Identification and Restoration of Forest Degradation Areas in Shaanxi Province Based on the LandTrendr Algorithm

Abstract

As an important ecological barrier in Northwest China, the health of forest ecosystems in Shaanxi Province is crucial to regional ecological balance and sustainable development. However, forest degradation has become increasingly prominent in recent years due to both natural and anthropogenic pressures. This study aims to identify the spatio-temporal pattern of forest degradation in Shaanxi Province, construct an ecological network, and propose targeted restoration strategies. To this end, we first built a structural-functional forest degradation (SFD) assessment system and used the Landsat-based detection of trends in disturbance and recovery (LandTrendr) algorithm to identify degraded areas and types; subsequently, we used morphological spatial pattern analysis (MSPA) and the minimum cumulative resistance (MCR) model to construct a forest ecological network and identify key restoration nodes. Finally, we proposed a differentiated restoration strategy for near-natural forests based on the Miyawaki method as a conceptual framework to guide future ecological recovery efforts. The results showed that (1) in 1991–2020, the total area of forest degradation in Shaanxi Province was 1010.89 km², which was dominated by functional degradation (98%) and structural degradation (87.15%), with significant regional differences; (2) the constructed ecological network contained 189 ecological source sites, 189 ecological corridors, 89 key nodes, and 50 urgently restored; and (3) specific restoration measures were proposed for different degradation conditions (e.g., density regulation and forest window construction for functional light degradation and maintenance of the status quo or full reconstruction for structural heavy degradation). This study provides key data and systematic methods for the accurate monitoring of forest degradation, the optimization of ecological networks, and scientific restoration in Shaanxi Province, which holds great practical significance for establishing a robust ecological barrier in Northwest China.

1. Introduction

Forests play a crucial role in providing forest products, maintaining species diversity, regulating the global climate, and balancing carbon sequestration [1]. However, the global forest degradation problem is becoming increasingly serious, threatening the ecology and sustainable human development. As an important ecological barrier in Northwest China, forest degradation in Shaanxi Province is a particularly prominent issue that significantly impacts the local ecological environment and economic development [2]. Therefore, it is crucial to develop effective forest degradation detection methods and restoration strategies. In response to global forest degradation, the United Nations Framework Convention on Climate Change (UNFCCC) has adopted the Reducing emissions from deforestation and forest degradation in developing countries (REDD+) program, which aims to promote the conservation and sustainable management of forests in developing countries [3]. As one of the world’s largest developing countries, China has implemented several large-scale ecological engineering measures, including returning farmland to forests and grasslands, afforesting mountainous regions, and conserving natural forest ecosystems [4,5].

Traditional forestry is based on forest resource surveys to evaluate forest degradation, which is time-consuming and costly [6]. Satellite remote sensing data are accurate and efficient and can reflect the spatial and temporal changes of forests at different scales [7]. Previously, vegetation cover was used as a baseline to measure degradation [8,9]. However, this method could not identify degradation in cases where the land cover type remained unchanged [10,11]. Subsequently, based on various remote sensing sources—including optical satellite imagery (e.g., Landsat data with 30-m resolution) [12], synthetic aperture radar (SAR), and light detection and ranging (LiDAR)—three main approaches have been developed for forest degradation assessment: (1) comparative analysis of spectral indices [13,14], (2) trajectory-based time series analysis [15,16,17], and (3) structural metric extraction from 3D data [18]. Types of detection methods include biomass, which is commonly used by scholars as an indicator of forest degradation [19], but this method involves a large amount of data and workload, making it difficult to be popularized and applied on a large scale in the region.

Forest degradation leads to the weakening of the connection between ecological patches, and the construction of a forest ecological network can promote the migration of species and energy flow in the region while maintaining the integrity and continuity of ecological habitats [20]. The study of ecological networks has gradually shifted from reducing species extinction to repairing damaged habitats and species homes, forming the “ecological substrate analysis—ecological source extraction—corridor node extraction—ecological network construction and optimization” construction model. The model of “ecological substrate analysis—ecological source extraction—corridor node extraction—ecological network construction and optimization” was formed [21]. In the process of ecological network construction, various approaches are used at different stages of ecological network construction. In studies identifying ecological source sites, scholars have widely used strategies such as existing ecological reserves, ecosystem service provisioning capacity [22], and landscape connectivity [23] to identify these sites. Some scholars have attempted to use morphological spatial pattern analysis (MSPA) to identify critical ecological habitats to improve the science of ecological source site selection. In the construction of ecological resistance surfaces, researchers often construct resistance surfaces by directly assigning values according to the current land type or constructing an evaluation system by selecting the corresponding indexes [24], which features a simple data structure that enables efficient computation [25]. The ecological corridor is constructed by using the minimum cumulative resistance (MCR) model [21] and the circuit theory model [26]. The circuit theory model simulates species migration and energy flow by modeling the travel of electric currents but requires high accuracy of land use type data [27]. In contrast, the minimum resistance model is based on the ecological factor resistance surface to obtain the minimum resistance value between the ecological source and destination to obtain the shortest migration path [28]. Constructing a forest ecological network based on the MSPA model in combination with the MCR model and targeting the restoration of degraded forest nodes throughout the ecological network can achieve the benefits of region-wide degraded forest restoration more quickly [21]. This integrated approach of utilizing different models provides an effective pathway for ecological network research in China.

Research on the restoration of degraded forests has gradually evolved from the initial vegetation management to comprehensive research on ecosystem restoration and rehabilitation [29,30]. China mainly adopts the methods of forest gap regulation and regeneration, sealing and restructuring [31,32], vegetation reconstruction [33], and ecological restoration of near-naturalization [34], etc. Among these methods, the Miyawaki reforestation method [30,35] is considered one of the most important. Among them, the Miyawaki method [36] offers an innovative and sustainable solution for forest ecosystem restoration by simulating the natural succession process.

In this study, we developed an integrated assessment framework for forest degradation, distinguishing between structural and functional degradation (SFD). Functional and structural degradation were identified and quantified using the Landsat-based LandTrendr algorithm. Based on the results, we constructed a forest ecological network using the MSPA–MCR model to identify key degraded nodes in Shaanxi Province that require ecological restoration. Finally, we proposed regionally targeted near-natural restoration strategies based on the Miyawaki method. Using the LandTrendr algorithm, the forest ecological network was constructed with the MSPA-MCR model, and the degraded forest nodes in Shaanxi Province in urgent need of restoration were identified at the regional scale. Utilizing the Miyawaki method, a restoration strategy of near-natural forests was proposed, and the differentiated programs were formulated for the different degradation levels, which improved the relevance and effectiveness of restoration efforts.

