How Has the Recent Climate Change Affected the Spatiotemporal Variation of Reference Evapotranspiration in a Climate Transitional Zone of Eastern China?

Abstract

Reference evapotranspiration (ET₀) is essential for agricultural production and crop water management. The recent climate change affecting the spatiotemporal variation of ET₀ in eastern China continues to still be less understood. For this purpose, the latest observed data from 77 meteorological stations in Anhui province were utilized to determine the spatiotemporal variations of ET₀ by the use of the Penman–Monteith FAO 56 (PMF-56) model. Furthermore, the Theil–Sen estimator and the Mann–Kendall (M–K) test were adopted to analyze the trends of ET₀ and meteorological factors. Moreover, the differential method was employed to explore the sensitivity of ET₀ to meteorological factors and the contributions of meteorological factors to ET₀ trends. Results show that the ET₀ decreased significantly before 1990, and then increased slowly. The ET₀ is commonly higher in the north and lower in the south. ET₀ is most sensitive to relative humidity (RH), except in summer. However, in summer, net radiation (R_n) is the most sensitive factor. During 1961–1990, R_n was the leading factor annually, during the growing season and summer, while wind speed (u₂) played a leading role in others. All meteorological factors provide negative contributions to ET₀ trends, which ultimately lead to decreasing ET₀ trends. During 1991–2019, the leading factor of ET₀ trends changed to the mean temperature (T_a) annually, during the growing season, spring and summer, and then to R_n in others. Overall, the negative contributions from u₂ and R_n cannot offset the positive contributions from T_a and RH, which ultimately lead to slow upward ET₀ trends. The dramatic drop in the amount of u₂ that contributes to the changes in ET₀ in Region III is also worth noting.

1. Introduction

Evapotranspiration (ET) is a crucial portion of the hydrologic cycle. It participates in surface runoff, groundwater recharge and other key processes, and plays a pivotal role in climate change, hydrological research and irrigation water management [1,2,3]. Reference crop evapotranspiration (ET₀) is the potential evapotranspiration that is further specified in terms of crop characteristics. Doorenbos et al. [4] defined ET₀ as the ET of vast grassland with uniform and normal growth, completely covering the surface and providing sufficient water at a height of 8–15 cm. Subsequently, Allen et al. [5] introduced the concept of ET₀ and defined it as the ET of an ideal 12 cm crop with a fixed canopy resistance of 70 s·m⁻¹ and an albedo of 0.23 (very similar to green grassland with an open surface, uniform height, vigorous growth, complete coverage of the ground and a sufficient supply of soil moisture). Since then, ET₀ has been widely used in the fields of agronomy, agriculture, irrigation and ecology [6].

Currently, the Penman–Monteith FAO (the Food and Agriculture Organization of the United Nations) 56 (PMF-56) model, a modified Monteith equation [7], has been used broadly for estimating ET₀ worldwide based on its solid theoretical base and wide applications [8]. In the PMF-56 model, the climatic factors (i.e., temperature, humidity, wind speed and radiation) are the main influencing elements for ET₀. However, in recent decades, climate change, especially global warming, has stimulated worldwide concerns [9,10,11], which have also led to changes in ET₀ in different parts of the world [12,13,14,15,16]. Although under the influence of global warming, ET₀ is not only affected by temperature, but also by other elements, and the coupling of multiple factors ultimately determines the increasing or decreasing trends of ET₀. Therefore, quantifying the impact of meteorological factors on ET₀ trends is very essential. In recent studies, quantitative methods have mainly been used to assess the effects of meteorological factors on ET₀ trends, for example, the multiple regression analysis [17], partial correlation analysis and stepwise regression [18], detrending method [2,19,20,21,22,23], sensitivity coefficient method and differential method [1,3,14,24,25]. As the aforementioned methods, the differential method can effectively quantify the actual contributions of climate factors to ET₀ trends; therefore, it has been successfully applied in earlier studies.

