Grasslands are highly diverse ecosystems providing important ecosystem services, but they currently face a variety of anthropogenic stressors simultaneously. Quantifying grassland responses to global change factors (GCFs) is crucial for developing effective strategies to mitigate the negative impacts of global change on grassland communities and to promote their resilience in the face of future environmental challenges. We conducted a field experiment in the Songnen grassland, northeastern China, to test the combined effects of 0, 1, 2, 4, 6, and 8 GCFs, including fungicide, herbicide, insecticide, antibiotic stress, heavy metal pollution, light pollution, microplastic pollution, nitrogen deposition, tillage disturbance, and increased precipitation. We found that within one year, the increasing number of GCFs negatively impacts both the productivity and diversity of grassland communities. In comparison to exposure to a single GCF, exposure to 8 GCFs led to a reduction in productivity and species richness by 42.8% and 42.9%, respectively. Furthermore, these negative effects seem to be linked to the reduction of dominant species and the concurrent increase in neonative species (i.e., species that have expanded their geographic range into a new area without direct human assistance, but as an indirect consequence of human-induced environmental changes). The results of hierarchical diversity-interaction modeling suggested that the adverse impacts of an increasing number of GCFs on community productivity and diversity are attributable to both the specific identities of GCFs involved and their unique pairwise interactions. The results suggest that grasslands may quickly lose stability and degrade more rapidly in response to multiple co-occurring GCFs. Greater efforts should be made to conserve the functions and services of grassland ecosystems by reducing the impacts of human activities.

Plant and soil microbial communities jointly sustain ecosystem multifunctionality (EMF) in temperate grasslands, yet their relative contributions to EMF under grazing management remain poorly understood. We simultaneously investigated three temperate grasslands to assess the effects of grazing management, climate, edaphic properties, and plant and microbial communities (diversity and community composition) on EMF (quantified by potential soil nitrogen (N) mineralization, arbuscular mycorrhizal fungal infection rate, phospholipid fatty acid, soil total carbon (C) and N, inorganic N, and plant biomass). Using random forest modeling, we identified important predictors, followed by structural equation modeling (SEM) to disentangle their relative roles. The results showed consistent declines in plant diversity and EMF with increasing grazing intensity, while soil bacterial and fungal diversity exhibited minimal responses. Heavy grazing management significantly reduced the abundance of perennial forbs and rhizome grasses, but increased that of annuals and legumes. Concurrently, we observed a significant decrease in copiotrophic Proteobacteria abundance accompanied by an increase in oligotrophic Gemmatimonadetes abundance. Random forest modeling identified grazing intensity, climate, soil properties, plant diversity and community composition, and bacterial community composition as important predictors of EMF. SEM revealed that plant diversity was the dominant biotic predictor of EMF, exceeding the influence of microbial communities across all grasslands. Notably, aridity indirectly influenced EMF through plant diversity rather than direct regulation. These findings demonstrate that plant diversity primarily maintains EMF under grazing pressure, highlighting the importance of biodiversity-focused management strategies in temperate grassland conservation.

In grassland ecosystem management, mowing influences the tolerance mechanism of plants by modifying their growth and reproductive traits; however, the specific processes involved remain unclear. This study focused on the Elymus species (Elymus nutans ‘Aba’, Elymus sibiricus ‘Qingmu No.1’, Elymus submuticus ‘Tongde’, Elymus breviaristatus ‘Tongde’ and E. sibiricus ‘Tongde’) and systematically evaluated the effects of different mowing intensities (no mowing, light, moderate and heavy mowing) at three growth stages (jointing, booting and flowering) on plant tolerance and the role of growth and reproductive traits in this mechanism. The results revealed that mowing generally reduced plant height and the reproductive branch quantity, while significantly increasing the tiller number, seedling number and relative growth rate. However, the responses of rhizome length and vegetative branch height varied across the growth stages. Mowing during the jointing stage had the most significant effect on morphological traits, with vegetative reproduction contributing the most to tolerance and increasing with mowing intensity. Overall, the plant response to mowing timing was more pronounced than its response to changes in individual traits. Moderate mowing at the jointing stage significantly increased growth rate, tiller number and seedling number, thereby enhancing mowing tolerance. In contrast, heavy mowing at the booting and flowering stages markedly reduced reproductive branch quantity and rhizome length, resulting in diminished mowing tolerance. The study indicated that differences in the mowing stage and forage species regulated adaptive changes in growth and reproductive traits, thereby influencing tolerance mechanisms. Grassland management should fully consider the effects of mowing at different growth stages to optimize the utilization and management of the Elymus grasslands.

