Effects of Effective Precipitation and Accumulated Temperature on the Terrestrial EVI (Enhanced Vegetation Index) in the Yellow River Basin, China

Abstract

To identify the vegetation dynamics and relationship with the hydrothermal conditions in the Yellow River basin (YRB), the spatial–temporal variations of EVI, effective precipitation (Epr), accumulated temperature (At), and their relationships were obtained based on the MODIS EVI data and meteorological data from the YRB during 2001–2020. The results indicate that EVI trends increased during 2001 to 2020, especially in the farmland, forestland, and grassland ecosystems. Epr and At have also increased over the last 20 years. Epr mostly increased faster in the grassland, and water bodies and wetland ecosystems. At mostly increased faster in the water bodies and wetland, desert, and forest ecosystems. Affected by Epr and At, the correlation between the EVI and hydrothermal conditions varied under different hydrothermal conditions. Compared to the At, the Epr was the restrictive factor for the EVI variations in the terrestrial ecosystem in the YRB. In addition, the dynamical thresholds of the EVI, Epr, and At were confirmed. This study can improve the understanding of vegetation variations and their response to regional climate change, which is critical for ecological conservation and the high-quality development of the YRB.

1. Introduction

Vegetation is a vital component of the terrestrial ecosystem. It is not only the link between the atmosphere, soil, and water bodies but also the basis for the survival of other organisms [1,2,3]. Climate is a direct driver of vegetation change and determines the distribution of surface vegetation. In turn, vegetation greatly influences the carbon and water cycles of terrestrial ecosystems. It serves as an indicator of the relationship between terrestrial ecosystems and climate change [4]. Therefore, it is necessary to study the interactions between the climate and terrestrial ecosystems. Some scholars argued that climate warming was the direct cause of vegetation variations in the northern hemisphere [5,6], while Braswell pointed out that warming showed a suppressive effect on vegetation growth [7].

Numerous studies have different emphases on various scales. On a global scale, the NDVI was significantly positively correlated with temperature in the northern middle- and high-latitude zones, and it was significantly negatively correlated with precipitation in some semiarid regions [8]. However, increasing precipitation promoted the growth of grassland except in the Arctic region [9,10,11,12]. In China, vegetation distribution patterns have significantly altered under the climate changes over the last half century [13,14,15,16]. In the Tibetan Plateau, vegetation changes largely coincided with the regional and seasonal characteristics of temperature and precipitation [3]. In the Qilian Mountains, temperature was the major factor affecting vegetation [17]. However, in the Loess Plateau, precipitation was a key factor in promoting vegetation growth [18]. Therefore, precipitation and temperature acted differently on vegetation changes [19].

Most previous studies largely concentrated on precipitation and temperature which affected vegetation growth and patterns. Although surface vegetation was directly influenced by hydrothermal conditions, there were cumulative effects and lags in the response of vegetation to climate driving forces [20]. Considering hysteretic effects and the actual vegetation growth, in contrast with other studies, this study selected accumulated temperature (At) and effective precipitation (Epr) as the key indicators to analyze the relationship between hydrothermal conditions and vegetation. At was the sum of all values of the daily average temperature ≥10 °C for 365 days of the year [21,22,23]. Epr was the net water quantity that can be determined using precipitation and the effective utilization coefficient [24]. For vegetation, the normalized difference vegetation index (NDVI) and enhanced vegetation index (EVI) were used as critical indicators of vegetation growth [25,26,27]. They are also effective indicators to study global and regional vegetation growth and change in surface vegetation cover [28,29]. However, compared with NDVI, EVI is more sensitive to topographic effects and has more advantages in the monitoring of sensitive and high biomass conditions [30,31]. Meanwhile, water and heat conditions were the main factors driving vegetation variations [32].

The YRB is very sensitive to global warming [33]. Its vegetation improved over the past 32 years and will mainly remain either unchanged or continuously improve in the future [27,34,35,36,37]. In order to better understand vegetation variations in response to hydrothermal conditions, based on the MODIS EVI data and meteorological data from the YRB during 2001–2020, this study aimed to: (a) explore the spatiotemporal pattern of EVI in the YRB since 2001; (b) identify the spatiotemporal distribution of the Epr and AT over 20 years; (c) explore the main hydrothermal conditions affecting vegetation variations and patterns; and (d) confirm the dynamical thresholds of EVI, Epr and At. Our study improves the understanding of terrestrial vegetation variations and their response to regional climate change, which is critical for regional ecological conservation and high-quality development. To investigate the response relationship between regional vegetation and climate change should provide scientific theory and a basis for the sustainable development of the regional ecological environment.