In recent decades, most scholars have quantified the contributions of meteorological factors to ET₀ trends worldwide, such as China [26,27,28,29,30,31], Slovenia [16], Spain [32], Iran [33], Bangladesh [34], etc. However, the dominant factors of ET₀ may shift under the changing climate [14,35]. Li et al. [36] pointed out that pan evaporation exhibited a distinct downward trend before 1993 and then reversed in Northwest China. Similarly, Han et al. [35] found a downward trend of ET₀ before 1991 and an upward trend after in the Jing-Jin-Ji region of North China, and the dominant factor contributing to ET₀ shifted from wind speed to mean temperature. Wang et al. [14] revealed that the increasing ET₀ trend after the 1990s over China could be attributed to the increasing air temperature, and the most sensitive factor to ET₀ was specific humidity. However, in Southwest China, the sunshine duration was the main contributor to ET₀ trends in the growing season from 1961 to 1996, and the relative humidity was the dominant variable for the increasing ET₀. Although research scholars have conducted relevant studies on ET₀ [14,37,38], ET₀ varies significantly among diverse regions and the trends of meteorological factors are essential to analyze the variation of ET₀. Thus, it is necessary to conduct regional studies of ET₀, especially in eastern China, where there is still a certain knowledge gap in the systematic study of ET₀. This work intends to close the knowledge gap in the existing literature. Our study also offers a rational theoretical basis for regional agricultural water management and irrigation planning.

The Yangtze River Delta urban agglomeration is one of the six major urban agglomerations in the world. It is an active economic development region, with the highest degree of openness and the strongest innovation ability in China. In 2019, the State Council of the People’s Republic of China (PRC) issued the planning scope of the Yangtze River Delta, which officially extended to all the cities in the four provinces of Jiangsu, Zhejiang, Anhui and Shanghai. Anhui province is adjacent to the Yangtze River Delta and is also one of the four provinces mentioned above to have witnessed rapid economic development over the past few decades. Furthermore, Anhui province is one of the 13 major granary provinces, with a total grain output of 34.15 billion kilograms, ranking sixth in China. Among them, the growth rate of the total grain output ranks first among 13 major grain provinces [39]. Thus, an accurate estimation of ET₀, a precise evaluation of its spatiotemporal distribution characteristics and variation trends, as well as the exploration of its influencing factors have scientific implications for agricultural production planning, water resource management, and ecological protection.

Based on the above discussion, we propose the hypothesis that the ET₀ trends and its dominant factors in Anhui province have changed over the past 59 years. To verify this hypothesis, the goals of this research are (1) to investigate the spatiotemporal characteristics of ET₀ and meteorological factors in Anhui province; (2) to clarify the sensitivities of ET₀ to meteorological factors; (3) to determine the dominant factors of ET₀ trends and their internal mechanisms driving ET₀ variations. The outcomes of this research would enhance our understanding of climate change and provide theoretical support for agricultural production and crop water resource management in similar regions worldwide.

2. Materials and methods

2.1. Study Area

Anhui province (114°54′~119°27′ E and 29°41′~34°38′ N), located in the lower Yangtze River Basin and middle Huai River Basin of eastern China, is a transitional zone between the warm temperate zone and the subtropical zone. North of the Huai River belongs to the warm temperate zone with a subhumid monsoon climate, while the south belongs to the subtropical humid monsoon climate zone. Anhui province has a warm and humid climate with four distinct seasons, and the average annual temperature and precipitation are 14~17 °C and 800~1600 mm, respectively. The precipitation is characterized by more south and less north, more mountains and fewer plains and hills. The summer precipitation is abundant, accounting for 40~60% of the annual precipitation. The total area of the province is 139,600 km², accounting for about 1.45% of the land area and ranking third in eastern China, as well as twenty-second in China [40]. The terrain in the study area is higher in the south and lower in the north. The southern region is mainly hills and mountains, while the northern is principally plain land (Figure 1a). The Yangtze River and Huai River run through Anhui province for 416 km and 430 km, respectively, dividing the province into three natural regions, namely, the Huaibei Plain, the Jianghuai Hills and southern Anhui mountains. The primary land use type is croplands, with a proportion of approximately 58% of the province’s area, followed by grasslands, forests, etc. (Figure 1b).

