2.1.

Study Region

The Wenchuan earthquake was measured 8.0 Ms (surface wave magnitude scale). It occurred on May 12, 2008, in the Sichuan province of China. It is the most damaging and influential of all earthquakes that have occurred in China since 1949. Tremors were also felt in Vietnam, Thailand, Mongolia, Bangladesh, Nepal, India, Pakistan, and Russia. Ten counties that were worst hit were Wenchuan, Beichuan, Dujiangyan, Pengzhou, Shifang, Mao, An Xian, Pingwu, Qingchuan, and Mianzhu.¹⁹ A sketch of this epicenter and region is showed in Fig. 1. Mean altitude is 2034 m and landform is complex, with many hills in this region. According to the vegetation classification, there are about 18 kinds of plants.²⁰ SI varied from SI7 to SI11, with the higher values occurring mainly in the east and south of this region. About 9.8%, 11.23%, and 23.38% of this region experienced SI11, SI10, and SI9, respectively.

Fig. 1

Seismic intensity, geographical profile, and vegetation classification of the study region.

2.2.

Materials

In order to clearly show vegetation restoration, normalized difference vegetation index (NDVI) images were collected from 1998 to 2011. A regional vegetation classification map, digital elevation model (DEM), and map of county boundaries were also collected. DEM data were downloaded from http://datamirror.csdb.cn/, they were acquired with shuttle radar topography mission (SRTM), produced by International Center for Tropical Agriculture (CIAT), and spatial resolution is 90 m.

NDVI are satellite images with spatial resolution of 1 km acquired by SPOT VGT, they are the maximum-value composites synthesized from daily images, with the highest NDVI of an individual pixel for any given geographic coordinate, which helps to ensure coverage of all landmasses worldwide with a minimum cloud-cover effect. The platforms of SPOT VGT sensor are SPOT satellites (SPOT-4 and SPOT-5), which have an equator-crossing time of 10:30. The 10-day synthesis products until January 2003 are available at http://free.vgt.vito.be/origin from SPOT-4 (VGT1 sensor) and those after that from the SPOT-5 (VGT2 sensor). Real NDVI values are retrieved from original NDVI product images according to the equation $y=ax+b$, where $y$ is the NDVI value, $x$ is the digital number of the image, and $a$ and $b$ are 0.004 and $-0.1$, respectively.²¹

2.3.

Methods

Changes of many parameters were used to reflect vegetation restoration, such as growth, stability, and diversity. These can be recognized by the statistical features of long time series of NDVI images. Spatio-temporal features of vegetation before and after earthquake were analyzed by annual means of NDVI from 1998 to 2007 and from 2009 to 2011, respectively. Before calculating the statistic features of NDVI, average values of NDVI per year were made from 10-day NDVI images over entire year. One year covers vegetation growth cycle in general, which can avoid errors caused by time differences of sprouting and deciduous of different vegetations. Based on these annual NDVI images, spatio-temporal features of vegetation restoration and variation were analyzed. After that, statistic images, including arithmetic mean, standard deviation and linear trend estimation of annual NDVI before and after earthquake, are analyzed with overlay methods based on the different counties, seismic intensity, and vegetation classification. By contrast, vegetation restoration and variation have been recognized and identified.

Arithmetic mean and standard deviation are the most commonly used statistical parameters. Arithmetic mean is the sum of a collection of numbers divided by the number of numbers in the collection,²² it is often used to report central tendencies. In this case, arithmetic mean was used to find the global central feature of annual NDVI in some regions or several years. Standard deviation is the square root of the variance of the random variable, statistical population, data set, or probability distribution. It shows how much variation or dispersion from the average exists.²³ Standard deviation reflects individual dispersion in statistical samples, and thus indicates the statistical distribution and fluctuation amplitude in any region. According to the statistical signification, standard deviation was used to show the stability of vegetation.

It is not enough to analyze the vegetation restoration with arithmetic mean and standard deviation, because their values only showed the statistical distribution state, which does not indicate the global change in trend of variables with time. In order to recognize vegetation change and restoration state, linear trend estimation method was used to analyze the change trend. Linear trend estimation shows the change in trend of the temporal series of statistical variables. The linear trend estimation method analyzed the change in trend of interannual mean values using Eq. (1):²³

3.1.