2. Materials and Methods

The Yellow River, originating in the Bayankara mountain range in Qinghai province, covers an area of approximately 752,443 km², flows through nine provinces, and empties into the Bohai Sea. As showed in Figure 1, YRB is located at 96°–119° E, 32°–42° N across four geomorphic units, including the Qinghai–Tibet Plateau, Inner Mongolia Plateau, Loess Plateau, and Yellow Huaihai Plain. The basin lies in the mid-latitude zone, which is affected by atmospheric and monsoon circulation in a complex manner [38,39]. This leads to significant climatic variability in different regions of YRB. The YRB spans three climatic divisions from north to south: arid (annual precipitation 0–200 mm), semiarid (annual precipitation 200–400 mm), and semihumid (annual precipitation 400–800 mm), with most of the area located in the arid and semiarid zones. The YRB has a large population and complex geographical and climatic conditions, and the sustainable development of its population, resources, and environment has been a major concern.

In this study, linear regression, Kendall test and correlation analysis were used to investigate the spatiotemporal patterns and variations of hydrothermal conditions and vegetation, on both basin scale and pixel scale. Then relationship between EVI and hydrothermal conditions were discussed. For hydrothermal conditions, Epr and At were selected as the main factors influencing the EVI changes (Figure 2).

2.1. Data Sources and Data Processing

2.1.1. MODIS EVI Data

The EVI was used as an indicator to reveal vegetation patterns [36,40,41]. The MODIS EVI data were derived from the NASA MODIS global vegetation index product (https://www.nasa.gov/, accessed on 30 September 2021) with a resolution of 250 m and a time scale of 16 days. EVI data from January 2001 to December 2020 were selected for the YRB, with 23 periods per year and a total of 460 images. The annual maximum synthesis method was used to obtain the 2001–2020 scale EVI series using ArcGis 10.7 software platform.

2.1.2. Meteorological Data

In this study, daily temperature and precipitation data of the YRB from 100 meteorological stations for the period 2001–2015 and 93 meteorological stations for the period 2016–2020 (the distribution of stations is shown in Figure 1) were selected from the National Meteorological Science Data Sharing Service Platform (http://www.resdc.cn, accessed on 5 October 2021). The relevant AML language was compiled using the ArcInfo Workstation 10.2 platform, and the IDW (inverse distance weighted) spatial interpolation method was used to graticule the data into 250 m resolution daily temperature and precipitation data from 2001 to 2020. The raster layers were obtained for the annual Epr and At in the YRB from 2001 to 2020.

2.1.3. Terrestrial Ecosystems Data

The data of terrestrial ecosystem types for 2015 in the YRB were obtained from the Resource and Environment Science Data Center of the Chinese Academy of Sciences (http://www.resdc.cn/, accessed on 5 October 2021). As indicated in Figure 3, the ecosystem types were classified into seven major categories: grassland, farmland, forest, settlement, desert, water bodies and wetlands, and others, accounting for 47.5%, 25.9%, 13.4%, 5.9%, 2.8%, 2.7%, and 1.8% of the total area, respectively. It can be concluded that the terrestrial ecosystem of the YRB is dominated by farmland, forests, and grasslands.

2.2. Methods

A linear trend analysis and the Kendall test were used to study the extent of hydrothermal conditions and EVI variations at different scales in the YRB. Effects of the hydrothermal conditions on the EVI were studied based on a correlation analysis.

2.2.1. Linear Regression Method

The annual series data of Epr, At, and EVI in the YRB from 2001 to 2020 were used for the trend analysis of the elements using the linear propensity rate method [42,43]. $x_i\left(i=1, 2, \dots , n\right)$ indicates a certain Epr, At or EVI;$t_i\left(i=1, 2,\dots ,n\right)$ indicates the time corresponding to x_i. The one-dimensional linear regression equation is:

$$x_i=a+b{t}_i, i=1,2,\dots ,n$$

(1)

where $a$is a constant term, $b$ is the tendency rate, $b$ > 0 means that the study object exhibits an increasing trend in the statistical time period and vice versa, and the absolute value indicates the degree of change.

