Study on the Evolution Law of Surface Landscape Pattern in Earthquake-Stricken Areas by Remote Sensing: A Case Study of Jiuzhaigou County, Sichuan Province

Abstract

Earthquakes not only cause the destruction of surface buildings and a large number of casualties, but also have an important impact on regional land-use change. Timely understanding of land cover and its changes before and after earthquakes is of great scientific significance for studying the restoration and reconstruction of the natural ecosystem and the sustainable development of the social economy in disaster areas. At present, there are few studies on land cover changes before and after earthquakes in the earthquake-stricken areas, especially the quantitative assessment of land cover changes before and after earthquakes at the micro scale of landscape spatial distribution. Therefore, this article is based on remote sensing data and uses an earthquake in Jiuzhaigou, Sichuan, as an example. By calculating the land-use degree comprehensive index and its rate of change, the transfer matrix of land-use, quantitative expression methods of the landscape via the landscape pattern index are analyzed from the points of view of the spatial–temporal evolution law of the landscape pattern of land cover before and after the earthquake. The experimental results showed that the area of forest, cropland, shrub. and grassland types in the study area changed greatly, and the area of grassland types in Shuanghe Township changed most obviously, as they increased from 6.57 km² in 2015 to 7.43 km², and then decreased to 6.21 km² in 2020. During the earthquake recovery period, the comprehensive index of land-use degree for most towns in the study area showed an upward trend, and the land cover types were improved. After 2017, the PD and LSI of the study area showed a downward trend, among which the bare land type changed greatly, with the PD value decreasing from 0.38 in 2017 to 0.21 in 2020, and the LSI value decreasing from 58.29 in 2017 to 40.69 in 2020. This indicates that the degree of landscape fragmentation and the spatial heterogeneity of the land surface in the study area are reduced in the later stage of earthquake recovery. After 2017, the AI value of the study area showed an upward trend, and the AI value of the bare land type increased from 63.97 in 2017 to 78.25 in 2020, indicating that the connectivity between landscape patches in the study area was enhanced in the later stage of the earthquake. This also reflects that the ecological environment of the study area gradually showed a good trend in the later stage of the earthquake. The results of this paper not only reveal the spatial–temporal evolution of land cover in Jiuzhaigou County before and after the earthquake, but also help the relevant national decision-making departments to formulate corresponding policies and measures.

1. Introduction

The deterioration the of ecological environment caused by sudden natural disasters has become a global problem and has attracted more and more attention from scientists [1,2]. Earthquakes, as one of the highly destructive sudden natural disasters, not only cause a large number of casualties and huge economic losses, but also cause great damage to the land cover and ecosystem [3,4,5]. For example, (1) earthquakes can cause landslides and lead to changes in vegetation on mountains, (2) earthquakes can block rivers, leading to the formation of dammed lakes, causing floods and other changes in water resources, and (3) earthquake makes geological structures change and rocks fracture, induced by heavy rain. Thus, dissolution is enhanced, making it easier for collapses to occur, resulting in the collapse of surface buildings and other phenomena. However, land cover change objectively records the spatial–temporal dynamic process of the surface landscape before and after the earthquake [6,7,8,9,10]. Therefore, the spatial–temporal evolution of land cover can be used to dynamically monitor the post-earthquake ecosystem recovery. It is of great scientific significance to study the restoration and reconstruction of the natural ecosystem and the sustainable development of the social economy in disaster areas [11,12,13].

Remote sensing technology, as an important part of contemporary high and new technology, is the main means to obtain land-use and land cover change information due to its large monitoring range, accurate and real-time data, intuitive and convenient use, and due to the fact that it can cope with various complex environments [14,15,16]. Remote sensing technology has been widely used in ecology, geoscience, catastrophology, and other fields [17,18,19]. Remote sensing images can objectively record the spatial–temporal dynamic process of land cover before and after an earthquake, which is an important data source for earthquake disaster prediction, disaster situation and intensity assessment, and post-disaster recovery and reconstruction [20,21,22,23].