General Features of Vegetation Variation

According to statistic means of arithmetic mean, its value mainly reflects the center tendency of statistical variables, that is to say, mean value of annual NDVI over years. In detail, the bigger mean value of annual NDVI is, the better vegetation growth is. Figure 2 indicates a generally stable spatial distribution of mean NDVI values before and after earthquake. Higher values are mainly in the north, and lower values mainly distribute over the southeast, which is in good correlation with elevation and regional landform. In regions with higher elevation and complex landform, vegetation grows better. Specifically, the highest NDVI value is in Pingwu and the lowest value is in Shifang, corresponding to no difference in NDVI before and after the earthquake. Also, there is a minute change in the range of NDVI values. To be precise, the NDVI value varies from 0.0025 to 0.6666 before the earthquake, and from 0.0163 to 0.6807 after the earthquake. It is clear that both maximum and minimum are slightly increased by 0.0138 and 0.0141, respectively. On the other hand, NDVI mean value over the whole region is decreased from 0.4553 to 0.4417. The reduction is 0.0137, about 3% of the NDVI mean value before the earthquake. Thus, there is an overall slight decrease in the general growth of vegetation in this region.

Fig. 2

NDVI mean value of vegetation restoration before and after the Wenchuan earthquake, (a) spatial distribution of NDVI mean value for 1998 to 2007, before the earthquake, (b) spatial distribution of NDVI mean value for 2009 to 2011, after the earthquake, and (c) static feature of NDVI value before and after the earthquake.

Regions differ in geological and geographical features, which result in spatial variation in the ability to withstand seismic activity. Therefore, it is reasonable to find that vegetation damage after earthquake is at different levels in different regions. In order to define this spatial variation, several types of maps are used to overlay the analysis: a map of political boundaries of counties, vegetation regionalization maps, and a seismic intensity distribution map. Figure 3 shows variation in mean values of NDVI before and after the earthquake (a) with seismic intensity, (b) by county, and (c) with vegetation type.

Fig. 3

Spatial variation of NDVI mean value of vegetation restoration before and after earthquake, (a) plot of NDVI mean value against seismic intensity, (b) plot of NDVI mean value in different counties, and (c) plot of NDVI mean value for different classes of vegetation.

As well as the region where earthquake damage to vegetation occurs, the impact of seismic intensity on the vegetation growth must also be taken into account, and so the NDVI mean value is plotted against the variation in seismic intensity before and after the earthquake. In Fig. 3(a), it is seen that the higher the seismic intensity of the region, the worse the damage to vegetation. Typically, the greatest decrease in vegetation appears in regions of SI11 and SI10. It is about 9.8% in the SI11 region, which covers the earthquake center. According to the graph for the different counties [Fig. 3(b)], average NDVI has been in decline for many years, for example in An Xian, Mao Xian, Pingwu, Wenchuan, and more especially in Mianzhu, Shifang, Pengzhou, and Dujiangyan. The decline trend is more obvious in the northeast of this region, where the land is mainly cultivated. In general, a decline in different classes of vegetation is not observed after the earthquake, as shown in the plot of NDVI mean value against vegetation classification before and after the earthquake [Fig. 3(c)].

3.2.

Stability Variation in Different Regions

The stability of a regional ecosystem is indicated by the stability of spatial distribution. Standard deviation is calculated with annual NDVI and used as the indication index: the bigger the standard deviation and the less stable the regional ecosystem. Adversely, the regional ecosystem is more stable. This index is clearly and closely linked to the area under assessment. Figure 4 shows the variation in standard deviation with (a) seismic intensity, (b) county, and (c) vegetation type. In general, the standard deviation is bigger after than before the earthquake. Both absolute and relative changes in stability increase with seismic intensity, being most obvious in the SI11 region [Fig. 4(a)].

Fig. 4

Spatial variation of NDVI standard deviation of vegetation restoration before and after earthquake, (a) plot of NDVI standard deviation against seismic intensity, (b) plot of NDVI standard deviation in different counties, and (c) plot of NDVI standard deviation for different classes of vegetation.

As Fig. 4(b) indicates, differences in stability before and after the earthquake are evident in different counties. The biggest absolute changes are in Qingchuan and Beichuan from 0.0424 to 0.5815 and from 0.0599 to 0.5355, respectively. The greatest relative changes appear in Qingchuan and Anxian, standard deviations after the earthquake being 12.7303 and 13.3132 times those before the earthquake, respectively. These values indicate that there are obvious changes in Qingchuan, Beichuan, and Anxian, the ecosystem being less stable after the earthquake. In general, ecosystem stabilities in different counties are similar before the earthquake; stability variation is greater after the earthquake. In different counties, variation of ecosystem stability is more obvious after the earthquake, increase of standard deviation is approximately 6.5 times. Overall, the ecosystem is less stable in the north than in the south. Absolute changes show the same spatial distribution.

