Identifying Critical Meteorological Elements for Vegetation Coverage Change in China

Intensifying global climate change has a significant influence on the vegetation, which is the basis of most of Earth’s ecosystems. It is urgent to identify the critical meteorological elements of vegetation coverage changes to address the problems induced by climate change. Many studies, ranging from theoretical advances to data-driven analyses, have been devoted to investigating meteorological elements’ roles in changing vegetation coverage. However, little has been considered in the aspect of the meteorological elements’ seasonal scale in data-driven studies. Herein, taking China as an example, we collected satellite-derived vegetation coverage data from 2000 to 2020. We then analyzed the meteorological elements, on a seasonal scale, that affect the vegetation coverage change in terms of temperature, precipitation, and solar radiation. We revealed that the critical meteorological elements facilitating vegetation coverage area change differ in both time and space and gave a detailed analysis in line with such findings. Moreover, an apparent seasonal delay effect of meteorological elements on the vegetation coverage change is uncovered.


INTRODUCTION
Changes in ecosystem structure and function are caused by climate change, topography, and human activities, such as afforestation [1,2]. Vegetation is the medium of land-air interaction, and it purifies the air and provides food [3][4][5][6][7]. Vegetation growth requires three processes: photosynthesis is a process in which green plants absorb water and CO 2 through leaf stomata, produce organic matter, and release O 2 under the action of visible light and enzyme catalysis; transpiration is a process in which plants absorb water from roots, only 1%-5% of which is used for photosynthesis, and the remaining is emitted into the air through leaf stomata; respiration is the process of oxidative decomposition of plants to release energy, water, and CO 2 [8,9]. The vegetation mainly interacts with the outside world through leaf stomata; the leaf stomatal state and conductance has remarkable difference under different climatic conditions [10,11]. Most vegetation only opens stomata under light and interacts with the external environment; stomatal conductance is determined by temperature, moisture, humidity, and CO 2 concentration [12][13][14][15]. Thus, temperature, precipitation, humidity, solar radiation, and CO 2 concentration will affect vegetation growth [16][17][18][19][20].
The scale has always been a research hot spot in the field of ecology [1]. In recent years, most scholars have been carrying out the influence of climate change on vegetation in different temporal and spatial scales based on remote sensing data [21][22][23][24][25]. On the global scale, it is proposed that the growing season vegetation change in the high latitudes of the northern hemisphere is governed by temperature, the arid and semi-arid areas are dominated by precipitation, and Amazon and South and East Asia are dominated by solar radiation [16,17]. Chen et al. [26] studied the impact of different climatic periods on vegetation change in the Northern Hemisphere from 1982 to 2013 and found that the impact of temperature on vegetation gradually decreased from spring to autumn. Conversely, the impact of solar radiation on vegetation increased. On the regional scale, Qu et al. [3] studied the key meteorological factors affecting vegetation growth during the growing season in China, indicating that northern China is mainly affected by precipitation, and other regions are affected by temperature. Piao et al. [21] studied the correlation between NDVI and climate variables for temperate grassland in China from 1982 to 1999, suggesting that the trend change of the NDVI caused by climate change is different between different vegetation types and seasons. Zhou et al. [27] analyzed the relationship between climate variability and NDVI in eastern China through correlation analysis, which showed that the NDVI in the arid area was negatively correlated with temperature and positively correlated with precipitation. The dominant factor of vegetation change is the temperature in southern China [23].
We know that the critical meteorological factors affecting vegetation change are different among different regions, decades, seasons, and vegetation types. The previous research mainly studies how temperature and precipitation affect vegetation change through correlation and trend analyses. Still, little is known about how vegetation growth responds to solar radiation change and quantifies the impact of seasonal meteorological factors on vegetation change in China. In addition, studies have shown that the impact of meteorological elements on vegetation growth has a delayed effect [28][29][30], Saatchi et al. [31] investigated that low-frequency drought events in the Amazon cause continuous change of forest canopy. Vegetation growth is driven by the current climate conditions and depends on early climate conditions. Therefore, the delay effect must be considered when studying the impact of meteorological elements on vegetation. The seasonal effect of meteorological elements on vegetation coverage is unclear in China. Understanding how vegetation change responds to meteorological elements is conducive to predicting and evaluating future vegetation changes.
This study aims to identify the critical meteorological element (temperature, precipitation, and solar radiation) periods that influence vegetation growth in different regions of China. Using reanalysis of meteorological elements and satellitederived vegetation coverage data in China during 2000-2020, we examined the relationship between meteorological elements and vegetation coverage change. We explored the delay effects of meteorological elements on vegetation coverage change on a seasonal scale. More importantly, we analyzed how meteorological elements in different seasons (winter, spring, summer, and autumn) have influenced vegetation coverage in the growing season using least absolute shrinkage and selection operator (Lasso) regression analysis.