At present, scholars at home and abroad have carried out related research work on land cover change caused by earthquakes and their secondary disasters from different angles [24,25]. For example, Wang et al. [26] took the Wenchuan earthquake as an example, fully considered the dynamic process of vegetation restoration, analyzed and compared the changes in the normalized difference vegetation index (NDVI) before and after the earthquake, and defined the COI index to identify areas with difficulty in vegetation restoration. Odriguez et al. [27] used remote sensing images to monitor hydrological changes in earthquake-stricken areas, linked vegetation, hydrology and agriculture, analyzed the relationship between the three before and after the earthquake, and summarized the methods to assess the impact of the earthquake on hydrology and agriculture, so as to improve the management of water resources in earthquake-stricken areas. Motamed et al. [28] worked from the perspective of spatial optimization management of land cover and fully considered the impact of earthquake disasters on land and nature, and used the improved mixed integer quadratic programming model to carry out reasonable planning of land cover types according to dangerous conditions. Chen et al. [29], using Sentinel-2 remote sensing images, took Jiuzhaigou in Sichuan Province as an example to study the changes in vegetation coverage before and after earthquakes in this area. The results show that the proportion of high vegetation cover area decreases, while the proportion of medium and low vegetation cover areas increases. Xiang et al. Additionally, [30] took Beichuan County, which was most severely affected by the Wenchuan earthquake, as the research area. Based on remote sensing and GIS technology, the binary pixel model and vegetation cover transfer matrix were used to explore and analyze the vegetation cover in the disaster area. The results showed that the interaction of elevation, temperature, and rainfall significantly increased the spatial differentiation of fractional vegetation cover, which indicated that fractional vegetation cover was the result of multiple factors. Li et al. [24] discussed and analyzed the impact assessment of far-field earthquakes in Taiwan on buildings in Southeast China. In this study, 43.29 million mobile phone location records collected after the Ms6.1 earthquake in Hualien, Taiwan Province, on 18 April 2021 were used to quantitatively analyze the density of 331 buildings using communication check-in data at different spatial and temporal scales after the earthquake, aiming at the bandwidth selection in kernel density estimation in different regions. Pandey et al. [31] analyzed the evaluation of vegetation loss and restoration caused by the Gorkha earthquake (7.8 Mw) in 2015. Remote sensing images from 2015 to 2021 were used to calculate vegetation restoration rates and public finance analysis data were used to compare the effects of artificial and self-ecological restoration. Experiments found that during the Gorkha earthquake, landslides destroyed about 8651.58 hectares of vegetation and, so far, about 4442 hectares of vegetation has been restored.

However, existing studies mainly focus on the dynamic monitoring and analysis of a single land cover type in the earthquake area, such as a single vegetation type and building type. As such, there is a lack of comprehensive dynamic monitoring research on the total factor land cover types in earthquake-stricken areas, especially in terms of the quantitative analysis of the changes in land surface landscape patterns before and after the earthquake at the micro scale.

Therefore, the purpose of this study is to take Jiuzhaigou County in Sichuan Province as the research area, guided by the relevant knowledge of ecology, geosciences, and catastrophology, and based on remote sensing image data, we studied the dynamic changes in the land cover landscape pattern in Jiuzhaigou County after the earthquake, and revealed the impact of the earthquake and secondary disasters on the changes in the land surface landscape pattern in the study area. The results not only facilitate the relevant government departments to achieve a timely and effectively grasp on the land cover changes in the disaster areas before and after the earthquake, but also provide important scientific basis for guiding the recovery and reconstruction of the disaster areas and formulating medium- and long-term ecological restoration plans.

2. Materials

2.1. Study Area

Jiuzhaigou County (32°54′ N~33°19′ N, 103°46′ E~104°4′ E) is located in the Aba Tibetan and Qiang Autonomous Prefecture, Sichuan Province, China, bordering the Wenxian, Zhouqu, and Dibu counties of Gansu Province. The south and west of the study area are bordered by Pingwu, Songpan, and Zoige of Sichuan Province, with a total area of about 5290 km². The valley of Jiuzhaigou County is vertical and horizontal, and the terrain is high in the northwest and low in the southeast. It is mainly dominated by mountains, and some mountains are plain and flat dams. The altitude difference is up to 2000 m. Jiuzhaigou is located in the transition zone from the Qinghai–Tibet Plateau to the Sichuan Basin, with a complex geological background and strong neotectonic movement. According to altitude, the study area is divided into warm temperate semi-arid, middle temperate, and cold temperate monsoon climates. The average annual temperature is 12.7 °C and the average annual precipitation is 550 mm. At 21:19 on 8 August 2017, a 7.0 magnitude earthquake occurred in Jiuzhaigou County, Aba Prefecture, Sichuan Province (33.20 N, 103.82 E), with a focal depth of 20 km. The maximum intensity of the earthquake was nine degrees. The geographical location of the study area is shown in Figure 1.

2.2. Data and Preprocessing

The land cover data used in this study came from the annual land cover data of 30 m in China produced by Professor Huang Xin’s team at Wuhan University [32]. This data collected the training samples by combining stable samples extracted from China’s land-use/cover datasets (CLUDs) and visually interpreted samples from satellite time-series data, Google Earth, and Google Maps. Using 335,709 Landsat images on the GEE, several temporal metrics were constructed and fed to the random forest classifier to obtain classification results. Considering the land cover distribution in China [33], this data defined a classification system including nine major LCs, as follows: cropland, forest, shrub, grassland, water, snow and ice, barren, impervious, and wetland. This classification system is similar to that of FROM_GLC [34], and can be conveniently remapped to the FAO (Food and Agriculture Organization) and IGBP systems. After obtaining three periods of land cover data, namely pre-earthquake (2015), earthquake year (2017), and earthquake recovery period (2020), WGS UTM projection was used as the reference frame for the analysis in this paper. Figure 2 shows the land cover data of each period covering the study area.