2.2.2. Mann–Kendall Test

The Mann-Kendall test was used to test the significance of the trends in Epr, At and EVI in the Yellow River basin [40,44]. The test statistic U was constructed according to Equation (2), and given the significance level $\alpha$, the trend was considered insignificant when $\left|U\right|<U_{a/2}$ and significant otherwise.

$$U=\tau /\sqrt{\mathrm{Var}\left(\tau \right)}$$

(2)

Among them:

$$\tau =\frac{4p}{n\left(n-1\right)}-1$$

(3)

$$\mathrm{Var}\left(\tau \right)=\frac{2\left(2n+5\right)}{9n\left(n-1\right)}$$

(4)

where: n is the length of the sequence, text $n=16$; $p$ is the number of occurrences of $x_i<x_j$ in the given sequence $\left\{x_i\right\}\left(i=1,2,\cdots ,n\right)$ for all pairs of values $\left\{x_i,x_j,i<j\right\}\left(i=1,2,\cdots ,n-1;j=i+1,i+2,\cdots ,n\right)$.

2.2.3. Correlation Analysis

The correlation between EVI and Epr or At in the YRB from 2001 to 2020 was analyzed at the image element level and on a regional scale. The influence of the hydrothermal conditions on the EVI in the YRB was studied based on the results of the correlation analysis [22,45]. The correlation coefficient was calculated as follows:

$$r_{xy}=\frac{{\displaystyle {\sum }_{i=1}^n(x_i-\overline{x})(y_i-\overline{y})}}{\sqrt{{\displaystyle {\sum }_{i=1}^n{(x_i-\overline{x})}^2}}\sqrt{{\displaystyle {\sum }_{i=1}^n{(y_i-\overline{y})}^2}}}$$

(5)

where $r_{xy}$ is the correlation coefficient between factors$x$ and y; $x_i$ and$y_i$are the values of factors $x$ and y in year $i$, respectively; $\overline{x}$ and $\overline{y}$are the multiyear averages of factors $x$ and y in the study period, respectively. $r_{xy}>0$ indicates that factors $x$ and y are positively correlated and vice versa; the larger $\left|r_{xy}\right|$, the stronger the correlation.

3. Results and Discussion

3.1. Spatiotemporal Patterns of EVI in the YRB

3.1.1. Temporal Variation of EVI

The variations of EVI exhibited an increasing trend in different ecosystems in the YRB, but the change slopes varied (Figure 4). The variation slope of EVI from 2001 to 2020 was about 0.0040 yr⁻¹ over the whole of the YRB. The trends of EVI were approximately 0.0049 yr⁻¹ and 0.0045 yr⁻¹ in the farmland and forest ecosystems, respectively; about 0.0038 yr⁻¹ in the grassland ecosystems; and approximately 0.0014 yr⁻¹ and 0.0017 yr⁻¹ in the water and wetland ecosystems and desert ecosystems, respectively. The slope of the EVI in the farmland and forest ecosystems was higher than the average slope in the YRB, and the slope of the others was lower than the mean slope over the whole basin.

The artificial cover in the agroforestry ecosystem consists primarily of crops (accounting for 71%). Their growth cycles are short and perform the significant seasonal variations. When lacking precipitation, substantial irrigation still meets the water requirements for crop growth. This could explain the greater slope of EVI in the farmland ecosystems. In addition, a series of ecological constructions and projects have been implanted in the YRB since the 1990s [11,46]. These artificial activities intervened and changed the land-use types of the YRB, especially the project of returning farmland to forests and grassland in 1999. Specific interventions included the conversion of: arable land to woodland, scrub and grassland; grassland to woodland and scrub; and desert to woodland, scrub, and grassland in local areas. As a result, the EVI of the farmland, grassland, and desert ecosystems dramatically increased after 2000.

3.1.2. Spatial Pattern of EVI

Figure 5 shows the spatial distribution of the multiyear mean values of the EVI for the YRB. Overall, the mean value was 0.39, and the EVI gradually increased from the northwest to southeast. Based on the existing research results, when the EVI was between 0.3 and 0.45, the vegetation cover can be classified as low to medium [12,47]. Therefore, the overall vegetation coverage of the YRB has been moderately low over the past 20 years. The low-value areas (EVI < 0.3) of the YRB accounted for 33% of the area and are predominantly located in the Inner Mongolia Plateau, Loess Plateau, Qinghai–Tibet Plateau, and west of the Taihang Mountains. The research indicates that the 400 mm annual precipitation line was the dividing line between semiarid and semihumid areas [48,49]. The majority of the Inner Mongolian Plateau, Loess Plateau, Qinghai–Tibet Plateau, and areas to the west of the Taihang Mountains located west of the dividing line is where precipitation was less than 400 mm. Therefore, low precipitation can lead to low EVI. For the Qinghai–Tibet Plateau, the low temperature was also an important reason for the low EVI due to its high altitude [50]. In addition, the vegetation cover in the northwestern part of the basin was dominated by desert and decertified grassland, whereas the Yellow River source area belonged to the typical mountain-type plateau and was widely covered with bare rocks and sparse grassland [51]. This also explained the low EVI in the northwestern part of the basin to some extent. Therefore, the vast low EVI areas were also responsible for the overall low vegetation cover level in the YRB.

