Spatiotemporal Differentiation and Attribution Analysis of Ecological Vulnerability in Heilongjiang Province, China, 2000–2020

Abstract

Heilongjiang Province, a major grain-producing region in China, faces ecological vulnerabilities that directly affect its sustainable development. A scientific assessment of the spatiotemporal characteristics of ecological vulnerability and its influencing factors in Heilongjiang is crucial for a deeper understanding of environmental issues and provides theoretical support for enhancing regional ecological governance capabilities. The SRP model, combined with the AHP-CRITIC weighting method, was employed to assess Heilongjiang Province’s ecological vulnerability’s temporal and regional differentiation trends between 2000 and 2020. The aggregation kinds of ecological vulnerability were examined using spatial autocorrelation. GeoDetector was used to determine the main elements affecting ecological vulnerability in the province. Additionally, the ecological vulnerability status in 2030 was predicted using the CA-Markov model. The findings indicate that (1) the average EVI values for Heilongjiang Province during the three periods were 0.323, 0.317, and 0.347, respectively, indicating a medium level of ecological vulnerability across the province; the ecological vulnerability initially decreased and then worsened. Spatially, the distribution followed a pattern of “high in the east and west, and low in the north and south”. (2) Spatial agglomeration is evident, with high-high (H-H) aggregation primarily occurring in heavily and extremely vulnerable areas characterized by high human activity, while low–low (L-L) aggregation is mainly found in mildly and marginally vulnerable areas with a favorable natural background. (3) Biological abundance, net primary productivity, dry degree, and PM_2.5 were the main drivers of ecological vulnerability, with interactions between these factors amplifying their impact on ecological vulnerability. (4) The CA-Markov model prediction results indicated an upward trend in the overall ecological vulnerability of Heilongjiang Province by 2030, reflecting a decline in the ecological environment. The study indicates that the ecological vulnerability of Heilongjiang Province is closely linked to its natural geographic conditions and is influenced through the interplay of several environmental elements. Based on the vulnerability zoning results, this paper proposes governance recommendations for regions with different vulnerability levels, aiming to provide theoretical support for future ecological restoration and sustainable development.

1. Introduction

The ecosystem is essential to human survival and development, providing a critical foundation for the sustainable development of society and the economy [1]. In recent years, due to the impact of global climate change and the increasing population [2,3,4,5], the restoration and self-purification capacity of ecosystems has been steadily weakening. Numerous extreme climate events have occurred globally, and ecological issues such as energy shortages, soil erosion, land desertification, and the sharp decline in biodiversity have become more frequent [6,7,8]. Human greenhouse gas emissions also have a significant impact on climate and environmental changes [9]. Because of this, assessing ecological vulnerability has steadily emerged as a major area of interest and concentration for numerous international organizations and institutions [10]. Implementing a holistic system view of the living community, which includes mountains, waters, forests, lakes, grasslands, and deserts, depends on identifying and scientifically evaluating the formation mechanisms of ecosystem vulnerability. This approach is extremely important for ecological environmental construction and the sustainable development of human society in China [11].

Ecological vulnerability is defined as the ecological response of ecosystems to external environmental disturbances under specific spatial and temporal conditions [12,13]. It results from the combined influence of natural environmental factors and anthropogenic activities [14] and has gradually become one of the most important indicators of ecological environmental quality and sustainable development. In the 20th century, the International Biological Program (IBP) initially researched ecological vulnerability. As research deepened, many domestic and international scholars developed a series of evaluation models based on different principles and objectives. These include the “driving force-pressure-state-impact-response” (DPSIR) model [15], the “sensitivity-resilience-pressure” (SRP) model [16], the “remote sensing ecological index” model [17], the “vulnerability scoping diagram” (VSD) model [18,19], and the “pressure-state-response” (PSR) model [20], among others, for selecting indicator factors. The SRP model, which emphasizes the interplay between natural and social forces, is built around the fundamental idea of ecosystem stability. It assesses ecological vulnerability from three dimensions: ecological sensitivity, ecological resilience, and ecological pressure, providing a more comprehensive understanding of the characteristics of regional ecological vulnerability. It is applicable to urban settings and has been extensively utilized in studies on ecological vulnerability [21,22]. In calculating the weights of indicators, most scholars primarily use methods such as artificial neural networks [23], analytic hierarchy processes [24], the entropy weight method [25], principal component analysis [26,27], and the comprehensive evaluation method [28] to measure the contribution of each indicator. However, single evaluation methods have various limitations. As research progresses, evaluation approaches have gradually shifted towards a combination of methods. This integrated approach leverages the strengths of both methods, making the results more scientifically robust and objective. The CRITIC weighting method effectively considers the differences and correlations among indicators, thereby minimizing errors caused by information overlap between them [29]. As a result, the indicator weights in this study are determined by combining the AHP with the CRITIC weighting approach. Few studies have focused on modeling projections for conditions of future environmental change, with most recent studies primarily focusing on the dynamic assessment of the geographical and temporal evolution of regional ecological vulnerability [30]. In terms of study areas, Current evaluations of ecological vulnerability are primarily focused on arid and semi-arid zones [31,32,33], extreme disaster areas [34,35], the Himalayas [36,37], the Tibetan Plateau [38,39,40], karst regions [41,42], wetlands [43], forests [44], and other ecologically fragile areas. The ecological fragility of China’s Heilongjiang Province, a significant grain-producing region, is, nevertheless, poorly understood.

