Forest disturbance detection is essential for monitoring ecosystem dynamics, as it provides critical insights into the impacts of climate change, natural disasters, and human activities on forest structure and function. Characterizing the spatiotemporal patterns of disturbances is key to assessing carbon sink loss, biodiversity decline, and ecological recovery potential, thereby offering scientific evidence to support global sustainability goals. In this study, we propose the GRU-LPETransformer model, which integrates local temporal feature extraction using a Gated Recurrent Unit (GRU) with global dependency modeling via a Transformer enhanced by learnable positional encoding (LPE). Using Landsat-derived Normalized Burn Ratio (NBR) time series, we conducted forest disturbance detection and spatiotemporal analysis. Training and validation were based on a reference dataset constructed through visual interpretation. Experimental results show that the model achieves an accuracy of 82.59

As global climate change intensifies, the urban heat island (UHI) has become an increasingly significant challenge, impacting both urban environments and public health. We used the SSP-RCP model to analyze the spatiotemporal dynamics of heat and cold island patches in Fuzhou under projected climate change, integrating spatial analysis and network evolution theory. The results show that: (1) Cold island patches are generally decreasing in size, while heat island patches are expanding. The most significant increase in heat island area occurs under the SSP585 scenario; (2) The spatial distribution of heat and cold sources has shifted, with cold sources moving from southwest to northeast and heat sources from south to north; (3) Climate change scenarios exert a substantial influence on the urban thermal environment (UTE). In low-intensity scenarios, the reduction of cold islands and the expansion of heat islands may exacerbate the heat island effect. Conversely, in high-intensity scenarios, the intensification of the heat network may further aggravate the UTE; (4) Synergistic development of cities within metropolitan areas, strengthening the connectivity of cold networks, and enhancing the resilience of heat networks are crucial strategies for mitigating the negative impacts of UHI and improving cities’ capacity to cope with climate change. This study reveals the spatial and temporal evolution of heat and cold networks in the Fuzhou metropolitan area, providing a theoretical foundation for UHI mitigation under varying development scenarios and for the optimization and adaptation of UTE in response to future climate change.

Individual culm locations underpin culm-density estimation, spatial-pattern analysis, and precision management in moso bamboo forests. In dense stands, canopy occlusion, repetitive culm arrangements, and weak texture constrain manual surveys, Unmanned Aerial Vehicle (UAV) imagery, and LiDAR-based mapping in terms of efficiency, understory visibility, or cost. RGB-D cameras simultaneously capture image observations and metric depth during near-ground mobile acquisition, while continuous sequences provide multi-view three-dimensional constraints. This study therefore develops an RGB-D visual simultaneous localization and mapping (visual SLAM) method for individual culm localization. Built on ORB-SLAM3, the method introduces a semantic partitioning strategy that uses ground keypoints to sustain camera tracking and a BambooPoint keypoint-proposal module for culm regions. BambooPoint redirects generic texture-driven keypoint detection toward culm-structure responses, suppressing non-target features from foliage, shadows, and ground texture, and producing more continuous sparse three-dimensional support along the culm height. Individual culm centers are subsequently estimated through height slicing and cross-layer aggregation. Validation against 258 culms from five plots in Fujian and Jiangxi showed that the method correctly matched 224 culms, with precision, recall, and F1-score values of 93.7%, 86.8%, and 0.901, respectively, and a mean horizontal localization error of 0.179 m. Compared with generic keypoint-proposal modules, BambooPoint markedly reduced spurious culm detections while preserving a high true-culm detection rate, limiting false positives to 15. These results demonstrate its potential for near-ground mapping of individual culms and for supporting culm-density estimation and spatial-pattern analysis in complex bamboo forests.

