Quantitative Analysis of Vegetation Dynamics in Response to Climatic Factors and Human Activities in Southwest China


 We determined the time scale of normalized difference vegetation index (NDVI) response to drought and used trend and correlation analyses to explore the spatial and temporal variability characteristics of the NDVI and SPEI and their sensitivity to climatic factors in southwest China from 2000 to 2020. We used a partial derivative approach to calculate the contributions of six climatic factors and human activities to the interannual variation in the NDVI. The results demonstrated that from 2000 to 2020, the annual mean NDVI in southwest China showed a slight decreasing trend at a rate of 0.0001 y−1. The NDVI had the highest sensitivity to the standardized precipitation and evapotranspiration index on a 12-month time scale. The NDVI exhibited a 1-year delayed response to drought. The SPEI has the highest sensitivity to precipitation. The percentage of pixels with a positive correlation between NDVI and precipitation, mean temperature, temperature difference, mean relative humidity, mean wind speed, and sunshine duration in the study area was 31.73%, 46.81%, 35.49%, 25.76%, 39.36%, and 39.89%, respectively. The average contributions of these six climatic factors to the interannual variation of NDVI were 0.00029, 0.00046, −0.00007, 0.00007, 0.0008, and 0.00001 y−1, respectively. The NDVI had the highest sensitivity to mean temperature and the lowest sensitivity to mean relative humidity. The average contributions of climatic factors and human activities to interannual variability in southwest China were 0.00156 and 0.00012 y−1, respectively. The positive influence of climatic factors on the NDVI was stronger than that of human activities. This study provides a theoretical basis for the sustainable management of the regional ecological environment.


Introduction 32
Vegetation is an important component of ecosystems (Lizaga et al., 2018;Christopher et al., 2019), 33 with significant sensitivity to climatic factors and human activities. Vegetation growth responds to 34 changes in environmental factors (Zhao et al., 2019). In particular, drought can affect the change 35 in vegetation growth on a regional scale. Changes in vegetation growth have been explored to 36 Penman-Monteith formula to calculate potential evapotranspiration (ET0), which can more 143 accurately characterize the drought condition, especially in regions with significant warming 144 (Zhang et al., 2015). The specific calculation method is found in Ruspini et al. (1965). 145

Trend analysis 146
To quantitatively reflect the spatial and temporal variation of the NDVI in the study region, 147 interannual trends in the NDVI were calculated at the pixel level using least squares regression, 148 and the F-test was used to determine the significance of each pixel. The calculation equation was 149 as follows： 150 where λslope is the slope of regression, n is the total number of years monitored (n = 21), and Ci is 152 the mean NDVI value in year i. When λslope > 0, the NDVI had an increasing trend during the 153 study period; otherwise, it had a decreasing trend. 154

Correlation analysis 155
The correlation between NDVI and SPEI and the correlation between NDVI and climate factors 156 In this study, the Sobol global sensitivity analysis method (Aovk, et al., 2017), which analyzes the 165 independent and total sensitivity of the SPEI to meteorological factors, was selected as a global function of individual variables and combinations between variables, and then, the total variance 168 of the model output is decomposed as a combination of individual input variables and correlations 169 between variables as follows: where V is the total variance value of the model output, Vi is the variance value caused by the 172 input variables, and Vij is the variance value caused by the correlation between the input variables i 173 and j.

174
The formula for calculating the sensitivity coefficient is as follows: where Eq. (5) is used to calculate the first-order sensitivity coefficient, which represents the 179 independent sensitivity of SPEI to meteorological factors, Eq. (6) is used to calculate the 180 second-order sensitivity coefficient, which represents the correlation between meteorological 181 factors, and Eq. (7) is used to calculate the total order sensitivity coefficient, which represents the 182 total sensitivity of the SPEI to meteorological factors. The variance of these equations was 183 estimated by approximating Monte Carlo values. SPre, STem, STmax−Tmin, SRh, SWs, and SSd represent 184 the sensitivity coefficients of the SPEI to precipitation, mean temperature, temperature difference, 185 mean relative humidity, mean wind speed, and sunshine duration, respectively. 186

Contribution of each driver to the interannual variation of the NDVI 187
The partial derivative method has been widely used to quantify the contribution of climatic factors 188 to evapotranspiration (Meng et where λslope is the interannual variability of NDVI, which is obtained from Equation (2), and 197 precipitation, mean temperature, temperature difference, mean relative humidity, mean wind speed, 198 and sunshine duration are expressed by Pre, Tem, Tmax − Tmin, Rh, Ws, and Sd, respectively. 199 Con(Pre), Con(Tem), Con(Tmax − Tmin), Con(Rh), Con(Ws), and Con(Sd) denote the 200 contributions of precipitation, mean temperature, temperature difference, mean relative humidity, 201 mean wind speed, and sunshine duration to the interannual variation of the NDVI, respectively; t 202 is time.

Sd t
  denote the slope of linear regression of 10 interannual variation of the NDVI indicated that the driver promoted an increase in the NDVI.

214
3 Results and analysis 215

216
The results of the linear trend analysis demonstrated that the mean NDVI value in southwest    global sensitivity method to analyze the sensitivity of the SPEI to precipitation, mean temperature, 284 temperature difference, mean relative humidity, mean wind speed, and sunshine duration in coefficients of the relationship between the SPEI and the six climatic factors were nearly identical, 287 that is, the SPEI had the highest sensitivity to precipitation, followed by average relative humidity,   . We analyzed the correlation between the NDVI and climatic factors (Fig. 5). Positive and negative correlations between the NDVI and precipitation, mean air temperature, temperature the NDVI and mean wind speed were mainly in northeastern Sichuan (Sichuan Basin), which is 331 related to the special topography of the basin. The NDVI was significantly correlated with 332 sunshine duration mainly in Guizhou Province The correlation between the NDVI and mean 333 relative humidity was negative in most areas. On longer time scales (6 and 12 months), the area with relatively high correlation coefficients 406 between the NDVI and SPEI in southwest China was extended, with 9.89% of the total pixels 407 passing the significance test for correlation on the 12-month scale (Fig. 2d), a phenomenon that

415
The response of vegetation to drought varied with time scale. In this study, we determined that the 416 NDVI had the slowest response to the SPEI-12 and the fastest response to the SPEI-1 (Fig. 2). In 417 previous studies, the response of vegetation to drought has been reported to differ among regions.

Response of vegetation to climatic and nonclimatic factors
study area (Fig. 4). However, the contribution of mean wind speed to the NDVI was the largest at 449 0.0008 y −1 , and the effect of mean wind speed on the interannual trend of the NDVI in the study 450 area was significantly spatially heterogeneous, with the largest effect mainly in eastern Sichuan

451
(Sichuan basin) and southern Guangxi (Guangxi hills) (Fig. 5-e). This may be related to the

504
All data generated or analysed during this study are included in this published article. 505

506
The authors declare no conflict of interest.