Figure 6 reveals the spatial average EVI along the different elevations. When the elevation was greater than 3500 m or less than 1000 m, the average EVI decreased with elevation. When 1000 < elevation < 3500, the average EVI and elevation increased at the same time. From the watershed perspective, the EVI in the upper and lower reaches of the YRB were higher than those in the middle reaches [52,53].

The upstream areas were less disturbed by anthropogenic activities, with large areas of primary forests (e.g., Sanjiangyuan National Park and Qilianshan National Park) and nature reserves (e.g., Qilianshan Nature Reserve, Xinglongshan Nature Reserve, and Helanshan Nature Reserve) [13,16]. In contrast, the downstream areas were predominantly artificial vegetation (crops) [54]. Therefore, the overall vegetation cover level was better. Furthermore, the midstream areas were mostly located in arid and semiarid climate zones, while the upstream and downstream areas were mostly located in humid and semihumid zones [14,15]. As a result, due to the influence of water and heat conditions, their vegetation was relatively the least suitable.

3.1.3. Spatial Distribution of EVI Tendency

Figure 7 illustrates the spatial characteristics of the EVI trends in the YRB from 2001 to 2020. As shown in the figure, the overall EVI trend is increasing. The proportion of the area with vegetation growth was 87%, which was much larger than the area with a decrease in vegetation (13%). The vegetation growth areas were predominantly located in the Qilian Mountains, Ningxia Plain, Loop Plain, Loess Plateau, northern Guanzhong Plain, southern Gansu province, and western Shanxi province. However, areas with significant changes (p < 0.05) accounted for approximately 13% of the total area and were primarily distributed in the areas with reduced vegetation. The possible reasons for the significant decrease were that these areas were mainly concentrated on both banks of the Yellow River in Qinghai, Henan, and the Shandong province and on both banks of the Wei River, which are relatively densely populated. The vegetation is vulnerable to human activities, and the accelerated urbanization process exacerbated the vegetation degradation to some extent, resulting in some reduction in vegetation cover.

3.2. Spatiotemporal Distribution of Epr and At in the YRB

3.2.1. Temporal Variation of Epr and At

As shown in Figure 8, the Epr of both the YRB and each ecosystem has increased over the past 20 years. The rate of increase in the Epr was 3.35 mm/yr, with the rates of change in different ecosystems as follows: grassland (4.12 mm/yr) > water bodies and wetland (3.89 mm/yr) > forest (2.62 mm/yr) > farmland (2.49 mm/yr) > desert (1.38 mm/yr). The Epr growth rate of grassland, and water bodies and wetland ecosystems was greater than that of the whole basin, while the growth rate of the forest, farmland, and desert ecosystems was less than that of the whole basin. Among them, the Epr in the YRB sharply increased in 2003 and 2006. These could have been mainly affected by meteorological changes, such as atmospheric circulation and tropical cyclone activity. In 2016, the Epr increased compared with previous years, occurring mostly during the flood season. In June, influenced by the low-pressure trough from northeast China to Chongqing, the precipitation in the central-eastern part of the YRB was higher than normal. In July, influenced by the short-wave trough of the westerly wind belt and the northwest Pacific high, the overall precipitation in the YRB was high. In August, the southern part of the basin was controlled by continental high pressure, while the subtropical high pressure was weak, the westward extension of the ridge point shifted eastward, and more precipitation occurred in the north with lower amounts in the south [11,55,56].

As shown in Figure 8, the multiyear At in the YRB increased at the rate of 7.86 °C/yr, with the increases of various ecosystems as follows: water bodies and wetland (11.12 °C/yr) > desert (10.30 °C/yr) > forest (9.33 °C/yr) >farmland (8.25 °C/yr) > grassland (7.23 °C/yr). The At increase in the water bodies and wetland, and desert ecosystems was greater than that of the entire basin, while the changes in the forest, farmland, and grassland ecosystems were smaller than that of the entire basin. The Kendall test indicated (Figure 9) that the hydrothermal conditions of the major terrestrial ecosystems in the YRB did not significantly vary.