2.2. Data Sources

The meteorological datasets adopted here were the routine meteorological observation data from 77 weather stations in Anhui province during 1961–2019, which were provided by the China Integrated Meteorological Information Sharing System (CIMISS) of the China Meteorological Administration (CMA), including daily average temperature (T_a, °C), maximum and minimum temperature (T_max and T_min, °C), relative humidity (RH, %), sunshine duration (SD, h) and wind speed at 10 m height (u₁₀, m·s⁻¹). Quality control of the datasets was already conducted by the National Meteorological Information Center (NMIC) of CMA. In addition, the whole year was divided into five study periods, namely, the growing season (April to October), spring (March to May), summer (June to August), autumn (September to November) and winter (December to February of next year).

2.3. ET₀ Calculation Procedure

The PMF-56 model recommended by the FAO was employed to calculate the ET₀ in this research [5], which is an international accepted method for calculating ET₀. The specific Equation (1) was as follows:

$${\mathrm{ET}}_0=\frac{0.408\Delta \left({\mathrm{R}}_{\mathrm{n}}-\mathrm{G}\right)+\mathsf{\gamma }\frac{900}{{\mathrm{T}}_{\mathrm{a}}+273}{\mathrm{u}}_2\left({\mathrm{e}}_{\mathrm{s}}-{\mathrm{e}}_{\mathrm{a}}\right)}{\Delta +\mathsf{\gamma }\left(1+0.34{\mathrm{u}}_2\right)}$$

(1)

where ET₀ refers to the daily reference evapotranspiration (mm·d⁻¹), $\Delta$ refers to the slope of the curve of the saturation vapor pressure at air temperature T_a (kPa·°C⁻¹), R_n refers to the net radiation (MJ·m⁻²·d⁻¹), G refers to the soil heat flux density (MJ·m⁻²·d⁻¹), γ refers to the psychrometric constant (kPa·°C⁻¹), T_a refers to the daily air temperature at 2 m height (°C), u₂ refers to the wind speed at a height of 2 m (m·s⁻¹) and e_s and e_a refer to the saturation vapor pressure and actual vapor pressure, respectively (kPa).

Due to the missing observation data of solar radiation (R_s, MJ·m⁻²·d⁻¹), Angstrom formula [41] was adopted in this work to estimate R_s through SD data. The detailed Equation (2) was as follows:

$${\mathrm{R}}_{\mathrm{s}}=\left({\mathrm{a}}_{\mathrm{s}}+{\mathrm{b}}_{\mathrm{s}}\frac{\mathrm{SD}}{{\mathrm{SD}}_{\max }}\right){\mathrm{R}}_{\mathrm{a}}$$

(2)

where R_s and R_a denote total daily solar radiation and extraterrestrial radiation, respectively (MJ·m⁻²·d⁻¹), SD and SD_max denote daily sunshine duration and maximum possible sunshine duration, respectively (h), and a_s and b_s are regression coefficients, according to the research of Chen et al. [42], with the values of 0.19 and 0.53, respectively.

Meanwhile, while lacking the SD data, the radiation formula put forward by Hargreaves et al. [43] was adopted in this study to calculate R_s through T_max and T_min; the specific Equation (3) was as follows:

$${\mathrm{R}}_{\mathrm{s}}={\mathrm{k}}_{\mathrm{RS}}\times \sqrt{\left({\mathrm{T}}_{\max }-{\mathrm{T}}_{\min }\right)}\times {\mathrm{R}}_{\mathrm{a}}$$

(3)

where k_RS denotes empirical coefficient, and the value in the inland area was usually 0.16 [43,44,45]. Moreover, the performance of this method was verified in our earlier research [1].