Accurate assessment of site quality in coastal Casuarina equisetifolia (C. equisetifolia) plantations is essential for enhancing the protective function of shelterbelts and implementing site-specific afforestation strategies. However, traditional ground-based surveys are limited in spatial coverage and efficiency, hindering effective forest management. To overcome these limitations, this study developed an integrated assessment framework that couples ground-based modeling with remote sensing inversion to achieve large-scale site quality mapping. Field investigations on Pingtan Island, Fujian Province, China, were used to establish a ground-based evaluation model. Soil fertility was quantified using Principal Component Analysis (PCA), and principal components were classified into discrete fertility grades through K-means clustering. These grades, together with topographic variables, were incorporated into a site quality classification model constructed using Quantification Theory I. The point-based model was subsequently extrapolated using Landsat 9 imagery to generate a spatially continuous site quality map. Spatial autocorrelation (Moran’s I) and LISA clustering were further employed to interpret spatial patterns. Results indicate that coastal sandy soils in the study area are generally nutrient-poor, with tree growth primarily constrained by total nitrogen, organic matter, available phosphorus, and total phosphorus. The five most influential site factors, ranked by importance, are soil fertility, distance from the coastline, aspect, slope gradient, and elevation. Optimal conditions for C. equisetifolia growth include fertile soil, location >1000 m from the coastline, south-facing or semi-sunny slopes, slope gradients <15°, and elevations between 10–100 m. Only 11.94

Accurately determining the age of Moso bamboo and establishing a reasonable age structure for bamboo forests are essential prerequisites for the scientific management and productivity maximization of Moso bamboo forests. Visible light images offer the advantages of low acquisition cost and abundant information. These images reveal significant phenotypic differences in Moso bamboo of varying ages, allowing for the use of such images in age determination. However, the phenotypic recognition of Moso bamboo is often influenced by environmental conditions, leading to variations in image brightness and high similarity in the features of Moso bamboo across different age groups, thereby reducing age determination accuracy. This study analyzes the phenotypic characteristics of Moso bamboo at different ages and their susceptibility to brightness variations, proposing a novel age determination method using visible light images. The YOLO target detection model and the Segment Anything Model (SAM) are employed to extract Moso bamboo regions from images. Phenotypic features under abnormal brightness are corrected using four image enhancement techniques. Subsequently, various age models, constructed using different backbone networks and loss functions, are compared and analyzed to identify the optimal combination for Moso bamboo age determination. The best image enhancement method and model were selected to discriminate the age of Moso bamboo. The results show that brightness significantly affects the phenotypic characteristics of Moso bamboo in the images. In this study, adaptive histogram equalization was used to enhance the images, and an age determination model was built using ResNet-101 with Focal Loss and Label Smoothing Regularization (LSR). The accuracy of age determination for Moso bamboo reached 88.6 %, representing a 10.5 % improvement over the baseline model's accuracy of 78.1 %.

Spatial differentiation of urban natural basement conditions leads to significant differences in urbanization development patterns and land evolution patterns in different regions. Taking Fuzhou, a typical coastal basin city located in the Minjiang River Estuary, as the study area, this paper analyzes the spatiotemporal evolution characteristics of land use/cover change (LUCC) and quantifies its driving mechanism from 1990 to 2020, by using the land use transition matrix (LUTM), the center-of-gravity model (CGM), the standard deviation ellipse (SDE), and the optimal parameters-based geographical detector (OPGD). The results show that (1) the land use structure has undergone drastic restructuring, the built-up land has increased significantly, the grassland has decreased significantly, and the cropland and forest land have shown phased evolution characteristics: a light increase from 1990 to 2000 and a continuous decline from 2000 to 2020. Water exhibited a fluctuating pattern: shrinking from 1990 to 2000, expanding from 2000 to 2010, and shrinking again from 2010 to 2020. (2) Constrained by the terrain of the Minjiang Estuary Basin, the gravity centers of cropland and grassland shifted northwestward, forest land moved southeastward, water shifted northeastward, and built-up land expanded northward. (3) Driving factors exhibited stagewise differences: socioeconomic factors played a dominant role from 1990 to 2000, with population density (q = 0.4029) and nighttime light (q = 0.3639) being significantly higher than other factors. From 2000 to 2010, the terrain constraint effect continued to intensify, with GDP (q = 0.4470), nighttime light (q = 0.3658) and DEM (q = 0.3638) as the dominant factors. From 2010 to 2020, urban land pattern evolution was jointly driven by multiple factors. This study clarifies the land use evolution mechanism of coastal basin cities during urbanization, providing a scientific reference for the sustainable development of similar coastal basin cities.