The growth rates of EVI, Epr, and At indicated some differences. The annual Epr and At growth rates of farmland and forest ecosystems were smaller than those of the grassland, and water bodies and wetland ecosystems. However, the interannual EVI growth rates of the farmland and forest ecosystems were the highest of all the ecosystem types. This is because, although the growth rates of the Epr and At of the grassland, and water bodies and wetland ecosystems were large, the base was small, and the overall level was lower than that of farmland and forest ecosystems, especially the At [37,57]. The annual At of the farmland and forest ecosystems was greater than 3000 °C on average, while that of the grassland, and water bodies and wetland ecosystems were below 3000 °C. In addition, although the growth rates of the Epr and At for the forest ecosystems were higher than those of the farmland ecosystems, the EVI growth rate in farmland ecosystems was higher than that of forest ecosystems because, in addition to precipitation, farmland fields were also irrigated and fertilized [58,59].

3.2.2. Spatial Patterns of Epr and At

During the period of 2001–2020, the multiyear average Epr and average At of the YRB were 428.70 mm and 3084.78 °C, respectively.

As demonstrated in Figure 10a, the multiyear average Epr increased from the northwest to the southeast. The regions with Epr above 600 mm were mainly located in the northern Sichuan province, southern Xi’an City, northwestern Henan province, and southern Jinan City in the Shandong province. From Figure 11a, the mean value of the multiyear Epr in the YRB exhibited a decreasing–increasing–decreasing trend with various elevations. In the areas below 1000 m above sea level in the YRB, the average Epr significantly decreased with elevation increasing by approximately 14.3 mm per 100 m increase in elevation. In the area between 1000 and 3500 m, the precipitation gradually increased by approximately 5 mm for every 100 m increase in elevation. In the area above 3500 m, the precipitation increased with increasing elevation. The reason for the two distinct dividing points between 1000 m and 3500 m in the Epr is probably because the area below 1000 m in elevation in the YRB is located predominantly in the central and eastern parts of the basin and is further inland with higher elevation. Further inland, the water vapor is blocked, resulting in less rainfall [60,61]. When the altitude is between 1000 m and 3500 m, the air rises owing to the high terrain and readily condenses into rain [62,63]. In addition, the temperature decreases as the altitude rises, and evaporation decreases, resulting in higher Epr. Therefore, the precipitation increases with the increase in elevation [64,65]. When the altitude reaches 3500 m, the precipitation decreases as the altitude continues to rise because the altitude is too high. The water vapor in the air decreases due to the large amount of precipitation after reaching a certain altitude. At the same time, the effect of rainfall reduction by evaporation on the total rainfall is very slight because the temperature is too low. Therefore, the precipitation decreased with the increase in elevation [66,67].

As shown in Figure 10b, the multiyear average At in the YRB gradually increases from west to east, decreasing with elevation increasing by the approximate linear trend in Figure 11b. The multiyear average At decreases by approximately 77 °C for every 100 m with elevation increasing. When the altitude exceeds 4000 m, elevation increases and the speed of At decreasing becomes slower. The annual At is the sum of the daily average temperature during which the temperature ≥10 °C [66,67,68]. Usually, in highland mountainous areas (at roughly the same latitude), the relationship between temperature and altitude is that the higher the altitude, the lower the temperature. This is because at high altitudes, the air pressure is low, the air is thin, and the atmosphere is poorly insulated. This leads to a large amount of heat dissipation and lower temperatures; for every 1000 m of elevation rise, the temperature drops 6 °C. The At is the sum of all values of the daily average temperature ≥10 °C for 365 days of the year. Therefore, the At also decreases with increasing altitude.

An analysis of the trends of Epr and At in the YRB over the past 20 years (Figure 12) indicates that the spatial variability of the Epr changes was substantial, with 12% of regions experiencing a decrease and 88% experiencing an increase. The areas with a decrease were concentrated in the middle reaches of the Inner Mongolia Autonomous Region, the Weihe River basin, and the lower reaches of the Yellow River. Those with an increase were predominantly concentrated in the Yellow River source area of Qinghai province, the Ningxia Plain, the central part of Gansu Inner Mongolia Plateau, Loess Plateau, Guanzhong region, and the central and western regions of the Shanxi province (Figure 12a). According to the significance (p < 0.05) analysis, the Epr in the YRB was mostly increasing but not significantly. Less than 20% of the area exhibited a significant change (p < 0.05) (Figure 12b), mostly with a significant increase, predominantly located in the upper reaches of the Yellow River in the Qinghai province and parts of the Sichuan province.