Then, the R_n could be calculated in the following Equations (4)–(6):

$${\mathrm{R}}_{\mathrm{n}}={\mathrm{R}}_{\mathrm{ns}}-{\mathrm{R}}_{\mathrm{nl}}$$

(4)

$${\mathrm{R}}_{\mathrm{ns}}=\left(1-\mathsf{\alpha }\right){\mathrm{R}}_{\mathrm{s}}$$

(5)

$${\mathrm{R}}_{\mathrm{nl}}=\mathsf{\sigma }\left(\frac{{\mathrm{T}}_{\max ,\mathrm{K}}^4+{\mathrm{T}}_{\min ,\mathrm{K}}^4}{2}\right)\left(0.34-0.14\sqrt{{\mathrm{e}}_{\mathrm{a}}}\right)\left(1.35\frac{{\mathrm{R}}_{\mathrm{s}}}{{\mathrm{R}}_{\mathrm{s}0}}-0.35\right)$$

(6)

where R_ns and R_nl are the incoming net short wave radiation and the outgoing net long wave radiation, respectively (MJ·m⁻²·d⁻¹), α is the reference crop albedo (with value of 0.23), $\mathsf{\sigma }$ is the Stephen Boltzmann’s constant (4.903 × 10⁻⁹ MJ·K⁻⁴·m⁻²·d⁻¹), T_max,K and T_min,K are the maximum and minimum absolute temperature within 24 h (K = °C + 273.16) and R_s0 is the clear sky radiation (MJ·m⁻²·d⁻¹); for the specific calculation procedure, please refer to work of Allen et al. [5].

As the routine observation data adopted in this study only included u₁₀, in order to obtain the u₂ and the convenience of calculation, our work adopted the wind speed conversion Equation (7) proposed by Allen et al. [5]:

$${\mathrm{u}}_2={\mathrm{u}}_{\mathrm{z}}\frac{4.87}{\ln \left(67.8\mathrm{z}-5.42\right)}$$

(7)

where u₂ denotes the wind speed with the height of 2 m above the ground plane (m·s⁻¹).

In addition, detailed calculations using Equations (8)–(12) of $\Delta$, $\mathsf{\gamma }$, ${\mathrm{e}}_{\mathrm{s}}$ and ${\mathrm{e}}_{\mathrm{a}}$ were as follows:

$$\Delta =\frac{4098\times \left[0.6108\exp \left(\frac{17.27{\mathrm{T}}_{\mathrm{a}}}{{\mathrm{T}}_{\mathrm{a}}+237.3}\right)\right]}{{\left({\mathrm{T}}_{\mathrm{a}}+237.3\right)}^2}$$

(8)

$$\mathsf{\gamma }=\frac{0.00163\mathrm{P}}{\mathsf{\lambda }}$$

(9)

$$\mathrm{P}=101.3{\left(\frac{273+{\mathrm{T}}_{\mathrm{a}}-0.0065\mathrm{h}}{273+{\mathrm{T}}_{\mathrm{a}}}\right)}^{5.26}$$

(10)

$${\mathrm{e}}_{\mathrm{s}}=0.6108\exp \left(\frac{17.27{\mathrm{T}}_{\mathrm{a}}}{{\mathrm{T}}_{\mathrm{a}}+237.3}\right)$$

(11)

$${\mathrm{e}}_{\mathrm{a}}={\mathrm{e}}_{\mathrm{s}}\times \mathrm{RH}$$

(12)

where $\mathsf{\lambda }$ is the latent heat of vaporization with the value of 2.45 (MJ·kg⁻¹), $\mathrm{P}$ is the atmospheric pressure (kPa) and $\mathrm{h}$ is the elevation above the sea level (m).

2.4. Sensitivity Coefficient

The differential equation method developed by McCuen [46] was adopted to calculate the sensitivities of ET₀ to meteorological factors in the following Equation (13):

$$\mathrm{S}\left({\mathrm{v}}_{\mathrm{i}}\right)=\underset{\Delta {\mathrm{v}}_{\mathrm{i}}\to 0}{\lim }\left(\frac{\Delta {\mathrm{ET}}_0/{\mathrm{ET}}_0}{\Delta {\mathrm{v}}_{\mathrm{i}}/{\mathrm{v}}_{\mathrm{i}}}\right)=\frac{\partial {\mathrm{ET}}_0}{\partial {\mathrm{v}}_{\mathrm{i}}}\times \frac{{\mathrm{v}}_{\mathrm{i}}}{{\mathrm{ET}}_0}$$

(13)

where S(v_i) denotes the sensitivity of ET₀ to meteorological factor (v_i), the positive (negative) sensitivity represents the ET₀ increases (decreases) with v_i and the absolute value of S(v_i) denotes the influence of v_i to ET₀. For the detailed calculation processes, please refer to work of Chu et al. [2].