2. Materials and Methods

2.1. Study Area

Shaanxi Province (105°29′–111°15′ E, 31°42′–39°35′ N) is located in northwestern China, with a total area of 205,600 km² and a distance of about 500 km from east to west and 800 km from north to south. It has a total area of 205,600 km², with an east-west distance of about 500 km and a north-south distance of about 800 km and is bounded by the Qinling Mountains in the south and north of Shaanxi Province, straddling the two major water systems of the Yellow River and the Yangtze River. The terrain is high in the north and south, with a basin in the center and favorable natural conditions. According to the terrain gradient, the province can be divided into three major natural regions—the Loess Plateau region in northern Shaanxi, the Guanzhong Plain region, and the mountainous region in southern Shaanxi. Along the Great Wall in northern Shaanxi is a mesothermal semi-arid climate, the Guanzhong area and most of northern Shaanxi has a warm temperate semi-humid climate, while southern Shaanxi has a northern subtropical humid climate. The terrain is complex, comprising plateaus, mountains, hills, plains, and basins. The complex combination of topography and climatic gradients shapes the diverse forest ecosystems.

As of 2023, Shaanxi Province has a forested area of 140,370 km², divided into the Qinling Forest (Bashan) and the Loess Plateau Forest (Guanshan, Ziwuling, Huanglongshan) [37]. Zonal vegetation shows latitudinal zonal differentiation: northern Shaanxi is dominated by warm temperate deciduous broad-leaved forests (with Liaodong oak (Quercus wutaishanica Blume) and poplar (Populus spp.) as the dominant species), Guanzhong is transitioned to sidecarp forests and artificial acacia (Robinia pseudoacacia) forests, and southern Shaanxi develops northern subtropical evergreen-deciduous broad-leaved mixed forests (with cork oak (Quercus variabilis Blume), sharp-toothed quercus oak, and oil pine as the dominant species). The forest area is large, dominated by pure forests, with complex and diverse ecosystems and high vulnerability. As a result of over-exploitation and climate change, Shaanxi Province has experienced a gradual increase in the area of forest degradation over the past two decades (Figure 1) and faces increasing ecological stress [38].

2.2. Key Data

Five types of data, including land cover type, remote sensing imagery, elevation, roads, and normalized difference vegetation index (NDVI), were selected for this study (

Table 1. Data sources.

Data Types	Data Sources	Resolution (m)	Collection Time
Land use	Google Earth Engine (https://developers.google.cn/ accessed on 1 October 2022)	30	1990–2021
Remote sensing image	Google Earth Engine (https://developers.google.cn/ accessed on 1 October 2022)	30	1991–2020
Elevation	National Aeronautics and Space Administration (NASA) (https://www.nasa.gov accessed on 1 May 2023)	30	2020
Road	OpenStreetMap (OSM) (https://wiki.openstreetmap.org/wiki/Use_OpenStreetMap accessed on 1 May 2023)	1000	2020
Normalized difference vegetation index (NDVI)	Moderate Resolution Imaging Spectroradiometer (MODIS)	250	2020

).

2.3. Research Methodology

2.3.1. Identification and Quantification of Forest Degradation

Forest degradation is a reverse succession process in which natural or human-induced disturbances lead to an imbalance in the structure of forests and a reduction in their service functions [39,40,41]. According to the conditions and processes, it can be divided into Structural Forest Degradation and Functional Forest Degradation, with the former referring to the process of conversion of forests to non-forests, with obvious changes in the original structure, and the latter referring to the loss of forest ecosystem functions in the case of retaining it as forest land. Functional Forest Degradation. This classification method makes up for the single perspective of the traditional method that relies only on land cover change and incorporates hidden degradation, such as forest stand quality decline, into the assessment system, thereby improving the comprehensiveness of degradation identification. To detect forest degradation, we used the LandTrendr algorithm on the Google Earth Engine platform, applying NBR (normalized burn ratio) as the primary change indicator. We used the Landsat Collection 1 Surface Reflectance (SR) data, which has undergone standardized radiometric calibration and atmospheric correction to ensure consistency across TM, ETM+, and OLI sensors. This cross-sensor harmonization helps reduce systematic biases in long-term change detection. Degradation thresholds were determined based on a combination of previous studies and visual interpretation of high-resolution remote sensing imagery. While this approach enables flexible adaptation to different forest types and landscape contexts, we acknowledge that visual interpretation may introduce confirmation bias, particularly when analysts are guided by pre-existing hypotheses or expected outcomes [42]. This subjectivity can affect the interpretation of NBR value changes and reduce the objectivity of the automated detection process. Future research should consider applying statistically grounded thresholding techniques, such as quantile analysis or clustering methods, to improve reproducibility and reduce bias [43,44,45].

Functional forest degradation identification begins with the preprocessing of remotely sensed data. Landsat 8 Operational Land Imager (OLI) data were converted to reflectance or radiance brightness temperature by applying calibration and correction conversion factors to minimize errors. The image quality assessment bands with “QA_PIXEL” are selected by the select function, and the C Function of Mask (CFmask) algorithm is used to mask the image clouds, cloud shadows, and snow to reduce the influence on the subsequent algorithms [46]. The forest distribution data of Shaanxi Province were obtained from the China Land Cover Dataset (CLDC), combined with Google Earth high-definition images and the data of the Second Land Survey of China, verified and supplemented the accuracy of the remote sensing data through field research. The data were analyzed and integrated to extract the forest coverage area of Shaanxi Province using ArcGIS 10.6.

In addition, information on forest degradation and restoration was extracted by identifying mutations in normalized burn ratio (NBR) values in the time series:

NBR = (NIR − SWIR)/(NIR + SWIR),

(1)

In Equation (1), NIR is a near-infrared light wave, and SWIR is a short-wave infrared light wave.

Finally, based on the previous research [47,48], we set the control parameters of the LandTrendr algorithm (

Table 2. LandTrendr parameter.

Parameters	Parameter Description	Value
Spectral index	Detection of vegetation growth	NBR
Max segment	Maximum number of divisions	6
Spike threshold	Maximum threshold	0.9
Vertex count overshoot	Number of nodes	3
Prevent one-year recovery	Recovery period greater than one year	Ture
Recovery threshold	Recovery threshold	0.25
p-value threshold	Vertex threshold, no change is considered if the value is not exceeded	0.056
Best model proportion	Optimal model scale. Continue to simplify if this value is exceeded	0.75
Min observations needed	Minimum number of cartographic observations	6

), including spike threshold removal, segmentation point identification, and model simplification (e.g., p-value threshold and best model proportion) to ensure robust trajectory fitting and reduce noise. In order to further mitigate the influence of potential artifacts caused by sensor changes, we also visually inspected the temporal profiles of NBR to distinguish between true ecological trends and sensor-induced discontinuities.

Structural forest degradation identification was first processed by splicing, cropping, and transforming projections of CLDC land cover products on the GEE platform. The projection used the World Geodetic System—1984 coordinate system (WGS1984) with a uniform resolution of 30 m to ensure the consistency and comparability of the data in space and time. Subsequently, the overlay analysis tool in ArcGIS10.X was used to transform and analyze the 30-period land cover maps to obtain the spatial distribution of each type of land cover in each period and to understand the evolution process of land use in different areas. Finally, the reclassification and raster calculation tools in ArcGIS10.X are used to generate land cover change maps and corresponding transfer matrices.

According to the natural law of forest succession, the degree of forest degradation corresponds to different stages of succession. In the segmentation results of the LandTrendr algorithm, the bands record the magnitude of changes in forest degradation in different regions. Referring to the existing studies and the visual interaction method using high-resolution remote sensing images in Google Earth, the NBR value range was used to calculate to determine the thresholds of different degradation degrees. The band was reclassified to obtain the degree of functional degradation (

Table 3. Classification of forest degradation levels.