Except for the Inner Mongolia Plateau and parts of the Guanzhong Plain, the At increased in most of the basin, accounting for approximately 87% of the total area, while the area with a decrease accounted for only 13% (Figure 13a). The area of the basin with a significant (p < 0.05) increase in the At was 41% (Figure 13b). It is mainly in the Qinghai–Tibet Plateau and the Yellow–Huaihai Plain, the areas along the banks of the Yellow River in Ningxia and the Qinling area in the south of the basin on both sides of the Weihe River.

Different regions in the YRB have different warming rates. The annual At and warming rate in the Loess Plateau and the eastern edge of the Qinghai–Tibet Plateau are significantly lower than those in the agricultural and livestock staggered zone of the Gansu province. These differences may be related to the increasing concentration of greenhouse gases [69,70]. It leads to the imbalance between the absorption and emission of energy in each region. The rising trend of CO₂, CH_4, and other greenhouse gases is directly proportional to the change in temperature, and the warming of the climate promotes the promotion of multicrops and the improvement of forest productivity but also increases the uncertainty of agricultural production.

3.3. Relationship between EVI and Hydrothermal Conditions

3.3.1. Variations EVI with Various Epr and At Zones

In the past 20 years, there has been a decreasing–increasing–decreasing relationship between the EVI and multiyear Epr in the YRB. As indicated in Figure 14a, when the Epr per year is less than 200 mm, the EVI in the basin decreases by approximately 0.33 per 100 mm with Epr increasing. Areas with an Epr of less than 200 mm are mainly located in the northwestern part of the basin. Most of these lands are unused and are saline lands. The evaporation intensity of saline soils is strong. The evaporation is much greater than the precipitation, thus inhibiting the growth of vegetation [71,72]. When the Epr is between 200 and 600 mm, the EVI increases approximately 0.1 per 100 mm with Epr increasing. When the Epr is greater than 600 mm, the EVI slowly decreases to about 0.03 per 100 mm with Epr increasing. As the surface slope of these areas is large, and the soil is very loose, it will lead to soil erosion and cause bare ground if there is too much rainfall, thus reducing the vegetation cover [73].

From Figure 14b, the trend and magnitude of the EVI varied in different At ranges. When the annual At was less than 1500 °C, the EVI significantly decreased with the increase in the At by 0.03 per 100 °C increase. When the annual At was between 1500 and 3500 °C, the EVI decreased relatively slowly with the increase in At by 0.01 per 100 °C increase. When the At was greater than 3500 °C, the EVI increased with the increase in At by 0.0157 per 100 °C. The areas with At below 3500 °C in the YRB are mainly located in the Qinghai–Tibet Plateau, Inner Mongolia Plateau, and Loess Plateau areas. These areas are semiarid. In the more arid natural environments, forest and grass vegetation cannot get enough water from the soil. Precipitation is the only source of soil moisture recharge in the region, and the team of vegetation growth and development of the region plays a vital role. Therefore, as At increases, evaporation increases and Epr decreases. This leads to the worse growth of vegetation.

As indicated by the location of the red circle in the Figure 14b, there is a significant abrupt change in the EVI where the At is approximately 1500 °C. Areas with At of less than 1500 °C are primarily located in the Gansu province, which was a fragile ecological environment. It was an important part of an ecological security barrier in China. In recent years, ecological projects, such as “returning farmland to forest and grass”, “three northern protective forests” and “natural forest protection” were implemented, and the vegetation was effectively restored [74]. This may lead the EVI value to significantly increase. The area with At greater than 3500 °C is mainly located in the eastern part of the basin in the Yellow and Huaihai Sea plains, where the effective rainfall is relatively high, and the hydrothermal conditions are relatively good. Therefore, the EVI values significantly increased with increasing At.

3.3.2. Correlations between EVI and Hydrothermal Conditions

The correlations between the EVI and hydrothermal conditions in the YRB are presented in Figure 15a, with more areas showing a positive correlation (accounting for 86%) and those with a negative correlation (accounting for only 14%). The significance (p < 0.05) level is presented in Figure 16a. It indicates that the areas with a significant positive correlation between the EVI and Epr are predominantly located at the beginning of the middle reaches of the Yellow River, including the Loess Plateau area in the central part of the basin as well as some areas in Inner Mongolia and the Taihang Mountains.