2.5. Contributions of Meteorological Factors to ET₀

As shown in Formula (1), the ET₀ is a function of meteorological factors. Therefore, the T_a, RH, u₂ and R_n were selected as four main meteorological factors affecting ET₀. Moreover, this study employed the differential equation method to assess the contribution of four main meteorological factors to ET₀ based on PMF-56 model. Precise Equation (14) showed as follows:

$$\frac{{\mathrm{dET}}_0}{\mathrm{dt}}=\frac{\partial {\mathrm{ET}}_0}{\partial {\mathrm{T}}_{\mathrm{a}}}\frac{{\mathrm{dT}}_{\mathrm{a}}}{\mathrm{dt}}+\frac{\partial {\mathrm{ET}}_0}{\partial \mathrm{RH}}\frac{\mathrm{dRH}}{\mathrm{dt}}+\frac{\partial {\mathrm{ET}}_0}{\partial {\mathrm{u}}_2}\frac{{\mathrm{du}}_2}{\mathrm{dt}}+\frac{\partial {\mathrm{ET}}_0}{\partial {\mathrm{R}}_{\mathrm{n}}}\frac{{\mathrm{dR}}_{\mathrm{n}}}{\mathrm{dt}}+\mathsf{\epsilon }$$

(14)

Equation (14) can be abbreviated to the below Equation (15):

$$\mathrm{C}\_{\mathrm{ET}}_0=\mathrm{C}\left({\mathrm{T}}_{\mathrm{a}}\right)+\mathrm{C}\left(\mathrm{RH}\right)+\mathrm{C}\left({\mathrm{u}}_2\right)+\mathrm{C}\left({\mathrm{R}}_{\mathrm{n}}\right)+\mathsf{\epsilon }$$

(15)

where C_ET₀ denotes the ET₀ trend, C(T_a), C(RH), C(u₂) and C(R_n) refer to the contribution of T_a, RH, u₂ and R_n to ET₀, respectively, and $\mathsf{\epsilon }$ indicates the deviation between C_ET₀ and ET₀ calculated by using Theil–Sen estimator (T_ET₀). The contribution of each meteorological factor to ET₀ could be computed by multiplying the result of Equation (13), the trend of each meteorological factor during the study period and the length of the corresponding study period (i.e., 365 or 366 days for the annual calculation, 214 days for growing season, 92 days for both spring and summer, 91 days for autumn, 90 or 91 days for winter) [1,3].

2.6. Trend Analysis

The Mann–Kendall (M–K) test was recommended for hydrometeorological time series data analysis [47,48,49]. Thus, it was adopted here to estimate trends of ET₀ and the four main meteorological factors [3]. The null hypothesis H₀ was that in a series of data (x_i, i = 1, 2, …, n), x_i was independent and evenly distributed. The alternative hypothesis H₁ was that a monotonic tendency consisted of X. The statistical S and standardized test statistics Z were calculated in the following Equations (16) and (17):

$$\mathrm{S}={\displaystyle \sum }_{\mathrm{i}=1}^{\mathrm{n}-1}{\displaystyle \sum }_{\mathrm{j}=\mathrm{i}+1}^{\mathrm{n}}\mathrm{sgn}\left({\mathrm{x}}_{\mathrm{j}}-{\mathrm{x}}_{\mathrm{i}}\right)$$

(16)

$$\mathrm{sgn}\left({\mathrm{x}}_{\mathrm{j}}-{\mathrm{x}}_{\mathrm{i}}\right)=\{\begin{array}{cc}\hfill +1& \mathrm{if}\left({\mathrm{x}}_{\mathrm{j}}-{\mathrm{x}}_{\mathrm{i}}\right)>0\hfill \\ \hfill 0& \mathrm{if}\left({\mathrm{x}}_{\mathrm{j}}-{\mathrm{x}}_{\mathrm{i}}\right)=0\hfill \\ \hfill -1& \mathrm{if}\left({\mathrm{x}}_{\mathrm{j}}-{\mathrm{x}}_{\mathrm{i}}\right)<0\hfill \end{array}$$