Shrubland functions as an important carbon sink. However, uncertainties have still persisted regarding shrubland C storage and its underlying drivers. In this study, we conducted a field survey encompassing 45 sites to investigate all sectors of C stocks in shrublands distributed in northern China, in order to accurately estimate the regional C storage and to explore the potential drivers. Our results showed that the total C density of shrubland was 78.78 Mg C ha^–1, with soil C density, vegetation C density and litter C density contributing 75.16, 2.99 and 0.64 Mg C ha^–1, respectively. Distinct C density sectors were driven by different factors: vegetation C density was primarily driven by plant community richness, litter C density by shrub diversity and soil C density by total annual sunshine and soil total phosphorus in our study. Climate factors, plant community traits and soil properties independently explained 5.15%, 6.79% and 23.73% variation of the shrubland ecosystem C density, respectively. Furthermore, the interactions between community structural traits and climate factors, as well as between community structural traits and soil properties, can explain 10.44% and 18.50% of the variation, respectively. Our findings, based on direct field measurements, refined estimates of C storage in shrubland ecosystems in northern China, and these findings provided crucial data for the validation and parameterization of C models both within China and globally.

Alien plants exhibit varied performance and distribution patterns across latitudinal gradients depending on species invasiveness and target community invasibility. Although numerous researches have studied the latitudinal patterns of plant invasiveness, few have focused on community-level invasibility. We hypothesize that community invasibility increases with latitude due to a reduction in native species richness (diversity-resistance hypothesis) and stronger environmental filtering (pre-adaptation hypothesis) at higher latitude. We conducted a field survey at 18 sites across 6 latitudes in southeast China to explore how the community invasibility changes with latitude and identify the key drivers underlying these patterns. We found that the community invasibility positively correlated with latitude, primarily due to a decrease of native species diversity at higher latitude. Climate factors exerted indirect effects on community invasibility by shaping native species diversity. The mean pairwise phylogenetic distance between species did not change with latitude indicating minor effects of pre-adaptation. Our study emphasizes the importance of native species diversity in shaping latitudinal patterns of community invasibility. These findings highlight biodiversity conservation as an effective strategy to mitigate biological invasions, particularly in regions vulnerable to climate change.

Straw and coal gangue, primary wastes from agriculture and industry, respectively, have the potential to improve soil nutrients. The impact of plant species and microbial nutrient activation on this improvement warrants further investigation. A greenhouse experiment was conducted to examine how phosphorus-solubilizing bacteria (PSB) and mulch made from coal gangue and straw affect soil available phosphorus (AP), available silicon (ASi) and plant growth. These species include two economic crops, Gossypium hirsutum and Glycine max, and two plants used for mining remediation, Solanum nigrum and Medicago sativa. We found that the effects of mulching with coal gangue and straw on soil AP, ASi and plant growth were influenced by plant species. The soil AP and ASi contents were significantly positively correlated in G. max and S. nigrum and the mixed mulch significantly increased the soil AP and ASi contents for G. max and S. nigrum. The mixed mulch significantly increased the total biomass of G. max, with no significant effect on the biomass of other plants, reflecting that planting G. max could be an optimal strategy for improving soil with straw and coal gangue. The enhancement effects of the mixed mulch on soil AP and ASi of S. nigrum and the total biomass of G. max were negated by PSB, while PSB increased the soil AP and ASi contents of M. sativa with the mixed mulch. Overall, these results demonstrated the necessity of planting suitable species and judiciously using microbial inoculants during the use of waste.

