**4. Discussion**

### *4.1. Tissue- and Resource-Dependent Reallocation or Accumulation*

At the end of season, NSC reallocated from leaves to woody tissues (Figure 5a–d) and N accumulated only in stems (Figure 5g), whereas P and K seemed to store in stems and roots (Figure 5k,l,o,p), indicating tissue- and resource-dependent reallocation or accumulation (Table 4). *B. ermanii* leaves reallocated NSC to other tissues (Figure 5a; Table 4), leading to decreases in leaf NSC (Figure 1e vs. Figure 1a) but increases in NSC concentrations in shoots (Figure 1f vs. Figure 1b), stem sapwood (Figure 1g vs. Figure 1c), and roots (Figure 1h vs. Figure 1d), and thus an NSC accumulation in those storage tissues (Figure 5b–d). In line with previous studies, our results confirmed that leaf carbon compounds in deciduous trees were reallocated and stored in ligneous tissues, particularly in stems and roots [5,8]. Leaf carbon components recycled through remobilization from leaves to storage tissues before leaf senescence are particularly important for regrowth of new leaves in a leafless state of deciduous trees in early spring [5,9,10]. We found that the NSC remobilization efficiency of *B. ermanii* leaves ranged from 6% at high elevation to 33% at low elevation (Figure 5a). Chapin et al. (1990) proposed a carbon remobilization efficiency of ~10% based on an assumption that the loss of leaf weight was a direct result of carbon remobilization [5]. Similarly, Eckstein et al. (1998) proposed a carbon remobilization efficiency of 6%–13% for deciduous species [35].

Nitrogen (N) is the most important stored compound [36], paralleling to carbohydrates in supporting growth and reproduction [37]. Our results indicated that all tissue types decreased in N concentration during leaf senescence except stems (Figure 5e–h), suggesting that stems participated in the N accumulation rather than N translocation. Our results agreed with previous studies, showing that N remobilized from senescing tissues (e.g., leaves and shoots) for storage prior to leaf abscission [3,37,38]. As previously reported for *Betula pendula* Roth., the deciduous broadleaf tree tended to store N in roots and stem [11,12]. Our findings differ from coniferous evergreen species, which store N in the younges<sup>t</sup> age class of needle [11,13]. N stored in specific tissues with an individual species, dependent on leaf habit [37]. In addition, the mean leaf N remobilization efficiency of 40% (Table 5) was lower than that value reported by Aerts (1996) for deciduous shrubs and trees (mean 54%) [3]. In another study, Vergutz et al. (2012) showed that, based on a leaf mass loss correction, average nutrient resorption was 62% for N during senescence [39].

Apart from mobile carbohydrates and nitrogen (N), plants also store and remobilize macronutrients such as phosphorus (P) and potassium (K) [38,40]. By the process of leaf senescence, P and K is reallocated from leaves and shoots to and stored in roots and stems (Figure 5i–p). The mean efficiency of P remobilized from senescing leaves was 42% (Table 5), which is lower than that reported by Aerts (1996) for deciduous shrubs and trees (50% for N) [3]. Vergutz et al. (2012) even suggested that, in general, leaf average P resorption value was 65%, after correction for leaf mass loss [39]. Similar results were found for K. As K occurs in ionic form but not in macromolecule form [18], K is thus highly mobile and has been transferred to storage tissues (i.e., stems and roots) before leaf abscission. However, our remobilization efficiency from senescing leaves of 31% (Table 5) was much lower than published resorption efficiency. Chapin et al. (1990) and Vergutz et al. (2012) proposed 50 and 70%, respectively [5,39].

As previously reported for deciduous species, carbohydrate remobilization took place in the leaves, indicating that carbon components recycled from senescing leaves (Figure 5a) are used as a carbon source for new growth [5,41]. But unlike carbohydrates, nitrogen was recycled from senescing leaves (Figure 5e), shoots (Figure 5f) and roots (Figure 5h) to support new leaf and shoot growth, showing that senescing tissues play an important role as potential supplier of nitrogen. Besides, our findings were in agreemen<sup>t</sup> with other studies that demonstrate N storage in roots and stems in broad-leaved deciduous species [11,12]. P and K were reallocated from leaves and shoots to stems and roots at the end-season (Figure 5i–p), although these patterns differed from those of NSC and N mentioned above (Table 4). In *B. ermanii* trees, 6%–33% of the leaf mobile carbohydrates was recycled at the end-season. However, about half the N and P was remobilized from senescing leaves before leaves were shed, and 36% of the leaf K was resorbed during leaf senescence. Our results support the fact that resource remobilization is composed of a large storage of nutrients but a small storage of carbohydrates [5].

