Earthquake Research in China  2020, Vol. 34 Issue (1): 157-166     DOI: 10.19743/j.cnki.0891-4176.202001007
Vegetation Restoration Monitoring in Yingxiu Landslide Area after the 2008 Wenchuan Earthquake
HE Jing1,2, ZHANG Keke3, LIU Xiuju4, LIU Gang1,2, ZHAO Xuqiang1, XIE Zhongyuan1, LU Heng5,6     
1. College of Earth Sciences, Chengdu University of Technology, Chengdu 610059, China;
2. State Key Laboratory of Geohazard Prevention and Geoenvironment Protection, Chengdu University of Technology, Chengdu 610059, China;
3. China Petroleum Engineering and Construction Corp. Southwest Company, Chengdu 610017, China;
4. Chengdu Planning Information Technic Center, Chengdu 610042, China;
5. State Key Laboratory of Hydraulics and Mountain River Engineering, Sichuan University, Chengdu 610065, China;
6. College of Hydraulic and Hydroelectric Engineering, Sichuan University, Chengdu 610065, China
Abstract: A total of more than 50 000 landslides has occurred in Sichuan province since the "5·12" Wenchuan earthquake, resulting in serious damage to the surface vegetation in southwestern China. In this study, we select Yingxiu, the epicenter of Wenchuan earthquake, as the experimental area. The vegetation coverage information of the experimental area is extracted from the remote sensing images collected in the year of 2005, 2011 and 2013, respectively. The surface vegetation coverage in different periods is analyzed, and the vegetation recovery rate of the whole area is calculated. The experimental results show that in the first three years after the earthquake, the speed of vegetation restoration is slow, and the vegetation coverage rate is less than 20% better than 0.241, while in 2013, the vegetation coverage increases significantly.
Key words: Wenchuan earthquake     Yingxiu landslide area     NDVI     Vegetation restoration    


Wenchuan earthquake in 2008 has induced a large number of landslides. According to incomplete statistics, there are more than 50 000 landslides in Sichuan Province, causing extremely serious damage to the surface vegetation (Xu Chong et al., 2013; 2009). After years of growth, the vegetation in the vicinity of the landslide area has changed. Thus, it is of great significance to evaluate the process of vegetation restoration after the earthquake for a better understanding of the restoration and reconstruction of ecological environment.

Numerous previous investigations have been conducted in the study area. Liu Ying et al. analyze the soil recovery rate and its spatial heterogeneity in the process of ecological recovery in the damaged area after the earthquake, and finally indicate that the content of SOC, TN and TP in the damaged areas are not able to recover back to the level before the earthquake, the nutrient soil structure in the disaster area is seriously damaged, and the vegetation growth is relatively slow (Liu Ying et al., 2019). Xiong Junnan et al.(2018) analyze the vegetation restoration in the disaster area from the perspective of topography, and finally suggest that the increase proportion of low vegetation coverage is negatively related to elevation and slope; they also find that the vegetation in the area with elevation less than 500 m and slope less than 5° is greatly affected by the earthquake, leading to longer restoration period (Xiong Junnan et al., 2018). Sun Liwen et al compare and analyze the vegetation restoration of different governance modes from the perspective of Engineering Governance (Sun Liwen et al, 2015). Based on the MODIS NDVI data time series from 2000 to 2018, Huang Runqiu's team propose a new automatic quantitative evaluation algorithm of vegetation damage caused by strong earthquakes. Such evaluation provides quantitative indicators of the degree of vegetation restoration after the earthquake (Yunus A. P. et al., 2020). Based on leaf area index products, vegetation coverage products, and global 30m-land-use data from GLASS data in 2010, Zhang Tao et al.(2019) investigate the intensity area above 8° in the vicinity of Wenchuan earthquake in 2008 and the intensity area above 6° in the region near the "3·11" earthquake occurred in 2011 (Zhang Tao et al, 2019).

We conduct our study from the perspective of remote sensing. The study area is about 190 km2 around Yingxiu, which is the epicenter of landslide concentration area. Based on the remote sensing images collected in 2005, the vegetation coverage before the earthquake is interpreted. After the earthquake, the vegetation coverage in the experimental area is interpreted using the images of 2011 and 2013 as data sources. According to the interpretation results of these three images, the vegetation cover changes in the experimental area are compared, and the reasons for the changes are analyzed.


The experimental area is Yingxiu, which is the epicenter of the 2008 Wenchuan earthquake and has an area of about 190 km2, as shown in the shadow in Fig. 1. In this area, substantial landslides have been triggered and the specific distribution of landslides is shown in the low right corner of Fig. 1. Each green dot represents an individual landslide. The image scale precludes the exhibition of real distribution density. However, the average landslide density of 11.5 km2 is obtained based on the statistical calculation of landslide data.