A confusion matrix is a commonly used method to evaluate the classification accuracy of remote sensing images, which expresses the classification results of sample data and the comparison results of actual ground classes with an n × n dimensional matrix [35,36,37]. The columns in the confusion matrix represent the actual category and the rows represent the predicted category. In this paper, the overall accuracy (OA) and Kappa coefficient were used to evaluate the accuracy [38]. Furthermore, sample points for accuracy verification of land cover data in each period were collected based on Google Earth (Figure 3 is the sample points collected in 2020), and then the confusion matrix was used to calculate the overall accuracy and Kappa coefficient of land cover data in 2015, 2017, and 2020, respectively. Thus, the accuracy of land cover data is evaluated (

Table 1. Accuracy evaluation results of land cover data of 2015.

	Reference Data
Evaluation Data		Cropland	Forest	Shrub	Grassland	Water	Snow/Ice	Bare land	Built-up	Wetland	UA(%)	CE(%)
	Cropland	30	1	1	2			1	1		83.33	16.67
	Forest		81	7	4			1			87.10	12.90
	Shrub	2	7	43	7						72.88	27.12
	Grassland	4	8	1	55	1	2		2	1	74.32	25.68
	Water			1		22					95.65	4.35
	Snow/Ice				1		7				87.50	12.50
	Bare land	1			1			11	2		73.33	26.67
	Built-up			2		1		1	39		90.70	9.30
	Wetland				1					4	80.00	20.00
	PA(%)	81.08	83.51	78.18	77.46	91.67	77.78	78.57	88.64	80.00
	OE(%)	18.92	16.49	21.82	22.54	8.33	22.22	21.43	11.36	20.00
	OA(%)	82.02

). Table 1,

Table 2. Accuracy evaluation results of land cover data of 2017.

	Reference Data
Evaluation Data		Cropland	Forest	Shrub	Grassland	Water	Snow/Ice	Bare land	Built-up	Wetland	UA(%)	CE(%)
	Cropland	36	2	1	2			1	1		83.72	16.28
	Forest		107	7	5	3		2		1	85.60	14.40
	Shrub	2	7	37	4						74.00	26.00
	Grassland	3	7	4	52	2	2		1		73.24	26.76
	Water		1			27			1		93.10	6.90
	Snow/Ice				1		7				87.50	12.50
	Bare land	1						10	2		76.92	23.08
	Built-up				2				48		96.00	4.00
	Wetland				1					5	83.33	16.67
	PA(%)	85.71	86.29	75.51	77.61	84.38	77.78	76.92	90.57	83.33
	OE(%)	14.29	13.71	24.49	22.39	15.63	22.22	23.08	9.43	16.67
	OA(%)	83.29

and

Table 3. Accuracy evaluation results of land cover data of 2020.

	Reference Data
Evaluation Data		Cropland	Forest	Shrub	Grassland	Water	Snow/Ice	Bare land	Built-up	Wetland	UA(%)	CE(%)
	Cropland	30	3		1			1	2		81.08	18.92
	Forest		92	8	2	1					89.32	10.68
	Shrub		7	44	5						78.57	21.43
	Grassland	5	7	2	62		3		1		77.50	22.50
	Water					30					100.00	0.00
	Snow/Ice				2		9				81.82	18.18
	Bare land	1			1			14	2		77.78	22.22
	Built-up		1					0	56		98.25	1.75
	Wetland					1				4	80.00	20.00
	PA(%)	83.33	83.64	81.48	84.93	93.75	75.00	93.33	91.80	100.00
	OE(%)	16.67	16.36	18.52	15.07	6.25	25.00	6.67	8.20	0.00
	OA(%)	85.89

show that the overall accuracy of the land cover data of the three periods is higher than 82%. According to the classification accuracy relation corresponding to different quantity disagreement and allocation disagreement values [39] (Figure 4, Figure 5 and Figure 6), the land cover data of the three periods obtained in this paper have good accuracy and can meet the research needs of this paper.

3. Methods

3.1. Comprehensive Index of Land-Use Degree

The degree of land-use can not only reflect the natural attributes of land itself, but also reflect the comprehensive effect of the interaction between human activities and the natural environment. It includes not only the change in land-use mode, but also the change in land-use quantity [40,41]. The degree of land-use can be evaluated by the comprehensive index of land-use degree and its change rate [42]. This index can reflect the degree of land-use in a specific period. On the other hand, the change rate of this index can reflect the change in the degree of land-use. The formula for calculating the comprehensive index of land-use degree is as follows:

$$I_a=100\times {\sum }_{i=1}^n{A}_i\times C_i$$

(1)

where $I_a$ represents the land-use degree value of the study area in a certain period, $A_i$ represents the grading index of the level $i$ land-use degree in the study area, and $C_i$ represents the percentage of class $i$ land area in the total land area of the study area.

$$R=\frac{{\sum }_{i=1}^n(A_i\times C_{ib})-{\sum }_{i=1}^n(A_i\times C_{ia})}{{\sum }_{i=1}^n(A_i\times C_{ia})}\times 100$$

(2)

where $R$ represents the rate of change in land-use degree, and $C_{ib}$ and $C_{ia}$ represent the area percentage of the study area in the level $i$ land-use degree at time b and time a, respectively.

