Ground freeze-thaw dynamics critically affect carbon cycling and ecosystem stability in cold regions. In the frozen ground region of northeastern China (FGRN China), these dynamics are governed by synergistic biotic, climatic, physiographic, and anthropogenic drivers, making spatiotemporal characterization and causal attribution particularly challenging. We establish a ground freeze-thaw dynamic index (FTDI) based on the ground freezing index (GFI) and ground thawing index (GTI), to quantify ground freeze-thaw dynamics in FGRN China (1982-2020). Using geostatistics, geodetector, and structural equation model (SEM), we analyze spatiotemporal patterns, critical thresholds, and driving mechanisms. The results indicate that the area where FTDI <0 (i.e., GFI > GTI) is shrinking significantly at a rate of 0.45 x 10(4) km(2)/a, with its gravity center shifting from the sporadic permafrost region (SPR) toward the discontinuous permafrost region (DPR) across the Da Xing'anling Mountains. This indicates that ground warming will be more pronounced in DPR within FGRN China. Furthermore, critical thresholds were detected only for precipitation changes (approximate to -3.2 mm/a; beyond inhibiting thawing) and snow cover changes (approximate to -0.19 %/a; beyond promoting thawing). SEM revealed a succession of dominant controlling factors and mechanistic transitions across the frozen ground degradation gradient. Precipitation changes primarily promoted thawing in DPR. In SPR, the inhibitory effect of soil water changes became prominent, while precipitation changes shifted from promotion to inhibition (suggesting a threshold). In the isolated patch permafrost region, thawing was regulated by the inhibitory effect of precipitation and the promoting effect of altitude. In the seasonal frozen ground region, the snow cover changes shifted from inhibition to promotion of thawing (suggesting a threshold). These findings reveal the environmental complexity governing ground freezethaw dynamics and provide insights into ecosystem stability and climate change projections in cold regions.

Transformer models have been widely adopted for hyperspectral image (HSI) classification due to their exceptional long-sequence modeling capabilities. However, the self-attention mechanism in Transformers incurs quadratic computational complexity, posing challenges in both speed and memory consumption. Recently, a novel state-space model-the Mamba model-has emerged, overcoming the quadratic complexity of self-attention by achieving linear computational complexity while retaining powerful long-sequence modeling. Yet, the original Mamba design does not account for the unique spectral-spatial characteristics of HSI data, making it difficult to capture multiscale features. This limitation can lead to the loss of critical spectral-spatial cues at fine targets and complex boundaries, resulting in increased classification noise, blurred boundary segmentation, and reduced overall accuracy. To address the loss of fine-grained spectral-spatial information in HSI, we propose HyM3S: a hyperspectral multiscale spatial-spectral sequence model that integrates multiscale spatial-spectral convolutions with Mamba's linear sequence modeling.HyM3S first extracts multiscale spatial and spectral features in parallel along horizontal and vertical branches, and reinforces salient channels via channel-wise attention. Features are then adaptively fused across modality and directional dimensions to form a unified joint representation. Finally, this representation is fed into the Mamba module for long-range dependency modeling under linear complexity, thereby significantly improving classification accuracy and suppressing noise. Experiments on four benchmark HSI datasets-Pavia University (PaviaU), Houston2013, WHU-Hi-HanChuan, and WHU-Hi-HongHu-demonstrate the clear superiority of the proposed HyM3S model for HSI classification.