Degradation Type	Degree of Degradation	Characteristic	Characteristic
Functional degradation	mildly	Forests change from dense woodlands to open forests	200 < magnitude ≤ 500
	moderately	Forests have changed from dense woodland to shrubland and much of the forest cover has been destroyed	500 > magnitude
Structural degradation	severe	Decrease in forest cover	Forest-scrub, forest-grassland, forest-bare ground, forest-farmland, forest-water, and forest-impervious surface

) [49,50].

2.3.2. Construction of the MSPA–MCR Ecological Network

Ecological network is the use of linear ecological corridors to effectively connect landscape patches with high fragmentation, ensuring the integrity, diversity, and richness of the landscape space. In this study, we constructed a forest ecological network in Shaanxi Province based on the coupled MSPA-MCR model framework. First, ecological source sites were identified based on the MSPA model. The 2020 land use type map of Shaanxi Province was saved as binary raster data in Tag Image File Format (TIFF) format at a resolution of 30 × 30 m. Forest land was set as foreground data and assigned a value of 1 for classification; other land use types except forest land were set as background and assigned a value of 0 as a supplement to the foreground and were not involved in classification. The classified raster data were transformed into an 8-bit raster format and imported into GuidosToolbox to analyze its spatial morphology pattern and to create the forest landscape pattern classification. Based on the LandTrendr algorithm, when exporting the segmentation segments, the “gain” segmentation segment was selected. Other spatial analysis tools and mapping of ArcGIS were utilized to draw the spatial and temporal distribution map of forest restoration in Shaanxi Province (Figure 2). In this study, the forest restoration area in the core area was selected, and due to the large scale of the study area, regions larger than 10 km² were designated as an ecological source areas for further research and protection. The 10 km² minimum area threshold for ecological sources was selected after testing several scale options and considering landscape-level continuity at the provincial scale.

Second, the ecological corridor is constructed based on the MCR model, and the least-cost pathway from the ecological source to the target site is constructed by weighing the environmental factors and calculating the minimum cumulative resistance faced by the species in the movement process [51].

MCR = fmin∑i = mDij × Rij

(2)

In Equation (2), MCR is the value of minimum cumulative resistance; D ij is the spatial distance of a species from source j to landscape unit i; R is the resistance coefficient of landscape unit i to the movement of a species; and f denotes the positive correlation between the minimum cumulative resistance and ecological processes.

The ecological resistance surface is the degree of resistance of the ground surface to the flow and diffusion of species and energy. The selected resistance factors included land use type, elevation, slope, distance, and NDVI [52], with the weights of each resistance factor calculated using hierarchical analysis to be 0.25, 0.14, 0.13, 0.11, and 0.37 [53]. The maximum-minimum standardization method was used to calculate the resistance values. High resistance values indicate that species are more likely to be impeded in migration, while low resistance values indicate that the area is more suitable. For discrete data land use types, the assignment method was used, assigning forest land as 1, cropland as 2, grassland as 2, watershed as 3, impervious surfaces as well as bare land as 5. The ArcGIS raster calculator was utilized to generate the composite resistance surfaces by selecting the raster cells of 30 m*30 m to obtain the cost data for the minimum resistance value model.

Finally, the ecological restoration nodes for forest degradation are selected to connect the source areas to sustain the ecological network [54]. According to the ecological source and resistance surface, the cost connectivity tool in ArcGIS was used to generate ecological corridors, and the intersection of the two was extracted as the degraded forest ecological node in combination with the forest degradation area.

2.3.3. Restoration Strategy Based on the Miyawaki Method

Potential Natural Vegetation (PNV) refers to the stable top communities formed through primary succession under the condition of no human disturbance. Neo-succession Theory (NST) emphasizes the accelerated succession mechanism of “artificial initiation-natural completion”, which utilizes the genetic memory of the soil seed bank and the self-organizing ability of the community to shorten the restoration period to 1/3–1/5 of the traditional method. The Miyawaki method is based on field research, the selection of community-building species, the combination of anthropogenic and forest self-restoration, and the establishment of forest communities (Figure 3).

Firstly, the vegetation, such as the dominant species present in the area, is determined through field visits, vegetation surveys, and ecological surveys, while the type of natural vegetation to be retained is established by combining methods such as literature review and expert consultation. Secondly, the types of restoration areas are divided according to the dominant species of plant communities and habitat types, and the vegetation for forest restoration is determined through methods such as expert consultation and literature review, and the existing tree species and dominant species are determined based on the investigation and analysis of natural vegetation types. In addition, density regulation is carried out according to the type of forest stand, different forestation objectives, and the use of quantitative indicators within the stand to maintain the balance of the forest ecosystem, avoid excessive competition for species resources, and ensure that the forest stand can properly perform its ecological functions. Moreover, mixed planting and a reasonable vertical structure of trees, shrubs, and grasses are maintained to ensure that there are 2–3 seedlings per square meter. Finally, post-management is carried out to promote natural regeneration, with basic management and conservation tasks such as weed pulling, irrigation, and fertilization carried out within 1–3 years after saplings are planted. After 15–50 years, the forests evolve into mature forests similar to natural forests, which have a certain degree of natural resilience and require minimal human intervention. However, while the method shows potential for rapid canopy development, its universal application in ecologically complex regions such as Shaanxi requires critical evaluation. High-density planting may lead to excessive competition, unsustainable maintenance burden, or disrupted natural succession, particularly in large forest areas or nutrient-limited zones. Moreover, the assumption that Miyawaki forests reach natural forest maturity in 15–50 years lacks empirical support under Shaanxi’s climatic and topographic conditions, such as steep slopes, seasonal drought, and shallow soils. Therefore, the Miyawaki method should be seen as one of multiple tools. It is more appropriate for highly degraded patches with low regeneration potential, urban greenbelts, and community engagement or educational projects. In most natural forest restoration scenarios in Shaanxi, more flexible and ecologically adaptive methods should be considered.

3. Results

3.1. Analysis of Forest Degradation Patterns in Shaanxi Province

3.1.1. Precision Analysis

To evaluate the accuracy of the forest degradation classification results, we conducted an independent validation using 1000 randomly sampled points across Shaanxi Province. These validation samples were derived from the UMD/Hansen Global Forest Change v1.8 dataset, a widely recognized global reference for forest loss mapping, and were used as ground truth data. Each point was labeled as either “forest loss” (degraded) or “no forest loss” (non-degraded) and compared with the corresponding classification outcome from our model.

The resulting confusion matrix and accuracy metrics are presented in

Table 4. Confusion matrix for accuracy evaluation of forest degradation classification.

Type	Degradation	Nondegenerate	User’s Accuracy	Producer’s Accuracy	Overall Accuracy	Kapper
Degradation	247	16	78.4%	93.9%	91.6%	0.797
Nondegenerate	68	669	97.7%	90.8%

. The classification achieved an overall accuracy of 91.6% and a Kappa coefficient of 0.797, indicating substantial agreement between the model predictions and the reference dataset. The producer’s accuracy for degraded areas was 93.9%, while the user’s accuracy was 78.4%.