The key problem of applying the comprehensive index of land-use degree lies in the classification of land-use degree and the setting of the classification index. The grading of land cover is assigned with reference to the grading of land cover by Huang et al. [43], using a comprehensive analysis based on the natural stability state of the land system under the action of various other conditions, as shown in

Table 4. Land-use degree grading assignment.

Land Cover Type	Bare land, Snow/Ice	Forest, Grassland, Shrub, Water	Cropland	Built-Up
Grading index	1	2	3	4

.

3.2. Land-Use Transfer Matrix

The land-use transfer matrix not only includes the data of the local class area at a certain time point in a certain region, but also contains the information of the transfer of the local class area at the initial stage and the transfer of the local class area at the final stage. It reflects the dynamic process of the mutual transformation between the local class area at the beginning and the end of a period in a certain region [44]. Therefore, the land-use transfer matrix can be used to visually express the number of local classes transferred under the earthquake disturbance, which is of great significance to research into ecological environment restoration in post-earthquake disaster areas. The general form of the land-use transfer matrix is as follows:

$$P_{ij}=\left\{\begin{array}{c}{P}_{11} P_{12 }{P}_{13} \cdots P_{1n}\\ P_{21} P_{22 }{P}_{23} \cdots P_{2n}\\ \cdots \cdots \cdots \cdots \cdots \\ P_{n1} P_{n2 }{P}_{n3} \cdots P_{nn}\end{array}\right\} \left(i,j=1,2,3,\cdots ,n\right)$$

(3)

where $S_{ij}$ represents the area converted from the class $i$ land cover type to the class $j$ land cover type before transfer, $n$ represents the number of land cover types before and after the transfer, and $i$ and $j$ represent the land cover types before and after the transfer, respectively.

3.3. Landscape Ecological Index

The landscape index refers to a quantitative index that can reflect its landscape structure and spatial characteristics. It plays an important role in describing landscape patterns, establishing landscape connections, and explaining landscape functions. At present, it is widely used to quantitatively analyze landscape patterns and dynamic changes in land cover data [45]. After the earthquake, the landscape pattern was changed due to the impact of the earthquake and its secondary geological disasters (natural factors), and the human disturbance in post-disaster reconstruction (human factors) [46]. Due to the relative complexity of the regional landscape and the correlation between different landscape pattern indices, many studies select some redundancy or similarity in landscape pattern indices in the process of landscape pattern analysis, which cannot accurately reflect the spatial heterogeneity and diversity of the landscape. Therefore, on the premise of a comprehensive understanding of the ecological significance represented by each landscape pattern index and its reflected landscape structure, based on relevant studies [47] and considering the characteristics of the study area, this paper selected PD as a representative index of landscape spatial heterogeneity. The LSI was used as a representative indicator of shape complexity, and AI was used as a representative indicator of patch connectivity. The calculation and ecological significance of each landscape index are as follows [48]:

(1) The PD. The PD refers to the number of patches per unit area, and the greater the value, the greater the degree of fragmentation of the landscape in the region. The calculation formula of PD is as follows:

$$PD=\frac{N}{A}, PD>0$$

(4)

where $PD$ refers to patch density, $N$ refers to the total number of patches in the landscape, and $A$ refers to the total landscape area;

(2) The LSI. The LSI is the ratio of patch perimeter to square perimeter of an equal area. This index is used to describe the degree of irregularity or deviation of landscape patch shape from a square, which can reflect the regularity degree of a certain landscape type. The larger the value, the larger the difference between the plaque shape and the square, and the more irregular the shape. The calculation formula of LSI is as follows:

$$LSI=\frac{0.25E}{\sqrt{A}}, LSI\ge 1$$

(5)

where $E$ refers to the total length of all patch boundaries in the landscape, while $A$ refers to the total landscape area;

(3) The AI. The AI represents the aggregation degree of the same type of patches in the landscape area, and the value range is (0, 100). A small AI value indicates that patches in the landscape are discrete and fragmented. On the contrary, a large AI value indicates that patches of the same type in the landscape converge with each other and have a compact structure. The calculation formula of AI is as follows:

$$AI=\left[\frac{g_{ii}}{\max g_{ii}}\right]\times 100, 0\le AI\le 100$$

(6)

where, $g_{ii}$ refers to the number of similar connections between pixels of patch type $i$.