As shown in Figure 15b, the correlation between the EVI and At in the YRB has distinct spatial variability, in which the positively correlated areas account for 66% and negatively correlated areas account for 34% of the total. The former is primarily located in the source and downstream areas of the Yellow River, and the latter is concentrated in the areas along both banks of the middle section of the Yellow River, the areas along both banks of the Tudao River, a tributary of the Yellow River, and some areas of the Loess Plateau. Among them, from the significance (p < 0.05) analysis, as shown in Figure 16b, the areas exhibiting a significant positive correlation between the EVI and At were predominantly located in the Tibetan Plateau area in the western part of the YRB, in the Taihang Mountains area, and in part of the Yellow and Huai Plain in the eastern part of the basin.

3.3.3. Different Thresholds of Epr and At in Different Ecosystems

Combining the relationships among the multiyear Epr, At, and EVI (Figure 14), the turning points of Epr and At changes were selected as the corresponding variable thresholds to classify different hydrothermal climate zones. The annual Epr thresholds of 200 mm and 600 mm and At thresholds of 1500 °C and 3500 °C were selected. Furthermore, different ecosystem types in the YRB were superimposed, and the statistical results of the various hydrothermal climate zones of the ecosystems in the YRB were obtained, as shown in

Table 1. Area of different hydrothermal condition zones as a proportion of the basin area.

Hydrothermal Conditions	Area Share of Different Ecosystems %
	Farmland	Forest	Grassland	Water Bodies and Wetland	Desert
Epr < 200 mm, At < 1500 °C	0	0	0	0	0
Epr < 200 mm, 1500 °C ≤ At ≤ 3500 °C	0.07	0	0.23	0.02	0.02
Epr < 200 mm, At > 3500 °C	0.97	0.11	1.78	0.21	0.26
200 mm ≤ Epr ≤ 600 mm, At < 1500 °C	0.10	1.12	8.44	0.54	0.01
200 mm ≤ Epr ≤ 600 mm, 1500 °C ≤ At ≤ 3500 °C	9.49	4.12	20.30	0.80	0.95
200 mm ≤ Epr ≤ 600 mm, At > 3500 °C	14.22	6.92	12.96	0.62	1.41
Epr > 600 mm, At < 1500 °C	0	0.21	3.23	0.49	0
Epr > 600 mm, 1500 °C ≤ At ≤ 3500 °C	0.28	0.16	0.17	0.02	0.04
Epr > 600 mm, At > 3500 °C	0.80	0.72	0.37	0.05	0.14

. In this study, the relationship between the EVI and hydrothermal conditions was focused on these ecosystem types, whose area comprised more than 0.8% of the total area (relatively large area share and more uniform distribution) under different hydrothermal condition partitions, and the results are provided in

Table 2. Correlation coefficients of EVI and Epr on different hydrothermal condition partitions.

Hydrothermal Conditions	Correlation Coefficients of EVI and Epr in Different Ecosystems
	Farmland	Forest	Grassland	Water Bodies and Wetland	Desert
Epr < 200 mm, At < 1500 °C	-	-	-	-	-
Epr < 200 mm, 1500 °C ≤ At ≤ 3500 °C	-	-	-	-	-
Epr < 200 mm, At > 3500 °C	0.075	-	0.411	-	-
200 mm ≤ Epr ≤ 600 mm, At < 1500 °C	-	0.126	0.185	-	-
200 mm ≤ Epr ≤ 600 mm, 1500 °C ≤ At ≤ 3500 °C	0.404	0.273	0.408	0.158	0.294
200 mm ≤ Epr ≤ 600 mm, At >3500 °C	0.199	0.147	0.324	-	0.116
Epr > 600 mm, At < 1500 °C	-	-	0.146	-	-
Epr > 600 mm, 1500 °C ≤ At ≤ 3500 °C	-	-	-	-	-
Epr > 600 mm, At > 3500 °C	0.146	-	-	-	-

and

Table 3. Correlation coefficients of EVI and At (accumulated temperature) on different hydrothermal condition partitions.

Hydrothermal Conditions	Correlation Coefficients of EVI and At in Different Ecosystems
	Farmland	Forest	Grassland	Water Bodies and Wetland	Desert
Epr < 200 mm, At < 1500 °C	-	-	-	-	-
Epr < 200 mm, 1500 °C ≤ At ≤ 3500 °C	-	-	-	-	-
Epr < 200 mm, At > 3500 °C	−0.069	-	−0.158	-	-
200 mm ≤ Epr ≤ 600 mm, At < 1500 °C	-	0.275	0.335	-	-
200 mm ≤ Epr ≤ 600 mm, 1500 °C ≤ At ≤ 3500 °C	0.016	0.082	0.078	0.058	−0.061
200 mm ≤ Epr ≤ 600 mm, At > 3500 °C	0.102	0.136	0.079	-	0.022
Epr > 600 mm, At < 1500 °C	-	-	0.225	-	-
Epr > 600 mm, 1500 °C ≤ At ≤ 3500 °C	-	-	-	-	-
Epr > 600 mm, At > 3500 °C	0.075	-	-	-	-

.