Heilongjiang Province is a key region in China, both in terms of agricultural and ecological resources, and plays a critical role as the “ballast of national food security”, making it vital to the country’s food and ecological security. Since 2000, the unreasonable development and utilization of land resources has led to the increasing degradation of black soil, resulting in soil erosion and desertification, which in turn causes a decline in land fertility. Due to the unique geographical location of Heilongjiang Province, spring droughts and flood disasters are also significant ecological and environmental issues. Additionally, with the continuous advancement of urbanization, the discharge of water used for production and daily life has impacted the water resources of the Songhua River Basin. These factors highlight the immense pressure on the ecological environment of Heilongjiang Province, making the strengthening of ecological protection and construction a matter of urgent importance. With the development of technology, the capabilities of Geographic Information System (ArcGIS 10.8) software in data processing and spatial analysis have significantly improved. In their study on urban forest management and the application of tree databases in urban greening, Nattasit Srinurak et al. [45] defined the tree survey and analysis process by integrating GIS and a web-based interface. CA-Markov models were initially widely used for assessing future land use changes and have since yielded promising results in various other fields [46]. Using Heilongjiang region as a case study, the study examines the ecological vulnerability of the region in 2000, 2010, and 2020 using remote sensing and GIS software. By integrating the AHP-CRITIC method with a GeoDetector, a more comprehensive perspective on factor analysis is provided, revealing the interactions of indicator factors over different periods and further deepening the understanding of the spatiotemporal dynamics of ecological vulnerability. Additionally, the application of the CA-Markov model offers a new approach for predicting trends in ecological vulnerability. This study not only advances the theoretical understanding of ecological vulnerability but also provides forward-looking predictions and new scientific evidence for regional ecological environment management and the creation of policies for sustainable development.

2. Materials and Methods

2.1. Study Area

The northeastern Chinese province of Heilongjiang (43°26′–53°33′ N, 121°11′–135°05′ E) has a total land area of 473,000 km². It includes the Daxinganling region and twelve cities at the prefectural level (Figure 1). According to the Heilongjiang Provincial Statistical Yearbook, the population was 31.71 million by the end of 2020, the GDP was 1369.85 billion yuan, and the urbanization rate was 65.6%. Heilongjiang Province is characterized by numerous mountain ranges with an average altitude of 300 to 1000 m. The climate is cold-temperate and temperate continental, featuring four distinct seasons with concurrent rainfall and heat. The average annual temperature is between −5.3 °C and 5 °C, while the average annual precipitation is between 400 and 650 mm. Heilongjiang Province is rich in natural resources, with vast forested areas, diverse wetland types, and prominent ecological functions, holding significant ecological importance. Due to its proximity to Russia, the Heilongjiang River Basin forms a unique cross-border ecological corridor along the China–Russia border. The black soil arable land area of Heilongjiang Province, which makes up 56.1% of Northeast China’s total area, is 1.04 × 10⁵ km², making it the province with the biggest area of arable land in China. It is one of China’s main bases for grain production and is essential to the long-term viability of the country’s agricultural sector [47]. The study region excludes the districts of Songling and Gaghdach.

2.2. Data Sources

This study selects the years 2000, 2010, and 2020 as crucial time periods to investigate the spatiotemporal changes in ecological vulnerability in Heilongjiang Province. Data for these years are relatively complete in both domestic and international databases, providing accurate information for ecological vulnerability analysis. Additionally, these years coincide with significant policy implementation and environmental change periods in Heilongjiang. In 2000, the province adopted several important measures for ecological protection and environmental governance, while 2020 marks a crucial milestone in China’s comprehensive advancement of ecological civilization. Therefore, the study utilized topographic, meteorological, remote sensing, land use, socio-economic, and other data for Heilongjiang Province covering the years 2000, 2010, and 2020 (

Table 1. Data sources and uses.

Datatypes	Source	Resolution/m	Use
Topographic data	Resource and Environmental Science Data Platform (https://www.resdc.cn/) (accessed on 15 February 2024)	90 m	Obtain elevation, slope
Meteorological data	National Earth System Science Data Center (http://www.geodata.cn/) (accessed on 15 February 2024)	1 km	Obtain annual average temperature, annual precipitation, dry degree
Land use data	Resource and Environmental Science Data Platform (https://www.resdc.cn/) (accessed on 15 February 2024)	30 m	Obtain degree of land use, biological abundance, landscape diversity indices
Remote sensing data	NASA (https://www.nasa.gov/) (accessed on 15 February 2024)	1 km	Obtain NDVI, net primary productivity
Socioeconomic data	WorldPop (https://www.worldpop.org/) (accessed on 15 February 2024)	-	Obtain population density
	Heilongjiang Statistical Yearbook	-	Obtain GDP per capita
Other data	CHAP (https://data.tpdc.ac.cn/) (accessed on 15 February 2024)	1 km	Obtain PM2.5

). To ensure consistency and accuracy across all datasets, the raster data resolution was uniformly set to 1 km × 1 km, and the spatial data were projected using the Krasovsky coordinates Albers projection.