Data and Preprocessing
Vegetation coverage (VC) is a critical indicator of vegetation growth, which is often used in research fields of ecology, climate, hydrology, and so on [32]. The monthly VC data are calculated from the NDVI data in the 1-km monthly synthetic product of the moderate-resolution imaging spectrometer (MODIS) according to Eq. 1.
where NDVI s represents the pixel value without vegetation coverage, and NDVI v represents the pixel value of complete vegetation coverage. The monthly VC at a spatial resolution of 0.01°× 0.01°was observed during 2000-2020. Vegetation coverage less than 0.05 is considered as non-vegetated areas, which are not considered in the study [33]. VC images of the growing season (April-October) [33][34][35][36][37], spring (April-May), summer (June-August), and autumn (September-October) were obtained by calculating the mean of the corresponding months [21]. The meteorological elements (0.25°monthly 2m temperature (TEM), total precipitation (TPR), and surface net solar radiation (SSR)) for the period 1999-2020 are obtained from ERA Interim Data from the European Centre for Medium Range Weather Forecast (ECMWF). We resample the VC data to ensure the same resolution of the meteorological element data, and the linear trend of the VC and meteorological element time series was removed before statistical analysis [26,38].

Partial Correlation Analysis
We analyze the data for partial correlation to explore the relationship between VC and single meteorological elements (TEM, TPR, and SSR) after controlling the influence of other variables. We calculate the partial correlation coefficient between meteorological elements and VC. Strong partial correlation means that meteorological element exerts strongly impact VC change in the regions. The critical value of the partial correlation coefficient at the 5% significance level is 0.455. We can significantly correlate VC and meteorological elements if the calculated absolute values are more significant than the critical value. The partial correlation coefficient between vegetation coverage and SSR after controlling the two variables (TEM and TPR) is r SSR,VC,TEM,TPR r SSR,VC,TEM − r SSR,TEM,TPR · r VC,TEM,TPR where r SSR,VC,TEM r SSR,VC − r SSR,TEM · r VC,TEM where r SSR,VC,TEM is Pearson's correlation coefficient. Similarly, the partial correlation coefficient between VC and TEM and TPR can be obtained. The influence of meteorological elements on VC in different seasons is determined by partial correlation analysis. Similarly, partial correlation analysis also explains the seasonal delay effect of meteorological elements on VC.

Lasso Regression
To determine the critical meteorological element periods affecting growing season VC change in China, Lasso regression was used to study the relationship between seasonal meteorological elements and VC change in the growing season [39][40][41]. Lasso is a classical variable selection model which reduces the regression coefficient of insignificant variables to 0, retains only a few significant variables, and vastly reduces the influence of multicollinearity between variables. Lasso regression is suitable for defining the critical climate stages affecting VC change in the growing season in China. The regression coefficient of Lasso output can effectively interpret the influence of meteorological elements in different seasons on VC in the growing season. The mean vegetation coverage in the growing season is used as the index of annual vegetation growth condition; the meteorological elements (TEM, TPR, and SSR) in four seasons (previous year winter, spring, summer, and autumn) are used as independent variables, and the annual vegetation growth condition is used as dependent variables to construct the regression model. In order to eliminate the influence of different variable dimensions on the vegetation model, the data of the input are standardized at the time. In addition, to study the changes of meteorological elements affecting vegetation growth in different decades, we divided the data from 2000 to 2020 into two stages: 2000-2010 and 2011-2020.

Basic Characteristics of Vegetation and Meteorological Elements
The VC gradually increases from northwest to southeast from 2000 to 2020 (Figure 1 The annual average spatial distribution of meteorological elements, TEM and TPR, gradually increase from the northwest to southeast (Figures 2A1-C1). The spatial distribution of TEM and TPR is consistent with the VC characteristics; that is, the area with appropriate TEM and sufficient TPR has higher vegetation coverage. The annual average TEM in China is about −12-24°C, the lowtemperature area is mainly distributed in southern Xinjiang and Tibet, and the high-temperature area is distributed in South China. In northern China, the mean annual TPR is 0-200 mm, which belongs to arid and semi-arid area [42]. In southern China, the mean annual TPR is about 600 mm, which pertains to humid areas. The SSR remains in the range of (1000-1400) × 10 4 J/m 2 in other areas, except for Xizang and Yunnan during 2000-2020. The TEM shows a warming trend, but the Tarim Basin in Xinjiang tends to get colder during 2000-2020. Except for northeast, Qinghai, and Sichuan, the TPR of other provinces generally shows a downward trend. The SSR increased in the northwest and northeast and decreased in the middle (Figures 2A2-C2). Figure 3 and Table 1 show the partial correlation coefficients between VC change and meteorological elements (TEM, TPR, and SSR) in spring, summer, and autumn. Meteorological elements have obvious temporal and spatial heterogeneity on VC change [43,44]  growth. The partial correlations between VC and previous-season TEM, TPR, and SSR are provided in Figure 4.  Table 2). It shows that the seasonal effect of previouswinter TPR and SSR on spring VC change is stronger than that of TEM. However, the influence of TEM and SSR in spring and summer on VC change in summer and autumn is stronger than that of TPR.