3.1.2. Analysis of Functional Forest Degradation Patterns in Shaanxi Province

During 1991–2020, functional forest degradation in Shaanxi Province totaled 992.76 km², with an annual average of 33.09 km². The degradation was predominantly concentrated in northern Shaanxi. The peak years were 1991 (241.78 km²) and 2020 (154.80 km²), while the lowest degradation occurred in 2003 (3.04 km²). Curve smoothing analysis revealed three major periods of declining degradation: 1991–1993, 1995–2000, and 2013–2015. In contrast, degradation increased during 1993–1995, 2009–2013, and 2015–2020. The period from 2000 to 2009 showed relatively stable conditions (Figure 4). Overall, the temporal pattern indicates alternating phases of forest stress and recovery, with a recent upward trend in degradation intensity.

Among all cities in Shaanxi Province (Figure 5), Yan’an exhibited the largest area of functional forest degradation (340.70 km²) and the highest degradation rate (3.71%). Ankang also experienced considerable degradation (143.06 km²), while Yulin recorded both the smallest degraded area (0.14 km²) and the lowest rate (0.18%). Relatively high degradation rates were also observed in Xianyang and Tongchuan (both 2.16%).

3.1.3. Analysis of Structural Forest Degradation Patterns in Shaanxi Province

In 2020, the structural forest degradation area in Shaanxi Province, relative to 1991, was 18.13 km², accounting for 0.02% of the total forest area (Figure 6). Land cover transitions indicate that forest-to-farmland conversion contributed the most (16.11 km²), followed by conversion to grassland (1.35 km²), with a minimal shift to bare land (0.0003 km²) (

Table 5. Land cover transfer matrix for Shaanxi Province, 1991–2020 (unit: km²).

1991 2020	Impervious Surface	Grassland	Shrub	Bare Land	Farmland	Forest	Water	Total
Impervious surface	16.8	0.004	-	0.001	0.19	0.0005	-	0.67
Grassland	0.66	410.85	0.9	3.03	55.84	65.62	0.001	0.55
Shrub	0.001	1.03	0.9	0.001	0.77	13.93	-	-
Bare land	0.62	0.11	32.09	1.12	2.39	0.003	-	-
Farmland	31.77	105.07	0.21	0.09	428.59	70.77	-	1.51
Forest	0.11	1.35	0.54	0.0003	16.11	784.57	-	0.012
Water	0.01	0.12	-	1.01	1.30	0.04	-	4.04
Total	49.97	518.53	34.64	5.24	505.19	934.93	0.001	6.78

).

From 1991 to 2020, structural forest degradation in Shaanxi Province was highest in Hanzhong City (5.03 km²), followed by Ankang (4.79 km²), and lowest in Yulin (0.02 km²). Forest degradation was highest in Ankang (0.05%), followed by Tongchuan (0.035%), and lowest in Xi’an (0.004%) (Figure 7).

3.1.4. Degree of Forest Degradation Scales in Shaanxi Province

From 1991 to 2020, forest degradation in Shaanxi Province covered an area of 1010.89 km² (Figure 8 and Table 4), with functional degradation (992.76 km²) being dominant, while structural degradation (18.13 km²) was less prevalent. Additionally, light degradation was dominant (87.15%), followed by moderate degradation (9.23%), and intense degradation accounted for the smallest proportion (3.64%). Spatially, Yan’an City experienced the highest concentration of forest degradation, and the overall degradation rate in southern Shaanxi was lower compared to that in Guanzhong and northern Shaanxi.

3.2. Shaanxi Province Forest Ecological Restoration Network Construction

3.2.1. Spatial Distribution of Ecological Source Sites

According to the landscape pattern analysis of Shaanxi Province (Figure 9a), seven landscape categories were identified, including core area (1387.64 km²), bridging area (6394.58 km²), fringe area (52.25 km²), pore (29.37 km²), branch (77.49 km²), isolated island (492.15 km²), and loop (76.18 km²). Core areas are mainly distributed in Baoji, Xi’an, Xianyang, Hanzhong, Ankang, Shangluo, and Yan’an. Among these, patches in Hanzhong, Ankang, and Baoji are relatively large and structurally stable, whereas those in Xi’an, Xianyang, and Yan’an are more fragmented and poorly connected, indicating the need for ecological linkage enhancement.

Isolated island patches are predominantly found in Yan’an and Xianyang and may serve as temporary ecological habitats.

Core patches exceeding 10 km² were designated as ecological source sites (Figure 9b), resulting in the identification of 189 such sites, totaling 13,722 km². These sources are densely concentrated along the Qinling Mountains, forming the most continuous ecological network in the province. In contrast, ecological source sites in Weinan and Ankang are more scattered and spatially isolated. No source patches over 10 km² were found in Yulin and Tongchuan, indicating limited ecological connectivity in these areas.

3.2.2. Spatial Distribution of Ecological Corridors

The resistance values in southern Shaanxi were relatively low, while those in Yulin and Xi’an were high, and those in other prefectural cities were the next highest. By analyzing the minimum resistance surfaces, 189 potential ecological corridors were identified (Figure 10), with corridor widths of 1 km. Ecological corridors in southern Shaanxi are dense and more suitable for species migration and material-energy exchanges, whereas ecological corridors in Guanzhong and northern Shaanxi have poor connectivity and more pronounced faulting phenomena.

3.2.3. Extraction of Ecological Nodes

Using the forest ecological corridor as the basic data and combining the forest degradation area, 89 ecological nodes were identified by extracting and analyzing the intersection of these two (Figure 11), which were mainly distributed in Hanzhong City, Ankang City, and Baoji City. Ecological network restoration starts with protecting and repairing ecological nodes, enhancing the stability and resilience of the entire ecosystem, and promoting species migration and flow.

3.2.4. Forest Ecological Network Construction and Optimization

A total of 189 important ecological sources, 189 ecological corridors, and 89 important nodes were extracted for the construction of the forest ecosystem network in Shaanxi Province. According to the ecological obstacle points and forest degradation areas, 50 small forestry classes in urgent need of restoration were selected at the forest stand level (Figure 12) to improve the ecological network of Shaanxi Province through systematic restoration, provide scientific support for the planning and implementation of ecological security, and promote the healthy development and sustainable utilization of forest ecosystems.

3.3. Rehabilitation of Forest Ecological Patterns in Shaanxi Province

3.3.1. Potential Vegetation Survey

Collate the area, degradation degree and type, dominant vegetation, vegetation subtypes, and administrative districts of the forest compartments in which each ecological restoration area is located (

Table A1. Statistics table of potential vegetation in ecological restoration areas.

