Identification and evaluation of land use vulnerability in a coal mining area under the coupled human-environment

2.3.1 Land use classification based on a support vector machine

A support vector machine (SVM) is a supervised non-parametric statistical learning technique, therefore, there is no assumption made on the underlying data distribution 30. Compared with other classification techniques, an SVM classifier is a powerful supervised classification method that can solve some practical problems, such as high-dimensional data, small sample sizes, nonlinearity, and so on, and has high accuracy and stability for the classification of objects 10. An SVM based on supervised classification separates multi-spectral feature data into discrete predefined clusters consistent with the training dataset in the remote sensing image classification application, so that the classification accuracy and stability are higher and the error is relatively small.

Let the training area for the study area be {xi,  yi}i=1N,where x_i ∈ Rⁿ represents the input mode and yi∈{−1,  1}represents the target output. Set the optimal decisionmaking equation: ω^T x_i + b = 0, then the weight vector ω and bias b to meet the constraints:

Where ξ is a slack variable under the condition of indivisible linearity, which indicates the degree of deviation of the pattern in the ideal linear case. From the classification of training samples required to minimize the error, we can deduce the following optimization formula:

where C is the penalty coefficient, which indicates that the SVM penalizes the misclassified samples and reflects the balance between the misclassification ratio and the complexity of the algorithm 31. Using the Lagrange multiplier method, the optimal decision surface can be transformed into the following optimization constraints:

where A is the Lagrange multiplier, but also to meet the constraints:

K(x, x_i) is a kernel function that satisfies the Mercer theorem. The kernel function has many forms, including linear, radial basis function (RBF), and polynomial, and by setting different kernel functions and parameters, its classification effect is different.

According to field research results, texture characteristics and spectral information combined with visual interpretation of Hegang coal mining land use types, using the SVM classification method, the landscape types of the study area are divided into eight categories, which include grassland, arable land, woodland, waters, residential land, coal land, unused land, and transport land.

The evaluation unit is a series of factors that affect the quality of the environment, which can reflect the difference of a certain quality of the environment. The commonly used evaluation units at home and abroad are planar-based vector units or point-based grid units. In this study, landscape-based vector surface units were used as evaluation carriers, and 63,047 patches were used as evaluation units to perform land use exposure, sensitivity, and adaptability evaluation and analysis of the comprehensive vulnerability of land from 1994 to 2014. Finally, the land use vulnerability in the study area is divided into five levels and the vulnerability of spatialization.

2.3.2 AHP to determine the weight of the vulnerability assessment index

AHP is a decision-aiding method developed by Saaty 32. The strength of AHP is that it organizes tangible and intangible factors in a systematic way, and provides a structured, yet relatively simple, solution to decision-making problems 33. The AHP system analysis should be based on a certain amount of information and people make judgments about the relative importance of each unit in each level. These judgments are expressed numerically by the introduction of a suitable scale, written as a judgment of the matrix:

where A is the discriminant matrix. For a_mn, element m and element n represent the importance comparison results (Table 2).

Having made all the pair-wise comparisons, the consistency is determined by using the eigenvalue, λ_max, to calculate the consistency index, CI.

where n is the matrix size. WhenCI < 0.1, the representative comparison matrix satisfies the consistency requirement and the assignment data is valid.

Luers 34 believed that ecosystems and the environment were affected to varying extents and frequencies by varying degrees of vulnerability, which were defined as exposures. Turner 35 used the HR and PAR models to explain the exposure and concluded that stress on the environment is a major exposure function. The sensitivity of land use under the system of human–earth coupling is manifested as the disastrous power of the disturbed regional ecosystem, which is positively correlated with the vulnerability. In the context of studies on global change and eco-environmental vulnerability, adaptive capacity is defined as the ability to adapt to, or reduce, systemic vulnerability 36 and is an ability to reduce the extent of vulnerability. Therefore, the combination of qualitative and quantitative methods, multi-dimensional analysis of the study area of land use vulnerability formation mechanisms, and characteristics, firstly, establish the target layer, criterion layer, element layer, and index layer from three aspects of the ecosystem’s exposure, sensitivity, and adaptability, and establish an accurate and complete evaluation system, and then use AHP and expert consultation to establish the weight (Table 3).

2.3.3 The evaluation factor of normalization

Before data analysis, we usually need to normalize the data first, and map the data to the range of 0-1, which makes the calculation easy and fast 37 Vulnerability is positively correlated with the inverse index and negatively correlated with the positive index.

The images of 1994, 2006, 2010, and 2014 obtained the slop, soil erosion, vegetation coverage, net primary productivity of vegetation, landscape fragmentation, and reciprocal fractal dimension data. Other data were obtained from the four-year statistical yearbook combined with field research data and other supporting data (Table 4).

2.3.4 GIS spatial analysis and vulnerability evaluation model

Spatial analysis is a spatial data analysis technique about the characteristics of geographical forms and the location of their objects. It can obtain important information, such as spatial location, shape, distribution, and evolution related to geographical objects from numerous spatial data 38. GIS can achieve the basic functions of spatial analysis, including spatial query and measurement, buffer analysis, overlay analysis, path analysis, spatial interpolation, and statistical classification analysis. In this study, the mathematical model was integrated with GIS spatial analysis to analyze the vulnerability of land use.

(1) Vulnerability evaluation model

According to the concept models and equations proposed by Metzger 39 and Turner 35, the landscape ecological principles and system theory are used to synthesize the indicators of multiple sub-system parameters. According to the vulnerability of the human–land coupling system in the study area, exposure, sensitivity, and adaptive capacity of the Hegang coal mine land use the vulnerability of three models:

In the formula:

V is the land use vulnerability;

E is the environmental exposure of land use;

S is the environmental sensitivity of land use;

AC is the environment adaptive capacity;

n, m, p are the number of participating factors of sensitivity, exposure, and adaptive capacity; and W_i, W_j, W_k are the weights of the factors of sensitivity, exposure, andadaptive capacity.