2.3. Methodology

Based on remote sensing and GIS technology, this work uses the SRP model in conjunction with the AHP-CRITIC approach for weighted aggregation to examine the spatiotemporal differentiation characteristics of ecological vulnerability in Heilongjiang Province over three periods. Its motivating factors are investigated using the geographical detector model, and the distribution pattern of ecological vulnerability in 2030 is predicted using the CA-Markov model. Figure 2 shows the framework for the investigation.

2.3.1. Selection of Evaluation Indicators

A total of 13 specific indicators were chosen to build the assessment index system based on pertinent vulnerability research publications, the SRP ecological vulnerability evaluation model [48,49], and the actual ecological conditions in Heilongjiang Province (

Table 2. Indices for evaluating ecological vulnerability.

Standard	Element	Factor	Nature of the Indicator
Sensitivity	Topographic factor	Elevation (X1) Slope (X2)	+ +
	Meteorological factor	Annual average temperature (X3) Annual precipitation (X4) PM2.5 (X5) Dry degree (X6)	+ − + +
Resilience	Eco-vitality factors	Landscape diversity (X7) NDVI (X8) Biological abundance (X9) Net primary productivity (X10)	− − − −
Pressure	Anthropogenic stress factors	Population density (X11) GDP per capita (X12)	+ +

). The selection followed the principles of systematics, timeliness, scientific validity, and data accessibility while incorporating both natural and anthropogenic factors. “Ecological sensitivity” refers to the way an ecosystem responds to external disturbances [50]. Six indicators are selected from the perspectives of topography and climate: elevation, slope, annual average temperature, annual precipitation, PM_2.5, and dry degree. Ecological resilience is the capacity of an ecosystem to recover from external disturbances as long as the disruption does not rise above the recovery threshold [51]. Four indicators are selected from the perspective of ecological vitality: landscape diversity index, normalized vegetation index, biological abundance, and net primary productivity of vegetation. The extent of pressure from outside disturbances on an ecosystem, usually resulting from human activity, is referred to as ecological pressure [6]. It is represented by three indicators: population density, GDP per capita, and land use intensity.

2.3.2. Index Standardization and the Determination of Weights

(1)

Index Standardization

Due to differences in the nature and attributes of each evaluation indicator, positive and negative indicators must be standardized separately using the polar deviation standardization method [52]. The following formula is applied:

Positive indicators:

$$R_m={\displaystyle \frac{x_i-x_{min}}{x_{max}-x_{min}}}$$

(1)

Negative indicators:

$$R_n={\displaystyle \frac{x_{max}-x_i}{x_{max}-x_{min}}}$$

(2)

where R_m and R_n are the standardized values of indicator i, x_i is the original value of indicator i and x_max and x_min are the maximum and minimum values of the i-th indicator, respectively.

(2)

Determination of Weights

Diakoulaki initially suggested the entropy weighting approach, which is improved upon by the CRITIC weighting method [53]. The standard deviation and correlation coefficient, which are then utilized to generate weight values, show the variability and disagreement between indicators. A larger standard deviation results in a higher weight assignment, while a higher correlation coefficient indicates less conflict, leading to a smaller weight assignment. The calculation formulas are as follows:

$${\omega }_i=\sqrt{{\displaystyle \frac{1}{m}}}\sum _{i=1}^m{\left(X_{ij}-{\overline{X}}_j\right)}^2 i=\mathrm{1,2},\dots ,n$$

(3)

$${\rho }_{ij}={\displaystyle \frac{cov\left(X_i, X_j\right)}{\left({\omega }_i, {\omega }_j\right)}} i,\hspace{1em} j=\mathrm{1,2},\dots ,n$$

(4)

$$W_j={\omega }_j\sum _{i=1}^{n}1-{\rho }_{ij}$$

(5)

$$E_j={\displaystyle \frac{W_j}{\sum _{j=1}^n{W}_j}}$$

(6)

where ω_i is the standard deviation of the i-th indicator; ρ_ij is the correlation coefficient of the i-th and j-th indicators; W_j represents the information content of the j-th indicator in the evaluation indicator system; and E_j indicates the objective weight of the j-th indicator.

The AHP method was chosen for subjective weighting, while the CRITIC method was used for objective weighting. The AHP method relies primarily on expert judgment to assess the reasonableness of indicator weights, making it highly subjective [54]. Therefore, the two methods are combined, and the formulas are as follows:

$$W=\lambda {\beta }_i+\left(1-\lambda \right){\beta }_j$$

(7)

where W, β_i, and β_j represent the combined, subjective, and objective weight values, respectively. Based on the formula for determining the comprehensive weight of the indicators, λ is set to 0.5 [55] to guarantee the impartiality of the calculation findings.

2.3.3. Integrated Ecological Vulnerability Assessment

In this paper, the Ecological Vulnerability Index (EVI) is employed to quantify the degree of ecological vulnerability, using the following formula:

$$EVI=\sum _{i=1}^n{X}_j\times W_{ij}$$

(8)

where EVI is the Ecological Vulnerability Index, W_ij is the standardized data for each indicator, n is the total number of indicators, and X_j is the aggregate weight of the j-th indication. The EVI and the degree of ecological vulnerability in the area are positively correlated.