Small Class Number	Area (km2)	Type of Forest Degradation	Dominant Vegetation	Vegetation Subtype	Administrative District
23	318.40	Functional degradation: mild	Huashan pine (Pinus armandii Franch.), hemlock (Tsuga chinensis (Franch.) Pritz.), red birch (Betula albo-sinensis Burk.), and light-barked birch (Betula luminifera H. Winkl.) forests	Subtropical coniferous forest	Ankang
38	191.42	Functional degradation: mild
18	17.29	Structural degradation: severe	Oriental white oak (Quercus aliena Blume var. acutiserrata Maximowicz ex Wenzig)	Temperate deciduous broad-leaved forest
40	54.56	Functional degradation: mild
48	196.29	Functional degradation: mild
37	45.58	Functional degradation: moderate	Mixed cork oak and broadleaf evergreen forests	Subtropical mixed evergreen and deciduous broad-leaved forests
39	34.95	Functional degradation: mild
15	21.97	Functional degradation: mild	Cork oak	Temperate deciduous broad-leaved forest
47	90.93	Functional degradation: mild
21	27.37	Functional degradation: mild	Robinia pseudoacacia (Robinia pseudoacacia Linn.) forest	Temperate deciduous broad-leaved forest	Baoji
24	41.08	Functional degradation: moderate
1	361.01	Functional degradation: mild	Oriental white oak
20	80.10	Functional degradation: mild
26	69.66	Functional degradation: moderate
28	21.88	Functional degradation: mild
29	33.02	Functional degradation: mild
19	50.62	Functional degradation: mild	Cork oak
25	196.03	Functional degradation: mild
16	68.67	Functional degradation: mild	Cork oak	Temperate deciduous broad-leaved forest	Hanzhong
33	46.31	Functional degradation: mild	Huashan pine, hemlock, red birch, and light-barked birch forests	Subtropical coniferous forest
30	678.70	Functional degradation: mild	Oriental white oak	Temperate deciduous broad-leaved forest
31	406.73	Functional degradation: mild
32	19.11	Functional degradation: mild	Cork oak and sawtooth oak (Quercus acutissima Carr.) forests	Subtropical deciduous broad-leaved forest
17	44.61	Functional degradation: mild	Cork oak	Temperate deciduous broad-leaved forest
34	361.27	Functional degradation: moderate
41	350.29	Functional degradation: moderate
43	143.50	Functional degradation: mild
44	17.05	Functional degradation: mild
35	53.87	Functional degradation: mild	Oriental white oak	Temperate deciduous broad-leaved forest	Shangluo
42	41.91	Functional degradation: mild	Sawtooth oak forest
36	60.40	Functional degradation: mild	Oriental white oak
45	79.36	Functional degradation: mild
46	56.48	Functional degradation: mild
27	458.30	Functional degradation: mild	Cork oak		Xi’an
22	196.92	Functional degradation: moderate	Birch forest
0	31.82	Functional degradation: mild	Robinia pseudoacacia forest		Xianyang
13	689.23	Functional degradation: moderate	Liaodong oak forest
4	14.19	Structural degradation: severe	Lateral Berlin (Platycladus orientalis (Linn.) Franco)	Temperate coniferous forest	Yan’an
7	63.67	Functional degradation: mild	Robinia pseudoacacia forest	Temperate deciduous broad-leaved forest
14	27.51	Functional degradation: moderate
2	236.14	Functional degradation: moderate	Liaodong oak forest
5	103.63	Functional degradation: mild
8	221.15	Functional degradation: moderate
10	54.89	Functional degradation: mild
3	79.25	Functional degradation: mild	Aspen grove (Populus davidiana Dode)	Temperate deciduous broad-leaved forest
6	74.07	Functional degradation: moderate
9	22.99	Functional degradation: mild
11	36.66	Functional degradation: mild	Pinus sylvestris (Pinus tabulaeformis Carr.)	Temperate coniferous forest
12	165.70	Functional degradation: moderate

). The potential vegetation survey provides a scientific basis for ecological restoration by analyzing the natural vegetation types, habitat conditions, and dominant tree species in the area; formulating restoration plans that are compatible with local ecosystems; ensuring that the restoration measures align with the laws of natural succession; and improving restoration efficiency and ecological benefits. Within these regions, since some forests have similar habitats, similar forest types, and the same dominant tree species, similar restoration programs can be shared to improve the benefits of restoration.

3.3.2. Degraded Forest Restoration Programming

Based on vegetation survey data, expert consultation, and existing studies [55,56,57], we identified 15 target forest types for ecological restoration across varying habitat conditions (

Table A2. Plant configuration program for ecological restoration areas.