4. Results

4.1. Land Cover Type Structure

Figure 7 shows the spatial pattern distribution of land cover changes in each township in the study area before and after the earthquake. The areas with large changes in cropland types before and after the earthquake are mainly distributed in the eastern part of the study area. From 2015 to 2020, the cropland area of Guoyuan Township, Anle Township, and Baihe Township increased year by year, and the cropland area of Yonghe Township, Baohua Township, and Shuanghe Township, under the comprehensive impact of earthquake and human activities, decreased slightly in 2017, but also showed an increasing trend. Before and after the earthquake, forest area in the study area changed little, and the distribution of forest types was greater in the west and lesser in the east. Zhangzha Township and Dalu Township were the two areas with the largest distribution, accounting for 21.6% and 20.4% respectively. The earthquake had a great impact on the shrub types in Dalu Township, and the shrub area in this region increased from 9.31 km² in 2015 to 12.02 km² in 2017, and then decreased to 10.96 km² in 2020. In terms of grassland type, 12 of the 17 townships in the study area showed a trend of first increasing and then decreasing under the comprehensive impact of earthquake and human activities. The bare land type is mainly distributed in Zhangzha Town, and the bare land area decreases first and then increases before and after the earthquake. From 2015 to 2020, the built-up area in the study area showed an overall growth trend, which may be caused by the rapid economic development in the study area and the large number of residents in need of resettlement after the earthquake. The wetland type was mainly distributed in Zhangzha Town in the study area, and the area of the wetland type showed a decreasing trend before and after the earthquake.

4.2. Land-Use Degree Analysis

Figure 8 shows the spatial pattern distribution of land-use degree before and after the earthquake at the township scale in the study area. The comprehensive index of land-use in the three periods did not change much, and the values were all around 200. The main reason was that the study area had lush vegetation and a high proportion of forest types, among which the forest area of Caodi Township and Wujiao Township accounted for about 94%.

From 2015 to 2017 (

Table 5. Index of land-use degree during different periods at township scale.

Name	Composite Index			Composite Index Rate (%)
	2015	2017	2020	2015–2017	2017–2020
Yongfeng	206.97	206.70	206.34	−0.13	−0.18
Guoyuan	209.36	209.58	211.88	0.10	1.09
Luoyi	219.30	219.21	218.41	−0.04	−0.37
Anle	205.32	205.76	206.91	0.21	0.56
Caodi	203.66	203.51	203.72	−0.07	0.10
Majia	199.50	200.00	200.04	0.25	0.02
Yongle	219.39	219.53	220.62	0.06	0.50
Zhangzha	197.34	198.95	198.52	0.81	−0.22
Baihe	202.41	202.54	202.81	0.06	0.13
Dalu	200.15	200.12	200.09	−0.01	−0.02
Heihe	200.93	201.07	201.07	0.07	0.00
Yonghe	207.37	206.42	212.03	−0.46	2.72
Baohua	226.46	225.50	228.42	−0.42	1.29
Wujiao	201.65	201.60	201.51	−0.03	−0.04
Shuanghe	208.41	208.15	209.01	−0.12	0.41
Lingjiang	200.42	200.55	200.54	0.07	−0.01
Yuwa	200.56	200.54	200.52	−0.01	−0.01

), the comprehensive index of land-use degree of most townships in the southeast of the study area showed different degrees of decline, and the greatest degree of decline was in Yonghe Township, with a change rate of −0.46%. From 2017 to 2020 (Table 5), most of the land-use degree indices of the townships in the study area recovered, and the areas with a large increase included Yonghe Township, Guoyuan Township, Anle Township, and Yongle Town. The change rates were 2.72%, 1.09%, 0.56% and 0.50%, respectively. The reasons for the above phenomena include the following. First, the southeast part of the study area has convenient transportation, frequent human activities, and a good economic development level, which is greatly affected by the earthquake. Second, after the earthquake, due to the direct destruction of the earthquake and geological disasters, part of the cropland in the region was converted into shrubs or grasslands, which led to the reduction of land use degree. Third, the southeastern part of the study area has a high population density and strong human reconstruction activities. With the promotion of post-disaster reconstruction and the drive of relevant land policies of the government, the built-up area increased due to the large number of residents’ resettlement after the earthquake and the rapid restoration of cropland damaged by the earthquake, so the land use degree index rose rapidly.

4.3. Analysis of Land-Use Intertransfer

Table 6. Matrix of land-use transfer of Jiuzhaigou from 2015 to 2017.

2015 Land Cover Type	2017 Land Cover Type
	Cropland	Forest	Shrub	Grassland	Water	Snow/Ice	Bare land	Built-Up	Wetland
Cropland	93.3021	0.6237	0.1269	6.3729	0.0144	0	0	0.0162	0
Forest	3.5874	3416.2695	5.9958	0	0	0	0	0.0036	0
Shrub	0.6444	0.702	32.8833	3.1968	0	0	0	0	0
Grassland	2.5173	2.4183	0.8316	1662.0669	1.1304	0	4.0833	0.0189	0
Water	0.0072	0.0117	0	0.594	3.3507	0	0.0279	0	0
Snow/Ice	0	0	0	0.1089	0.0108	2.4102	0.2061	0	0
Bare land	0.0018	0	0	4.1823	0.0414	1.6704	18.801	0	0
Built-up	0	0	0	0	0.0009	0	0	2.3823	0
Wetland	0.0009	0	0	0	0	0	0	0	0.0063

and

Table 7. Matrix of land-use transfer of Jiuzhaigou from 2017 to 2020.

