ABSTRACT Previous studies often overlooked the dynamic effects of karst vegetation restoration on soil multi‐trophic biodiversity and the underlying community‐assembly mechanisms. To fill this gap, we performed a seasonal comparative analysis of soil biodiversity, co‐occurrence network architecture, and assembly processes across three representative karst restoration patterns (grass, forest‐grass, and forest). During the wet season, the grass pattern exhibited significantly lower biodiversity than both the forest‐grass and forest patterns by 17%–49%. In the dry season, the forest pattern achieved the highest biodiversity indices and displayed minimal seasonal fluctuation. Vegetation type and season jointly shaped soil biotic co‐occurrence networks: the forest–grass pattern generated the greatest network complexity and connectivity, whereas the dry season fostered tighter network structures. The bacterial community was predominantly driven by heterogeneous selection with a relatively high contribution in the wet season, whereas it was dominated by dispersal limitation in the dry season. In contrast, the fungal community was largely influenced by dispersal limitation across seasons. As for the protist and nematode communities, undominated accounted for a relatively high contribution in the wet season, while dispersal limitation became the dominant driver in the dry season. Moreover, soil moisture, nutrient status, biodiversity, and network properties emerged as the principal drivers of assembly, with their pathways varying by taxon and season. Collectively, our reveal integrated regulation of soil biotic communities by restoration pattern and seasonal dynamics in karst ecosystems, offering a robust theoretical basis for soil‐biota recovery.

Assessing the climate resilience of plantation versus natural forests is critical for climate-smart forestry, yet large-sample quantitative comparisons remain scarce. Here we integrated field biomass data from 1022 plots across 11 forest types in China with deep learning models (Transformer, LSTM, ResNet-50) to compare growth responses to temperature and drought stress. Plantations exhibited higher peak growth rates but were confined to a narrow thermal optimum, whereas natural forests maintained stable productivity across a much wider climatic range. Under drought, natural forests demonstrated significantly higher resilience (index: 8.47 +/- 0.08) than plantations (6.18 +/- 0.15), with this advantage widening under extreme aridity. The Transformer model achieved superior predictive accuracy (R2 = 0.86). Critically, future projections under RCP 8.5 indicate that by 2050, plantations will lose 40% of their current resilience by 2050 under RCP 8.5-a relative decline 1.4-fold greater than the 28% loss projected for natural forests, underscoring the disproportionate vulnerability of planted systems. Our results provide a quantitative basis for reorienting forest management from growth-centric metrics toward enhanced structural and compositional diversity to safeguard ecosystem functioning under accelerating climate change.

Peatlands in Eurasian permafrost regions are among the world’s most important soil carbon reservoirs and may strongly influence future climate through carbon release. Runoff-mediated dissolved carbon export represents a critical carbon-loss pathway in permafrost peatlands, yet observations from the southern margin of Eurasian permafrost remain scarce. Here, we investigated dissolved carbon export from two differently sized permafrost peatland catchments in the Greater Khingan Mountains, northeastern China. We estimated dissolved organic carbon (DOC) and dissolved inorganic carbon (DIC) export fluxes and evaluated hydrological controls on dissolved carbon dynamics using three fluorescence indices. DOC and DIC concentrations were closely coupled with discharge during seasonal hydrological fluctuations, indicating that runoff dynamics were the primary driver of dissolved carbon export. During the growing season, the Fukuqi River exported 703.84 t total dissolved carbon, equivalent to approximately 20% of peatland net ecosystem exchange. Hydrological processes regulated not only dissolved carbon fluxes and the relative contributions of DOC and DIC, but also DOC chemical characteristics. Flood-peak flows promoted the export of more humified DOC, characterized by higher humification index and lower fluorescence index and freshness index, whereas baseflow showed the opposite pattern. The vertical organic–mineral stratification of peatland soils, with contrasting hydraulic transmissivity and dissolved-carbon production potential, directly governed temporal variations in discharge and dissolved carbon concentrations. Active-layer deepening may increase runoff contributions from deeper mineral soils, thereby increasing the DIC fraction while reducing DOC concentration and the degree of humification. Fluorescence indices were consistently correlated with discharge across both catchments, suggesting their potential as robust indicators of hydrological processes. Overall, these findings highlight runoff-mediated dissolved carbon export as an underrepresented carbon-loss pathway and suggest that permafrost degradation will reshape both the magnitude and chemical composition of dissolved carbon exports under future warming.