The natural breaks method is a commonly used classification technique for spatial data. It automatically determines the breakpoints based on the fluctuations in the data distribution curve, reducing errors introduced by subjective selection. This method reflects the data’s variation trends, making the classification results more interpretable and visually intuitive. Based on the ecological vulnerability index’s standardization and the findings of earlier research [56], the vulnerability index was divided into five groups using the natural breakpoint technique [24]. These levels are slight vulnerability [0~0.231], light vulnerability (0.231~0.314], medium vulnerability (0.314~0.412], heavy vulnerability (0.412~0.510], and extreme vulnerability (0.510~1].

2.3.4. Spatial Autocorrelation

The degree of reliance between a given unit’s attribute values and those of its nearby units is known as spatial autocorrelation [57]. It is separated into local and global Moran’s I indices [58]. A certain attribute value’s spatial relationship throughout the entire study region is evaluated using the global Moran’s I. Global Moran’s I ∈ [−1, 1]; when Global Moran’s I > 0, it indicates spatial positive correlation; when global Moran’s I < 0, it indicates spatial negative correlation; and when Global Moran’s I = 0, it suggests no significant spatial autocorrelation. The local Moran’s I reveals the spatial distribution of local heterogeneity [59]. By analyzing spatial clustering using the local Moran’s I index, a LISA map can be generated. Ecological vulnerability is classified into five clustering patterns within a 95% confidence interval: high-high clustering (H-H), high-low clustering (H-L), low-high clustering (L-H), low-low clustering (L-L), and no significant clustering (no significant). The specific details are provided in

Table 3. The meaning behind various LISA clustering models.

Clustering Types	Connotation
High-high clustering (H-H)	Characteristics of spatial clustering where both the region and the surrounding regions have a fair amount of ecological risk.
High-low clustering (H-L)	Features of spatial clustering when the surrounding area’s ecological vulnerability is low, and the region’s is high.
Low-high clustering (L-H)	Features of spatial clustering where the ecological sensitivity of the surrounding area is great but that of the region is minimal.
Low-low clustering (L-L)	Characteristics of spatial clustering where both the region and the surrounding area have comparatively low ecological vulnerability.
Not significant	No significant spatial clustering characteristics.

[60].

$$Global Moran’s I={\displaystyle \frac{\sum _{i=1}^n\sum _{j=1}^n{w}_{ij}\left(x_i-\overline{x}\right)\left(x_j-\overline{x}\right)}{\sum _{i=1}^n\sum _{j=1}^n{w}_{ij}{\sum }_{i=1}^n\left(x_i-\overline{x}\right)}}$$

(9)

$$Local Moran’s I_i={\displaystyle \frac{\left(x_i-\overline{x}\right)}{S^2}}{\sum }_j{w}_{ij}\left(x_j-\overline{x}\right)$$

(10)

where x_i and x_j represent the landscape ecological risk values for the i-th and j-th evaluation units; $\overline{x}$ is the average landscape ecological risk value across all evaluation units; w_ij is the spatial weight matrix; and S is the sum of all elements in the spatial weight matrix.

2.3.5. GeoDetector

A statistical technique called GeoDetector is utilized to find the causes of geo-graphic events and detect their spatial heterogeneity [61]. It was put forth by Wang Jinfeng [62] and associates and consists of the ecological, factor, risk, and interaction detectors. Both a factor detector and an interaction detector are used in the investigations in this publication. The amount that the independent variable, represented by q, explains the dependent variable is revealed using a factor detector. Stronger explanatory power of the independent variable X in explaining changes in the dependent variable Y is indicated by a greater value of q.

$$q=1-{\displaystyle \frac{\sum _{h=1}^L{N}_h{\sigma }_h^2}{N{\sigma }^2}}$$

(11)

where the stratification of the independent variable or factor is denoted by h = 1, …, L; the number of cells in stratum h and the entire area is N_h; and the ecological vulnerability variances in stratum h and the entire area are denoted by σ_h² and σ_h², respectively.

To determine if the explanatory power of ecological vulnerability is increased or decreased when several independent variables interact, interaction detectors are utilized [63]. The types of interactions are presented in

Table 4. The type of interaction.

Source for Judging	Interaction Type
q(X1∩X2) < Min[q(X1),q(X2)]	Weakened, non-linear
Min[q(X1),q(X2)] < q(X1∩X2) < Max[q(X1),q(X2)]	Weakened, single factor non-linear
q(X1∩X2) > Max[q(X1),q(X2)]	Enhanced, double factors
q(X1∩X2) = q(X1) + q(X2)	Independent
q(X1∩X2) > q(X1) + q(X2)	Enhanced, non-linear

.

2.3.6. CA-Markov Model

The dynamic evolution of complex systems is frequently simulated using cellular automata (CA), a dynamic mathematical model that functions in discrete states in both time and space [64], as modeled by

$$S_{ij}^{t+1}=f_q\left(S_{ij}^t\right)$$

(12)

where the state of the ij-th cell is denoted by S, the time moments by t and t + 1, the transition function by f, and the neighborhood by q.

Markov models are used to simulate the prediction of future conditions based on the transfer probabilities between the initial and intermediate periods [53] and are modeled as

$$S_e=S_O{P}_e$$

(13)

where S_e is the state after e cycles; S_O is the initial state; e is the number of cycles; and P_e is the transition probability matrix.