Transformer models have been widely adopted for hyperspectral image (HSI) classification due to their exceptional long-sequence modeling capabilities. However, the self-attention mechanism in Transformers incurs quadratic computational complexity, posing challenges in both speed and memory consumption. Recently, a novel state-space model—the Mamba model—has emerged, overcoming the quadratic complexity of self-attention by achieving linear computational complexity while retaining powerful long-sequence modeling. Yet, the original Mamba design does not account for the unique spectral–spatial characteristics of HSI data, making it difficult to capture multiscale features. This limitation can lead to the loss of critical spectral–spatial cues at fine targets and complex boundaries, resulting in increased classification noise, blurred boundary segmentation, and reduced overall accuracy. To address the loss of fine–grained spectral–spatial information in HSI, we propose HyM3S: a hyperspectral multiscale spatial–spectral sequence model that integrates multiscale spatial–spectral convolutions with Mamba’s linear sequence modeling.HyM3S first extracts multiscale spatial and spectral features in parallel along horizontal and vertical branches, and reinforces salient channels via channel–wise attention. Features are then adaptively fused across modality and directional dimensions to form a unified joint representation. Finally, this representation is fed into the Mamba module for long–range dependency modeling under linear complexity, thereby significantly improving classification accuracy and suppressing noise. Experiments on four benchmark HSI datasets—Pavia University (PaviaU), Houston2013, WHU–Hi–HanChuan, and WHU–Hi–HongHu—demonstrate the clear superiority of the proposed HyM3S model for HSI classification.

Remote sensing-based landslide segmentation is of great significance for geological hazard assessment and post-disaster rescue. Existing convolutional neural network methods, constrained by the inherent limitations of spatial convolution, tend to lose high-frequency edge details during deep semantic extraction, while frequency-domain analysis, although capable of globally preserving high-frequency components, struggles to perceive local multi-scale features. The lack of an effective synergistic mechanism between them makes it difficult for networks to balance regional integrity and boundary precision. To address these issues, this paper proposes the Geometric Context and Frequency Domain Fusion Network (GCF-Net), which achieves explicit edge enhancement through a three-stage progressive framework. First, the Pyramid Lightweight Fusion (PGF) block is proposed to aggregate multi-scale context and provide rich hierarchical features for subsequent stages. Second, the Geometric Context and Frequency Domain Fusion (GCF) module is designed, where the frequency-domain branch generates dynamic high-frequency masks via the Fourier transform to locate boundary positions, while the spatial branch models foreground–background relationships to understand boundary semantics, with both branches fused through an adaptive gating mechanism. Finally, Edge-aware Detail Consistency Improvement (EDCI) module is designed to balance boundary preservation and noise suppression based on edge confidence, achieving adaptive output refinement. Under the joint supervision of Focal loss, Dice loss, and Edge loss, experiments on the mixed dataset and LMHLD dataset demonstrate that GCF-Net achieves OAs of 96.42% and 96.71%, respectively. Ablation experiments and visualization results further validate the effectiveness of each module and the significant improvement in boundary segmentation.

Semantic segmentation of high-resolution remote sensing imagery (RSI) is essential for applications such as land-cover mapping, urban monitoring, and disaster assessment, yet it remains challenging due to complex backgrounds, high inter-class similarity, and large object-scale variation. We propose MBS-Net (Multi-Scale Attention Boosted Segmentation Network), a Multi-Scale Attention-Enhanced Network built on an encoder-decoder architecture to address blurred boundaries, loss of fine structures during downsampling, and insufficient multi-scale context modeling. The encoder integrates a ResNet-50 backbone augmented with Squeeze-and-Excitation (SE) blocks to recalibrate channel responses and highlight discriminative features. To reduce semantic misalignment in skip connections, we introduce a Fine-Scale Edge Attention (FSEA) module that combines a spatial attention stream for boundary cues with a channel attention stream for small-scale structure preservation, enabling precise fusion of encoder and decoder features. In the decoder, an Efficient Multi-Scale Attention (EMA) module aggregates parallel contextual paths with varied receptive fields to balance global semantics and local detail. By jointly leveraging SE, FSEA, and EMA, MBS-Net simultaneously strengthens feature representation, preserves fine structures, and adaptively models cross-scale context. Experimental results on the ISPRS Potsdam and Vaihingen datasets demonstrate that MBS-Net achieves consistent and substantial gains in segmentation accuracy and boundary delineation.