Plant intraspecific trait variation (ITV) is critical for community assembly and functioning. However, its effect on rhizosphere microbes and the nematode micro-food web has rarely been explored. To investigate how ITV affects rhizosphere bacteria, fungi, and nematode communities, we selected twelve tree species representing arbuscular mycorrhizal (AM) and ectomycorrhizal (ECM) types in a tropical common garden and assessed their root traits in root economics space (RES) and root exudate, rhizosphere soil properties, and rhizosphere soil biota communities. Root ITV significantly altered the between-species trait variation in RES, which captures trade-offs in ‘collaboration’ and ‘conservation’ gradients that ranging from ‘DIY’ (do-it-yourself) to ‘outsourcing’ and ‘fast’ to ‘slow’ strategies. The soil properties were also captured by ‘nutrient’ and ‘pH’ gradients. Collectively, ECM trees exhibiting ‘DIY’ and ‘fast’ root strategies, as well as inhabiting soils with higher nutrient levels, were associated with increased rhizosphere fungal and nematode diversity. And AM trees inhabiting soils with elevated pH were associated with increased bacterial diversity. Both ECM and AM trees with a ‘fast’ root strategy enhanced cross-guild co-occurrence network complexity, and with higher nutrient levels and pH values increased within-guild network complexity. Notably, the intraspecific variations in root traits and soil properties exerted more shifts in rhizosphere soil biota diversity and network complexity in ECM trees relative to AM trees. This study demonstrates the role of plant ITV in structuring rhizosphere soil biota and highlights the importance of plant phenotypic diversity for soil community assembly and functioning.

Cloud fraction (CF) over the Tibetan Plateau (TP) critically modulates regional radiative forcing, yet uncertainties persist in satellite-derived cloud products due to heterogeneous earth-atmosphere interactions and temporal sampling limitations. Here, we present the evaluation of CF biases derived from the Advanced Himawari Imager (AHI) onboard Himawari-8/9, Moderate Resolution Imaging Spectroradiometer (MODIS) onboard Terra and Aqua, and ERA5 reanalysis by comparing them with ground-based observations. MODIS shows excellent accuracy with CF biases below 1% on daily and monthly scales. Its polar-orbiting, multi-spectral design effectively discriminates cloud tops and minimizes topographic effects over the rugged TP. ERA5 exhibits moderate biases (under 2%), likely due to its use of assimilated cloud observations. In contrast, AHI has larger regional errors (8-10%), which can be attributed to inherent limitations of geostationary viewing and restricted shortwave infrared channels. Cross-validation with CloudSat vertical profiles further shows that the cloud hit rate of AHI is lower than that of MODIS by about 5.82%. Moreover, there is an 11.61% underestimation when only "cloudy" is accounted for, while the combined "probably cloudy" classification overestimates CF by 17.14% compared to the cloud mask product of CloudSat. It highlights significant uncertainties brought by the AHI cloud classification threshold algorithm. These findings underscore the necessity of multi-sensor synergy in reconciling cloud-climate feedback uncertainties over the TP, advocating for optimized AHI cloud classification thresholds in radiative transfer models while prioritizing MODIS or ERA5 for long-term cloud feature studies. (c) 2026 COSPAR. Published by Elsevier B.V. All rights are reserved, including those for text and data mining, AI training, and similar technologies.

