2.3.2. Analysis of Correlation

Correlation analysis is a statistical method to study the correlation between two or more variables. In data analysis, it is often used to analyze the relationship between continuous independent variables and continuous dependent variables. When there are many features, Pearson correlation analysis is used. Pearson's correlation coefficient is a statistic reflecting the degree of linear correlation between two variables. The calculation formula goes as follows:

$$\tau\_{xy} = \frac{\sum\_{i=1}^{n} (\mathbf{x}\_i - \overline{\mathbf{x}})(y\_i - \overline{\mathbf{y}})}{\sqrt{\sum\_{i=1}^{n} (\mathbf{x}\_i - \overline{\mathbf{x}})^2} \sqrt{\sum\_{i=1}^{n} (y\_i - \overline{\mathbf{y}})^2}} \tag{2}$$

where, *rxy* is the correlation coefficient of NDVI precipitation or temperature, which is between −1 and 1. The larger the value, the greater the correlation, and the smaller the value, the smaller the correlation. *x*, *y* are the mean values of multi-year NDVI and precipitation or temperature, *xi*, *yi* are the NDVI values of the ith year and the temperature and precipitation values of the *i* year.

$$R\_{12,3} = \frac{r\_{12} - r\_{13}r\_{23}}{\sqrt{\left(1 - r\_{13}^2\right)\left(1 - r\_{23}^2\right)}}\tag{3}$$

where *R*12,3, *R*13,2, *R*23,1 is the partial correlation coefficients among variables; *R*12,3 is the partial correlation coefficient between *r*<sup>1</sup> and *r*<sup>2</sup> after fixing the variable *r*3. *R*12,3 > 0 indicates positive correlation, that is, the two factors are correlated in the same direction; *R*12,3 < 0 indicate negative correlation, that is, the two elements of heterotrophy correlation; the larger the partial correlation coefficient is, the stronger the correlation between the two elements at the pixel is.

#### 2.3.3. Coefficient of Variation

The coefficient of variation, also known as the "coefficient of dispersion", is a normalized measure of the degree of dispersion of a probability distribution. The calculation formula is shown below.

$$\mathbf{C}\_{\upsilon} = \frac{1}{\overline{\overline{\chi}}} \sqrt{\frac{\sum\_{i=1}^{n} (\boldsymbol{x}\_{i} - \overline{\boldsymbol{x}})^{2}}{n-1}} \tag{4}$$

*Cv* stands for the coefficient of variation of NDVI; *xi* stands for the NDVI value in the *i*-th year; *x* stands for the mean NDVI value in the *n* years. The higher the *Cv* value, the more discrete the data, the higher the variation degree of the corresponding NDVI value, and the greater the inter-annual variation. The smaller the *Cv* value is, the more the data is aggregated, the lower the variation degree of the corresponding NDVI value and the lower the inter-annual variation.

#### **3. Results**

#### *3.1. Trends in Time Scale and Spatial Change of NDVI*

In order to explore the influence of the construction and operation activities along the GH highway, the NDVI obtained along the GH highway was divided into the construction period (1986–1991) and the operation period (1992–2020) in the time scale. By piecewise fitting of time-series NDVI, the trend of NDVI change in each time period was obtained.

Within the construction period, the NDVI in both the core and contrast areas showed a clear upward trend (Figure 2a). The NDVI increased more significantly in the contrast area at 0.0170/a, while it increased in the core area at 0.0149/a (Figure 2a). The results revealed that the construction of the GH highway caused some damage to the surrounding vegetation, and the NDVI values decreased most significantly in the early stage of construction (1987).

During the operation period, the vegetation cover in the core and contrast areas of GH highway gradually improved and the NDVI showed a generally increasing trend. The growth rates in the core and contrast areas were 0.0024/a and 0.0027/a, respectively (Figure 2b). However, the change trend of NDVI obviously differed before and after 2000. The NDVI increased at a faster rate in the core area (0.0137/a) than in the contrast area (0.0130/a) during 1992 to 2000. After this period of growth, the NDVI reaching a relatively

stable state, with little fluctuation around 2000. However, after this, the NDVI in both the core and the contrast areas decreased (at rate of −0.0030/a and −0.0002/a, respectively), showing two significant decreases (in 2010 and 2018) and one significant increase (2012) between 2001 and 2020.