Reliable photovoltaic (PV) power forecasting is essential for the secure operation and optimal dispatch of power systems. Unlike deterministic methods that provide only point estimates, uncertainty forecasting characterizes the distribution of future generation and associated risks, thereby better supporting robust scheduling and risk management. However, PV output is influenced by multiple meteorological factors and exhibits pronounced volatility and multivariate coupling, posing challenges for uncertainty quantification. Inspired by the concepts of structured time-series decomposition and adaptive temporal-scale fusion, this study proposes STD-MoE, a structured time-series decomposition framework for PV uncertainty forecasting. STD-MoE decomposes PV power series into structural components and adopts component-adaptive feature-extraction branches to facilitate parallel and effective component-wise modeling. A sparse Mixture-of-Experts mechanism is introduced to adaptively select periodic patterns, while a polynomial interaction modeling scheme is employed to capture cross-component synergistic relationships. Based on the fused representation, multi-quantile prediction heads generate quantile forecasts, and conformal calibration is applied to improve the reliability of interval coverage. Experimental results demonstrate that STD-MoE delivers improvements of 17.8% in uncertainty forecasting (PL) and 19.2% in point forecasting (MSE) relative to the mean performance of the baseline methods, while maintaining a parameter-efficient and interpretable model architecture conducive to deployment in engineering systems.

In the hilly regions of southern China, complex terrain significantly affects UAV-LiDAR point cloud processing, leading to discrepancies in canopy structural parameters before and after ground point fitting—a phenomenon termed “canopy distortion.” This distortion poses a challenge for the high-precision extraction of three-dimensional (3D) canopy structures. Using Cunninghamia lanceolata as the study species, this research systematically analyzed how terrain factors influence single-tree canopy parameters. Ground points were fitted using the RANSAC algorithm to derive 3D slopes (X, Y, Z) and intercepts (X, Y, Z). These derived variables, along with conventional terrain attributes (slope, aspect, curvature), were employed to quantify terrain-induced effects on canopy parameters. A Bayesian-optimized random forest model, featuring automatic parameter tuning to enhance predictive performance, was applied to predict changes in canopy parameters and assess the relative contributions of each terrain factor. Following normalization, terrain effects were mitigated, though residual distortions persisted: mean treetop offsets in X and Y directions remained below 0.1 m; changes in crown width and projected area were negligible; mean tree height decreased by 0.77 m; crown length increased by 0.05 m; while crown surface area and volume decreased by 0.54 m2 and 0.62 m3, respectively. Tree height was primarily affected by intercept Y, crown volume and surface area by slope X, and crown length by curvature and slope. Model accuracy was highest for treetop X offset (R2 = 0.730) and lowest for crown surface area (R2 = 0.143). These findings elucidate how terrain induces distortions in canopy parameters and underscore the necessity of systematically quantifying terrain effects using enriched topographic variables and predictive modeling. This study provides a theoretical foundation for accurately extracting 3D canopy structures in complex terrains and supports intelligent forest monitoring and site quality assessment.

Rapid and effective estimation of soil organic matter (SOM) is crucial for the scientific management of Moso bamboo forests. This study investigated Moso bamboo forest soils in Yongan City, Fujian Province, and systematically evaluated the synergistic adaptation strategies coupling spectral preprocessing methods, feature extraction strategies, and machine learning models based on visible and shortwave near-infrared (Vis-NIR) spectroscopy. The results indicated that: (1) Conventional preprocessing algorithms attenuated the SOM spectral feature signals dominated by soil color within the limited wavelength range of field in situ spectral data, resulting in a general decline in the accuracy of the estimation models. (2) Feature extraction and modeling algorithms exhibited distinct adaptability across different content intervals. Within the low-content interval (<15 g/kg), simple physical indices combined with random forest (RF) achieved effective estimation at a lower computational cost (RPD = 2.18). Within the high-content interval (>25 g/kg), the synergistic strategy of the CARS algorithm combined with support vector regression (SVR) yielded the optimal estimation performance (R2 = 0.83, RPD = 2.45) and effectively mitigated the underestimation of high values caused by data imbalance. In conclusion, this study proposed a feature-model synergistic estimation approach, validating its feasibility for estimating SOM in Moso bamboo forests under the specific constraints of the current study area, thereby serving as a valuable reference for forest soil SOM monitoring in specific regions.