The trait-based approach has long been crucial for understanding and predicting seedling dynamics in forest ecosystems. However, the complex architecture of trait networks governing these dynamics has been scarcely explored. The relationships among plant traits, reflecting plants’ overall life strategies, can offer deeper insights into plant performance than traditional trait-based analyses. Focusing on trait network architecture and traits themselves helps better understand the seedling-to-sapling transition. During a 12-year period, we carried out a field census in a subtropical forest. Using a demographic progress model, we evaluated two key population-level metrics for 26 species: the time for seedlings to reach the 2-m sapling stage and their early-growth survival period. We constructed plant trait networks with 16 leaf, stem and root traits and explored their connections with population vital rates. Our results showed that newly recruited seedlings took 17–81 years to become saplings, with a survival span of 1–4 years over a 5-year observation period. Network centralization was negatively correlated with the transition time but did not explain early-stage survival variation. Species with acquisitive strategies exhibited shorter transition time and a shorter survival period. Hub traits with high connectivity, such as root tissue density and leaf dry matter content, were more influential in determining the transition time. Overall, our study highlights the importance of trait network centralization and hub traits. By linking population characteristics derived from long-term individual-based observations with plant traits, this study advances our understanding of trait-population relationships and provides theoretical implications for forest management and conservation.

The intricate relationships between plants and their rhizobacteria are crucial for plant success, yet our understanding of these associations, particularly in diverse alpine natural grasslands, remains limited. Here, we investigated two widespread grass genera (Stipa Linn and Poa Linn) and their core rhizobacteria across a vast 2161 km transects on the Qinghai–Tibetan Plateau. Compared to Stipa L., which has a broader niche breadth, Poa L. displays higher above- and below-ground biomass. This characteristic reflects a more resource-conservative and stable growth strategy, consistent with traits commonly observed in K-strategists. However, Poa’s core rhizobacteria (5458 species, 9.51% of total amplicon sequence variants (ASVs)) were enriched with here r-strategists, while Stipa’s (5193 species, 9.05% of total ASVs) were dominated by K-strategists. These findings highlight contrasting life-history strategies between grasses and their associated core rhizobacteria. Notably, only 633 core rhizobacteria overlapped between these two grasses. Functionally, Poa’s r-strategist microbiota likely prioritizes rapid resource acquisition for high biomass production, while Stipa’s K-strategist-dominated community might enhance stress tolerance in their resource-limited habitat. The observed pattern of life-history differences between grasses and rhizosphere microbes supports plant survival in alpine ecosystems. Our study advances understanding of rhizosphere ecology and its importance for ecosystem health in natural environments.

Microorganism-mediated nitrogen (N) denitrification and dynamics are crucial ecosystem functions that influence N losses. However, the drivers and microbial mechanisms underlying seasonal variations in denitrification across elevations in alpine forest mountain ecosystems remain poorly understood. Here, we assessed the variations in potential denitrification rates and denitrifier communities using ¹⁵N-labeling techniques and high-throughput sequencing and examined soil properties across an elevational gradient in alpine forest mountains. Our findings demonstrated that soil potential denitrification rates decreased with increasing elevations, exhibiting lower levels during the wet season (0.095 ± 0.005 mg kg⁻¹ d⁻¹) compared to the dry season (0.12 ± 0.007 mg kg⁻¹ d⁻¹) (P < 0.05). Soil substrates, including NH4+-N, dissolved organic nitrogen, and total nitrogen, were identified as pivotal regulators of soil denitrification during the dry season, indicating substrate-driven control. Conversely, microbial attributes (nirS and nosZ genes abundance), were the primary factors governing soil denitrification during the wet season, reflecting microbial regulation. Additionally, Bradyrhizobium emerged as the dominant genus contributing to denitrification rates in our study. Overall, our study underscores the diverse factors driving the seasonal dynamics of soil denitrification and provided critical insights to improve ecosystem models to better predict N losses under increasingly pronounced wet-dry precipitation patterns.

Alpine treeline ecotone is considered an ideal “warning line” for monitoring global climate change. However, the shifts in the nonstructural carbohydrates (NSCs) of tree organs within alpine treeline ecotones along the gradient from timberline–treeline–tree species line remain elusive. This study aimed to quantify the soluble sugar, starch and NSC contents of Qinghai spruce and their driving factors across six typical treeline ecotones of the Qilian Mountains, northeastern Tibetan Plateau. The soluble sugar, starch and NSC contents were the highest in leaves, followed by fine roots and twigs. The NSC contents in tree leaves and twigs peaked with an increase in elevation from the timberline to the treeline. At the tree species line, the NSC content was the highest in fine roots. Namely, the aboveground carbon investment decreased, whereas the underground carbon allocation increased after exceeding the upper limit of treeline. The NSC contents in leaves and twigs were all higher, and in fine roots were lower in the western part of the Qilian Mountains (an extremely arid region) than in the eastern and central parts (semi-arid region). The NSC contents were positively correlated with mean annual temperature and negatively correlated with mean annual precipitation. The redundancy analysis results showed that the soil factors contributed to 66.3% of the variation in tree NSC contents, with soil NO₃^–-N (26.6%) and NH₄⁺–N (24.2%) being the dominant factors influencing the changes in tree NSC in the alpine treeline ecotone. This study provides novel insights into carbon metabolism and growth adaptation strategies of trees in alpine arid mountainous ecosystems.