**Figure 2.** Inter annual dynamic change of average NDVI within construction (**a**) and operation (**b**) periods in the study area.

The overall NDVI level in the operation period was higher than that in the construction period, while the fluctuation of NDVI was smaller than that in the construction period. Within the construction period, the MAT decreased slightly, while the MAP and NDVI increased significantly (Figure 3a). The MAP increased at a rate of 12.04 mm/a, while the MAT decreased at a rate of −0.0180 ◦C/a.

The trend of MAT generally increased, while the MAP decreased (Figure 3b). In the fitting of temperature and precipitation from 1992 to 2020, the temperature increased at a rate of 0.0288 ◦C/a, while the precipitation decreased at a rate of −16.44 mm/a.

Considering that there are many other traffic routes, cities, towns and villages along the GH highway. The core area was divided into four buffer zones with a distance of 1 km in order to measure the spatial impact level of GH highway. In the temporal dimension, the NDVI showed an overall increase due to self-healing of the environment. In the spatial dimension, the construction and operation of the GH road negatively affected the vegetation within 2 km of the route. This impact was inversely proportional to the distance from the GH highway route. The NDVI values were higher within the 1 km buffer than the 2 km buffer, but did not increase outward at the 3 km and 4 km ranges (Figure 4a,b).

**Figure 3.** Trend fitting of MAT (**a**), MAP (**b**) and NDVI within the construction and operation periods.

**Figure 4.** NDVI values at different distances within construction (**a**) and operation (**b**) periods.

Within the construction and operation period, the slopes of the regression equations in the core and contrast areas are concentrated between −0.05–0.15 and 0.002–0.015, respectively (Table 1).

**Table 1.** Classification statistics of different trends in core and contrast areas within construction and operation periods.


Within the construction period, the extent of NDVI damage in the core area was greater than that in the contrast area. However, the total trend of NDVI change was still overall slightly increasing, with the proportion accounting for 92.42% and 92.99% in core and contrast areas, respectively. Within the operating period, the NDVI showed an increasing trend as before. However, the proportion showing a slight increase became larger, accounting for 94.10% and 94.84% in the core and contrast areas, respectively.

Spatial differences in the increase or decrease in NDVI were observed along the GH road. Slight and significant decreases were dominant near the road and in urban areas (Figure 5a), while slight increases were dominant elsewhere. In particular, the increase was more significant in the mountain forest area (Figure 5b).

**Figure 5.** Spatial distribution of the trend of NDVI from 1986–2020. Trends were separated into the following classification: significantly decrease (<−0.005), slightly decrease (−0.005–0), slightly increase (0–0.005), non-significant increase (0.005–0.010), significantly increase (0.010–0.015), more significantly increase (>0.015).

#### *3.2. Coefficient of Variation Analysis*

The construction and operation of the GH freeway has had a negative impact on the stability of vegetation along the route. The fluctuations of NDVI in the construction period were greater than those in the operation period, and all fluctuations in the core area were greater than those in the contrast area. The NDVI, in terms of both periods and area were dominated by lower fluctuations, with lower fluctuations in the construction and operation periods of 845.12 km2 and 993.80 km2, respectively, and lower fluctuations in the core and contrast areas of 793.49 km<sup>2</sup> and 968.78 km2, respectively. The high fluctuation of NDVI in the construction period were larger than those in the operation period, as well as were larger in the core than in the contrast area. The high fluctuation areas for the construction and operation periods were 280.64 km2 and 121.98 km2, respectively, while the high fluctuation sizes in the core and contrast areas were 225.54 km<sup>2</sup> and 177.08 km<sup>2</sup> respectively (Figure 6).