Traditional gravity models typically employ a power-law function derived from biological migration patterns, which differs intrinsically from the physical mechanisms of heat transfer. Given that heat flow exhibits exponential decay characteristics associated with thermodynamic gradients, this study introduces an exponential function to replace the power-law function, proposing a modified gravity model to accurately identify key corridors for mitigating the urban heat island (UHI) effect. Based on this modified model, we compared the average resistance differences between key and general corridors, along with the network topology following the removal of key corridors, to validate the effectiveness of the modification. Subsequently, a segmented linear regression model was employed to investigate the trends and statistical inflection points regarding the influence of 2D and 3D indicators of patches and corridors on the corridor thermal intensity (CTI). The results indicate that: (1) The modified gravity model exhibited superior performance in the cold island (CI) network, whereas the traditional model was more suitable for the heat island (HI) network. (2) The optimal λ values for the CI network were 1.3 and 1.4, while the value for the HI network was 0.1. (3) In the HI network, the key factors were the proportions of impervious surfaces and cropland within the corridor, and the mean SVF; in the CI network, the key factors were the proportions of water and cropland within the corridor. This study provides a theoretical framework more aligned with thermodynamic principles for quantifying the thermal regulation functions of urban corridors. Practically, the identified key corridors and morphological thresholds offer precise scientific guidance for optimizing the thermal environment in the central urban area of Chongqing.

The aging of Casuarina equisetifolia (C. equisetifolia) in coastal shelterbelts and the depletion of soil nutrients has led to a decline in their protective capabilities, with significant differences among various soil types. This study investigates the effects of soil nutrient availability on the growth mechanisms of C. equisetifolia shelterbelts in red soil and sandy soil regions, aiming to offer guidance for the establishment and functional improvement of coastal shelterbelt systems. Shelterbelts of C. equisetifolia at different growth stages were selected for analysis using ANOVA, Pearson correlation analysis, stepwise regression, and redundancy analysis (RDA) to elucidate the relationship between soil nutrients and the growth of C. equisetifolia stands. Stand age and soil type had significant effects on soil nutrient status, as total potassium (TK) and available potassium (AK) exhibited divergent trends depending on soil type. Soil nutrients were closely associated with growth indicators of C. equisetifolia, and phosphorus consistently exerted a significant influence on growth in both soil types. However, the effects of soil pH, total nitrogen (TN), TK, and organic matter (OM) varied by soil type. RDA results further indicated that soil nutrients explained 88.5

Unmanned aerial vehicle light detection and ranging (UAV–LiDAR) is a new method for collecting understory terrain data. The high estimation accuracy of understory terrain is crucial for accurate tree height measurement and forest resource surveys. The UAV–LiDAR flight altitude and forest canopy cover significantly impact the accuracy of understory terrain estimation. However, since no research examined their combined effects, we aimed to investigate this relationship. This will help optimize UAV–LiDAR flight parameters for understory terrain estimation and forest surveys across various canopy cover. This study analyzed the impacts of three flight altitudes and three canopy cover on the estimation accuracy of understory terrain. The results showed that when canopy cover exceeded a specific value, UAV–LiDAR flight altitudes significantly affected understory terrain estimation. Given a forest canopy cover, the reduction in ground point coverage increased significantly as the flight altitude increased; given a flight altitude, the higher the canopy cover, the more significant the reduction in ground point coverage. In forests with a canopy cover≥0.9, there were substantial differences in the accuracies of understory digital elevation models (DEMs) generated using UAV–LiDAR at different flight altitudes. For forests with a canopy cover <0.9, the mean absolute error (MAE) of understory DEMs from UAV–LiDAR at different flight altitudes was ≤ 0.17 m and the root mean square error (RMSE) was ≤ 0.24 m. However, for forests with a canopy cover≥0.9, the UAV–LiDAR flight altitude significantly affected the accuracy of understory DEMs. At the same flight altitude, the MAE and RMSE of the estimated elevation for forests with a canopy cover≥0.9 were approximately twice those of the estimated elevation for forests with a canopy cover <0.9. In forests with low canopy cover, it is possible to improve data collection efficiency by selecting a higher flight altitude. However, UAV–LiDAR flight altitudes significantly affected understory terrain estimation in forests with high canopy cover, it is essential to adopt terrain-following flight modes, reduce flight altitudes, and maintain a consistent flight altitude during long-term monitoring in high canopy cover forests.