The correctness of the spatial patterns between the expected and actual findings is verified using the Kappa coefficient, which is computed using the CROSSTAB module in IDRISI Selva software (v17.02). When Kappa < 0.4, it indicates a poor consistency between the simulated and actual results, with low simulation accuracy. The results of the simulation are average at best when 0.4 < Kappa < 0.75. A high degree of consistency between the simulated and real results, together with good simulation performance, is indicated when Kappa is greater than 0.75 [65]. The following formula can be used to determine the Kappa coefficient:

$$Kappa={\displaystyle \frac{p_0-p_c}{p_p-p_c}}$$

(14)

where p₀, p_p, and p_c represent the percentages of correct modeling under actual conditions, ideal conditions, and random conditions, respectively.

3. Results

3.1. Features of Heilongjiang Province’s Ecological Vulnerability Changes over Time

The percentage of the area in Heilongjiang Province corresponding to the three vulnerability levels, as shown in

Table 5. Change in area of vulnerability.

Level of Vulnerability	2000		2010		2020
	Area/km2	Percentage/%	Area/km2	Percentage/%	Area/km2	Percentage/%
Slight vulnerability	118,463	27.04	142,884	32.61	92,439	21.1
Light vulnerability	130,248	29.73	89,835	20.5	120,293	27.46
Medium vulnerability	75,469	17.22	76,714	17.51	88,887	20.29
Heavy vulnerability	66,386	15.15	95,008	21.68	66,728	15.23
Extreme vulnerability	47,609	10.87	33,736	7.7	69,749	15.92

, reveals the temporal changes in ecological vulnerability. The average Ecological Vulnerability Index (EVI) values from 2000 to 2020 were 0.323, 0.317, and 0.347, in that order. This indicates a trend of initial stabilization followed by a worsening of ecological vulnerability. The areas categorized as having mild, slight, and moderate vulnerability predominantly dominated the province, accounting for 73.98%, 70.62%, and 68.85% of the total area, respectively. Although the percentage of the study area classified under various ecological vulnerability classes varies significantly over time, these numbers suggest that more than half of the region constantly falls under these vulnerability categories. Overall, Heilongjiang Province’s ecological vulnerability has been gradually increasing, as evidenced by a decline in the percentage of places with high vulnerability grades and an increase in the percentage of areas with lower vulnerability classes. Between 2000 and 2010, the areas classified as slightly, moderately, and severely vulnerable increased by 5.57%, 0.28%, and 6.53%, respectively. In contrast, the proportion of areas with mild and extreme vulnerability declined by 9.22% and 3.17%, respectively. Notably, the reduction in extremely vulnerable areas contributed to a relative decrease in the overall ecological vulnerability of Heilongjiang Province during this period. From 2010 to 2020, the regional vulnerability index of Heilongjiang Province increased. The slightly vulnerable area decreased from 142,884 km² to 92,439 km², a reduction of 11.51%. Meanwhile, the moderately vulnerable and extremely vulnerable areas increased by 2.78% and 8.22%, respectively. This trend indicates that ecosystems in the province were under growing pressure during this period.

To further analyze the area transformation characteristics across different vulnerability levels, the area transfer matrix was visualized. An ecological vulnerability level transfer map from 2000 to 2020 in Figure 3 displays the findings. Between 2000 and 2010, areas with slight vulnerability predominantly transitioned to mild vulnerability, with a smaller proportion shifting to moderate vulnerability. Mild vulnerability mainly transitioned to slight vulnerability, with the remainder moving to moderate vulnerability. Moderate vulnerability primarily transitioned to severe vulnerability, while severe vulnerability largely shifted to moderate and extreme vulnerabilities. Approximately half of the extreme vulnerability area transitioned to moderate vulnerability. The total area transitioning out of slight vulnerability was 17,855.8 km², with 9706 km² shifting to lower vulnerability categories and 83,852 km² transitioning to higher vulnerability classes. This further suggests a reduction in ecological vulnerability between 2000 and 2010. Between 2010 and 2020, all vulnerability classes predominantly shifted towards higher vulnerability categories, further indicating an increase in ecological vulnerability during this period.

3.2. Features of Ecological Vulnerability’s Spatial Distribution

The results of the weighted indicators were summed up using ArcGIS 10.8 software to generate the ecological vulnerability distribution map of Heilongjiang Province for the period from 2000 to 2020 (Figure 4). As shown in Figure 4, the spatial heterogeneity is evident, with a distribution pattern of ‘higher in the east and west, lower in the north and south’. The north-western (Daxing’anling Region, Heihe City), north-eastern (Yichun City), and south-eastern (Mudanjiang City) regions of Heilongjiang Province are predominantly characterized by mild and slight vulnerability. These areas are marked by elevation, woodland land use types, high vegetation cover, strong soil water retention capacity, and low population density, contributing to favorable ecological conditions. In contrast, the western (Qiqihar City, Daqing City, Harbin City) and central (Suihua City) regions are dominated by heavy and extreme vulnerability. These areas are densely populated and frequently disturbed by human activities, with rapid economic development and industrialization exacerbating PM_2.5 pollution and increasing arid climate, causing the ecological environment to deteriorate. From 2000 to 2010, the mean EVI value in Heilongjiang Province decreased from 0.323 to 0.317. During this period, the Daxinganling area shifted from light to slight vulnerability, while part of the extreme vulnerability areas in Qiqihar City and Suihua City changed to heavy vulnerability areas, indicating a decrease in vulnerability. However, from 2010 to 2020, the mean EVI increased from 0.317 to 0.347, reflecting an overall increase in vulnerability. During this time, parts of Daxinganling transitioned from slight to medium vulnerability, the north-western Heihe region shifted from slight to light, Qiqihar City changed from medium to extreme vulnerability, and the southern part of Harbin City transitioned from light to medium vulnerability.

