansion in roots without differences in stems.
Previous studies in different species did not find differences in relative resilience between species (Granda et al., 2018), maybe due to the use of this component only in stems. Although, both organs in both stages showed that the drought effect persists after the event, roots in the transition zone showed the lowest drought legacy effect. Higher relative resilience can be reflecting a higher buffer capacity to recover or compensating positive effects of the impact due to a higher neighbour mortality that increase resource availability to surviving trees (Lloret et al., 2011). In this case, we showed that there is a higher buffer capacity to recover of roots in the transition zone ( Figure 4d). Our results showed the importance of measuring this component of resilience in both organs to avoid underestimation (in the transition zone) and overestimation (in mature forests) of the relative resilience of trees if only stems are used ( Figure 6).

| CON CLUS IONS
Forest growth and dynamics are increasingly driven by drought, which are raising their frequency and intensity in many already dry regions of the world. Moreover, forests are expanding and becoming denser in the Northern Hemisphere, mainly due to land abandonment. Hence, there is a need to understand the mechanisms of forest responses to drought in this type of 'new forests'. This study focused on the different responses of roots and stems in trees to drought events in different stages of Juniperus thurifera expansion gradient. Our results show that secondary growth is greater in both organs in the transition zone which could be causing a higher relative investment in roots in this stage. Moreover, our findings reveal that tree strategy to cope with drought differs between organs and stages. In mature forests, a trade-off between organs in resistance and recovery promotes a decrease in resilience whereas high values in these traits in both organs in the transition zone promote an increase. These differences could be due to the benefits of land-use legacies. Our results show the importance about measuring all secondary growth resilience components at stem and root levels since differences between stages only appear in relative resilience of roots. Stems do not show differences in relative resilience between stages but roots in mature forests are less relatively resilient which could cause a high vulnerability of trees established in this stage.
Thus, our study extends the knowledge of the resilience in forests undergoing expansion and highlights the potential of land-use legacies to reduce the negative impact of climate change by promoting increased root recovery after drought events in trees established in past agricultural lands.

AUTH O R CO NTR I B UTI O N S
Belén Acuña-Míguez contributed to the methodological approach, performed the field and laboratory work, led the analysis and the writing of the manuscript; Jose Miguel Olano contributed to analysis and to the manuscript production; Fernando Valladares contributed to the methodological approach and to the manuscript production; Miguel García-Hidalgo contributed to laboratory work and analysis and manuscript production; Andrés Bravo-Oviedo led the methodological approach and contributed to field work, data analysis and the writing of the manuscript. We also wish to thank Antonio Mas Barreiro, David López Quiroga and Marina Piquer Doblas for their valuable support in field, laboratory an statistical analyses.

CO N FLI C T O F I NTE R E S T
The authors declare no conflicts of interest.

PEER R E V I E W
The peer review history for this article is available at https://publo ns.com/publo n/10.1111/1365-2745.14024.

DATA AVA I L A B I L I T Y S TAT E M E N T
Datasets generated in the current study are archived on DIGITAL.