Introduction:Chinese fir (Cunninghamia lanceolata) is the fastest-growing timber species in China. investigating its spatial structure and influence on aboveground biomass allocation is crucial for understanding its adaptability to environmental conditions, enhancing carbon sequestration, and maintaining forest ecosystem stability. Methods:In this study, airborne LiDAR technology was used to derive forest structural metrics, and weighted Voronoi diagrams were constructed to extract spatial configuration metrics. Biomass models for different components of Chinese fir were developed using 20 harvested trees, and stem mass fraction (SMF), branch mass fraction (BMF), and leaf mass fraction (FMF) were calculated. Path analysis quantified the effects of stand structure variables on biomass allocation among different organs. Results:The openness ratio (OP), angle competition index (UCI), forest layer index (S), and openness (K) were identified as the primary spatial structural factors influencing aboveground biomass allocation. Stem biomass accumulation is maximized when 0.75 < OP ≤ 1, 0 < UCI ≤ 0.25, 0 < S ≤ 0.25, and 0.4 < K ≤ 0.5, with SMF reaching its highest value. Branch biomass peaks when 0.5 < OP ≤ 0.75, 0 < UCI ≤ 0.25, 0.75 < S ≤ 1, and 0.4 < K ≤ 0.5, maximizing BMF. Leaf biomass is highest when 0 < OP ≤ 0.25, 0.5 < UCI ≤ 0.75, 0.5 < S ≤ 0.75, and 0.2 < K ≤ 0.3, leading to the maximum FMF. Discussion:The results of this study not only reveal the survival strategy of Chinese fir in environmental change, but also provide a theoretical basis for understanding ecosystem carbon sequestration and sustainable management of Chinese fir plantations.

Coastal shelterbelts serve as ecological safety barriers in coastal areas, and the effective evaluation of stand quality (SQ) is crucial for enhancing their development potential and supporting decision-making for sustainable management and environmental improvement. However, improving SQ and accurately monitoring coastal shelterbelts remains challenging due to factors such as limited soil resources, tree biological characteristics, and coastal storm surges. This study focuses on the typical coastal shelterbelt species Casuarina equisetifolia. Firstly, based on field survey data, using the Analytic Hierarchy Process, stand parameters such as mean tree height, diameter at breast height, crown width, canopy closure, and stand density were selected from both stand growth status and structure for the quantitative evaluation of SQ in Casuarina equisetifolia shelterbelts. Secondly, by combining drone-based multispectral and LiDAR point cloud data, the study explores the applicability of Individual Tree Integration and Stand Regression methods in the inversion of stand parameters for Casuarina equisetifolia shelterbelts. Ultimately, precise remote sensing monitoring of Casuarina equisetifolia SQ was achieved at the stand scale. The results of this study provide technical references and a theoretical basis for the efficient monitoring of SQ and management of coastal shelterbelts, which are significantly essential for their sustainable development.