Relatively, vegetation regionalization is the smallest analysis unit. As a result, differences between before and after the earthquake are not obvious. However, stability variation is the most important in understanding the ecosystem change in this earthquake region. As Fig. 3(c) indicates, there are obvious changes in MNF_CTT, NF_T, NF_ST, MNF_STT, MMF_ST, MBF_ST, BEF_ST, BFS_STT, BS_STT, BDS_SA, and GM_T. The greatest absolute changes appear in MNF_CTT and MMF_ST: 0.0235 and 0.0343, respectively. The greatest relative changes are in MMF_ST and BFS_STT. In general, regions of obvious change are mainly found at higher altitudes, that is to say, in the west of the study region.

3.3.

Variation in Restoration of Vegetation in Different Regions

Although a vegetation ecosystem undoubtedly fluctuates between different years, this change is minute in a normal year. When a major natural disaster happens, the path of general change is from good to bad and then back to good. In this process, vegetation growth change is one of the most useful indices for monitoring and assessing the restoration of vegetation. In order to show the variation in restoration after the earthquake by area, the linear trend estimation coefficient is taken as the indicator. The variation of linear trend estimation coefficient over the different regions is shown in Figs. 5 and 6. Over the whole region, linear trend estimation coefficient increases from 0.0079 before the earthquake to 0.0359 after the earthquake. Such a large change indicates that the plant ecosystem is in the process of rapid restoration from 2009 to 2011. As with the spatial distribution before the earthquake, the biggest and smallest linear trend estimation coefficients are found in the eastern, cultivated, vegetation; and the western hill vegetation, which reflects the natural hill flora, is stable. However, the biggest linear trend estimation coefficient is mainly scattered over the hill region, indicating that plant restoration and increase after the earthquake are good in this region.

Fig. 5

NDVI linear trend estimation coefficient of vegetation restoration before and after earthquake, (a) spatial distribution of linear trend coefficient of annual NDVI mean value before the earthquake and (b) spatial distribution of linear trend coefficient of annual NDVI mean value after the earthquake.

Fig. 6

Linear trend estimation coefficient of vegetation restoration after earthquake, (a) plot of linear trend estimation coefficient of annual NDVI mean value in different counties, (b) plot of linear trend estimation coefficient of annual NDVI mean value against seismic intensity, and (c) plot of linear trend estimation coefficient of annual NDVI mean value for different vegetation classes.

In order to clearly show the variation in restoration in different regions, graphs of linear trend estimation coefficient in different counties (a), with seismic intensity (b) and with vegetation type (c) are shown in Fig. 6. Bigger line trends and good plant restoration are shown in these figures; however, fluctuation in linear trend estimation coefficient in different counties is more obvious after the earthquake than before the earthquake. Except in Qingchuan, the vegetation shows a bigger linear trend estimation coefficient value and better growth in all counties. The changes are more obvious, especially in Wenchuan and Maoxian, than before the earthquake; the absolute increases are 0.0447 and 0.0350, and the relative increases are 9.8887 and 10.5814, respectively. Among the regions of different seismic intensity, restoration of vegetation is the most obvious in SI10 and SI11 regions. However, the biggest absolute and relative increases occur in areas with SI8 and SI7: 0.0330 and 0.0292, respectively.

In contrast, the graphs of linear trend estimation coefficient of annual NDVI mean values for different classes of vegetation show greater fluctuation before the earthquake. The highest and lowest values are for different vegetation classifications. Vegetation growth is best for NF_ST before and MMF_ST after the earthquake, and worst for SV_A before and GM_T after the earthquake, the increase being 0.0132, 0.0379, 0.0323, and 0.0123, respectively, for these vegetation types. Over all the regions, the growth shows the greatest change for BESS_SB, i.e., 0.0453, 72.0954 times the growth before the earthquake. There is also a large increase in MNF_STT and SV_A. According to Fig. 6(c), vegetation types MNF_STT, MMF_ST, BDS_SA, BESS_SB, and M_AK are growing well.