Drivers of Growing Season Vegetation Cover Change
Meteorological element change is thought to affect growing season VC change, including TEM, TPR, and SSR in four seasons (previous winter, spring, summer, and autumn). We study the key drivers of vegetation coverage change in different regions of China and efforts to quantitatively analyze the contribution of each variable to vegetation coverage increase in a different decade. Precipitation-driven VC increase is most evident in northwest China and Inner Mongolia. The regions belong to arid and semi-arid areas and lack water. Therefore, the increase of precipitation can promote VC increase. SSR is a necessary condition for photosynthesis, and the increase of solar radiation in an appropriate range is conducive to vegetation growth. As  Frontiers in Physics | www.frontiersin.org January 2022 | Volume 10 | Article 834094 7 the SSR in North China and central China is low, the increase of solar radiation is conducive to vegetation growth. In Yunnan, the SSR positively impacted VC change in the 2000s, but the impact of SSR gradually decreases and the impact of precipitation gradually increases in the 2010s. In the coastal area of South China, it is mainly affected by temperature in the 2000s, but the influence of temperature on it gradually weakens and precipitation gradually increases in the 2010s. From the 2000s to 2010s, the impact of precipitation on the change of vegetation coverage in China gradually increased. The effects of radiation and temperature gradually weakened. It is also found that the vital meteorological factors promoting vegetation growth are different in different decade periods. Among all meteorological elements, the driving factors of vegetation change in the growing season in China are mainly precipitation in spring and summer, followed by radiation in spring. In 2000s, 35% of Chinese vegetation coverage was dominated by precipitation, mainly concentrated in northern China. From the 2000s to 2010s, the area dominated by temperature increased from 32% to 36%, but the radiation decreased from 32% to 29% ( Figure 5). There are apparent inter-decadal changes in temperature and radiation.

DISCUSSION AND CONCLUSION
This article analyzes the relationship between meteorological elements and vegetation on a seasonal scale in China. The influence of meteorological elements on growing season vegetation change has obvious seasonal shift and inter-decadal difference. The conclusions are as follows: In China, the high-coverage areas gradually increased and the medium-and low-coverage areas gradually decreased from 2000-2005 to 2016-2020. Low-coverage areas are mainly distributed in Xinjiang and Inner Mongolia as well as Tibet, and high-coverage areas are in southern and northeast China; TEM and TPR gradually increase from the northwest to southeast (Figures 2A1-C1). The spatial distribution of TEM and TPR is basically consistent with the VC characteristics.
The results show that the key meteorological factors affecting vegetation change are different in different regions and seasons. In arid and semi-arid area, especially Inner Mongolia and Xinjiang [47,48], the VC change mainly depends on TEM, whereas the relationship between VC change and TPR and SSR is weak during spring. There is a strong positive correlation between VC change and TPR during summer and autumn, but the relationship between VC change and TEM is negatively correlated (Figure 3). TPR increase can promote vegetation growth and warming will aggravate water loss and inhibit vegetation growth [45,46]. These findings are consistent with the response of vegetation to climate in arid and semi-arid areas [3,16,17]. The vegetation growth in spring mainly depends on the previous-winter TPR and SSR, and the vegetation growth in autumn depends on the summer TPR ( Figure 4). In addition, our study also shows that the increase of TPR in spring and summer is conducive to the increase of annual vegetation coverage (Figures 5C1-C3). In northeast China, spring TEM and summer VC change has a strong relationship. In the North China Plain, the VC change during spring is significantly positively correlated with TEM and TPR. During summer and autumn, it is significantly positively correlated with TPR, but the correlation with TEM is weak. Vegetation growth in summer mainly depends on spring TPR; spring SSR has little effect on summer VC. This is consistent with previous studies [3,22,49]. The increase of autumn SSR and winter and spring TPR is conducive to the increase of VC. In humid areas, such as South China and East China, the VC change mainly depends on the change of TEM in every season, and there is a statistically significant correlation between VC change and SSR (Figure 3). These findings are consistent with those of previous studies [3,23]. TEM has a strong seasonal delay effect on VC change, but TPR is relatively weak (Figure 4).
Through the research, we can know the critical meteorological elements affecting vegetation change in different regions of China. Critical meteorological element modeling can more accurately predict future vegetation coverage. In the course of this study, the scale of vegetation and climate data does not match, and interpolation may cause errors in the data set. On the other hand, the study ignored the impact of human activities, such as ecological engineering and crop irrigation.

DATA AVAILABILITY STATEMENT
The original contributions presented in the study are included in the article/Supplementary Material; further inquiries can be directed to the corresponding author.

AUTHOR CONTRIBUTIONS
HB is responsible for data processing, drawing, and article writing. GS, LL, and YW are responsible for raising scientific issues. ZG and GF provided data.