Small object detection in unmanned aerial vehicle (UAV) aerial imagery is challenged by small object size, sparse feature information, and complex backgrounds. This paper presents AeroVision-DET, an aerial small object detection algorithm based on dynamic convolution and hierarchical attention fusion. We construct the Adaptive Receptive Field Network (ARFNet) based on the Multi-Scale Adaptive Feature Module (MSAFM) as the backbone, which achieves adaptive receptive field adjustment through dual-path feature enhancement blocks and multi-shape adaptive convolution. We design the Semantic-Spatial Fusion Module (SSFM), which adopts a multi-scale contextual attention mechanism to achieve semantically consistent and spatially sensitive feature fusion. We further propose the Efficient Feature Encoding Layer (EFEL), which integrates polarized linear attention and a frequency-modulated feed-forward network to model long-range dependencies under linear complexity. Experiments on VisDrone2019 show that AeroVision-DET achieves 24.2% AP, a 3.4% improvement over the RT-DETR-r18 baseline, with a 3.5% gain on small-object AP and a 27.0% parameter reduction while computational complexity remains essentially unchanged. With only 14.57 M parameters and 56.9 GFLOPs, AeroVision-DET delivers 69.5 FPS real-time inference and is well-suited for deployment on resource-constrained edge devices such as embedded UAV platforms, while maintaining detection accuracy competitive with or superior to state-of-the-art (SOTA) models.

As geologically unique landforms characterized by extreme micro-topographic heterogeneity (peak-depression coupling) and intense human-land contradictions, karst peak-cluster depressions represent globally significant yet ecologically fragile landscapes fundamentally distinct from non-karst ecosystems in their water-soil-nutrient coupling processes. Understanding ecosystem service (ES) relationships in these geomorphologically complex regions is critical for optimizing ecological restoration. This study analyzed spatiotemporal dynamics and drivers of five ES (water yield, sediment retention, nitrogen export, carbon sequestration, and habitat quality) in a representative karst peak-cluster depression from 2005 to 2020. Results revealed pronounced temporal fluctuations in water yield and soil retention compared to stable carbon sequestration and habitat quality, alongside marked spatial heterogeneity governed by topography and land use. Significant trade-offs dominated between habitat quality and nitrogen export (56.14% pixel ratio)/sediment retention (54.02%), water yield and sediment retention (53.94%), and carbon sequestration and nitrogen export (53.09%), primarily in southern high-altitude and northern lowland regions. Conversely, synergies prevailed between sediment retention and nitrogen export (61.29%) and habitat quality and carbon sequestration (55.93%). Geodetector analysis identified annual precipitation, evaporation, and slope as primary drivers, with two-factor interactions (particularly precipitation-evaporation) exhibiting substantially stronger explanatory power (q-values up to 0.857) than individual factors. These findings highlight that ES relationships in karst depressions are shaped by coupled climatic-topographic controls rather than single determinants. We advocate spatially differentiated zoning strategies-including hydrogeological nitrogen interception, slope-threshold land-use planning, and climate-adaptive vegetation configuration-to balance inherent trade-offs between water yield and soil/carbon retention, thereby supporting sustainable development in vulnerable karst regions.

The sustainability of mountains, among Earth's most dynamic and fragile ecosystems, is pivotal for global SDGs. A core challenge, however, is the profound overlap between human actions and natural systems, which dissolves the conventional boundaries between natural and urban systems. We synthesize research on classic ecotones and propose a novel coupling index of social-ecological systems as a unified metric for identifying multiple transition types. Applied to a Chinese mountainous urban agglomeration, our framework successfully delineates urbanrural, agropastoral, and terrain transition zones. This study identifies a complex coupling ecosystem (CCE) at the grid scale, a critical transition area where socio-economic and natural systems intensely interact, covering 17.28 % of the study area. It is the transition zone from the basin-periphery mountainous area (The average elevation of the western mountainous region exceeds 3000 m, with the highest peak reaching 7500 m) to the plain area. It concentrates the drastic transitions of multiple elements, including topography, climate, economy, and land use. The CCE-type areas exhibit a profound spatial conjunction between ecological sensitivity and socioeconomic vulnerability, evidenced by nine out of the 10 counties/districts with the most CCE-type areas in Sichuan Basin being previously considered poverty-stricken counties in China at the national level. With a relatively large population and scarce high-quality land resources, this area is the core region facing the trade-off dilemma between economic development and ecological conservation. These findings contribute to our understanding of the mutual feedback within mountainous urban agglomeration systems, thus supporting sustainable land management and government policies.