A further comparison of the differences in ecological vulnerability among the municipalities was made based on the average ecological vulnerability values for the three periods, as shown in Figure 5. The mean ecological vulnerability values for each city ranged from 0.21 to 0.49, with Qiqihar having the highest three-period average and Yichun having the lowest. Among them, Yichun City, Daxing’anling Region, Qitaihe City, and Mudanjiang City had EVI values between 0.2 and 0.3. In contrast, Shuangyashan City, Qitaihe City, Hegang City, Jixi City, and Jiamusi City had EVI values between 0.3 and 0.4. The remaining municipalities had average EVI values greater than 0.4 across the three periods.

3.3. Analysis of Ecological Vulnerability via Spatial Correlation

For 2000, 2010, and 2020, the corresponding Moran’s I values were 0.916, 0.927, and 0.919, respectively. This suggests that ecological vulnerability in Heilongjiang Province is significantly spatially clustered and positively correlated. Between 2000 and 2020, Heilongjiang Province’s degree of spatial agglomeration and correlation first strengthened before subsequently diminishing. The LISA map was created using the Moran’s I value, as seen in Figure 6. High-high (H-H) and low-low (L-L) agglomeration are the two main agglomeration types in the research region; low-high (L-H) and high-low (H-L) agglomeration make up a comparatively lesser percentage of the total. The H-H areas are mostly found in the eastern (Hegang City, Jiamusi City, Shuangyashan City) and southwestern (Qiqihar City, Daqing City, Suihua City, Harbin City) regions of the study area, where severe and extreme ecological vulnerability predominates. The percentage of this area increased from 27.28% in 2000 to 28.67% in 2020, with a reduction in area of 1835.27 km². The L-L aggregation area is primarily distributed in the northeastern (Yichun City), northwestern (Heihe City), and southeastern (Mudanjiang City) parts of the study area, where ecological vulnerability is mainly dominated by mild and slight vulnerability. The area share decreased from 37.55% in 2000 to 35.00% in 2020, with a reduction in area of 11,480.82 km². This suggests that Heilongjiang Province’s total area of ecological vulnerability, both in the high and low value, is shrinking. Less than 1% of the entire land is made up of the high-low and low-high regions, which are dispersed unevenly.

3.4. Driving Factor Analysis

The study organizes 13 indicator factors into a unified format spatial dataset and applies the natural breaks method for discretization. These factors are treated as independent variables. The study region is divided using a 5 km × 5 km grid, and the dependent variable is the mean EVI values for the three periods. The q-values of each factor are calculated to assess its explanatory power regarding the spatial distribution of ecological vulnerability. Factors with q-values greater than 0.1 are considered to have a strong explanatory influence on ecological vulnerability. Building on the single-factor effects, the interacting q-values are calculated to determine whether significant synergistic effects exist between factors.

The factor detection results from the GeoDetector (Version 2017) are presented in

Table 6. Findings from the identification of the main causes of ecological vulnerability.

Driving Factor	q-Value			Mean q-Value	Ranking of Mean q-Values
	2000	2010	2020
Elevation	0.291	0.42	0.285	0.332	6
Slope	0.245	0.3	0.238	0.261	8
Annual average temperature	0.275	0.433	0.331	0.346	5
Annual precipitation	0.245	0.148	0.143	0.178	10
PM2.5	0.46	0.598	0.516	0.525	4
Dry degree	0.558	0.597	0.486	0.547	3
Landscape diversity	0.064	0.073	0.07	0.069	13
NDVI	0.388	0.206	0.146	0.247	9
Biological abundance	0.586	0.625	0.561	0.591	1
Net primary productivity	0.515	0.652	0.58	0.583	2
Population density	0.049	0.12	0.047	0.072	12
GDP per capita	0.049	0.085	0.117	0.084	11
Degree of land use	0.329	0.326	0.281	0.312	7

, which shows the findings from the identification of the main causes of ecological vulnerability. Each indicator factor’s explanatory capacity for Heilongjiang Province’s ecological vulnerability over the three phases is significant, as indicated by the p-value of 0. The top five q-values in 2000 were biological abundance (0.586), dry degree (0.558), net primary productivity (0.515), PM_2.5 (0.460), and population density (0.408), with biological abundance exhibits the strongest explanatory power in relation to ecological vulnerability. In 2010, the top five q-values were net primary productivity (0.652), biological abundance (0.625), PM_2.5 (0.598), dry degree (0.597), and annual average temperature (0.433), with net vegetation productivity emerging as the dominant factor of ecological vulnerability, followed by biological abundance. The top five indicators in 2020 were the same as those in 2010, although the magnitude of the q-values slightly changed. It can be observed that biological abundance, net primary productivity, dry degree, and PM_2.5 are the primary factors affecting the shifts in Heilongjiang Province’s ecological vulnerability’s regional distribution, with q-means of 0.591, 0.583, 0.547, and 0.515, respectively. In contrast, the q-means of population density, GDP per capita, and landscape diversity index are all less than 0.1, indicating their relatively minor influence. Additionally, the q-mean for the degree of land use is 0.312, making it the anthropogenic factor with the strongest explanatory power for ecological vulnerability in Heilongjiang Province.