### *4.2. Elevational Effects on Resource Remobilization*

In the present study, we calculated the remobilization efficiency using the resource concentration within the same tissue type measured at two time points (end-season vs. peak growth season), which may provide insights into resource remobilization processes during leaf senescence. This calculation revealed that the end-season remobilization efficiency of NSC, N, P, and K from leaves to storage tissues tended to decrease with increasing elevation (Figure 5a,e,i,m), which does not support our hypothesis that expected an increase in remobilization efficiency at higher elevations, especially at the upper limit. In our previous study [25], we calculated the remobilization efficiency using the end-season difference in the resource concentrations between leaves (source) and shoots (sink), and found that the remobilization efficiency from leaves to shoots increased with increasing elevation. Using that method [25], we re-calculated the end-season remobilization efficiency from leaves to shoots for the present study and found that NSC remobilization efficiency had a bimodal curve of R% with a bigger remobilization efficiency (smaller R% value) at the upper distribution limit (Figure 6a). Leaf to shoot K reallocation efficiency did not show any response to elevational gradients (Figure 6d), and leaf to shoot N and P reallocation efficiency tended to be lower above the treeline compared to below the treeline (Figure 6b,c). This comparison suggests that a standard method for calculating the remobilization efficiency and thus for understanding the reallocation processes

of resources is urgently needed. Previous studies proposed that resource remobilization efficiency can be higher at low nutrient availability or in low temperature condition [14,16], representing an adaptation to stress environment [17]. However, Kutbay and Ok (2003) have demonstrated that the absolute and proportional N and P resorption efficiency did not change significantly along elevational gradients [22,29,31,42]. Many reports have revealed the importance of temperature on regulating leaf senescence and fall [18]. For example, leaf senescence and fall were delayed, in response to warmer temperature [43–45]. The timing or phenology of leaf abscission is considered as one of the reasons leading to altered resorption patterns of nutrients. To minimize such leaf phenological effects on resource remobilization, we did not collect end-season leaves from trees, instead, we collected the fallen leaves as end-season samples in the present study. However, *B. ermanii* in low temperature at the upper limit may have less time to complete nutrient resorption from leaves, which may result in a lower resorption of resources [18], as discussed in our previous study [25].

**Figure 6.** Carbohydrates, N, P and K remobilization efficiency (Mean ± 1SE) (R%) in *Betula ermanii* trees along the altitudinal gradients in Changbai Mountain (*n* = 5 for each elevational site). R% was calculated using *R*% = ((*Cl* − *Cs*)/*Cl*) × 100% = (1 − *Cs*/*Cl*) × 100%, where *Cl* and *Cs* represent concentrations in leaves and shoots at the end of growing season, respectively. Different letters display significant differences at the 0.05 level among elevations as determined by Tukey's HSD test (The larger the *R*-value, the lower the remobilization efficiency is).

Stem and shoot NSC accumulation increased but root NSC accumulation decreased with increasing elevation (Figure 5b–d). Given that the NSC accumulation increased in stems and shoots up to the upper limit, this process may act to resist the harsh cold alpine environment for survival in winter [46], because soluble sugars participate in cell osmotic regulation and prevent intracellular ice formation by decreasing the freezing point of the cytoplasm [47]. Moreover, accumulation of soluble sugars would prime stems for recovery after the stress is alleviated, and stem is an important organ not only for long-distance carbon transport, but also for the regulation of the tree's carbon balance [48].

Similarly, our results showed that N, P, and K were also accumulated in stems, especially in trees grown close to or at their upper limit (Figure 5g,k,o). Oleksyn et al. (1998) and Körner (1999) have reported that nutrient accumulation and conservation are adaptive responses that enhance metabolic activity and growth rates in harsh cold environment [49,50]. In such situations, the tree would internally store more recycling nutrients to recover from harsh environment because of losses in respiration. The reserve storage is a risk-aversion function to minimize risk of catastrophic loss [5]. The greater the risk (high frequency or large losses), the more the tree should invest in internally stored reserves.