Forest Class Number	Tree Species That Form Part of a Group		Pioneer Tree Species (Dominant Species)	Intermediate Stage	Advanced Stage
38	Huashan pine, hemlock, red birch, light birch	Tree layer	Cork oak	Green Pressure Maple (Acer davidii Franch.), Chinkin Elm (Carpinus cordata Bl.), Spiny Leaf Oak (Quercus spinosa David ex Franch.), and Lesser Veined Tilia (Tilia paucicostata Maxim.)
		Shrub layer	Mountain leech (Desmodium), Lonicera caprifolium (Lonicera hispida Pall.ex Roem.et Schult.), Lonicera japonica (Litsea pungens Hemsl.), and Emei’s rosemary (Rosa omeiensis)	Sarsaparilla (Smilax china Linn.), Magnolia multiflora (Indigofera amblyantha Craib), Rudbeckia (Abelia biflora Turcz.), and Sambucus nigra (Symplocos paniculata (Thunb.) Miq.)
		Herbaceous layer	Forsythia (Deyeuxia arundinacea (Linn.) Beauv.), Ogi (Triarrhena sacchariflora (Maxim.) Nakai), wild strawberries (Fragaria vesca Linn.), and Artemisia ossificans (Artemisia dubia Wall.ex-Bess.)	Rubia cordifolia (Rubia cordifolia Linn.), Epimedium (Epimedium brevicornu Maxim.), and Downy Matsumoto (Thalictrum aquilegifolium Linn.var.sibiricum Regel et Tiling), Ogi	Arctostaphylos (Arthraxon hispidus (Trin.) Makino) and East Asian pinnatifid ferns (Gymnocarpium oyamense (Bak.) Ching)
18, 40, 48, 30, 31	Oriental white oak (Quercus aliena)	Tree layer	European hornbeam (Carpinus turczaninowii Hance)	Huashan pine, jack elm, and Sizhao flower (Dendrobenthamia japonica (DC.) Fang var.chinensis (Osborn) Fang)
		Shrub layer	Hickory (Rubus corchorifolius Linn.f.) and Weeping Spear (Euonymus alatus (Thunb.) Sieb.)	Jack elm, wood ginger, grosbeak maple (Acer grosseri Pax), wild rose (Rosa multiflora Thunb.), and hairy cherry (Cerasus tomentosa (Thunb.) Wall.)	Hydrangea (Spiraea), Pipevine (Viburnum utile Hemsl.), and Corky-winged Weeping Spear (Euonymus phellomanes Loes.)
		Herbaceous layer	Bloodthirsty (Glechoma longituba (Nakai) Kupr.), longleaf tussock (Carex giraldiana Kukenth), and three-spike tussock (Carex tristachya Thunb.)	Rubia cordifolia, broadleaf tussock (Carex siderosticta Hance), and dogbane (Cucubalus baccifer Linn.)	Oatmeal (Ophiopogon japonicus (Linn.f.) Ker-Gawl.) and wild strawberries
37, 39, 15, 47, 19, 25, 32, 16, 17, 34, 41, 43, 44, 27	Cork oak	Tree layer	Albizia kalkora (Albizia kalkora (Roxb.) Prain) and Pistacia chinensis (Pistacia chinensis Bunge)	Mountain Pepper (Lindera glauca (Sieb.et Zucc.) Bl.) and Saltbush (Rhus chinensis Mill.)	Brachypodium distachyon (Quercus serrata Murray var.brevipetiolata (A.DC.) Nakai) and European hornbeam
		Shrub layer	Forsythia (Forsythia suspensa (Thunb.) Vahl)	Hanging hook (Rubus)	Shikotana, White Sandalwood
		Herbaceous layer	Wild Chrysanthemums (Chrysanthemum indicum L) and Fine Leaf Carex (Carex duriusata C.A.Mey.subsp.stenophylloides (V.Krecz.) S.Y.Liang et Y.C.Tang)		Carex macrocephalus (Carex lanceolata Boott), Fine Leaf Carex, and Astilbe chinensis (Astilbe chinensis (Maxim.) Franch.et Savat.)
21, 24, 0	False acacia	Tree layer	Masson pine (Pinus massoniana Lamb.)		Big-leaf beech (Zelkova schneideriana H.-M.) and Huashan pine
		Shrub layer	European hornbeam, three boughs of Uva ursi (Lindera obtusiloba Bl.), Rudolfia, thorns (Vitex negundo Linn.var.heterophylla (Franch.) Rehd.), and sea buckthorns (Hippophae rhamnoides Linn.)	Double Shield Wood (Dipelta floribunda Maxim.) and Wood Ginger
		Herbaceous layer	Pisum sativum (Elymus dahuricus Turcz.) and white lambsquarters (Bothriochloa ischcemum (Linn.) Keng)	Long-stalked Mountain locust (Podocarpium podocarpum (DC.) Yang et Huang) and stinkweed (Melica scabrosa Trin.)	Carex broadleaf sedge (Carex siderosticta Hance)
1, 20, 26, 28, 29	Oriental white oak (Quercus aliena)	Tree layer	Chinese red pine	Huashan pine, sharp-toothed oak, and hemlock	Birch
		Shrub layer	Honeysuckle (Lonicera japonica Thunb.) and Hanging hook	Bamboos (Phyllostachys heterocycla (Carr.) Mitford cv.Pubescens Mazel ex H.de leh.) and southern snakeroot (Celastrus orbiculatus Thunb.)	Magnolia vine (Schisandra chinensis (Turcz.) Baill.)
		Herbaceous layer	Tussock (Carex tristachyaThunb.), Cliff Palm (Carex siderosticta Hance), and Wire Ferns (Adiantum capillus-veneris Linn.)		Early morning glory (Poa annua Linn.)
33	Huashan pine, hemlock, red birch, light birch	Tree layer	Chinese red pine	Paintbrush (Toxicodendron vernicifluum(Stokes)F.A.Barkley), green maple, sharp-toothed oak, and Chinese linden (Tilia chinensis Maxim.)
		Shrub layer	Ginger, amber (Swida macrophylla (Wall.) Sojak), willow (Salix), and hazelnut (Corylus heterophylla Fisch.ex Trautv.)	Rose (Rosa multifolora Thunb), Hairy Cherry, and Pearl Plum (Sorbaria sorbifolia (Linn.) A.Br.)	Hydrangea (Neillia thrysiflora D.Don), Hanging hook, Hydrangea, and Hoodia (Lespedeza bicolor Turcz.)
		Herbaceous layer	Tussock grass, bromeliad (Paris), fescue (Festuca ovina Linn.), and early morning glory	Downy Pineweed, Deerfoot (Pyrola calliantha H.Andr.) and Mountain Edelweiss (Oxalis acetosella Linn.subsp.griffithii (Edgew.Et Hook.f.)Hara)	Artemisia capillaris (Artemisia dubia Wall.ex Bess.) and Aster trituberculatus (Aster ageratoides Turcx.)
35	Oriental white oak (Quercus aliena)	Tree layer	Cork oak	Chinese sumac (Rhus chinensis)	Chemical incense (Platycarya strobilacea Sieb.Et Zucc.), hawthorn (Crataegus pinnatifida Bge.), and oil pine
		Shrub layer	Forsythia and forsythia	Hanging hook	Chinese rose (Viburnum dilatatum Thunb.)
		Herbaceous layer	Goat’s Beard Grass (Eriophorum) and Ugly Vegetable	Wild Chrysanthemum and Aristolochia (Aristolochia debilis Sieb.et Zucc.)
42	Sawtooth oak	Tree layer	Liaodong oak (Quercus liaotungensis)		Quercus serrata and maple (Acer ginnala Maxim.)
		Shrub layer
		Herbaceous layer	Wolfsbane (Sophora moorcroftiana (Benth.) Baker) and Conifer Carex (Carex onoei Franch.et Sav.)
36, 45, 46	Oriental white oak (Quercus aliena)	Tree layer	Chinese red pine	Chinese chestnut (Castanea mollissima Bl.)	European hornbeam
		Shrub layer	Hanging hook	Honeysuckle, Southern snakeroot, and Magnolia vine
		Herbaceous layer	Tussock grass and morning glory	Longleaf tussock and three-spike tussock
22	Birch	Tree layer	Menorah (Bothrocaryum controversum (Hemsl.) Pojark.)
		Shrub layer	Chinese rose (Podocarpus indicus)	Rose	Indian ginger (Gingerbread)
		Herbaceous layer	Goat’s Beard Grass and Ugly Vegetable	Lonicera, tussock grass, and early morning glory
4	Chinese cypress	Tree layer	sea-buckthorn
		Shrub layer
		Herbaceous layer	Sheep’s beard grass, Artemisia annua, and white fescue
7, 14	False acacia	Tree layer	Chinese red pine		Various trees of genus Populus
		Shrub layer	Chinese cypress	Rose, Lonicera
		Herbaceous layer	White Goat Weed, Iceweed (Agropyron cristatum (Linn.) Gaertn.), and Artemisia lilaca	Sheep’s beard and along the step-grass (Ophiopogon bodinieri Levl.)
2, 5, 8, 10	Liaodong oak (Quercus liaotungensis)	Tree layer	Almonds (Armeniaca sibirica (Linn.) Lam.) and large-fruited elms (Ulmus macrocarpa Hance)		False acacia
		Shrub layer	Side-oak and spindly maple (Acer stenolobum Rehd.)	Yellow Rose (Rosa hugonis Hemsl.) and Cotoneaster (Cotoneaster multiflorus Bge.)
		Herbaceous layer	Early morning glory (Graminia przewalskii)
3, 6, 9	aspen	Tree layer	Birch (Betula platyphylla Suk.) and Liaodong oak	Pine, Tea Strip Maple, and Mountain Apricot
		Shrub layer	Yellow rose, caraway, and sharp-leaved tetrapod (Dendrobenthamia angustata (Chun) Fang)		Juniper (Pyrus betulifolia Bge.), elm (Ulmus), and lilac (Syzygium aromaticum(L.)Merr.Et Perry)
		Herbaceous layer	Lanceolate tussock (Carex lanceolata), whitehead (Pulsatilla chinensis (Bunge) Regel), long manzanita (Stipa bungeana Trin.), and Artemisia ferruginea (Artemisia sacrorum Ledeb.)	Iceweed and Dioscorea bulbifera (Dioscorea nipponica Makino)
11, 12	Chinese red pine	Tree layer
		Shrub layer	Yellow Roses, Sour Dates (Ziziphus jujuba Mill.var.spinosa (Bunge) Hu ex H.F.Chow.), and Hoarfrosts	Lonicera japonica (Lonicera ferdinandii Franch.)	Tujang Hydrangea (Spiraea pubescens Turcz.)
		Herbaceous layer	Lanceolate tussock, windweed (Saussurea japonica (Thunb.) DC.), and dragon’s toothwort (Agrimonia pilosa Ledeb.)	Aster (Aster tataricus Linn.f.) and Magnolia multiflora

).