2017 Land Cover Type	2020 Land Cover Type
	Cropland	Forest	Shrub	Grassland	Water	Snow/Ice	Bare land	Built-Up	Wetland
Cropland	91.0512	3.5739	0.4338	4.9023	0.0279	0	0	0.072	0
Forest	3.2157	3411.8919	4.9086	0	0	0	0	0.009	0
Shrub	0.8784	4.9203	31.1526	2.8908	0	0	0	0	0
Grassland	12.8007	6.8805	2.1744	1641.2742	1.5372	0	11.8116	0.0432	0
Water	0.0027	0.0054	0	0.8892	3.2013	0	0.4482	0.0018	0
Snow/Ice	0	0	0	0.234	0.0189	1.3032	2.5245	0	0
Bare land	0	0	0	7.1019	0.0504	0.3573	15.6087	0	0
Built-up	0	0	0	0	0	0	0	2.421	0
Wetland	0	0	0	0.0036	0	0	0	0	0.0027

show the mutual transfer of land cover types before and after the earthquake in the study area from 2015 to 2017 and 2017–2020, respectively. It can be seen from Table 3 that, during 2015–2017, all the table cover types in the study area were transferred and transferred out, and the transfer amount of cropland to grassland was the largest, with a transfer area of 6.37 km². The transfer of forest land type to shrub type was also obvious, and the transfer area was 6.00 km². Combined with Figure 7, it can be seen that the area of the cropland type in Yonghe Township was transferred out more, while that of the forest in Yuwa Township was transferred out more, which may be due to the damage of cropland and forest in these areas caused by the earthquake and secondary disasters.

It can be seen from Table 7 that, during 2017–2020, with the ecological restoration and artificial restoration over time in the study area, a large number of grassland types were transferred out, and the area of grassland converted to cropland and forest was 12.80 km² and 6.88 km², respectively. The results show that, after the earthquake, under the artificial repair and reconstruction, the earthquake-damaged farmland is restored again. In addition, part of the grassland has been turned into forest, indicating that the ecological environment damaged by the earthquake is gradually being repaired.

4.4. Analysis of Landscape Pattern Evolution

4.4.1. Temporal Variation Characteristic

Figure 9, Figure 10 and Figure 11 show the temporal variation trend of landscape PD, LSI, and AI indexes before and after the earthquake. Compared with 2015, the PD of cropland, forest, shrub, snow, bare land, and built-up land in the study area showed an increasing trend after the earthquake. From 2017 to 2020, except for the PD of the built-up type of land, the PD of other land types in the study area showed a downward trend. Compared with 2015, except for the LSI of wetland, the LSI of other landscape types in the study area showed an upward trend after the earthquake. From 2017 to 2020, except for the built-up type, the LSI of other land types showed a downward trend. Compared with 2015, the AI of cropland, forest, grassland, and bare land in the study area showed a downward trend after the earthquake. From 2017 to 2020, the AI of cropland, forest, grassland, and bare land in the study area showed an increasing trend.

In general, before and after the earthquake, the PD and LSI of the ground class in the study area increased first and then decreased, and the AI decreased first and then increased. The results showed that the earthquake and secondary geological disasters caused serious landscape fragmentation, complex landscape shapes, and reduced connectivity among landscape patches. In the three years after the earthquake, with the gradual restoration of the ecological environment, post-disaster reconstruction and the influence of relevant government policies, the degree of fragmentation of the surface landscape in the study area was reduced, the shape of the landscape became regularized, and the connectivity between landscape patches was enhanced.

4.4.2. Spatial Variation Characteristic

Figure 12, Figure 13 and Figure 14 show the spatial pattern change and distribution of the PD index, LSI index, and AI index before and after the earthquake. The southeast of the study area and the surrounding Heihe and Baihe River basins had higher PD and LSI values, but lower AI values. From the perspective of spatial pattern distribution, the PD, LSI, and AI values changed little before and after the earthquake and were mainly concentrated in the southeast of the study area and around the Heihe and Baihe River basins. From 2015 to 2020, the higher values of PD and LSI in the south of Zhangzha Town showed a trend of first increasing and then decreasing, while the lower values of AI showed a trend of first decreasing and then increasing.

The main reasons for the above phenomena before and after the earthquake are as follows. Firstly, the surface landscape in the southeast of the study area and around the Heihe and Baihe River basins is complicated and greatly disturbed by human activities. Secondly, the ground shaking caused by seismic shear waves will directly lead to the destruction of buildings, and the loose materials generated by earthquakes in areas greatly affected by human reconstruction activities are more likely to cause landslides, debris flow, collapse, and other geological disasters, thus, leading to the change in surface cover (such as forests, water, etc.).

4.5. Analysis on the Influence of Natural Factors on Land Cover Change in Earthquake Stricken Areas

Based on remote sensing data, this paper analyzed the spatial–temporal evolution of land cover in Jiuzhaigou County before and after the earthquake by using the methods of land-use degree change, the land-use transfer matrix, and the landscape ecological index. The experimental results show that land cover in Jiuzhaigou County exhibits destructive deformation at both macroscopic and microscopic scales.

