This study was conducted in the Xilinji Forestry Bureau, which belongs to Mohe city in the Great Xing’an Mountain area of northeastern China. The climate is characterized by a long and severe winter, with mean annual air temperature ranging from –6°C to 1°C (Hu et al., 2017), and the mean annual precipitation is 500 mm. The dominant tree species are the Larix gmelinii (Rupr.) Kuzen, Pinus sylvestris Linn. var. mongholica Litv., Picea koraiensis Nakai, Betula platyphylla Suk., and two species of aspen (Populus davidiana Dode and Populus suaveolens Fisch.). Understory shrubs are dominated by Ledum palustre Linn., Vaccinium vitis-idaea Linn., Rhododendron dauricum Linn., Vaccinium uliginosum Linn., and Eriophorum angustifolium Honck (Meng et al., 2017). Soils are classified as brown soil (Wrb Iwg, 2015). The Great Xing’an Mountains are usually affected by natural wildfire disturbances due to the accumulation of combustible matter on the forest floor. Fire regimes are characterized by surface fires mixed with stand-replacing crown fires (Liu et al., 2012a). Records of fire provided us with the background to undertake experiments in this area.

Fire severity refers to the severity of organic material consumed or vegetation mortality directly caused by fire (Lentile et al., 2006). Fire severity follows the standard of Composite Burn Index (CBI) assessment protocol (Key, 2006; Lentile et al., 2006). Specifically, we visually estimated the changes in coarse woody debris, black carbon, char height, mortality rates of tall trees, and the proportions of fallen trees (Key, 2006; Boby et al., 2010; Fang and Yang, 2014). After the investigations, we found that the burn proportion was 41–60% and was viewed as moderate fire severity. Based on the precise historical records, we selected five treatments with different recovery periods, including four burned treatments (the fire occurred in 2015, 2007, 1997, and 1987, respectively) and one unburn treatment, and each treatment had three replications (three plots per treatment). The interval of each plot was more than 100 m to avoid the spatial autocorrelation between plots. During July and August 2017, 15 plots (20 × 20 m) were established, with 12 plots in the burned treatments (2a, 10a, 20a, and 30a, respectively) and the other 3 plots in the unburned treatment. The treatments were presented as at year 2, year 10, year 20, and year 30 after recovery and unburned treatment hereafter. In our study, 2a and 10a were viewed as early recovery periods, 20a and 30a as the medium recovery periods, and unburned treatment as the long recovery period. Considering the effects of environmental characteristics, there were no significant differences in soil bulk, slope aspect, slope position, and altitude among the selected plots (more details of the data information in Supplementary Table 1). The study area is located in the cold temperate continental climate. The forest type of treatment was L. gmelinii forest. These treatments had similar topography and environmental factors.

Five litter samples were collected from five points (four vertices and the center) in each plot. The aboveground plant material and live roots were removed before soil sampling. Three soil cores were collected from each point at a depth of 0–20 cm using a metal auger with an inner diameter of 5 cm. Fresh samples were placed in polyethylene ziplock bags, stored in a cooler with ice, and transported to the laboratory within 8 h, where they were preserved at 4°C. The chemical analyses of the soil and litter samples were completed in 10 days after sampling.

Compared with the unburned treatment, the soil N concentration decreased after a wildfire (Figure 2A, P < 0.05), whereas soil P concentration significantly increased (Figure 2B, P < 0.05), especially in early periods of recovery (2a and 10a). The litter N concentration decreased at 2a and was the highest at 20a recovery (Figure 2C, P < 0.05). The litter P concentration significantly decreased at 2a compared with the unburned area and then increased at 10a (Figure 2D, P < 0.05).