Due to large-scale commercial logging and prolonged anthropogenic disturbances over the past decades, large tropical secondary forests (SFs) have been converted into rubber plantations (RPs), which have substantial impacts on soil bacterial community via altering soil properties. However, how forest conversion affects soil bacterial community composition and diversity is still poorly understood. To address this, we compared the soil bacterial communities and physicochemical properties between typical SFs and RPs in Hainan Island. The results showed that SF exhibited higher soil organic matter, total nitrogen, total carbon and pH compared with RP, particularly in the 0–10-cm layer. Besides, the soil bulk density in SF was lower than that in RP. The bacterial community composition and microbial networks were significantly varied between SF and RP. The dominant soil bacterial phyla in SF were Proteobacteria (27.4%–28.7%), followed by Acidobacteria (19.0%–19.8%), while Chloroflexi prevailed in RP (27.9%–28.5%), followed by Acidobacteria (20.9%–24.4%). Furthermore, SF exhibited more complex microbial networks owing to its higher edges, degree and links compared with RP. Meanwhile, bacterial community assembly in SF was primarily governed by deterministic processes, while stochastic processes dominated RP. The soil bacterial α-diversity in SF was higher than that in RP, which was primarily dominated by pH. Our findings demonstrate that converting SFs to RPs does not favor soil microbial diversity and stability, as it decreases soil pH, suggesting that sustainable management strategies should prevent soil acidification in RPs.

Alpine cushion plants are recognized as keystone species essential for sustaining plant communities and biodiversity. However, their contributions to the structures of plant–plant co-occurrence networks remain poorly understood. This study constructed plant–plant co-occurrence networks within cushion microhabitats at both regional and local scales in the Himalaya–Hengduan Mountains, focusing on the influences of multiple cushion plants in shaping plant–plant network structures. Results reveal that cushion plants significantly influence the network structures. Compared to random expectations, these networks display lower linkage density, weighted connectance and weighted nestedness (wNODF), but higher modularity, implying distinct organizational principles driven by cushion plant facilitation. Non-cushion plants show stronger associations with cushion microhabitats than with open ground microhabitats. Additionally, the spatial influence of cushion plants extends beyond their immediate canopies, highlighting their role in structuring surrounding plant assemblages. Moreover, the facilitation strength of cushion plants positively correlates with network metrics, indicating that network complexity increases with increasing facilitation by cushion plants. Simulations suggest that the loss of cushion plants would likely trigger cascading extinctions of associated non-cushion plants, particularly when high-degree centrality cushion species—those with the greatest network connectivity—are removed. These results highlight the keystone roles of cushion plants’ facilitation in supporting biodiversity via enhancing network complexity and robustness. Given the vulnerability of alpine ecosystem to environmental disturbances, our study emphasizes the urgent need to prioritize the conservation of cushion plant diversity. Future conservation strategies should adopt an integrated approach that protects not only individual cushion species but also the micro-communities they support.

Nutrient enrichment can negatively impact the stability of various ecosystems. However, whether these impacts depend on biotic and abiotic factors, such as plant diversity and climate, remains unclear. Here, we investigated how climatic factors and plant diversity affected temporal stability of aboveground net primary production (ANPP) in responses to nitrogen (N) and phosphorus (P) additions and its biotic mechanisms over 4 years across seven Chinese grasslands spanning a climatic gradient (mean annual temperature: −0.38 to 8.75 °C; annual precipitation: 134.60 to 462.40 mm). We showed that N and NP additions significantly reduced temporal stability of ANPP, while P addition marginally reduced the temporal stability of ANPP across all sites. Importantly, nutrient effects at each site were inconsistent and were closely related with sites with initial richness and temperature seasonality. The structural equation modeling clarified that N and P addition alone mainly decreased community stability via decreasing dominant species stability. Notably, sites with higher species richness exhibited lower stability of dominant species and lower community stability under N or P enrichment. Taken together, our results highlight that effect of nutrient amendment on the stability of grassland ecosystems depends on local plant diversity shaped by specific climatic factors. This understanding is crucial for effective grasslands management in the context of global environmental change.