This paper draws on the existing research results [49,50,51]. Taking the study area of the Tazang fault zone as the center, the change rules of land cover before and after the earthquake were quantitatively expressed in the 3 km, 6 km, 9 km, and 12 km regions (Figure 15). The results showed (Figure 16) that the land cover change area within 0–3 km, 3–6 km, 6–9 km, and 9–12 km of the Tazang fault zone accounted for 9.16%, 9.15%, 10.91% and 9.14% of the total land cover change area, respectively, mainly distributed in the southeast of the study area. The results show that, under the influence of the Tazang fault zone, the closer it is to the Tazang fault, the greater the land cover change and the more serious the land damage, except for at the range of 6–9 km. In the range of 6–9 km, human activities are frequent, and the surface landscape is more complex, resulting in greater land cover change and more serious land destruction.

In conclusion, the fault zone is an important natural factor affecting land cover change in the earthquake-stricken area of Jiuzhaigou. Our experiments show that the closer the fault is to the surface, the more serious the earthquake damage is. Therefore, we should pay more attention to the activity status, scale, and temporal and spatial distribution characteristics of the fault zone in the future, so as not to cause great changes in the surrounding land cover.

5. Discussion

5.1. Analysis on the Influence of Human Factors on Land Cover Change in Earthquake Stricken Areas

Government policy is an important factor affecting land cover change [52,53]. After the Jiuzhaigou earthquake, the Ministry of Land and Resources and the People’s Government of Sichuan Province issued a series of relevant policies to support the restoration and reconstruction of Jiuzhaigou, thus, affecting the land cover change in Jiuzhaigou. The impact of government policies on land cover change is mainly reflected in the following aspects:

(1)

Land policies of government departments directly affect land cover change. After the Jiuzhaigou earthquake, the government departments introduced land policies, such as post-disaster recovery and reconstruction of easy land, re-optimization and adjustment of basic farmland, and reclamation and utilization of abandoned industrial and mining land in the disaster area, which will lead to a change in land cover structure. According to the experimental results in Part 4.1 of this paper, from 2015 to 2020, the land cover types of all villages and towns in Jiuzhaigou have changed. For example, before and after the earthquake, the water area in Majia Township of Jiuzhaigou changed significantly and kept increasing from 0.43 km² in 2015 to 1.06 km² in 2020;

(2)

Ecological restoration and protection policies are also important factors affecting land cover change. After the Jiuzhaigou earthquake, the government issued a policy to include eligible arable land on steep slopes in the quake-hit areas in a new round of conversion projects, which will lead to the conversion of some arable land into forest land. For example, it can be seen from the experimental results in Section 4.3 of this paper that during 2015–2017, the transfer of cropland land type to grassland type was the largest, with a transfer area of 6.37 km²;

(3)

As the first nature reserve in China whose main purpose is to protect natural scenery, policies related to scenic spot restoration and industrial development are also influencing land cover change in Jiuzhaigou. After the Jiuzhaigou earthquake, the government departments were required to support the revitalization and development of tourism in the disaster area, to support the development of modern agriculture in the disaster area, and to support the transformation and upgrading of the industry in the disaster area, green development, and other policies. For example, according to the spatial change in land-use degree calculated in this paper, from 2017 to 2020, the growth trend of land-use degree in Yonghe Township, Guoyuan Township, Anle Township, and Yongle Town in the study area was obvious. Yonghe Township, in particular, had a rate of change as high as 2.72%;

(4)

Infrastructure policies also affect land cover change. After the earthquake in Jiuzhaigou, the government departments also introduced relevant infrastructure construction policies, such as constructing the section of road from the main temple of the national highway 544 to the county seat of Jiuzhaigou in order to build a safe, smart and green eco-tourism new demonstration road. The construction of built-up land will make parts of cropland, forest, grassland, and other types of land shift to built-up types. During 2017–2020, the area of the study area changed from other land types to built-up was 0.12 km².

In addition, social and economic development is also an important factor affecting land cover change. Economic development is accompanied by the growth of construction land and cropland, so it will inevitably affect the change in land cover [54]. After the earthquake, according to the “8·8” Jiuzhaigou Earthquake Recovery and Reconstruction Master Plan, Jiuzhaigou County actively promoted the recovery and reconstruction work after the earthquake, actively developed the ecological industry, and gradually made the traditional agriculture more ecologically sound. In 2019, the per capita disposable income of urban residents and rural residents in Jiuzhaigou County was 35,475 yuan and 14,254 yuan, respectively, 1.26 times and 1.32 times of that in 2016. The income of urban and rural residents both exceeded the levels before the earthquake. Making full use of the opportunity of post-earthquake reconstruction in Jiuzhaigou and optimizing the spatial layout of land cover in earthquake-stricken areas according to the “8·8” Jiuzhaigou Earthquake Recovery and Reconstruction Master Plan will have long-term social significance for promoting the sustainable development of land cover in earthquake-stricken areas.