Fig. 1 Research Area

Vegetation coverage refers to the vertical projection area of vegetation (including leaves, stems and branches) on the ground as a percentage of the total area of the statistical region. In general, the area of vegetation can be extracted by using the normalized vegetation index (NDVI) constructed in the near-infrared band of remote sensing image, which can well separate vegetation from water and soil. According to relevant statistical research, the accuracy of vegetation extracted using this index is more than 85%, and the specific calculation formula (Justice C. O. et al., 1985) is shown in Eq.(1).

$ {\rm{NDVI = }}\frac{{{\rm{NIR - RED}}}}{{{\rm{NIR + RED}}}} $ (1)

In Eq.(1):NIR represents near-infrared band. Generally, the fourth band of ordinary remote sensing image is NIR, and the band range is 760-900 nm; RED is red band which ranges from 620 to 760 nm and is the third band of image. The value range of NDVI is [-1, 1], where the negative value indicates that the ground is covered with clouds, water, snow, highly reflecting the visible light; 0 indicates the area is covered with rock or bare soil, etc. On the contrary, the positive value indicates that there is vegetation coverage, which has a positive linear relationship with the corresponding value.

NDVI index can be used to extract vegetation area from remote sensing images, providing us the approach of setting up new indexes ωv, which stand for vegetation coverage and vegetation recovery rate(Lin Chaoyuan et al., 2002; 2004; Lin Y. F. et al., 2004), and the calculation Equation are (2) and (3) respectively.

$ \overline \omega = {\left({\frac{{1 - {\rm{NDVI}}}}{2}} \right)^{1 + {\rm{NDVI}}}} $ (2)
$ v = \frac{{{{\overline \omega }_1} - {{\overline \omega }_2}}}{{{{\overline \omega }_1} - {{\overline \omega }_0}}} $ (3)

The range of ω is [0, 1]. Smaller value indicates higher vegetation, and vice versa. v represents the vegetation recovery rate, which can be calculated only when there are at least three stages of vegetation coverage index. Among them, ω0 is the first stage data, which can also be understood as local data, ω1ω2 are the second and the third stage data, respectively. According to the calculation, the size of the value of v is measured, the vegetation restoration rate can be divided into four levels, and the specific standards are shown in Table 1.

Table 1 Vegetation restoration rate

For the vegetation restoration rate, if it is less than 0, the local vegetation restoration conditions will be relatively poor, which is not conducive to the growth of vegetation. It is possible that the vegetation decreases instead of increasing in this period. In this case, first of all, it is necessary to check whether there are any calculation errors in the data. If not, further analysis of the cause of this situation is needed. For example, whether there is a large-scale geological disaster in the area, leading to the destruction of surface vegetation, or whether tree numbers are reduced due to man-made deforestation, etc. Under normal conditions, the value is between 0 and 50, indicating that the vegetation in this area has little change in this period. When the value exceeds 50, it means that the vegetation has increased significantly in this period, and the growth conditions have improved significantly as well. In this case, most of them are caused by human factors, such as afforestation, construction of water conservancy projects to improve vegetation growth conditions, and other control measures.


In order to accurately extract the vegetation part from the experimental area, the selected satellite data are required to have high resolution, good performance in the near infrared band, and favorable picking period which is generally from April to October. In this period, the vegetation growth is vigorous and its features in the near-infrared band are more obvious, which is more conducive to the extraction of information. Based on the cloud amount in the image, we finally select the SPOT5 satellite data in July 2005 as the local image data of the experimental area, the worldview 2 image in April 2011 as the first phase data after the earthquake, and the Pleiades image in April 2013 as the second phase data. The image of each phase is shown in Fig. 2. In the local image of 2005, there is a small amount of cloud cover in the experimental area, which should be removed when calculating the ground vegetation coverage.

Fig. 2 Image of experimental area (a) 2005 image; (b) 2011 image; (c) 2013 image
3.2 Calculation of Multi-Year Vegetation Coverage in the Experimental Area

In 2005, SPOT5 satellite data was selected as the image of the experimental area. According to the data parameters of the satellite, the corresponding near-infrared band is 780-890 nm, B3 band, while the red band is 610-680 nm, B2 band. Therefore, the corresponding NDVI index calculation formula is shown in Eq.(4)

$ {\rm{NDVI = }}\frac{{{B_3} - {B_2}}}{{{B_3} + {B_2}}} $ (4)

There are 8 multi-spectral bands in the corresponding worldview2 data in 2011. The near-infrared band is 770-895 nm, and the red band is 630-690 nm. Other new bands have limited effect on vegetation information extraction. The corresponding Pleiades satellite data in 2013 has less band information than those of worldview 2. The number of multispectral bands is the same as SPOT5. The near-infrared band range is 750-950 nm, and the red band range is 600-720 nm. The NDVI vegetation index can be calculated by taking the corresponding wave bands of each image into Eq.(4). Then, according to Eq.(2), the vegetation coverage index distribution map of each year in the experimental area can be obtained, and the specific calculation results are shown in Fig. 3.