Restoration measures are taken accordingly for the three different types of degradation. Mild degradation can be achieved by adjusting the density of the forest stand, cutting down 20–40% of the trees to form a forest window, introducing positive pioneer tree species at the initial stage, adding fast-growing native tree species after 10 years, and gradually forming a top community after 20 years. Moderate degradation requires a moderate reduction in density, improved water and fertilizer conditions, and promotion of vegetation recovery before target transplantation. Severe degradation should be dominated by shrubs and grasses, and ecosystem health should be gradually restored through micro-establishment transformation and moderate management. The overall restoration process should follow the law of natural succession, from pioneer plants to positive plants, and ultimately forming a stable top community (

Table 6. Density regulation of degraded forests, mixed forest plantations, and natural regeneration.

Forest Class Number	Degree of Degradation	Density Control	Construction of a Mixed Forest	Natural Regeneration
0, 1, 3, 5, 7, 9, 10, 11, 15, 16, 17, 19, 20, 21, 23, 25, 27, 28, 29, 30, 31, 32, 33, 35, 36, 38, 39, 40, 42, 43, 44, 45, 46, 47, 48, 49	Functional degradation: mild	Adjusting stand density by cutting 20–40% of trees in the middle of the forest to form forest windows	Transplantation of pioneer tree species, the first three years of fixed-value nurturing, about 10 years in the community of pioneer tree species to adapt to the habitat, into the forest canopy; adding fast-growing native shallow-rooted tree species to stabilize the forest community style; about 20 years of slow-growing negative species growth stabilization, the forest is gradually close to the top of the community. (Miyawaki method limited, only in isolated degraded patches).	Transplanting seedlings, sealing, or nurturing management
2, 6, 8, 12, 13, 14, 22, 24, 26, 34, 37, 41	Functional degradation: moderate	Moderate land preparation, proper water and fertilizer management, and selective logging based on water carrying capacity	Transplantation of pioneer tree species, the first three years of fixed-value nurturing, about 10 years in the community of pioneer tree species to adapt to the habitat, into the forest canopy; adding fast-growing native shallow-rooted tree species to stabilize the forest community style; about 20 years of slow-growing negative species growth stabilization, the forest is gradually close to the top of the community. (The Miyawaki method is partially applicable for small degraded microsites).	Transplanting of seedlings and nursery management
4, 18	Structural degradation: severe	Cut down all kinds of trees	The first step is to improve the microtopography by transplanting fast-growing trees to rapidly form the forest canopy in the first 10 years, adding shrubs and herbs to acclimatize the community to the habitat conditions, transplanting slow-growing, shady trees in the third year or so, and gradually approaching the top of the forest community in the 20th year or so. (Miyawaki method applicable, but requires soil improvement and site-specific design)	Not recommended to update

).

3.3.3. Restoration Program—Example of Acacia Forests in Clear Plateau Township

Located in Qingyuan Town, Xunyi County, Xianyang City, Shaanxi Province (longitude 108°77′–111°15′ East, latitude 35°43′–39°35′ North), the acacia forest of Qingyuan Town covers a total area of 31.82 km². With mountainous topography, gently sloping slopes, and high elevations (909–1472 m), it belongs to a temperate continental climate. The average annual precipitation is 537–650 mm, the average annual temperature is 9.0–13.2 °C, the average annual sunshine hours are 2017.2–2346.9 h. The forestry subclass number is 0, the forest degradation area is 10.55 km², the dominant species of acacia forest is acacia, and the common vegetation is thorn, sea buckthorn, and white sheep grass.

Taibai Mountain National Forest Park is located in the middle of the North China Plate (107°22′–107°51′E longitude, 33°49′–34°35′N latitude), with an altitude of 1000–1500 m above sea level and a temperate continental climate, with an average annual temperature of about 10 °C and an annual average of about 2500 h of sunshine. Annual precipitation is about 534 mm, and there is a rich variety of vegetation types. Tibai Mountain and Qingyuan Town have similar physical conditions, are less disturbed by human beings, and have more complete vegetation communities, which can be used as target communities.

Vegetation restoration in Clear Plateau Township was carried out in three stages: pioneer stage, intermediate stage, and top stage (

Table 7. Main plants for vegetation rehabilitation in Clear Plateau Township.

Community Name	Tree Layer		Shrub Layer		Field Layer
	Dominant Species	Accompanying Species	Dominant Species	Accompanying Species	Dominant Species	Accompanying Species
Pioneer stage	Masson pine (Chinese red pine and horsetail pine)	-	Thorn and sea buckthorn	Gooseberry, three-branched ocotillo, and Rokudoumu	wire grass (Eleusine indica)	lover
Intermediate stage	-	-	Indian ginger (Gingerbread)	double-shielded wood	-	Long-stalked Mountain locust, and stinkweed
Top-level stage	beech	Chinese red pine	-	-	-	Carex broadleaf sedge

), which is in line with the natural community succession from pioneer plants to positive fast-growing tree species and then to the top community.

4. Discussion

4.1. Identify the Effectiveness and Limitations of Monitoring Methods

This study employed the LandTrendr algorithm on the Google Earth Engine (GEE) platform to monitor forest degradation in Shaanxi Province, using the normalized burn ratio (NBR) as the primary indicator. Parameters were adjusted based on forest type. Compared with traditional statistical time series approaches, this method captures temporal dynamics more effectively and can detect minor degradation and disturbances [58]. Limitations remain. Only 200 sample points were selected for visual interpretation using Google Earth, which may omit certain forest changes and affect validation accuracy. The method supports long-term forest change analysis and has potential for carbon sink studies, although a standardized national carbon stock assessment system is lacking. In addition, pollution indicators (e.g., NO₂, SO₂) and climate variables (e.g., temperature, precipitation, greenhouse gas concentrations) were not included. These factors may influence degradation processes but were excluded due to data availability and resolution mismatch with remote sensing analysis. Although the use of Landsat Collection 1 Surface Reflectance data mitigates cross-sensor differences, residual inconsistencies between TM, ETM+, and OLI sensors may still influence long-term trend interpretation. One limitation of the current approach is that the LandTrendr-based temporal segmentation relies on spectral curve fitting, which may be affected by temporal autocorrelation within the time series. Although the algorithm includes spike and overshoot controls to reduce overfitting, it does not incorporate formal time-series modeling. Future research could apply autoregressive or Bayesian models to better account for natural spectral fluctuations and improve the robustness of trend detection. Future work could benefit from additional normalization techniques or sensor-specific bias correction models. This study acknowledges several methodological limitations that may influence the interpretation of results. First, the use of MSPA for ecological source identification does not inherently account for ecological quality or biodiversity value. The selection of a 10 km² threshold for source designation was not based on a universal ecological criterion but rather chosen to balance spatial scale, model tractability, and landscape continuity in the context of Shaanxi Province’s large extent. Second, while the resistance surface was derived using AHP based on expert knowledge, the lack of empirical calibration or sensitivity analysis introduces uncertainty in the resistance values. Future studies should consider incorporating species-level movement data [59] or landscape functional indicators [60,61] to improve the objectivity and accuracy of connectivity modeling. These limitations are acknowledged and suggest directions for future work.