To sum up, government policies and regulations, as well as social and economic development, comprehensively affected the land cover change in the earthquake-stricken area of Jiuzhaigou. Some constructive suggestions are given as follows. First, more strict monitoring should be carried out on the Jiuzhaigou nature reserve, which has great influence. Second, the vegetation of Jiuzhaigou was investigated and restored, and the forest’s self-regulation ability and artificial assistance were used to restore the ecological environment. Third, the efforts to return farmland to forest should be strengthened. The earthquake damaged the ecosystem, so the conversion of farmland to forest can be used to restore the function of vegetation and water retention in the whole region to control soil erosion. Fourth, we will adjust and improve the overall plan for land-use and optimize the scale, structure, and distribution of all types of land-use.

5.2. The Results of This Study Were Compared with Those of Other Literatures

Table 8. Related research on land cover change in earthquake-stricken areas.

Source	Region	Method	Conclusion
Peng et al. [55]	Lushan, Southwest China	landscape ecology method, GIS technology	It is concluded that there exists a positive correlation between the land-use change in the earthquake-stricken areas and the intensity of human disturbance, and between the change in the landscape pattern index and the intensity of human disturbance.
Xiang et al. [30]	Beichuan County in Sichuan Province	Vegetation Coverage, Vegetation Coverage Transfer Matrix, Geodetector	The artificial measures have promoted eco-environment restoration and accelerated the restoration speed of regional eco-environment.
Balamurugan et al. [7]	Western Coast of Gujarat, India	Comparative analysis of land cover area in different periods	The output shows a decline in the agricultural and scrubland, an increase in fallow land and saltpans, and exponential increases in built-up regions in the area.

shows the analysis and research of land cover change in several earthquake-stricken areas. The result of Peng et al. [55] shows that there exists a positive correlation between the land cover change in the earthquake-stricken areas and the intensity of human disturbance, and between the change in the landscape pattern index and the intensity of human disturbance. This is consistent with the trend that human factors have a greater impact on land-use change in earthquake-stricken areas, as analyzed in Section 5.1. Additionally, the conclusions of Xiang et al. [30] showed that artificial measures promoted the restoration of the ecological environment and accelerated the restoration speed of the regional ecological environment. All these conclusions verify that the research results of this paper are reliable. Balamurugan et al. [7] obtained the change trend of shrub reduction and cultivated land increase in the study area, which is consistent with the conclusion of this paper that the areas of forest, cropland, and shrub types in the study area changed greatly. The results of previous studies are basically consistent with those of this paper.

6. Conclusions

Based on the theories of landscape ecology, landscape geology and catastrophe, and using GIS and RS technical means. This paper explored and analyzed the spatial–temporal evolution characteristics of land cover before and after the earthquake in Jiuzhaigou County. The conclusions of this study are as follows:

(1)

Before and after the Jiuzhaigou County earthquake, the types of forest, cropland, shrubland, and grassland in the study area changed greatly. Affected by the earthquake, the comprehensive index of land cover degree of most towns decreased first and then increased. The southeast of the study area has convenient transportation, high population density, and frequent human activities, and the comprehensive index of land cover degree changes more obviously;

(2)

Affected by the earthquake and secondary disasters, the cropland and forest were damaged to a large extent. However, due to the special climate of Jiuzhaigou, some vegetation can easily take root and sprout, and become grassland or shrubland. After the earthquake, under the artificial repair and reconstruction, the cropland damaged by the earthquake is restored again. In addition, part of the grassland has been turned into forest, indicating that the ecological environment damaged by the earthquake is gradually being repaired;

(3)

In terms of time, the PD and LSI of most landscape types before and after the earthquake increased first and then decreased, and the AI decreased first and then increased. The results showed that the degree of landscape fragmentation was aggravated and the connectivity between patches was reduced. With the gradual recovery of the ecological environment after the earthquake, the degree of landscape fragmentation was less, the connectivity between patches was improved, and the ecological environment was gradually improved. From the perspective of space, the spatial changes in the PD, LSI, and AI landscape indexes were mainly concentrated in the southeast of the study area and around the Heihe and Baihe River basins, where the surface landscape was more complex and greatly disturbed by human activities;

(4)

From the perspective of overall change trend, the closer to the fault, the greater the land cover change and the more serious the land damage. Moreover, human activities are frequent in the southeastern region, which is more significantly affected by the fault zone.

However, there are some problems in using remote sensing to monitor the environmental changes in earthquake-stricken areas. For example, when an earthquake occurs, the disaster area is often accompanied by rainfall, debris flow, and other natural disasters, resulting in a large amount of cloud and fog coverage in some parts of the disaster area, which affects the timeliness of remote sensing data acquisition.

The changes in land cover before and after an earthquake further suggest that there are two typical time frames after an earthquake that affect resilience. The first is the short-term period after a disaster, in which people must be resilient in the face of a disaster response. It may be a time when people are focused on surviving an event and taking care of themselves or their community in the days that follow. The second period is much longer and includes recovery periods, which can stretch from days to weeks or even years, resulting in a long recovery period that is a challenging time for resilience. Resilience and sustainability become intertwined during this period, because people seek to restore their community, to become more resilient (for example, the built-up planning is designed to be more able to adapt to future adverse events), and to focus on long-term sustainability (for example, long-term land cover planning to ensure that the offspring survive and develop).