(17)

where x_i and x_j are the value of year i and j, respectively, and n is the data length. The S obeyed normal distributions (n ≥ 8), the calculation of average value E(S) and variance Var(S) were given below in Equations (18)–(20):

$$\mathrm{E}\left(\mathrm{S}\right)=0$$

(18)

$$\mathrm{Var}\left(\mathrm{S}\right)=\frac{1}{18}\left[\mathrm{n}\left(\mathrm{n}-1\right)\left(2\mathrm{n}+5\right)-{\displaystyle \sum }_{\mathrm{p}=1}^{\mathrm{q}}{\mathrm{t}}_{\mathrm{p}}\left({\mathrm{t}}_{\mathrm{p}}-1\right)\left(2{\mathrm{t}}_{\mathrm{p}}+5\right)\right]$$

(19)

where q refers to same group number and t_p represents to the value in pth step.

$$\mathrm{Z}=\{\begin{array}{ll}\frac{\mathrm{S}-1}{\sqrt{\mathrm{Vas}\left(\mathrm{S}\right)}}& \mathrm{if}\mathrm{S}>0\hfill \\ 0& \mathrm{if}\mathrm{S}=0\\ \frac{\mathrm{S}+1}{\sqrt{\mathrm{Vas}\left(\mathrm{S}\right)}}& \mathrm{if}\mathrm{S}<0\hfill \end{array}$$

(20)

in which Z is the change trend of time series’ data and Z > 0 (Z < 0) denotes the increasing (decreasing) trend. If $\left|\mathrm{Z}\right|>{\mathrm{Z}}_{\left(1-\mathsf{\alpha }/2\right)}$, then the hypothesis was rejected and the time series data showed a significant changing trend. Moreover, ${\mathrm{Z}}_{\left(1-\mathsf{\alpha }/2\right)}$ was the standard normal deviation in the standard normal distribution chart. When the α = 5% and α = 1% were the significant levels, the ${\mathrm{Z}}_{\left(1-\mathsf{\alpha }/2\right)}$ values were 1.96 and 2.58, respectively.

The Theil–Sen estimator was used to determine the magnitude of the variation trends of ET₀ and meteorological factors [50,51]. Detailed calculation Equation (21) was as follows:

$$\mathsf{\beta }=\mathrm{Median}\left(\frac{{\mathrm{x}}_{\mathrm{j}}-{\mathrm{x}}_{\mathrm{i}}}{\mathrm{j}-\mathrm{i}}\right),\forall 1<i<j$$

(21)

where β is the calculated slope of data series; ${\mathrm{x}}_{\mathrm{i}}$ and ${\mathrm{x}}_{\mathrm{j}}$ represent the sequence data corresponding to time i and j, respectively; the positive (negative) β indicates the increasing (decreasing) trend. Spatial distribution map was prepared by inverse distance weighting (IDW) method within ArcGIS platform (version 10.2).

3. Results

3.1. Spatiotemporal Variations of ET₀

3.1.1. Temporal Scale

As shown in Figure 2, the ET₀ exhibited a significant decreasing trend before 1990 (−3.89 mm·a⁻²) and then increased slowly (0.62 mm·a⁻²) throughout the entire region. In subregions, the ET₀ presented significant decreasing trends before 1990, and the decrease magnitudes ranked in the order: Region I > Region II > Region III. After 1990, the ET₀ showed a slowly decreasing trend in Region I, while it exhibited slow increasing trends in other regions, being especially higher in Region III. Because of the definition of the World Meteorological Organization (WMO) of the standard climate standard (i.e., the 30-year average) and the opposite change trend of ET₀ around 1990 in this study, we divided the entire study period into two periods, namely, 1961–1990 and 1991–2019.

Detailed temporal trends of ET₀ are also shown in

Table 1. Temporal trends of ET₀ during 1961–1990 and 1991–2019 in Anhui province.