**Figure 6.** *CV* for the core (**a**) and contrast (**b**) areas within the construction period divided into five levels; namely, low fluctuation (0–0.2), lower fluctuation (0.2–0.4), medium fluctuation (0.4–0.6), higher fluctuation (0.6–0.8) and high fluctuation (0.8–1). *CV* in the core (**c**) and contrast (**d**) areas within the operating period was divided into five levels, namely, low fluctuation (0–0.1), lower fluctuation (0.1–0.2), medium fluctuation (0.2–0.3), higher fluctuation (0.3–0.4) and high fluctuation (0.4–1).

#### *3.3. Correlation Analysis of NDVI with Temperature and Precipitation*

The correlation coefficient between NDVI and precipitation from 1986 to 2020 was mainly concentrated between −0.45 and 0 (mean = −0.03), showing a low negative correlation, while, its correlation coefficient with air temperature mainly ranged from 0–0.5 (mean = 0.08), showing a low positive correlation. The mean partial correlation coefficients of NDVI with precipitation and temperature were 0.05 and 0.17, respectively, which were both less than 0.5 (Figure 7).

**Figure 7.** Correlation between temperature (**a**), precipitation (**b**), and NDVI from 1986 to 2020. The distribution range of correlation coefficients were divided into high negative correlation (<−0.6), lower negative correlation (−0.6–−0.4), low negative correlation (−0.4–−0.2), pianissimo negative correlation (−0.2–0), pianissimo positive correlation (0–0.2), lower positive correlation (0.2–0.4), low positive correlation (0.4–0.6) and high positive correlation (>0.6).

The correlation of NDVI with temperature was mainly low positive, while the correlation with precipitation is mainly low negative; however, the correlation between NDVI and precipitation was positive, while the correlation between NDVI and temperature was negative in where the underlying surface was artificial, especially in urban areas and along roads (Figure 7).

#### *3.4. Analysis of Study Area LUCC*

Wetlands were re-classified as water bodies before the calculations, due to their small size. The LUCC decreased by 0.39% for forest, 6.04% for cultivated land, 0.97% for water, 0.48% for shrub, and 0.97% for grass from 2000 to 2020, while the artificial cover increased by 8.67% (Table 2).


**Table 2.** LUCC Classification statistics from 2000 to 2020.

The trend of NDVI decreased significantly along the GH highway (Figure 8), especially in urban areas along the route (Figure 8). The construction and operation of the GH highway and other later roads drove the development of towns along the route, leading to an expansion of artificial cover along the route and exacerbating the decline of NDVI (see Figure 5).

**Figure 8.** 2000 LUCC classification (**a**) and 2020 LUCC classification (**b**).

#### **4. Discussion**

#### *4.1. Characteristics and Reasons for Change in NDVI during the Construction Period*

Within the construction period, the annual average value and growth rate of the NDVI in the core area along the GH highway were smaller than those in the contrast area, and the coefficient of variation was larger than that in the contrast area, due to by the destruction of the original land-cover caused by the construction of the GH highway. The closer to the road, the greater the damage to the vegetation. This slowed the NDVI growth rate in the surrounding 4 km from the road, and breaks the environment for a certain distance along the line, thus increasing the variation in fluctuation of vegetation along the line. The fluctuation caused by highway construction activities on vegetation along the road were also larger closer to the road, indicating that the construction activities of GH highway had a negative effect on the stability of the surrounding environment. The construction and operation of the GH highway have increased the intensity of human activities in towns along the route, resulting in higher ecological fluctuations around the towns than other areas. The influence of the GH highway is mainly within 2 km, as the highway along the road are mainly within 2 km, making this the area with the strongest human activities. In the area far from the GH highway, the vegetation was weakly affected by the highway, and the heterogeneity of the surface was observed to have a greater impact on NDVI than the highway and human activities. In the early stage of construction (1987), the NDVI in both the area and the contrast areas declined sharply. This was due to the large-scale destruction of the surface vegetation in the early stage of the project construction, which led

to a sharp decline in NDVI in this year. In addition, the rocky desertification was serious in this period, and there was a lack of relevant control work, leading to damage of the fragile environment [43]. The engineering construction in the core area can easily affect involve the vegetation cover in the contrast area, causing further damages.