Understanding the evolution of metropolis-wilderness transitional landscape in the context of the Anthropocene remains a major challenge. This study constructs an Urban–Rural–Natural (U–R–N) landscape gradient using the Human Activity Intensity Index (HAII) from 2010 to 2020 to identify and classify multiple types of transitional zones (TZs) across Southwest China. Results indicate that while the total area of TZs slightly declined from 22.20 × 104 to 21.73 × 104 km2, the Transition Significance (TS) increased, signaling intensified internal restructuring. A clear gradient-based evolutionary trajectory emerged: urban and natural zones expanded while rural areas contracted. TZs mirrored these trends—areas adjacent to urban centers expanded, whereas those associated with traditional rural settlements declined. This yield fragmented and functionally differentiated spaces titled towards urban and natural landscape. These asymmetric dynamics drive social-ecological system spatial restructuring, shaped by the uneven diffusion of human expansion and limited ecological restoration. Despite variations in explanatory power, road, HAII, and POI consistently emerge as dominant drivers of TS change. Based on these factors, the most critical transitional zones are identified, whose total area declined from 1.34 × 104 km2 to 0.86 × 104 km2 showing a shift from widespread low-intensity to fragmented high-intensity transitional landscapes. By identifying the typology and drivers of TZs evolution, this study lays a scientific basic for optimizing sustainable landscape management and planning.

In the highly heterogeneous and fragile karst forest ecosystem, existing research on single trophic levels fails to explain the maintenance of ecosystem multifunctionality (EMF). Moreover, integrated studies adopting multitrophic perspectives remain scarce. Therefore, this study focuses on investigating the driving mechanisms of multitrophic biodiversity on EMF. Our study explores the link between multitrophic diversity and EMF in karst forests, assessed the diversity of six trophic groups: plants, bacteria, fungi, protists, nematodes, and arthropods. We established both belowground and aboveground-belowground multitrophic co-occurrence networks, concurrently assessing eight key ecosystem functions associated with material cycling and biomass production. Biodiversity (particularly that of plants, soil bacteria, and nematodes) showed significant positive correlations with EMF, and the multitrophic diversity index exhibited stronger predictive power than any single-trophic-level index. The topological properties of multitrophic co-occurrence networks (such as negative edge count, edge density, centralization, and relative modularity) are significantly closely linked to EMF. Notably, the predictive and explanatory power of network complexity integrating aboveground and belowground organisms for EMF is significantly stronger than that of networks containing only soil organisms. Structural equation modeling revealed that plant diversity and multitrophic network complexity had a direct positive effect on EMF, while multitrophic diversity primarily enhanced EMF indirectly by promoting network complexity. Environmental factors, in turn, indirectly influenced EMF by modulating these biological components. Multitrophic network complexity exhibited the largest standardized total effect among all factors, highlighting its central role in maintaining EMF. We conclude that maintaining EMF in karst forests should prioritize conserving the overall complexity of cross trophic biotic interaction networks rather than focusing solely on species richness.