To assess the extent of the explanatory power of ecological vulnerability after the interaction of any two factors, an interaction analysis was conducted, and the results are presented in Figure 7. The analysis of indicator interactions across the three phases consistently revealed either a two-factor enhancement or a non-linear enhancement. This indicates that the explanatory power of ecological vulnerability, when accounting for the interaction between any two factors, surpasses that of any single factor alone. It implies that rather than being purely determined by individual elements or their straightforward cumulative implications, Heilongjiang Province’s ecological vulnerability is shaped by the combined effects of several driving forces.

With a q-value of 0.723, biological abundance (X9) and PM2.5 (X5) showed the highest correlation in 2000. In 2010, the strongest interaction occurred between net primary productivity (X10) and PM_2.5 (X5), with a q-value of 0.789. In 2020, the strongest interaction was between net primary productivity (X10) and desiccation (X6), with a q-value of 0.710. These findings further confirm that biological abundance, net primary productivity, dry degree, and PM_2.5 are key factors with strong explanatory power for ecological vulnerability in Heilongjiang Province.

3.5. CA-Markov Predictive Analytics

Using the CA-Markov model in the IDRISI program, ecological vulnerability data from 2010 in Heilongjiang Province served as the baseline. The ecological vulnerability of Heilongjiang Province in 2020 was simulated using the ecological vulnerability transition matrix from 2000 to 2010. The proportional error was set to 0.15 and the iteration coefficient was 10. The simulation results were then compared with the 2020 data for accuracy verification. The simulated and real 2020 data showed a high degree of consistency, as indicated by the computed Kappa coefficient of 0.77, which is higher than 0.75. This suggests that the model is suitable for predicting the ecological vulnerability of Heilongjiang Province in 2030. The ecological vulnerability data for 2010 and 2020 were used to predict the vulnerability of Heilongjiang Province in 2030, with the results presented in Figure 8. Overall, in 2030, Heilongjiang Province is expected to be predominantly characterized by light and medium vulnerability, with the percentages of slightly vulnerable, lightly vulnerable, mediumly vulnerable, heavily vulnerable, and extremely vulnerable areas being 16.37%, 25.99%, 22.23%, 15.74%, and 19.68%, respectively. While the area proportion of low vulnerability grades fell, the area share of high vulnerability grades rose compared to 2020. Specifically, the area of slight vulnerability decreased by 20,729 km², a reduction of 4.73%, and the area of light vulnerability decreased by 3968 km², a decrease of 0.91%. Conversely, the area shares of heavy and extreme vulnerability increased by 0.51% and 3.67%, respectively. This indicates that by 2030, the ecological vulnerability of Heilongjiang Province will have worsened, highlighting the need for strengthened ecological protection measures.

4. Discussion

The study combines the SRP model with the AHP-CRITIC method for weighted analysis to examine the spatiotemporal distribution characteristics of ecological vulnerability in Heilongjiang Province across three periods. Additionally, the study uses the geographical detector model to explore the driving factors and employs the CA-Markov model to predict ecological vulnerability distribution in 2030.

The results of the study indicate that the ecological vulnerability of Heilongjiang Province follows a distribution pattern of “high in the east and west, low in the north and south”. This pattern is primarily attributed to the higher altitudes in the northern and southern regions of Heilongjiang Province, where the land use is predominantly forested. Additionally, these areas host several national nature reserves (e.g., the Mudanfeng National Nature Reserve and the Fenglin National Nature Reserve), which provide a favorable natural environment with a strong ecological recovery capacity, leading to a predominance of slight and mild vulnerability. The western part of Heilongjiang Province, located in the western section of the Songnen Plain, is characterized by land use primarily dominated by arable land. This area experiences frequent human interference and a drier climate, leading to significant soil moisture loss. Consequently, there is a higher prevalence of saline and infertile soils, which reduces the region’s resistance to external disturbances, resulting in a higher degree of ecological vulnerability. Over the past two decades, the ecological vulnerability in Heilongjiang Province has generally followed a pattern of improvement followed by deterioration. The improvement observed from 2000 to 2010 may be closely related to the implementation of ecological restoration projects, such as the “Three-North Shelterbelt Program”, initiated in 1980, and the “Grain-for-Green Program” and wetland restoration initiatives, which began in 2000. During this period, Heilongjiang Province focused on promoting the development of light industry, while also undergoing economic restructuring and industrial upgrading. In addition, the implementation of the “Environmental Protection Regulations of Heilongjiang Province” contributed to improvements in environmental quality. From 2010 to 2020, ecological vulnerability in Heilongjiang Province predominantly shifted towards higher vulnerability levels, with areas of environmental degradation mainly concentrated in the Daxing’anling Area, Heihe City, and Mudanjiang City. This trend could be associated with the growth of the tourism industry [66,67]. These regions are rich in natural resources, with some natural attractions developed into tourist sites. The growth of tourism often leads to issues such as land reclamation and infrastructure development, which put pressure on the ecological environment and contribute to the exacerbation of ecological vulnerability. Meanwhile, with the acceleration of urbanization, the land use structure in Harbin has undergone significant changes [68], causing ecological vulnerability to rise and ecosystem stability to diminish. Using the GeoDetector to analyze the driving factors behind the spatial distribution of ecological vulnerability in Heilongjiang Province, the results indicate that both natural and human factors influence ecological vulnerability. However, natural factors are the primary drivers of the spatial distribution of ecological vulnerability, compared to human factors. This finding differs from previous studies, such as that by Zhang Huilin et al. [69], who identified population density and GDP as the primary drivers of ecological vulnerability in Shanxi Province. This may be related to the level of economic development; compared to more developed regions, Heilongjiang Province has a relatively lower level of economic development. As a result, population density and per capita GDP have a lesser impact on ecological vulnerability in the province. This study finds that biological abundance is the most significant factor in explaining ecological vulnerability in Heilongjiang. Biological abundance is influenced by land use types, with larger areas of forests, grasslands, and water bodies corresponding to higher levels of biological abundance. For example, the Daxing’anling region, with its high forest coverage and rich biodiversity, exhibits lower ecological vulnerability. The discovery not only provides an important basis for the ecological vulnerability assessment in Heilongjiang Province, but also a valuable reference for other regions to conduct ecological vulnerability research under similar environmental backgrounds. Compared to 2020, the CA-Markov model predicts an increase in ecological vulnerability by 2030. It is recommended that relevant authorities focus on areas with rising vulnerability, strengthen ecological protection measures tailored to local conditions, promote green development, and ensure regional ecological security.