Abstract Quantitative analysis of blue carbon stored in mangrove ecosystems is essential for understanding their role in climate change mitigation. However, the impacts of edge effects in fragmented mangrove forests are often overlooked in large‐scale blue carbon assessments. In this study, we present the first global‐scale quantitative assessment of how edge effects influence carbon storage in Ramsar‐protected mangrove forests, emphasizing both the necessity of incorporating these effects into carbon assessments and their variability across different fragmentation gradients. Specifically, we identified three distinct patterns of edge effects along fragmentation gradients, each further divided into six scenarios based on biomass decay rate dynamics. Our results indicate that the estimated edge‐effect penetration distances were less than 100 m in approximately 63.37% of the studied Ramsar‐protected mangrove forests. In these edge zones, biomass declines by an average of 5.15% compared with interior zones. Neglecting such edge effects has likely led to an overestimation of Ramsar‐protected mangrove carbon stocks by about 2.08 Tg C. These findings underscore the importance of accurately accounting for edge effects in mangrove carbon assessments, as their omission can substantially bias carbon storage estimates and, in turn, hinder effective management of blue carbon ecosystems.

In the context of accelerating global urbanization and sustainable development challenges, impervious surfaces, as a key component of urban land cover, are significantly associated with regional economic development. This study takes Harbin, a typical cold region city, as a research object and constructs a three-level analytical framework of “land surface classification-economic simulation-mechanism analysis.” By innovatively integrating multi-source remote sensing, demographic, and economic data, the research addresses gaps in understanding urban sustainability in cold environments. An enhanced XGBoost algorithm was employed to achieve high-precision classification of ten land surface materials, resulting in a high overall accuracy. Furthermore, a gridded GDP spatialization model developed using high-resolution population data demonstrated superior performance compared to traditional methods. Machine learning-assisted analysis revealed that asphalt and metal surfaces are the most significant impervious materials driving economic output, reflecting the respective influences of transportation infrastructure and industrial agglomeration. Spatial pattern analysis indicates that Harbin’s impervious surfaces exhibit a lower fractal dimension and a distinct grid-like morphology compared to the typical subtropical city of Guangzhou, underscoring urban form adaptations to cold climatic constraints. The strong spatial coupling between gradients of GDP intensity and the attenuation of impervious surface density is quantitatively confirmed. This study provides a quantitative basis and a transferable technical framework for optimizing land use intensity and infrastructure planning in cold cities, thereby offering a scientific foundation for sustainable, intensive land utilization in climate-vulnerable urban systems.

Phenological development is intricately linked to climate change, and existing evidence suggests that the frequent occurrence of extreme weather events (EWEs) exerts a more profound influence on phenology. However, most prior studies have primarily focused on revealing the impact of overall extreme weather conditions over a period on specific phenological transition dates. There is an urgent need to obtain short-term, continuous development processes to deeply understand the underlying response mechanisms. In this study, we developed and assessed the Daily Phenological Development Model (DPDM) to explore the continuous influence relationship between remotely sensed vegetation phenology and EWEs in the Lesser Khingan Mountains of northeastern China from 2000 to 2022. The DPDM effectively captures daily vegetation phenological development during both spring and autumn, with R2 values of 0.84 +/- 0.06 and 0.78 +/- 0.08, respectively. Additionally, our findings indicate that increased precipitation and more frequent hot events accelerate leaf greening and slow down leaf browning. In contrast, heavy rainfall, frost events and droughts negatively impact vegetation growth. Among these factors, frost events have the most significant inhibitory effect on spring vegetation growth. Furthermore, evergreen needleleaf forests demonstrate the strongest resistance to EWEs in all vegetation types. These insights provide valuable contributions to the accurate assessment and forecasting of vegetation phenology under climate change.