Coastal areas face challenges in updating and enhancing the quality of coastal shelterbelt forests due to limited soil resource utilization, the biological characteristics of tree species, and the impact of coastal storm surges. This study, based on clarifying the connotation of stand quality (SQ) for Casuarina equisetifolia (C. equisetifolia), a typical coastal shelterbelt species, integrates stand growth conditions and structure, applying the AHP-EWM method to construct an SQ evaluation model. The model identifies key factors influencing the quality of C. equisetifolia stands and explores the mechanisms driving their growth processes. The results indicate that mean tree height, mean diameter at breast height, and stand density are key indicators for assessing the SQ of C. equisetifolia stands. The SQ of C. equisetifolia coastal shelterbelt stands varies across different land-sea positions, mainly influenced by stand age and soil nutrient levels. As stand age increases, the SQ initially improves and then declines. Additionally, C. equisetifolia trees growing in nutrient-rich soils exhibit better growth and higher SQ than those in poor soils. The findings of this study provide a theoretical foundation for the management and quality enhancement of coastal shelterbelt forests.

The analysis of visibility in urban parks is an essential component of landscape spatial analysis, and it holds significant importance for human well-being, health, and the transition to sustainable urban development. LiDAR point clouds offer highly detailed and accurate depictions of the urban park environment, and the calculation of visual volume can effectively quantify the visual perception indicators of urban parks. However, current methods often overlook the sparsity of ground point clouds, leading to inaccuracies in visual volume calculations. In light of this, we propose a theory of "boundary-ground-air" integration based on the "point cloud-ray-polyhedron" method to characterize the three-dimensional visibility of urban parks. The visual volume is divided into two major parts: ground and sky. Our method optimizes the calculation of visual volume for the ground part by supplementing missing point clouds based on ground continuity to enhance the accuracy of visual volume calculations. The method involves 5 key steps: identifying the boundary between ground points and non-ground points, voxelization of point clouds, calculation of aerial visual volume, calculation of ground visual volume, and volume index calculation. This method not only enables the calculation of three-dimensional visual space at any viewpoint in different locations within the landscape but also addresses the issue of accuracy deviation in visual volume calculations caused by the sparsity of ground point clouds. Using Chating Park in Fuzhou, China as a case study, the results demonstrate that our proposed method can accurately simulate the visibility measurement of urban parks at a resolution of 1m x 1m. This research achievement can provide technical support for landscape architecture planning and smart city development.

Aim The Pilot Area for Subtropical Ecological Civilization in Southern China has made significant strides in ecological conservation through ecological projects and forest rights reform. This study assesses FVC dynamics to evaluate the relative impacts of climate change and human activities on vegetation restoration, aiming to inform optimised management strategies. Location Fujian Province, China. Time Period 2000-2023.Major Taxa StudiesFractional vegetation cover (FVC). Methods We utilised the pixel dichotomy method to derive FVC from MODIS13A2 data (2000-2023) within the Google Earth Engine platform. To evaluate the impacts of climate change and human activities on vegetation restoration, we applied slope trend analysis in conjunction with the Mann-Kendall mutation test. Results (1) From 2000 to 2023, Vegetation Restoration in Fujian Province exhibited pronounced spatial heterogeneity. Approximately 72.65% of the region exhibited an increasing trend in FVC, with over 80% of the study area maintaining moderate to high levels of vegetation cover. In contrast, the southeastern coastal areas showed slower gains. (2) Approximately 69.55% of the changes in vegetation cover were attributed to the combined influence of human activities and climate change, with human activities contributing more significantly to vegetation restoration than climate change (67.88% vs. 64.14%). (3) Within the 40%-100% contribution range, the proportion of areas where human activities predominantly influenced changes in FVC was higher than that influenced by climate change (69.89% vs. 51.12%). (4) Although the total area of forests, shrublands and grasslands in Fujian Province declined during this period, the overall increase in FVC underscores the effectiveness of ecological restoration programs such as the Grain for Green Initiative. These findings indicate that even under substantial human disturbances, well-targeted and effectively implemented ecological policies can act as key drivers of vegetation recovery. Main Conclusions This study highlights that even under intense human disturbance, well-targeted and robust ecological policies remain the primary driving force behind vegetation recovery in subtropical ecological civilisation pilot zones. It underscores the importance of integrating climate adaptation strategies with human interventions to achieve effective ecological management, offering valuable insights and replicable pathways for vegetation restoration in other ecologically sensitive regions.