Fig. 3 Vegetation coverage index distribution (a)2005; (b)2011; (c)2013

In Fig. 3, red area indicates less or no vegetation coverage. By overlaying the calculated vegetation coverage image with the original image, we find that most of these areas are river water or landslides caused by earthquake damage. Among the data in 2005, according to the calculated results, the vegetation coverage of the area under the cloud cover is also extremely low, however, the vegetation coverage reaches a high level after the combination with surrounding vegetation coverage, indicating a kind of misjudgment. The possible reason is that most of the energy is reflected and absorbed when the red band and near-infrared band pass through the cloud layer, and the actual energy weakly reaches the ground, leading to a large error in the calculation of the area under the cloud.

Fig. 4 is the statistical comparison of the area of vegetation coverage in each stage. The blue line represents the distribution of vegetation coverage in 2005, the red line represents the distribution of vegetation coverage in 2011, and the green line stands for the distribution of vegetation coverage in 2013. It can be seen from the figure that the areas of vegetation coverage of the three years are relatively close within a range of 0.241 and 0.32. In 2005, the vegetation coverage mainly concentrates on the left side of 0.241-0.32, accounting for more than 95% of the total area of the experimental region, of which the total area less than 0.11 is up to 134 km2, accounting for more than 70% of the total area.

Fig. 4 Statistics of vegetation coverage area in the experimental area

In 2011, the area on the left side of 0.241-0.32 accounts for less than 20%, indicating that only small amount of vegetation area reaches the growth level in 2005, and the vegetation coverage is relatively low. However, in the range of 0.321-0.44, the area accounts for more than 70%, which indicates that the vegetation coverage in most areas is not optimistic and the vegetation growth is not ideal. Combined with the original satellite image, we can see that there are large white areas in the image. Field survey data confirm that most of these areas are landslides and their accumulated deposits caused by earthquakes. The resulting sporadic shrub vegetation is associated with several reasons regarding its unfavorable growing environment. Firstly, most of the deposits on the landslide surface are rocks, leading to limited soil area suitable for the vegetation growth. Furthermore, the soil area is often eroded by rain wash during rainy seasons. As a result, the vegetation coverage is not optimistic in the three years after the earthquake.

In 2013, the area of the left section of 0.241-0.32 increases in varying degrees, showing that after 5 years of natural ecological restoration, vegetation coverage in most areas increases. The low shrubs on the original landslide has the function of conserving water and soil after 2 years of growth, which constantly enhances the growth environment on the landslide and induces massive new vegetation growth and development. As a consequence, the vegetation patch effect is formed and the green area starts increasing with a corresponding decrease of white area.


On the basis of the vegetation coverage ω0, ω1, ω2, combined with the three time nodes in the experimental region, the vegetation recovery rate of the area can be calculated according to the Eq.(3), and the calculation results can be classified according to the level in Table 1. Besides, the vegetation recovery rate map of the experimental area can be obtained, as shown in Fig. 5. It can be seen from the figure that the recovery rate in most areas stays at a general level, and only the rate in a few areas is at the level of fast or very fast. Through the investigation of the areas with recovery rate of more than 50, we observe that more than half of the areas are due to the inaccurate extraction of some information in the three periods of data, and the rest areas are verified to recover efficiently. The area with fast recovery is mostly located at the slope smaller than 15° and close to the water source. Meanwhile, it is mostly sunny slope where the accumulation body is mainly composed of mudstones and fine gravels, accompanied by some humus soil; such environment creates excellent water, air, soil and other conditions for vegetation growth. After investigation, the areas with a recovery rate smaller than 0 are basically rivers. Besides, a small portion of these areas are exposed bedrocks with a slope gradient larger than 40°, resulting in adverse growing conditions and zero vegetation.

Fig. 5 Vegetation recovery rate

In this paper, the vegetation coverage information in the epicenter area of "5.12" Wenchuan earthquake is extracted from the image data of 2005, 2011 and 2013. Then, the vegetation restoration in the whole experimental region is analyzed. In general, after 5 years of natural ecological restoration, the vegetation coverage in the experimental region has an obvious rising trend. However, the restoration rate is slow. The main reasons for this situation are as follows:

(1) A majority of slope gradients are larger than 30°, such highly oblique slopes lead to unstable soil that may experience strong erosion by rain wash.

(2) Most part of the landslide area is covered by gravels with large particle size, but the humus suitable for vegetation growth is relatively less, which brings considerable difficulties to the natural growth of vegetation.