The net effects of permafrost degradation on soil organic carbon (SOC) stock and the underlying mechanisms are unclear due to poor understanding of the response of particulate (POC) and mineral-associated organic carbon (MAOC) pools. To explore changes in POC, MAOC, and SOC across permafrost degradation, we collected 72 soil samples at 0-10, 10-20, and 20-30 cm from three permafrost types characterized by an increasing degradation. While the overall SOC content remained stable, POC content decreased by 45-77 g kg(-1) by permafrost degradation and increased MAOC content by 57-96 g kg(-1) across the three soil depths. Specifically, permafrost degradation decreased the portion of POC in SOC by 15-37 %, but increased the portion of MAOC in SOC by 16-37 %. Using random forest modeling, we identified that the content of total nitrogen (TN), microbial biomass carbon (MBC), and microbial biomass nitrogen (MBN) are key factors influencing the contribution of POC and MAOC to the overall SOC. These variables together explained 45 % and 46 % variations of POC/SOC and MAOC/ SOC, respectively. Altogether, this study demonstrates how POC and MAOC content and their contributions to SOC change across permafrost degradation, and highlights the distinct roles of POC and MAOC in sustaining SOC storage under climate change.

Global warming has accelerated permafrost degradation, destabilizing soil organic carbon (SOC) in wetlands. Soil microorganisms play a crucial role in SOC decomposition. Despite increasing studies on soil microorganisms and their associated impacts on SOC decomposition in wetlands, few studies have simultaneously investigated the impacts of soil physical properties and nutrient stoichiometry on soil microbial communities across diverse swamp wetlands. By selecting three types of swamp wetlands in the Xiaoxing’an Mountains, China, this study explored the relationships between soil physical properties, nutrient stoichiometry and microbial communities across different soil profiles. Our results showed that there were significant differences on soil physical properties and soil pH across the three types of swamp wetlands, leading to divergent patterns on soil carbon, nitrogen and phosphorus content, as well as on microbial diversity. Soil depths also significantly affected carbon, nitrogen, phosphorus content and bacterial diversity. Particulate organic carbon (POC) strongly correlates with microbial community composition and diversity, making it a key indicator. Dominant bacteria included Proteobacteria, Acidobacteriota and Actinomycetes, while over 80% of fungi were Ascomycetes, Basidiomycetes and Mortierellomycota. These findings will enhance the understanding of the relationship between microbial community distribution patterns and carbon components in different swamp wetlands, providing both theoretical and empirical insights into the role of soil microorganisms in the carbon cycle.

The forest swamp ecosystem, as a special wetland ecosystem, is a key link in the material cycle and an important carbon sink in the carbon cycle. The global carbon cycle is of great significance, but the impact of forest swamp succession and soil depth on soil active organic matter and nematode community structure and diversity is unclear. This study used the “space instead of time” method to investigate the succession process of forest swamps from grasslands (WC) and shrubs (WG) to forests (WS) in national nature reserves. The results showed that during the forest succession process, the dominant nematode communities in the WC and WG stages were dominated by the genera Apis and Labroidei, while the dominant genera increased in the WS stage. The total abundance of nematodes increased, and the number of groups was ordered WG > WC > WS. The diversity in soil nematode communities according to Shannon–Wiener (H′), Pielou (J), and Trophic diversity (TD) was WS > WG > WC, which is related to vegetation, soil physical and chemical properties, and microbial community structure. The maturity index (MI) was WG > WS > WC. The soil food web was dominated by bacterial channels and had characteristics in forest metabolic activity and regulation ability. At different soil depths, there were significant differences in the community, with species such as the spiny cushioned blade genus being key. The number and group size of nematodes varied from 0–10 cm > 10–20 cm > 20–30 cm. The relative abundance of feeding nematodes changed with depth, while diversity indices such as H′, J, and TD decreased with depth. Ecological function indices such as MI and PPI showed depth variation patterns, while basic indices (BI) and channel indices (CI) showed significant differences. In terms of soil variables, during the forest succession stage, soil organic carbon (SOC), soluble organic nitrogen (DON), easily oxidizable organic carbon (ROC), microbial biomass carbon (MBC), and microbial biomass nitrogen (MBN) showed a gradually increasing trend with WC-WG-WS, while total nitrogen (TN), soluble organic carbon (DOC), soil temperature (ST), and soil moisture (SM) showed opposite changes. There were significant differences in soil ST, SM, and DON values with succession (p < 0.05). At different soil depths, except for DON and ROC, which increased first and then decrease with depth, the values of other physical and chemical factors and active carbon and nitrogen components at depths of 0–10 cm were higher than those at other depths and decreased with depth. An analysis of variance showed significant differences in MBC and MBN values at different soil depths (p < 0.05), which is of great significance for a deeper understanding of the mechanism of soil nematode community construction and its relationship with the environment.