As shown in Table 2, the EVI and Epr in the YRB exhibited a strong positive correlation, especially in areas with Epr greater than 200 mm, which include: all areas in the upper and lower reaches of the basin; all areas in Henan, Shaanxi, and Shanxi in the middle reaches of the basin; and some areas in Ningxia, Gansu, and Inner Mongolia. The YRB is predominantly located in arid and semiarid areas, with a relatively fragile ecological environment and water shortage [75]. The soil water content, transpiration level, and air humidity affect plant growth [76,77]. When water is lacking, plant cell division and elongation are affected, which, in turn, affects vegetation growth [78,79]. In arid and semiarid regions of the basin, vegetation growth is more sensitive to Epr because of low rainfall and high evaporation; therefore, the Epr in this region is the key factor limiting vegetation growth in the YRB.

As can be seen in Table 3, in the area with At less than 1500 °C, all the ecosystems have positive correlation with the EVI, with the highest in the grassland ecosystems. When the annual At was between 1500 and 3500 °C and the Epr was between 200 mm and 600 mm, only the desert ecosystem showed a negative correlation between the EVI and At, and the other ecosystems showed a positive correlation. In areas with annual At greater than 3500 °C, where the Epr was relatively small, such as the farmland and grassland ecosystems, the corresponding EVI was negatively correlated with the At, whereas in areas with more Epr, the EVI was positively correlated. In terms of the relationship between crops and temperature, different crops have different requirements for the AT, such as 2500–3500 °C for corn, 1000–2500 °C for rice, 2500 °C for potatoes, and 2500 °C for wheat. Insufficient At will reduce the crop yield, but if the At is too high, exceeding the maximum temperature to which the crops are adapted, it will also cause crop malnutrition [80]. Typically, crops do not require At higher than 3500 °C. Therefore, the EVI in the farmland ecosystems exhibited a negative correlation with the At in areas where the basin-wide At exceeded 3500 °C. For the grassland ecosystems, the temperature is relatively low in areas with vegetation cover and relatively high in areas without vegetation cover owing to solar radiation and subsurface reflection, and precipitation is the key constraint [81]. Therefore, when the At was greater than 3500 °C and the Epr was less than 200 mm, the EVI of the grassland ecosystems showed a negative correlation with the At.

4. Conclusions and Limitations

This study obtained EVI dynamics and their responses to hydrothermal conditions in the YRB. The primary conclusions are as follows:

(1)

The multiyear EVI increased in the YRB from 2001 to 2020. The EVI decreased–increased–decreased with the elevation increasing. The elevation thresholds (1300 m and 3500 m altitudes) affecting EVI variations were clarified. EVI varied diversely during the different elevation threshold zones.

(2)

For major ecosystems and the whole YRB, the EVI, Epr and At have all increased. Their increasing rates most significantly changed in the farmland, grassland, and water bodies and wetland ecosystems, respectively.

(3)

The EVI varied with the variation of hydrothermal conditions in the YRB over the past 20 years. Nine dynamical threshold zones combining Epr (200 mm or 600 mm) with At (1500 °C or 3500 °C) were confirmed. Compared to the At, the Epr was the main driving factor of the EVI variations in the YRB.

However, with regards to driving factors, accepting the hydrothermal conditions of the vegetation dynamics considered in this study, anthropogenic activities and socioeconomic factors also have effects on vegetation growth. Climate changes and human activities were regarded as major driving factors, which could contribute to the understanding of vegetation variation attribution. In the future, the comprehensive impact and relationship of their factors on vegetation changes need to be quantified and assessed. In addition, evapotranspiration, snow, sunshine, and radiation all should be considered as part of the water and heat conditions. In terms of methods, cross-wavelet analysis could show the time delay between climate and vegetation that arise from land cover type and soil characteristics. Therefore, this method should be applied to investigate the relationship between vegetation variations and climatic changes.

In this study, the spatiotemporal effects of water and temperature on the EVI of the terrestrial ecosystem in the YRB were statistically analyzed at the basin scale. The results provide the basic data and technical support for the study of terrestrial ecosystem responses and adaptation to climate change as well as ecological protection and high-quality development in the YRB.