At the species level, the C, N, and P stoichiometries showed significant differences among five recovery periods (Figure 3, P < 0.05), and the concentrations of these elements were higher in the leaf than in the fine root. The leaf C concentration was significantly lower at the early recovery period (2a) than those at other recovery periods, whereas fine root C concentration was significantly lower at the medium recovery period (30a; Figure 3A, P < 0.01). Leaf N was higher at 10a recovery. Compared with unburned, fine root N concentration significantly decreased after a wildfire (Figure 3B, P < 0.05). The P concentration of leaf significantly decreased at 10a (Figure 3C, P < 0.05). The C:P ratio of leaf and fine root significantly increased at 10a (Figure 3E, P < 0.01). Moreover, the N:P ratio of the leaf significantly increased at 10a. The mean value of leaf N:P ratios was lower than 14 under the unburned treatment but was higher than 16 at 10a (Figure 3F, P < 0.01).

The C, N, and P stoichiometries at the community level showed variations among five recovery periods (Figure 4). Compared with the unburned treatment, leaf C concentration significantly declined at 2a, while the C concentration of fine root decreased significantly at 30a (Figure 4A, P < 0.01). Similarly, the leaf N concentration at the community level was the highest at 10a. However, fine root N concentration decreased significantly at 30a as compared with unburned treatment (Figure 4B, P < 0.05). The fine root P concentration decreased significantly at 10a compared with the unburned treatment (Figure 4C, P < 0.05). The leaf N:P ratios at the community level also increased significantly at 10a, and the mean value was higher than 16 (Figure 4F, P < 0.01).

The C, N, and P allocation slope exponents of fine root vs. leaf at the species level were significantly different among the recovery periods (Figure 4 and Supplementary Table 1). The slopes of C concentration showed an allometry pattern (b < 1) at 2a after when it exhibited an isometry pattern (b = 1; Figure 5A, P < 0.05), indicating that the allocation of leaf C would change more in the early recovery periods. All slope exponents of N concentration were larger than 1 (b > 1), indicating that the allocation of root N would change more than the leaf (Figure 5B, P < 0.05). As for the P concentration, the slope exponents transformed from b > 1 in early recovery periods (10a) to b < 1 in medium recovery periods (Figure 5C, P < 0.01). The slope exponent of C:N in the long recovery period (unburned) decreased as compared with those of early recovery periods. The slope of C:P and N:P was generally smaller than 1.

At the community level, C, N, and P allocation slope exponents between fine root and leaf also showed differences among the five recovery periods. The allocation slope exponents of C concentration between fine root and leaf showed allometry (b > 1) after burn (Figure 6A, P < 0.05). As for N and P concentrations, the slope exponents (b) were smaller than 1 at 2a then transformed to larger than 1 after 10a recovery (Figures 6B,C, P < 0.05). Interestingly, all slope exponents of plant C, N, and P concentrations at medium-term recovery (20a) showed clear allometry (b > 1), which further exhibited a transformation from b < 1 to b > 1 at 20a and 30a. This transformation indicated that more C, N, and P changes in fine root with an increased recovery period in comparison with those in leaf at the community level (Figures 6A–C, P < 0.05). The slopes of C:N and C:P all showed an allometric pattern, which transformed the slope from smaller than 1 to larger than 1 with increasing recovery periods.

In our study, the changes in soil and litter nutrient concentrations largely varied among the different recovery periods, for example, litter N, litter P, and soil N concentrations decreased at year 2 after recovery. Wildfire altered the soil nutrient pools and reduced litter nutrient concentrations (Tufekcioglu et al., 2010; Michalzik and Martin, 2013; Holden et al., 2016). Wildfire incinerates the majority of litter on the forest floor, where the amount of nutrients volatilizes with the destroyed vegetation (Toberman et al., 2014; Scalenghe et al., 2015). Limited nutrient supplement hampers aboveground growth and lowers the nutrient concentrations of litter (Nobles et al., 2009). Soil total N decreased in early recovery periods, which is consistent with a previous study that wildfires with moderate severity caused the declines in soil N pool (Wan et al., 2001). In contrast, significant increases in soil P concentration were observed in our study, which may be because the P is hard to volatilize under moderately severe fires, and fires can promote the P released from soil (Butler et al., 2018). The soil and litter N concentrations were increased in the medium recovery periods (Figure 2). With increasing recovery periods, the increases of soil and litter N may be caused by the restoration of the understory community, which is consistent with the increasing soil and litter nutrient concentrations with higher species diversity (Scalenghe et al., 2015).