Surface freezing and thawing processes pose significant influences on surface water and energy balances, which, in turn, affect vegetation growth, soil moisture, carbon cycling, and terrestrial ecosystems. At present, the changes in surface freezing and thawing states are hotspots of ecological research, but the variations of surface frozen days (SFDs) are less studied, especially in the permafrost areas covered with boreal forest, and the influence of the environmental factors on the SFDs is not clear. Utilizing the Advanced Microwave Scanning Radiometer for EOS (AMSRE) and Microwave Scanning Radiometer 2 (AMSR2) brightness temperature data, this study applies the Freeze–Thaw Discriminant Function Algorithm (DFA) to explore the spatiotemporal variability features of SFDs in the Northeast China Permafrost Zone (NCPZ) and the relationship between the permafrost distribution and the spatial variability characteristics of SFDs; additionally, the Optimal Parameters-based Geographical Detector is employed to determine the factors that affect SFDs. The results showed that the SFDs in the NCPZ decreased with a rate of −0.43 d/a from 2002 to 2021 and significantly decreased on the eastern and western slopes of the Greater Khingan Mountains. Meanwhile, the degree of spatial fluctuation of SFDs increased gradually with a decreasing continuity of permafrost. Snow cover and air temperature were the two most important factors influencing SFD variability in the NCPZ, accounting for 83.9% and 74.8% of the spatial variation, respectively, and SFDs increased gradually with increasing snow cover and decreasing air temperature. The strongest explanatory power of SFD spatial variability was found to be the combination of air temperature and precipitation, which had a coefficient of 94.2%. Moreover, the combination of any two environmental factors increased this power. The findings of this study can be used to design ecological environmental conservation and engineer construction policies in high-latitude permafrost zones with forest cover.

Download This Paper Open PDF in Browser Add Paper to My Library Share: Permalink Using these links will ensure access to this page indefinitely Copy URL Copy DOI

As the regulator of water and nutrient changes in the active layer after permafrost degradation, root signaling substances affect the plant–soil carbon allocation mechanism under climate warming, which is a key issue in the carbon source/sink balance in permafrost regions. To explore how plant root signaling substances regulate carbon allocation in plants and soils under permafrost degradation, the changes in carbon allocation and root signaling substances in the plants and soils of peatland in different permafrost regions at the time of labeling were studied by in situ 13C labeling experiments. The results showed that the fixed 13C of Larix gemlini, Carex schumidtii, and Sphagnum leaves after photosynthesis was affected by permafrost degradation. In regions with more continuous permafrost, the trend of the L. gemlini distribution to underground 13C is more stable. Environmental stress had little effect on the 13C accumulation of Vaccinium uliginosum. Nonstructural carbohydrates, osmotic regulatory substances, hormones, and anaerobic metabolites were the main root signaling substances that regulate plant growth in the peatlands of the three permafrost regions. The allocation of carbon to the soil is more susceptible to the indirect and direct effects of climate and environmental changes, and tree roots are more susceptible to environmental changes than other plants in isolated patches of permafrost regions. The physical properties of the soil are affected by climate change, and the allocation of carbon is regulated by hormones and osmotic regulators while resisting anoxia in the sporadic regions of permafrost. Carbon allocation in discontinuous permafrost areas is mainly regulated by root substances, which are easily affected by the physical and chemical properties of the soil. In general, the community composition of peatlands in permafrost areas is highly susceptible to environmental changes in the soil, and the allocation of carbon from the plant to the soil is affected by the degradation of the permafrost.