Microbial necromass carbon (MNC) contributes largely to soil organic C (SOC) pool in terrestrial ecosystems. However, the pattern and underlying mechanisms of MNC and their contribution to SOC along elevational gradients are controversial due to montane ecosystems subject to environmental change. Here, in this study, we investigated the seasonal variation of MNC, its contribution to SOC, the necromass accumulation coefficient and the influencing factors across different elevations in the mountain forests ecosystem of Southwest China. Soil microbial biomass rather than MNC showed seasonal variations, this decoupling pattern was mainly attributed to higher soil extracellular enzymes (i.e. N-hydrolyzing enzyme) and C:N ratio, which accelerated the decomposition of MNC especially bacteria necromass C (BNC) during the humid and warm wet season. In contrast, the drought and cold conditions in dry season inhibited microbial activities and conversion to MNC. During the dry season, the MNC and MNC/SOC exhibited hump-shaped pattern along elevational gradients. The fungal necromass C (FNC) was positive with fungal biomass, indicating that living biomass may have a greater influence on the accumulation of FNC than BNC. On average, MNC constituted about 15% of SOC, with the contribution from FNC (11.9%) surpassing that from BNC (3.1%). The joint effects of soil pH and clay composition significantly influenced MNC dynamics along elevational gradients. These findings demonstrate that the rapid decomposition of BNC is the main way of MNC loss in wet season in the mountain forests ecosystem and further highlight the importance of microbial traits in MNC accumulation.

As important components of soil organic carbon (SOC), soil amino sugars and lignin phenols serve as reliable biomarkers for microbial- and plant-derived C, and their relative contents provide critical insights into the SOC composition and stability. However, the dynamic changes of soil amino sugars and lignin phenols with restoration time in the widely planted subtropical Castanopsis hystrix plantations remain unclear, which limits the development of scientific management strategies. Here, we analyzed soil amino sugars, lignin phenols and the related biotic and abiotic factors at two soil depths (0–10 cm and 20–30 cm) across various stand age of Castanopsis hystrix plantations in southern China. The results showed a divergent accumulation of soil amino sugars and lignin phenols with stand age, in which the ratio of soil amino sugars to SOC exhibited a unimodal pattern with stand age in the topsoil (0–10 cm), while the ratio of soil lignin phenols to SOC decreased with stand age in both soil depths. Consequently, the ratio of soil amino sugars to lignin phenols increased with stand age, particularly in the subsoil (20–30 cm). Further analysis indicated that the divergent accumulation of soil amino sugars and lignin phenols with stand age was mainly related to changes in the litter biomass and quality, while the divergence with soil depth was primarily driven by microbial residue formation efficiency. These findings demonstrate the SOC composition shift from plant-derived C to microbial-derived C during plantation restoration, offering important implications for forest management practices and nature-based climate solutions.

Invasive plants often generate more positive plant–soil feedbacks (PSFs) than natives. The direction and strength of PSFs have been affected by resource availability and conspecific plant density, but their joint effects on PSFs of invasive plants have not been examined. We conducted a two-phase PSF experiment to examine how soil nutrient availability and planting density affect feedback with soil biota between a community of five invasive Asteraceae plants (two clonal species) and four co-occurring native plants (one clonal species). Soil biota from invaders did not inhibit plant growth, but soil biota from natives did so. The difference in PSFs between geographic origins was most pronounced under high-nutrient and high-density conditions where the biomass of natives in conspecific soils was much lower than that in soils conditioned by heterospecific invaders. Clonality and its inter-actions with nutrient and density did not affect PSFs. Soils from invaders had a higher diversity of arbuscular mycorrhizal fungi, and soils from invaders and natives differed in the composition of pathogenic fungi. Our results may explain why invaders but not natives often form dense monocultural stands. That is, as plants take advantage of increased nutrient supply and conspecific density increases, PSFs change little for invaders, but they change to be highly negative for natives, preventing them from forming dense monocultural stands. This invasion mechanism may be particularly pertinent to clonal invaders, as they can swiftly proliferate within habitats via clonal reproduction without encountering negative density dependence, thus establishing dense monocultural stands rapidly from just a few individuals.