Ozone pollution suppresses plant photosynthesis and reduces crop yields, yet many global assessments still rely on indicators or models that are difficult to compare across regions. We propose a concise and scalable remote sensing index, called ozone vegetation exposure intensity (VEIO3), which combines near-surface ozone concentrations with the normalized difference vegetation index (NDVI) to characterize the spatiotemporal distribution of vegetation ozone exposure. Using global observations from 2003 to 2019, we find pronounced spatial and seasonal contrasts, with higher exposure in industrialized and densely vegetated regions, particularly during spring and summer in the Northern Hemisphere. VEIO3 is consistently associated with independent vegetation metrics: regions of higher ozone exposure generally correspond to lower solar-induced chlorophyll fluorescence (SIF), indicating ozone-induced stress on vegetation activity. In site-level evaluations using the Community Land Model (CLM5.0) and FLUXNET eddy-covariance data, we identify a new potential critical ozone concentration of 42.38 ppb, which yields improved performance over the traditional AOT40 (accumulated exposure over a threshold of 40 ppb) in broad, noncropland contexts-suggesting that it can complement existing ozone exposure indices. VEIO3 can be operationalized for ozone risk management: for example, by screening hotspots to prioritize field inspections and expand monitoring networks; by serving alongside AOT40 as a metric for evaluating seasonal mitigation plans and issuing alerts during peak growing seasons; and by coupling with land-use and crop distribution data to support emission-control planning and scenario analysis. Key limitations of our current approach include reliance on monthly ozone and NDVI products (with uneven monitoring coverage), a predominantly linear statistical framework that does not capture time lags or nonlinear dose-response behavior, and the absence of explicit stomatal-flux representation. Priorities for future work include controlled experimental validation, integration of VEIO3 with stomatal flux schemes in land-surface models, expansion of ground observations in data-sparse regions, and testing of alternative vegetation indices with stricter cloud and reflectance screening. VEIO3 represents a new remote sensing index for detecting vegetation ozone exposure. The critical levels and model-improvement directions inferred here should be treated as potential References that, with further validation and refinement, can help bridge the gap between observations, models, and policy, ultimately providing a reliable remote sensing-based decision support tool for managing risks to ecosystem productivity and food security.

Current research has confirmed that forest productivity in non-karst regions is closely related to biodiversity. However, the relationship between forest productivity and multi-trophic biodiversity remains unclear in karst habitats, characterized by thin soil layers and limited soil nutrients. To better understand the productivity of natural and planted forests in karst regions and main impact factors on the forest productivity, we selected restored karst forests in the southwest of China. By using methods such as high-throughput sequencing, the Mantel test, random forest model, and structural equation model, we analyzed the differences in productivity, habitat factors, and multi-trophic biodiversity between natural forests and planted forests and explored the key influencing factors for productivity of karst forests. The findings indicated that the aboveground, belowground, and total productivity of planted forests were significantly higher than those of natural forests. Conversely, the soil nutrient stocks, multi-trophic biodiversity, and plant diversity of natural forests were significantly higher than those of planted forests. Moreover, the complexity of the multi-trophic co-occurrence network in planted forests was higher than that in natural forests. The Mantel test revealed significant positive correlations between soil nutrient stocks, soil moisture content, altitude, slope position, soil layer thickness, multi-trophic biodiversity, and forest aboveground productivity, belowground productivity, and total productivity. The random forest model identified soil nutrient stocks as the most significant factor influencing the karst forest productivity. The structural equation model showed that stand characteristics, soil nutrient stocks, and multi-trophic biodiversity collectively explained 69% of the total productivity variation. Forest type could not only directly impact the total productivity but also affect the total productivity along with the changes in soil nutrient stocks. This study is conducive to a deeper understanding of the formation mechanism of productivity in Southwest karst forests.

Peatlands in permafrost regions store vast carbon reserves; however, their responses to enhanced nitrogen availability remain insufficiently understood, particularly regarding the contrasting responses of rhizosphere and bulk soils. The field study investigated the impact of different nitrogen addition levels (0, 12, and 24 g N m–2 yr–1) on soil organic carbon (SOC) fractions in the 0–20 cm layer of both rhizosphere (with plants) and bulk (without plants) soils associated with Eriophorum vaginatum L. in a permafrost peatland located in the Greater Khingan Mountains, Northeastern China. Short-term nitrogen addition had no significant effect on total SOC in either rhizosphere or bulk soil. During the early growth stage, nitrogen addition significantly increased dissolved organic carbon content in both rhizosphere and bulk soils by 17.8–40.8