Elucidating pollution characteristics of polycyclic aromatic hydrocarbons (PAHs) in water and assessing the associated carcinogenic risks is crucial for improving public health. PAHs in the surface water of seven main river basins across China, compiled from 95 studies from 2004 to 2022, were used to investigate geographic variations of occurrence, source, and carcinogenic risk. Total PAH concentrations exhibited substantial geographic distributions ranging from 300 to 7552 ng·L−1. Low molecular weight PAHs predominated, showing three-ring PAHs abundant in the north, while two-ring PAHs dominated in the south due to distinctions regarding energy consumption. The northern basins exhibited higher concentrations of PAHs than the southern owing to the synergistic impacts of low temperature, increased energy consumption, and higher industrial activities. Coal combustion and industrial emissions were the primary contributors in the northern basins, accounting for 23–44% and 20–38%, respectively, which were associated with pollutants released from heavy industries and space heating during cold periods. In contrast, vehicle exhaust emissions and petroleum leakage from river transport constituted the principal sources in the relatively economically developed southern basins, accounting for 24–35% and 31–57%, respectively. A lifetime carcinogenic risk model revealed that the highest health risks existed in adults, followed by adolescents and children. Toxic concentrations of BaP and the daily intake of water directly enhanced the PAHs’ carcinogenic risks, while body weight featured negative correlations with the risks.

Understanding water quality is crucial for environmental management and policy formulation. However, existing methods for assessing water quality are often unable to fully integrate with multi-source remote sensing data. This study introduces a method that employs a stacking algorithm within the Google Earth Engine (GEE) for classifying water quality grades in the Songhua River Basin (SHRB). By leveraging the strengths of multiple machine learning models, the Stacked Generalization (SG) model achieved an accuracy of 91.67%, significantly enhancing classification performance compared to traditional approaches. Additionally, the analysis revealed substantial correlations between the normalized difference vegetation index (NDVI) and precipitation with water quality grades. These findings underscore the efficacy of this method for effective water quality monitoring and its implications for understanding the influence of natural factors on water pollution.

2020年初COVID-19疫情期间人类活动减少,对于各地植被的生长状况造成了不同程度的影响.本文利用2015—2020年MOD13Q1影像分析哈尔滨、北京、武汉、广州及纽约的NDVI时空变化特征.通过Theil-Sen Median趋势分析和Mann-Kendell的检验数据来分析NDVI空间上的变化特征,并结合对应的气温和降水数据分析其相关影响,利用残差分析研究人类活动减少对各地区植被生长状况的影响.研究表明:1)从时间变化上来看,除纽约外,其余城市NDVI值均较上一年增加.2)从变化趋势上来看,除广州植被改善的区域小于植被退化的区域外,其余城市均大于植被退化的区域.3)从气温降水方面分析,只有哈尔滨与气温正相关,其余城市均为负相关;只有广州与降水呈负相关,其余城市均正相关.4)通过残差分析,北京、武汉、广州人类活动的减少对植被的生长状况起促进作用;而哈尔滨、纽约人类活动的减少可能没有对植被的生长产生正面影响.