The distribution and assembly of plant species is a fundamental ecological question. Understanding how various drivers of community assembly vary along elevational gradients and differentially affect species occurrence versus abundance is critically important. Using an advanced tool, the joint species distribution model (JSDM), we aim to investigate shifts in the relative importance of climatic and non-climatic effects on forest community assembly along large elevational gradients, and to compare the elevational trends of these effects on species occurrence and abundance. We documented 243 forest plots, each with a size of 20 × 30 m, along elevational gradients (700–3650 m) of Qinling Mountains, the highest mountain in Central China. We performed JSDMs to quantify the relative importance of climatic and non-climatic effects on species occurrence and abundance along elevational gradients. Climatic and non-climatic effects exhibited distinct elevational trends and differed in their respective influences on species occurrence and abundance. The influence of climate on species abundance increased with elevation, whereas its effect on species occurrence showed a weak decline. At lower elevations, species occurrence was mainly determined by climate, while abundance was affected by both climatic and non-climatic drivers. At higher elevations, climate emerged as the dominant factor affecting both occurrence and abundance. Our study reveals different trends of climate and non-climate effects on species occurrence versus abundance. These findings underscore the importance of jointly considering both types of environmental drivers and both occurrence and abundance, which is critical for predicting community dynamics under climate change and guiding conservation strategies.

Silicon, the second most abundant element in the Earth's crust, affects plant defenses across large geographic scales, which is an interesting yet mostly unexplored issue in non-native species. We investigated silicon-mediated variation in defensive/nutritional traits and palatability to the generalist herbivores Mythimna separata and Locusta migratoria in 16 geographic populations of the invasive grass Spartina alterniflora collected from native and introduced ranges. We found that silicon supplementation generally increased silicon accumulation in Spartina leaves by 76%, and enhanced their physical/chemical defenses (jasmonic acid +28%, flavonoids +12%, toughness +9%) and reduced nutritional quality (carbon content −2%, C:N ratio +5%), ultimately impeding the growth of both generalists (M. separata −28%, and L. migratoria −43%). Introduced populations, independent of silicon treatments, had significantly higher silicon content in leaves than native populations. However, silicon supplementation increased more leaf silicon in low-latitude introduced populations (+81%) than in high-latitude introduced populations (+63%). Moreover, the mechanisms of silicon accumulation affecting generalists differed between plant provenances: enhancing quantitative and qualitative chemical defenses in native populations, but improving quantitative chemical or structural defenses in introduced ones. These results suggest a rapid evolution of defense strategies mediated by silicon in introduced populations. Our findings provide insights into how non-native plants utilize silicon to enhance defenses, emphasizing the importance of metalloid defenses in invasion success.

Understanding leaf phenology is essential for capturing forest dynamics, yet traditional monitoring methods fail to resolve vertically stratified phenology due to canopy occlusion and limited spatial coverage. To address this gap, we developed an integrated unmanned aerial vehicle and ground-fixed camera system enabling simultaneous monitoring of forest overstory and understory phenology. Deployed in a subtropical forest during 2017–2023, this system archived 0.075 m × 0.075 m resolution aerial imagery and hourly ground photography, tracking vegetation dynamics across community and species scales. Our system-derived Green Chromatic Coordinate was strongly correlated with Normalized Difference Vegetation Index (r = 0.82), Enhanced Vegetation Index (r = 0.91), Gross Primary Productivity (r = 0.95) and Leaf Area Index (r = 0.79 for overstory; r = 0.92 for understory) validating its effectiveness as a phenological proxy in subtropical forests. Critically, the understory exhibited delayed leaf maturation (16.2 days) and senescence (61.2 and 11.6 days for start and end of leaf falling, respectively) compared with the overstory, revealing a vertical ‘phenological escape’ phenomenon. These phenological mismatches buffered seasonal productivity fluctuates, by sustaining carbon uptake during overstory senescence. Our approach overcomes the limitations of fixed observation towers and satellite imagery by offering flexible, scalable and cost-effective monitoring of vertical stratification in forests. By quantifying vertical layer interactions, our approach advances predictive modeling of ecosystem–climate feedback and guides forest management under climate change.