4.2. Analysis of the Main Drivers of Forest Degradation in Shaanxi Province

The drivers of forest degradation in Shaanxi Province are multidimensional and spatial-temporal and can be attributed to the dual role of natural factors and human activities [2,62,63,64]. Soil erosion in areas with steep slopes (>25°) and high altitudes (>2000 m) is a major natural driver of forest degradation, as these topographic conditions intensify erosion and reduce forest stability [2]. Land use conversion has increased ecological pressure. Studies have shown that the expansion of built-up land driven by rapid urbanization has significantly crowded out ecological space, and plains with slopes <2° have been preferentially occupied by urban development needs. Areas converted to built-up land typically experience vegetation degradation and fragmentation of forests and farmland. Meanwhile, two-way conversions between forests and agricultural land affect forest stability by altering land cover patterns [2]. In addition, anthropogenic disturbances, such as energy development in the coal industry and extensive agricultural production in northern Shaanxi, have indirectly weakened the ecological functions of forests by altering the surface microclimate and soil structure [64]. It is worth noting that although ecological projects such as returning farmland to forests have partially offset degradation pressures, indicators of urban morphology show that the coexistence of agglomeration and fragmentation in urban sprawl still poses a long-term threat to the carbon sink capacity of forests [64].

4.3. Rationality and Optimization Direction of Ecological Network Construction

Compared with most of the previous studies that did not form a scientific and systematic framework for forest restoration, this study proposes the overall restoration idea of “restoring ecological obstacles—opening ecological corridors—increasing the number of ecological sources—improving the quality of ecological sources—optimizing the ecological network of forests”. From the perspective of Shaanxi Province as a whole, the first step is to increase the number of ecological sources and ecological corridors in northern Shaanxi, where forest degradation is more serious, so as to promote the construction of forest ecological networks and the restoration of degraded forests in northern Shaanxi. Secondly, the quality of forest ecological sources in southern Shaanxi should be further improved, leading to the enhancements in the quality of forest ecosystems along the Qinling Mountains.

4.4. Restoration Strategies and Adaptation of Degraded Forests

This study proposes a strategy for near-natural forest restoration based on the Miyawaki method and formulates differentiated restoration plans for different degradation levels. The restoration of mildly degraded forests is technically difficult and cost-controllable, and the local community can obtain short-term benefits by harvesting timber, which is highly acceptable, but the intensity of harvesting needs to be controlled to avoid soil erosion; moderate degradation requires sustained investment in capital and manpower, and the soil improvement may destroy the original microbial community. Heavily degraded areas are dependent on natural restoration, which is low-cost but time-consuming, while full-scale reconstruction is costly and carries the risk of species invasion. Potential vegetation survey is the basis of ecological restoration, relying on historical data, remote sensing technology, and field verification, with high sampling cost and technical threshold. Reasonable target communities can accelerate the recovery of ecological functions, but if we ignore the successional pattern or species adaptability, it may cause secondary degradation, and we need to comprehensively consider the ecological factors such as soil, climate, species interactions, etc. The heterogeneity of different habitats may lead to the selection bias of the target communities [65]. There is a significant difference in climate between the north and south of Shaanxi Province, and this study focuses on restoration in south Shaanxi and Guanzhong according to the forest ecological network, while the Loess Plateau area in north Shaanxi has an average annual precipitation of only 300–500 mm, with frequent droughts, and the soil is low in organic matter content, loose in structure, and weak in water-holding capacity, which makes it difficult to satisfy the high requirements of the Miyagi method for water and soil improvement.

4.5. Future Research Directions and Recommendations

Future research could combine multiple indices for monitoring, further analyze based on forestry survey data and specific physical surveys, increase validation samples to improve accuracy, and predict and estimate changes in forest carbon sinks through forest degradation levels and trends in order to support the development of forest management and REDD+ policies in countries. Further exploration of the drivers of forest degradation should focus on synergizing ecological restoration and land-use planning in terrain-sensitive areas to mitigate the overlapping effects of multiple drivers. Make full use of the ecological functions and benefits of ecological source areas and ecological corridors, carry out degraded forest restoration from the “surface line,” promote the overall restoration of degraded forests in Shaanxi Province, advance in phases, prioritize the restoration of lightly and moderately degraded areas, reduce overall risks, strengthen multi-stakeholder collaboration, compensate economic costs through ecological compensation or carbon sink trading, improve dynamic monitoring, employ Internet of Things (IoT) technology to track restoration effects, and adjust strategies in a timely manner. Combine regional climate and soil characteristics to make localized adjustments; for example, for the northern Shaanxi Loess Plateau, such as drought and low rainfall, soil infertility, and other constraints, by reducing planting density, giving priority to drought-resistant deep-rooted tree species with irrigation and a grass-first strategy, supplemented with micro-topography modification and mulching and moisture preservation technology to improve survival rates, retaining the advantages of the Miyawaki Method of “rapid forestation”, while overcoming the ecological constraints of arid areas. At the same time, it can overcome the ecological constraints of arid areas and achieve precise restoration through “low intervention- high adaptation”.

5. Conclusions

This study established an SFD-based forest degradation assessment system for Shaanxi Province, analyzed the spatial and temporal dynamics of forest degradation over the past 30 years, constructed a forest ecological network, and proposed targeted restoration strategies based on the Miyawaki method. The main conclusions are as follows: 1) In 1991–2020, the total area of forest degradation in Shaanxi Province was 1010.89 km², dominated by functional degradation (98%) and mild degradation (87.15%), with significant differences in degradation among prefecture-level cities. 2) The forest ecological network in Shaanxi Province contains 189 ecological source sites (13,722 km²), 189 ecological corridors, 89 important nodes, and 50 small forestry classes in urgent need of restoration. 3) Diversified restoration strategies were adopted according to the degree of forest degradation, with light degradation forming forest windows through density regulation and thinning, medium degradation with land preparation and water and fertilizer management, and heavy degradation opting for in-situ maintenance or full reconstruction.

Theoretically, this work provides a new approach for identifying and evaluating forest degradation. Practically, it offers cost-effective tools and strategies, especially through the GEE platform and open-access satellite data, to support ecological restoration in large areas and guide REDD+ implementation in resource-limited regions.