An important issue of plant growth is its response to the changes of N and P supplements in burned ecosystems (Tarvainen et al., 2016). The total C, N, and P stoichiometries of leaf and fine root showed variations among five recovery periods regardless of the levels (Figures 3 and 4). The most striking result was that the stoichiometry of the leaf and fine root at year 10 after recovery was significantly different from others. At the species and community levels, leaf N concentrations were the highest, and the P concentrations of leaf and fine root were the lowest at year 10 after recovery. Although soil nutrients substantially decreased after being burned, plants adopt a growth strategy in response to severe environmental changes (Song et al., 2021). These flexible growth strategies of the plant would change resource allocations among organs to enable them to adapt to the nutrient-insufficient environment (Wright and Sutton-Grier, 2012; Song and Liu, 2019). The results of our study show that the C, N, and P concentrations were higher in leaf than in fine roots. Plants usually promote the photosynthetic efficiency of the leaf and allocate more resources to the aboveground (Ordoñez et al., 2009), which promotes leaf N utilization in our study. Additionally, to fully use light after a wildfire, plants can promote leaf photosynthesis and allow more nutrient investments to leaf (Muqaddas et al., 2015), which can be demonstrated in the higher litter nutrient concentration in our results. Leaf N absorption efficiency rapidly increases during the period of nutrient insufficiency, and thus, plants significantly promote leaf N concentrations (Scoffoni et al., 2011).

The decline of plant P concentration at year 10 after recovery may be due to the massive appearance of regenerating species that have low P concentrations (Yu et al., 2011; Shenoy et al., 2013). Moreover, in early recovery periods, soil organic acid secretion increases soil acidification, causing a huge loss in soil available P (Butler et al., 2017), which then decreases root P concentrations (Hume et al., 2016). Compared with unburned treatment, leaf C concentration was significantly decreased at year 2 after recovery, whereas fine root C concentration increased (Figures 3 and 4). At the species and community levels, a previous study also observed that the aboveground biomass decreased during early recovery periods (Nave et al., 2011). In the context of soil nutrient deficiencies, the regenerating tree species will synthesize more C to root growth to absorb nutrients under disturbance conditions (Mo et al., 2010; Kong et al., 2015).

The plant C:nutrient ratios are not constant after wildfires (Cui et al., 2010; Pellegrini et al., 2015), which was also observed in our results with significant changes of C:P and N:P ratio, especially at year 10 after recovery. More importantly, the N:P ratio at the species and community levels showed a similar pattern that the leaf N:P ratio was higher than 16 at year 10 after recovery, suggesting that plant growth tended to be a P limitation. This result of leaf N:P ratios was consistent with the previous observation under fire (Bai et al., 2013). N and P are usually the limiting nutrients for plant growth in forest ecosystems (Bünemann et al., 2018). According to the eco-stoichiometry theory, the higher leaf N:P ratio leads to a lower concentration of mRNA, thereby suggesting that species growth may likely be limited by P deficiency (Güsewell, 2004). During the recovery period, the appearance of regenerating species enhances the leaf N concentration, but leaf P concentration tends to be lower (Certini, 2005). After a decade’s recovery from wildfire, the recovery of N and P suggests that community restoration attenuates P limitation in plant growth (Wood and Bowman, 2012; Dantas Vde et al., 2016). Thus, the nutrient changes in different recovery periods may contribute to the allocation strategies of plant C, N, and P.