Large-scale vegetation restoration has the potential to profoundly impact the ecosystem functions of the Karst region in southwest China. However, most existing studies primarily focused on individual aspects such as nutrient dynamics, water availability, biomass production, lacking comprehensive research on ecosystem multifunctionality (EMF). We collected 60 soil samples from three typical forest and grassland restoration patterns to investigate how plant restoration patterns influence EMF and soil microbial diversity. Based on correlation analysis and structural equation modeling, we explored the relationships between soil properties (e.g. soil pH, moisture, and microbial diversity in different soil layers) and EMF. The results demonstrate that, compared with the single forest and grass restoration models, the combined forest-grass model enhanced both EMF and the topsoil bacterial diversity. Across the three restoration models, microbial diversity decreased significantly with increasing soil depth, whereas fungal co-occurrence network complexity increased. Structural equation modeling identified soil bacterial co-occurrence network complexity as the key determinant of EMF in both top and subsurface soil layers. Moreover, soil pH, moisture, and microbial diversity affected EMF primarily through their effects on soil bacterial co-occurrence network complexity. Notably, subsurface soil properties explained a greater proportion of the variance in EMF (40%) than did topsoil properties (31%). These findings highlight the importance of multi-species integrated forest-grass restoration in enhancing EMF, while also emphasizing the crucial roles of bacterial co-occurrence network complexity and deeper soil layers in maintaining EMF.

High-resolution mapping of urban surface CO2 is essential for refined carbon monitoring, emission management, and low-carbon urban planning. Mobile monitoring provides dense street-level observations, but raw CO2 measurements are often affected by transient traffic disturbances, vehicle idling, and localized plume events, which limits their direct use as stable spatial mapping targets. This study developed an integrated framework for predicting, mapping, and interpreting stable surface CO2 patterns in Shenzhen by combining vehicle mobile observations, CSF processing, multiscale remote sensing predictors, machine learning. A CSF-based lower-envelope filter was used to suppress short-duration positive peaks and extract a more stable CO2 accumulation signal from mobile observations. Multiscale predictors representing transportation, urban activity, surface environment, and built form were constructed to characterize both local and surrounding urban contexts. Compared with raw CO2, the CSF-processed target substantially improved prediction performance. The best validation R2 across the candidate models increased from 0.59 to 0.90 in April and from 0.62 to 0.93 in November. The predicted maps identified persistent high-CO2 areas in central and southwestern Shenzhen. SHAP results showed that transport networks and urban activity reinforced surface CO2 accumulation, whereas vegetation and open-surface contexts weakened accumulation at broader spatial ranges. These findings provide an interpretable framework for high-resolution urban CO2 mapping and refined low-carbon governance.

Herbal medicines with dual medicinal and nutritional values, known as medicine-food homologous (MFH) herbs, are traditionally identified by practitioners using sensory evaluation and personal experience, which often leads to inconsistencies and inaccuracies. To address these challenges, near-infrared (NIR) spectroscopy combined with classification models offers a non-destructive and rapid approach for herb identification. However, the limited availability of NIR data presents a small-sample issue, impairing model accuracy and stability for practical applications. This study proposes a NIR denoising diffusion implicit model (DDIM-NIR) to generate high-quality species-labeled NIR spectral data for improving MFH herb identification. The model first performs forward diffusion, where Gaussian noise is progressively added to real spectra under a cosine schedule. The noisy data is then fed into a prediction network, enabling recovery of the original spectra. During inference, the model starts from pure noise and adopts DDIM sampling, retaining partial randomness to efficiently generate high-fidelity spectra with fewer iterations. Experimental results on 48 categories of MFH herbal data show that DDIM-NIR achieved superior results with a maximum mean discrepancy of 0.0939, spectral angle mapper of 0.0051 rad, and percent root mean square difference of 2.77%, outperforming the classical Generative Adversarial Network (GAN) and Wasserstein Generative Adversarial Network with Gradient Penalty (WGAN-GP) deep learning models. Furthermore, DDIM-NIR is combined with three classification models to identify MFH herbs, with OneDimensional Convolutional Neural Network (1D-CNN) achieving 98.82% accuracy, 98.44% recall, and 98.58% F1-score using a dataset expanded fivefold. This method significantly enhances small-sample NIR analysis, enabling robust herb identification in practical quality control and traceability systems.