When V > 1, it indicates that the potential impact outside the environment is greater than the adaptive capacity, the vulnerability highlights, and the ecological environment

have deteriorating trends. When V < 1, the vulnerability of the environment is invisible and is in a trend of recoverability.

(2) Vulnerability index calculation model and vulnerability classification

The ecological vulnerability is a comprehensive manifestation of the vulnerability under the structure and function of the landscape type. The regional environmental vulnerability index can reflect the ecological vulnerability of the whole region. The calculation formula of the vulnerability index (RVI) is as follows 40:

A_i is the area of the ith landscape type in the area; A is the total size of the area.

According to the research on the existing results of vulnerability classification 41 the vulnerability values are divided into five grades according to the actual situation of land use vulnerability in the study area, and the vulnerability values are classified into regions with different degrees of vulnerability according to the grades (Table 5).

3.5.1 Integrated quantitative analysis of land use vulnerability

The Hegang coal mine has its own special complicated urban system and the vulnerability of human–ground coupling is obvious. Under the common interference of human factors and natural factors, the vulnerability of the ecological environment in the study area has become increasingly evident. Based on the economic system, social system, and eco-environment system, this paper constructs a vulnerability assessment system and mathematical calculation model of the Hegang coal mine area from the aspects of exposure, sensitivity, and adaptive capacity, the study area exposure index, sensitivity index, and adaptive capacity index calculation. The use of ARCGIS software overlays the analysis of the results and spatial statistics, with the results shown in Figure 5.

Figure 5

Calculation results of comprehensive vulnerability of the land use in the Hegang coal mining area.

The order of overall fragility of landscape types in the Hegang coal mine area from 1994 to 2014 is: unused land > coal land > waters > residential land > transport land > grassland > arable land > woodland, among them, the value of vulnerability of woodland is less than 1, others are all more than 1, this shows that under the coupling of humans and land, the overall ecological environment of the study area has less self-adaptation and adjustment capacity than the combined effect of external exposure and sensitivity, and the ecological environment is in a worsening situation. Although the vulnerability index of woodland shows an increasing trend to 1, and although the adaptive capacity is slightly larger than the overall capability of exposure and sensitivity, it has little effect on the environmental regulation in the whole area. The vulnerability of land use in coal mining areas is obviously prominent, the ecological vulnerability is serious, and the ecological environment is deteriorating.

3.5.2 Spatial differentiation analysis of land use vulnerability in coal mining areas

According to the integrated value of land use vulnerability in the study area, the fragmented area is divided according to the fragile value area, and the vulnerability grade distribution map is generated after spatialization (Figure 6).

Figure 6

Spatial distribution of land use vulnerability in the study area of Hegang from 1994–2014.

As can be seen from Figure 6, from the perspective of the spatial distribution of vulnerability, the coal mine in Hegang from 1994 to 2014 was mostly in the fragile zone because of the vast majority of forest land in the northern and central regions with relatively high adaptive capacity. The basic trend of vulnerability in the central urban area ranged from mild vulnerability in 1996 to a region of extreme vulnerability in 2001, and it was a severe vulnerability in 2014. This result is consistent with the overall vulnerability assessment data.

In terms of the change of time, compared with 1994, the area of mildly vulnerable areas decreased, the moderately vulnerable areas gradually concentrated, the extremely vulnerable areas expanded, and the overall vulnerability obviously moved toward the dominant trend. Compared with 2006, the extremely vulnerable areas spread to the periphery in 2010, but their area has been reduced. In 2014, compared with 2010, the extremely vulnerable areas in 2010 weakened and their vulnerability to moderately vulnerable areas increased. Most of them were converted into highly vulnerable areas and the mildly vulnerable areas were transformed into moderately vulnerable areas. On the whole, although the index of vulnerability decreased gradually from 1994 to 2014, it was still far greater than 1, and the vulnerability was obviously highlighted. The serious areas of vulnerability were mainly distributed in the southern urban area with a strong human disturbance.

From the perspective of urban sustainable development, there are some contradictions between the environment and the economy. How to increase the economy while ensuring the environment is not damaged is something that we should study. Due to the complexity of the causes of vulnerability, each type of land-use sensitivity, exposure, and ability to resist natural or man-made interference need to be adapted to meet the needs of sustainable development. Therefore, it is necessary to quickly find out ways to mitigate and control the increasing vulnerability of different landscape types, and it is also a direction that needs to be studied for a long time.

As a special city group, there has been little research on the coal mine city, especially the land use vulnerability in the coal mine area. With the deterioration of the internal structure and bearing capacity of the land use landscape, human disturbance gradually shows the phenomenon of vulnerability. In order to solve the problems caused by the increasing environmental damage, the related research of coal cities should be strengthened and corresponding ecological remedies should be taken. Therefore, in future research, more efforts should be made with respect to the coal mine city to deeply explore and reveal the interference effect of the coupling relation between humankind and the natural environment.

Hegang coal mine land use shows the most prominent vulnerability, with the surrounding environment and natural factors being greatly affected by human factors. Therefore, ecological restoration should be taken as a priority strategy to utilize the means of vegetation restoration, soil quality improvement, and natural landscape maintenance in the process of natural evolution to give priority to improving the adaptability of natural ecosystems. Different landscape types have different security patterns and play a decisive role in their stability. They can effectively protect the landscape security pattern in order to maintain a healthy ecological landscape and the ecological environment in areas with fragile landscape restoration.