Based on the ecological vulnerability characteristics of different regions, relevant authorities should implement scientifically sound and region-specific protection measures. For example, the Daxing’anling and Xiaoxing’an mountains, which are primarily slightly and mildly vulnerable, serve as the ecological source of Heilongjiang Province and should focus on reducing anthropogenic interference while maintaining existing protective policies. The western region of Heilongjiang Province, which is primarily characterized by severe and extreme ecological vulnerability, is densely populated and industrialized. Since PM2.5 is a key factor contributing to ecological vulnerability in the province, strengthening air pollution control should be a key focus, in line with the “Air Pollution Control Regulations of Heilongjiang Province” issued by the provincial government. Ecological vulnerability is not the result of a single factor alone; therefore, the government should also consider the interaction of various factors and implement integrated management measures to coordinate their effects. The ecological environment is closely linked to human life and socioeconomic development. Local governments should strengthen the promotion of environmental protection policies, deepen public awareness of environmental conservation, and make ecological protection a shared value, thereby achieving harmony between humans and nature.

This study has some limitations. At present, there is no standardized index system in place for evaluating ecological vulnerability. The influence of government policies, natural disasters (such as heavy rainfall, floods, droughts, etc.), and other human interventions has not been taken into account when selecting the indicators. Additionally, due to data availability constraints, the time series used in this study is relatively short, which may affect the precision of the ecological vulnerability assessment results. Therefore, future research should focus on developing a more comprehensive vulnerability assessment system that incorporates multiple factors.

5. Conclusions

(1)

The total ecological vulnerability of Heilongjiang Province remained at a moderate level between 2000 and 2020, showing a trend of first slowing down and then growing. The distribution was “high in the east and west, low in the north and south.” At the municipal level, Daxing’anling, Heihe, Mudanjiang, and Hegang were categorized as less than moderately vulnerable, but Shuangyashan, Qitaihe, Harbin, Jixi, and Jiamusi showed moderate ecological sensitivity. Suihua, Qiqihar, and Daqing, on the other hand, displayed ecological vulnerability levels that were greater than moderate. This finding provides a scientific foundation for regional ecological management and protection by highlighting the variations in ecological vulnerability across Heilongjiang Province’s various regions.

(2)

Heilongjiang Province’s ecological vulnerability has notable spatial clustering features, mostly consisting of low-low and high-high clustering regions. While the low-low clustering areas are mostly found in locations with good natural conditions, such as moderate and lightly sensitive zones, the high-high clustering areas are predominantly found in severely and extremely vulnerable regions with strong human activity. The regional distribution of ecological vulnerability during the relevant period is consistent with these tendencies. The temporal and spatial stability of vulnerability distribution is further supported by the spatial distribution characteristics, which align with the patterns of ecological vulnerability changes over time.

(3)

Numerous factors impact Heilongjiang Province’s ecological vulnerability’s spatial distribution characteristics. The four most significant factors affecting the study area are biological abundance, net primary productivity, dryness, and PM_2.5. Moreover, the highest interactions between these factors occurred in different years when they were combined. This highlights the importance of ecological complexity and the interaction of multiple factors, suggesting that future ecological protection efforts should consider the synergistic effects of various factors.

(4)

Ecological vulnerability can be predicted using the CA-Markov model. By 2030, Heilongjiang Province’s total ecological vulnerability is predicted to rise, with a greater proportion of regions having severe and extreme vulnerability and a decrease in the proportion of regions with low and moderate vulnerability. In addition to creating more focused and targeted ecological protection measures, relevant authorities should focus more on high-vulnerability areas, especially the extremely susceptible areas in the province’s west and south.