The regulatory mechanisms of soil multifunctionality (SMF) in response to vegetation succession and seasonal drought and wetness fluctuations remain poorly understood in karst regions, limiting our deep understanding of ecological functions in naturally restored karst vegetation. Here, we investigated three natural restoration stages (forest, shrubland, and grassland) across a well-defined successional gradient in Southwest China, with paired sampling during wet and dry seasons. By integrating multi-trophic microbial sequencing (bacteria, fungi, protists, and nematodes) with co-occurrence network analyses, we systematically examined how SMF responds to vegetation succession and seasonal variation. SMF and enzyme-mediated decomposition functions increased significantly with vegetation succession, exhibiting pronounced seasonal divergence. Both forest and shrubland supported substantially higher SMF during the wet season compared to the dry season. Multi-trophic biodiversity across all groups (bacteria, fungi, protists, and nematodes) was significantly higher in the dry season and increased progressively along the successional gradient. Occurrence network of soil biology in shrubland exhibited the highest complexity (greatest number of nodes, edges, and average degree) during the dry season, whereas that in forest demonstrated the highest network modularity. Notably, soil biotic network in grassland showed significantly enhanced connectivity and density during the wet season relative to these of dry season. SMF was regulated by both biotic (multi-trophic diversity, network complexity) and abiotic factors (soil moisture and pH), yet these drivers operated distinctly across seasons. During the dry season, soil multi-trophic diversity indirectly drove SMF primarily through network complexity, whereas direct effects of soil moisture dominated under wet season conditions. Collectively, vegetation succession, soil moisture, soil pH, diversity and network complexity of multi-trophic biology jointly explained 64% and 71% of SMF variation in dry and wet seasons, respectively. These findings elucidate the synergistic biotic–abiotic mechanisms governing SMF during natural vegetation restoration in fragile karst landscapes and their critical dependency on seasonal moisture regimes, providing a theoretical foundation for understanding soil functional improvement in recovering karst ecosystems.

The official Landsat 8 surface reflectance (SR) product, generated by the Land Surface Reflectance (LaSRC) algorithm, is the most extensively utilized medium-resolution dataset and serves as a benchmark to cross-validate the accuracy of other SR products. However, the accuracy of the Landsat 8 SR products did not meet the expectations of the previous studies under specific conditions. Consequently, it is necessary to analyze the Urban Clean aerosol-type assumption implemented in the LaSRC algorithm and comprehensively re-evaluate the accuracy of the Landsat 8 SR. Therefore, this study leverages Landsat 8 data over 600 scenes acquired at 100 Aerosol Robotic Network (AERONET) sites globally and conducts a comprehensive analysis of how different dynamic aerosol types – MOD04-based (used in the Moderate Resolution Imaging Spectroradiometer (MODIS) Atmosphere Level-2 Aerosol Optical Depth Product), MOD09-based (used in MODIS Terra Atmospherically Corrected Surface Reflectance Product), and Urban Clean (used in LaSRC) – affect the accuracy of atmospheric correction (AC) for the first time. The results indicated that, in terms of aerosol optical depth (AOD), the MOD04 aerosol type exhibited the highest accuracy, with a coefficient of determination (R2_AerT) of 0.7236, Root Mean Square Error (RMSE) of 0.0437, and bias of 0.0052. The accuracy (A), precision (P), and uncertainty (U) of the four evaluated SR products ranged from −2.9754 × 10−4 to 3.0145 × 10−3, from 2.3184 × 10−2 to 2.6020 × 10−2, and from 2.3366 × 10−2 to 2.6040 × 10−2, respectively. The MOD04-based aerosol type demonstrated the highest overall accuracy in the visible and near-infrared (VNIR) bands. The MOD09-based aerosol type outperformed the others in the bright surface regions. The Urban Clean aerosol type showed a comparable but slightly inferior performance to that of the MOD09-based aerosol type, with limited advantages in specific reflectance ranges. Moreover, LaSRC-derived SR demonstrated higher stability and accuracy in the shortwave infrared (SWIR) bands compared to its inferior performance in the VNIR. These findings emphasize the critical importance of aerosol-type assumptions in AC workflow. A mixed strategic implementation framework is proposed as follows: (1) adopt MOD04-based aerosol types for AOD retrieval and VNIR SR retrieval, (2) use MOD09-based aerosol types for scenes dominated by very high-reflectance surfaces, and (3) leverage SWIR SR products derived by LaSRC. Our findings provide actionable guidelines for dynamic aerosol-type selection to enhance the AC performance across diverse environments.