As mentioned above, plants can adjust the elemental concentrations according to the nutrient supply, especially in early recovery periods. These stoichiometry shifts would further affect the nutrients’ reallocation between above- and belowground parts (Niklas, 2005; Schafer and Mack, 2010). Allocation slope exponents of fine root vs. leaf at the species and community levels were significantly different among recovery periods (Figures 5 and 6). At the species level, the allocation slopes of N and P concentrations were significantly larger than 1 during the recovery period of 2 years and 10 years, indicating that more nutrient changes occur in the fine root than a leaf in early recovery periods. The changes in the element distribution of leaf and fine root largely drive the pattern of organs’ metabolic activity and their functions (Kerkhoff et al., 2006). Leaf functioning usually depends on the nutrients offered by the roots, and at the same time, root growth relies on the carbohydrates produced by the leaf (Minden and Kleyer, 2014; He et al., 2016). Under low soil N concentrations, more nutrients will be allocated to the root to sustain vital physiological functions to mine nutrients (Fortunel et al., 2012; Yan et al., 2016). Another possible reason is that the leaf needs more nutrients to ensure photosynthesis, which requires higher nutrient investments to roots to mine nutrients (Marschnert et al., 1997). Increased soil P concentrations could promote root growth. Thus, at the species level, nutrients were more changed in fine root than a leaf in early recovery periods.

At the community level, the slope of C, N, and P allocation between leaf and fine roots was smaller than 1 at year 2 after recovery, indicating that more elements changed in leaf than fine root (Figure 6). Species diversity and community composition are being strongly influenced by the nutrient changes (Liu et al., 2012b), especially after a wildfire (Nave et al., 2011). In the early recovery periods, greater species diversity with the occurrence of regeneration species presents fast growth strategies and high nutrient utilization. Even under low soil nitrogen concentrations, more N and P were allocated to aboveground at the community level (Maliakal et al., 2000; Niklas, 2005). Further evidence shows that the increasing soil P concentration after wildfire promotes more nutrient changes in the leaf of regenerated species, such as herbs with a shorter leaf life span (Johnstone et al., 2010; Chen et al., 2013). The regenerated plants need to enhance their photosynthetic rate to obtain light during their short growing season (Elser et al., 2010). Compared to woody species, herbaceous species were characterized by more leaf nutrient changes and quick growth, which can boost community recovery (Adler et al., 2014).

At the community level, the N and P allometry slope between leaf and fine root showed a clear transformation with the increasing recovery periods. Specifically, all slope exponents of N and P concentration showed allometry (b < 1) in early recovery periods (i.e., 2a and 10a), after when the slope exponents were more than 1 at medium recovery period (Figure 6). This transformation of slope exponents indicated that more N and P would change in fine root in the medium recovery periods (i.e., 20a and 30a). At the community level, the nutrient allocation in early recovery periods showed more changes of nutrient in leaf than fine root, which is similar to the finding that shrubs would preferentially allocate P to leaf to maintain plant physiological functions (Fortunel et al., 2012). From early- to medium-term recovery, the species diversity would enrich with the community recovery. The nutrient allocation relationship would be changed at the community level (Chen et al., 2019). The dynamics of nutrients at the community level reflect the combined results of nutrient allocation in the ecosystem (Lavorel and Grigulis, 2012). In medium recovery periods (i.e., 20a and 30a), soil N and P concentrations were recovered to those of unburned treatment; therefore, more nutrients will be transported to nonphotosynthetic organs to promote plant growth and enhance plant competitiveness (Fortunel et al., 2012), especially in the regenerated shrub species (Lavorel, 2013). Plants may allocate more nutrients to the stems and roots than the leaf to survive in nutrient-limited conditions (Enquist, 2002). Even in areas where soil N content is relatively abundant, more N is transported to roots to promote community stability (Fortunel et al., 2012). Thus, more nutrients are allocated to the leaf in early recovery periods, and more nutrients are allocated to root in medium recovery periods, which is consistent with our hypothesis (H₂).

Nutrient allocation is important for plants to adapt to environmental changes, which may be of particular importance for plant resource distribution in response to nutrient changes (Ordoñez et al., 2009; Smithwick et al., 2012). Therefore, the findings in our study suggest that differentially elemental allocation of a plant is crucial for resource utilization during ecosystem recovery after the wildfire.