Several studies have proposed strategies to alleviate the urban heat island (UHI) amidst the challenges of global warming and rapid urbanization. Though, few have explored multi-network synergies of heat and cold islands and simulation of their effectiveness. This study constructed an urban thermal environment (UTE) network by 2D and 3D urban structural parameters from a perspective of connectivity. Cooling measures are proposed to mitigate the UHI by combining forward and reverse thinking. The thermal environment is simulated using the patch-generating land use simulation (PLUS) model to assess the effectiveness of the cooling network. Firstly, morphological spatial pattern analysis (MSPA) and connectivity analysis are used to identify urban cold and heat island sources. Then, thermal resistance is constructed by 2D and 3D structural parameters. A multi-level thermal environment network is generated through the minimum cumulative resistance (MCR) model and circuit theory. Subsequently, key nodes within this network are identified. Finally, the PLUS model simulates and compares the thermal environment before and after the implemented of cooling network. Taking Fuzhou City as an instance, we find that: (1) 43 cold island sources, 28 heat island sources, 196 cooling corridors, 64 heat island corridors, and 711 thermal environment spatial network pinch-points are identified. (2) The PLUS model demonstrates an overall accuracy of 78% in simulating the 2023 thermal environment, affirming the feasibility of UTE simulation. (3) After optimization, the growth rate of the low-temperature zone increased by 60.63% while the growth rate of the high-temperature zone decreased by 8.63%, indicating the effectiveness of the proposed multi-level cooling network in mitigating the UHI. The approach presented in this research provides new insights for sustainable urban development and climate adaptation planning, and is crucial for addressing the UHI.

Soil respiration (SR) is a vital indicator of soil quality. With global warming and soil fertility degradation, understanding the evolution of fertilization and SR along with their response to climate change is crucial. This study used specially formulated fertilizers (N : P : K 17 : 8 : 5) in Phyllostachys edulis plantations in Yongan city, Southern China, to explore changes in SR and its temperature sensitivity (Q10) under five different fertilization (0, 15, 30, 50, and 75 kg ha–1) covering all seasons. Within a certain range of fertilizer application, both SR and Q10 increased. However, when fertilizer application exceeded a threshold, both SR and Q10 decreased instead. At the seasonal scale, both SR and Q10 exhibited a seasonal pattern, with higher values during summer and lower values during winter. Furthermore, SR in spring consistently higher than that in fall. The primary driving factors of SR varied among the seasons, with soil temperature being the dominant factor in spring, while C/N was predominant during summer, fall, and winter. However, soil pH was the primary driving factor of Q10 in all seasons. These findings provide valuable insights for optimizing sustainable forest management practices to improve soil health.

As the main carbon reservoir of terrestrial ecosystems, small changes in forest ecosystems can lead to the destabilization of carbon stock, affecting the global carbon cycle and the climate, thus triggering ecological risks. Scientific assessment of forest carbon stock ecological risk and optimization of the security pattern of forest ecosystem carbon stock is an urgent need for scientific ecological environmental protection. In order to accurately assess forest carbon stock ecological risk, this study analyzes multiple sources of risk, synthesizes landscape patterns and ecological processes, and constructs a three-dimensional framework of the ecological adaptive cycle of "potential - connectedness - resilience" to assess forest carbon stock ecological risk. The ecological risk of forest carbon stocks in 2000, 2010 and 2020 was partitioned into two regions in collaboration with carbon stocks and ecological risk, which facilitated targeted regulation. The research results indicate that from 2000 to 2020, the forest carbon stock in the Minjiang River Basin showed a trend of initial increase followed by a decrease, with a reduction of 18.25 x 106 t over the 20-year period, The spatial evolution of high-risk areas for forest carbon stock ecological risk mainly extended from the central region to the southwest, with risk transfer primarily occurring between adjacent ecological risk levels. Forest carbon stock ecological risk is negatively correlated with carbon stock overall, with the spatial distribution characterized by a pattern of high carbon stock-low ecological risk and low carbon stock-high ecological risk clustering. The ecological risk assessment of forest carbon stocks is of great significance in enhancing regional carbon sink capacity, promoting rational land utilization and green low-carbon development.