Significant advancements have been achieved in recent years in the restoration of degraded vegetation in the karst regions of Southwest China. However, the effects of this restoration on soil multifunctionality (SMF) and soil multi-trophic organism remained largely unexplored. This study investigated SMF of three typical artificial vegetation restoration patterns (grass, forest, and forest-grass) in the Southwest karst region. High-throughput sequencing, morphological identification, and co-occurrence network analysis were utilized to examine the characteristics of multi-trophic organism communities in karst soil. Vegetation restoration patterns significantly affected single soil functions, SMF, and the biodiversity of soil multi-trophic organism. All co-occurrence networks of soil organism were predominantly positive relationships, with network in forest-grass pattern exhibiting greater complexity than that of forest and grass patterns. The topological parameters of the co-occurrence network showed a significant correlation with soil functions, while soil multi-trophic organism diversity did not. Structural equation models revealed that the restoration pattern, soil pH, soil moisture, soil multi-trophic organism diversity, and co-occurrence network complexity collectively explained 64 % of the SMF variation. The restoration pattern not only directly influenced SMF (p < 0.01) but also indirectly affected SMF by altering soil pH and the complexity of the co-occurrence network. These insights provide a better comprehension of the mechanisms underlying vegetation restoration effects on soil functions in karst regions, specifically from the perspective of soil multi-trophic organism.

IntroductionThe revelation of the assembly mechanism of plant communities in karst region has crucial implications for the restoration of degraded vegetation. Niche theory and neutral theory are the two main theories to elucidate community assembly of karst plant community. However, the relative significance of habitat filtration and biological action in community assembly remains a topic of debate.MethodsBy using measurement of plant functional traits, detection of phylogenetic signal (K value), and average shared variance, our investigation aimed to ascertain whether species coexistence in community assembly of primary forest is driven by habitat filtering or biotic constraints.ResultsIn all 10 plant functional traits, leaf carbon (LC) had the lowest variation coefficient, whereas leaf area (LA) exhibited the highest. Significant phylogenetic signals (P < 0.05) were identified for plant LC, LA, wood density (WD), leaf nitrogen (LN) and leaf phosphorus (LP). Phylogenetic signal strength (K < 1) of all traits indicated that the phylogenetic conservation of functional traits is relatively weak and may be influenced by environmental screening or convergent evolution. Both the phylogenetic net relatedness index (NRI) and nearest taxon index (NTI) were negative, indicating a divergent phylogenetic structure. Additionally, with the exception of LA and leaf length-width ratio (L/D), the mean pairwise trait distance indices (SES.PW) were greater than 0, suggesting a tendency towards aggregation in the functional trait structure. Furthermore, average shared variance demonstrated that variation in plant functional trait was predominantly influenced by soil fertility and topography of the sample DiscussionOur finding indicated that the community assembly of primary forest plant was dominated by habitat filtering, which could significantly promote a more profound comprehension of natural restoration in karst degradation region.