Using basic data from China’s 30 provincial regions from 2005 to 2015, the slacks-based measure of super-efficiency in data envelopment analysis (super-efficiency SBM-DEA) is used to measure industrial green economic efficiency (IGEE). Moreover, the temporal change and spatial equilibrium of IGEE are analyzed, and the panel quantile regression model (PQRM) is used to empirically analyze the impacts of the degree of agglomeration of pollution-intensive industries (PIIs) and other influencing factors on IGEE. The results reveal the following. (1) The kernel density distribution curve of China’s IGEE exhibits a double-peaked distribution, and the overall peak value of the curve changes relatively greatly, exhibiting obvious spatial inequality. (2) Based on their trajectories, the centers of gravity of IGEE are all distributed in the central region. The moving speed of the centers of gravity exhibits a trend of “accelerating–decreasing–quickly accelerating.” Judging from the changes of the standard deviation ellipse (SDE), the spatial distribution pattern of China’s IGEE generally presents a northeast–southwest pattern, and it has exhibited the trend of shifting to a north–south pattern in recent years. (3) The agglomeration level of PIIs has a significant positive effect on the development of China’s IGEE at each quantile level, and the most significant impact on China’s IGEE is at the 75th quantile level. This paper empirically analyzes the impact of the agglomeration level of PIIs on IGEE to aid in the greening of industrial economic efficiency and the sustainable development of the ecological environment in China and other developing countries.

Soil organic carbon (SOC) is a sensitive indicator of climate change, and small changes in the soil carbon pool will affect the carbon balance. Accurate and robust SOC quantitative prediction is of great significance to studying the carbon budget of swamp wetlands and its response to climate change. In this study, a new framework was proposed and assessed for predicting the SOC content based on Sentinel-2 (S2), Sentinel-1 (S1), and the digital elevation model (DEM) together with the extreme gradient boosting with random forest (XGBRF) model. The determination coefficient (R-2), root mean square error (RMSE), mean absolute error (MAE), and Lin's concordance correlation coefficient (LCCC) were applied to assess the performances of the models. The results revealed that the prediction performance of the XGBRF regression model was much better than that of extreme gradient boosting and random forest regression models. Compared with single sensor data, using multisensor data to predict the SOC content yielded more accurate results. The XGBRF model based on S1, S2, and DEM fusion yielded the highest prediction accuracy (R-2_testing = 0.6639, RMSE = 1.3236 g/kg, MAE = 1.2546 g/kg, LCCC = 0.7621). Regarding the importance of the variables, the S1 and S2 features were major contributors to the SOC content prediction (41% and 52%, respectively), followed by the topographic variables extracted from the DEM (7%). The proposed framework can be used for SOC prediction based on a small sample dataset, and it provides a method for long-term and rapid monitoring of the SOC contents in wetlands.

In the context of global warming, melting permafrost increases the instability of soil organic matter in wetlands, and soil microorganisms play an important role in the process of decomposing organic matter. However, how the relationship between soil physical properties and soil nutrient stoichiometry drives changes in the soil microbial community in different swamps under permafrost degradation remains unclear. Therefore, this study selected three kinds of swamp wetlands in a region with sporadic and isolated patches of permafrost in the Xiaoxing'an Mountains as the research area, and analyzed the characteristics and driving mechanisms of soil microbial communities (bacteria and fungi) in the 0–10, 10–20, and 20–30-cm soil layers of the three typical wetlands. The results showed that different habitats affected soil physical properties and pH, thereby driving changes in soil carbon, nitrogen, and phosphorus components and indirectly affecting bacterial diversity and fungal community composition. Soil depth had significant effects on soil carbon, nitrogen, and phosphorus ratios, storage, and bacterial diversity. Particulate organic carbon (POC) showed a strong correlation with microbial community composition, suggesting that POC could be used as an important indicator of microbial community composition. Proteobacteria, Acidobacteriota, and Actinomycetes predominated in soil bacterial communities, and more than 80% of soil fungal communities were composed of Ascomycetes, Basidiomycetes, and Mortierellomycota. The results of this study clarified the relationship between the distribution characteristics of soil bacteria and fungi and the carbon, nitrogen, and phosphorus components, and provided basic data and theoretical support for understanding the microbial driving mechanism of carbon cycling and the maintenance mechanism of the carbon sink function in wetland ecosystems.