Therefore, it can be seen that the task of vegetation restoration in the whole experimental region is still quite arduous. If there is no manual intervention in the sterile areas, the recovery efficiency will be very slow. Moreover, these areas, may experience different geological disasters such as rainstorm, mountain flood and debris flow in the rainy seasons. If the ecological environment of these areas can be restored as soon as possible and the vegetation coverage can be improved, the probability of geological disasters in these areas may be largely reduced, improving the safety factor of the life and property of local people.

Justice C.O., Townshend J.R.G., Holben B.N., Tucker C.JJustice C.O., Townshend J.R.G., Holben B.N., Tucker C.J. Analysis of the phenology of global vegetation using meteorological satellite data[J]. International Journal of Remote Sensing, 1985, 6(8): 1271-1318.
Lin Chaoyuan, Lin W.T., Chou W.CLin Chaoyuan, Lin W.T., Chou W.C. Soil erosion prediction and sediment yield estimation:the Taiwan experience[J]. Soil and Tillage Research, 2002, 68(2): 143-152.
Lin Chaoyuan, Lo H.M., Chou W.C., Lin W.ZLin Chaoyuan, Lo H.M., Chou W.C., Lin W.Z. Vegetation recovery assessment at the Jou-Jou Mountain landslide area caused by the 921 Earthquake in Central Taiwan[J]. Ecological Modelling, 2004, 176(1/2): 75-81.
Lin Y.F., Lin Chaoyuan, Chou W.C., Lin W.T., Tsai J.S., Wu C.FLin Y.F., Lin Chaoyuan, Chou W.C., Lin W.T., Tsai J.S., Wu C.F. Modeling of riparian vegetated buffer strip width and placement:a case study in Shei Pa National Park, Taiwan[J]. Ecological Engineering, 2004, 23(4/5): 327-339.
Liu Ying, He Jingwen, Wu Jianzhao, Cui Yu, Luo Qinghu, Lin Yongming, Wang Daojie, Li JianLiu Ying, He Jingwen, Wu Jianzhao, Cui Yu, Luo Qinghu, Lin Yongming, Wang Daojie, Li Jian. Soil recovery rate and spatial heterogeneity in the process of ecological restoration in an earthquake-affected area[J]. Chinese Journal of Applied and Environmental Biology, 2019, 25(4): 823-830 (in Chinese with English abstract).
Sun Liwen, Shi Changqing, Zhao Tingning, Li Danxiong, Lai Wenhao, Ding WanquanSun Liwen, Shi Changqing, Zhao Tingning, Li Danxiong, Lai Wenhao, Ding Wanquan. Vegetation recovery in the landslide-tackling area of Wenchuan earthquake[J]. Science of Soil and Water Conservation, 2015, 13(5): 86-92 (in Chinese with English abstract).
Xiong Junnan, Zhang Hao, Peng Chao, Fan Chunkun, Zhu Jilong, Gong YingXiong Junnan, Zhang Hao, Peng Chao, Fan Chunkun, Zhu Jilong, Gong Ying. Vegetation variations and correlations with topographic factors in Wenchuan earthquake area[J]. Bulletin of Soil and Water Conservation, 2018, 38(3): 24-31, 37 (in Chinese with English abstract).
Xu Chong, Dai Fuchu, Yao XinXu Chong, Dai Fuchu, Yao Xin. Incidence number and affected area of Wenchuan earthquake-induced landslides[J]. Science & Technology Review, 2009, 27(11): 79-81 (in Chinese with English abstract).
Xu Chong, Xu Xiwei, Wu Xiyan, Dai Fuchu, Yao Xin, Yao QiXu Chong, Xu Xiwei, Wu Xiyan, Dai Fuchu, Yao Xin, Yao Qi. Detailed catalog of landslides triggered by the 2008 Wenchuan earthquake and statistical analyses of their spatial distribution[J]. Journal of Engineering Geology, 2013, 21(1): 25-44 (in Chinese with English abstract).
Yunus A.P., Fan Xuanmei, Tang Xiaolu, Jie Dou, Xu Qiang, Huang RunqiuYunus A.P., Fan Xuanmei, Tang Xiaolu, Jie Dou, Xu Qiang, Huang Runqiu. Decadal vegetation succession from MODIS reveals the spatio-temporal evolution of post-seismic landsliding after the 2008 Wenchuan earthquake[J]. Remote Sensing of Environment, 2020, 236: 111476.
Zhang Tao, Tang HongZhang Tao, Tang Hong. Impacts of major earthquake disasters on land vegetation:comparison of China Wenchuan 512 earthquake and Japan 311 earthquake[J]. Geography and Geo-Information Science, 2019, 35(1): 44-49 (in Chinese with English abstract).