Time	Region	Annual		Growing Season		Spring		Summer		Autumn		Winter
		Z	β	Z	β	Z	β	Z	Β	Z	β	Z	β
1961–1990	Whole	−3.53	−3.953 ***	−2.96	−2.861 **	−1.07	−0.474	−2.85	−2.242 **	−1.61	−0.494	−1.64	−0.390
	I	−3.93	−5.254 ***	−3.75	−4.148 ***	−2.14	−1.256 *	−3.78	−3.066 ***	−1.57	−0.450	−1.50	−0.548
	II	−3.25	−3.567 **	−2.78	−2.672 **	−0.82	−0.362	−2.93	−2.233 **	−1.68	−0.499	−1.53	−0.343
	III	−2.96	−3.188 **	−2.32	−2.363 *	−1.03	−0.258	−1.96	−1.748 *	−1.93	−0.554	−1.89	−0.326
1991–2019	Whole	0.62	0.552	0.54	0.364	1.26	0.809	0.88	0.594	−1.44	−0.478	−0.58	−0.096
	I	−0.02	−0.059	−0.21	−0.286	0.73	0.364	0.06	0.046	−1.22	−0.555	−0.66	−0.198
	II	0.21	0.199	0.13	0.075	1.26	0.850	0.96	0.619	−1.67	−0.541	−0.88	−0.155
	III	1.59	1.339	1.26	1.120	2.04	1.029 *	0.73	0.551	−0.51	−0.139	−0.36	−0.091

Note: Z indicates the M–K test statistic; β is the estimated ET₀ slope, β > 0 (β < 0) denotes increasing (decreasing) trend; *, ** and *** denote the significance level of 0.05, 0.01 and 0.001, respectively.

. Before 1990, the temporal trends of ET₀ on an annual timescale, growing season, and summer were similar. During spring, autumn and winter, the downward trends of ET₀ were not significant in most regions. After 1990, ET₀ exhibited slow upward trends annually, during the growing season, spring and summer, while it showed downward trends during autumn and winter. It is worth noting that only the ET₀ trend in Region III in spring was significant.

3.1.2. Spatial Scale

As shown in Figure 3, the spatial distribution of ET₀ during 1961–1990 and 1991–2019 was basically consistent. The ET₀ was higher in the north and lower in the south annually, during the growing season, spring and summer. However, the ET₀ was higher in the southwest of Regions II and III in both autumn and winter. Before 1990, ET₀ annually, during the growing season and summer showed a significant decreasing trend in most regions. After 1990, the ET₀ in Region III showed a significant increasing trend annually, during the growing season and spring. Moreover, the ET₀ in Region II also exhibited a significant increasing trend in spring. These phenomena are echoed in Table 1.

3.2. Spatiotemporal Variations of Meteorological Factors

To evaluate the impact of meteorological factors on ET₀, we analyzed the change trends of meteorological factors. As shown in Figure 4, T_a decreased significantly at the rate of −0.021 °C·a⁻¹ before 1990, and then increased significantly with the rate of 0.034 °C·a⁻¹ (Figure 4a). RH first increased at the rate of 0.071 a⁻¹ before 1990, and then decreased slightly at the rate of 0.070 a⁻¹ (Figure 4b). u₂ declined significantly with the rate of −0.020 m·s⁻¹ before 1990, and then slowed down to −0.006 m·s⁻¹ (Figure 4c). Similar to u₂, R_n also declined significantly with the rate of −0.017 MJ·m⁻²·d⁻¹ before 1990, and then slowed down to −0.004 MJ·m⁻²·d⁻¹ insignificantly (Figure 4d).

From

Table 2. Temporal trends of meteorological factors during 1961–1990 and 1991–2019 in Anhui province.

Meteorological	Time	Region	Annual		Growing Season		Spring		Summer		Autumn		Winter
Factor			Z	β	Z	β	Z	β	Z	β	Z	β	Z	β
Ta	1961–1990	Whole	−2.57	−0.022 *	−1.64	−0.011	−0.86	−0.007	−1.82	−0.035	−0.54	−0.007	−0.75	−0.010
		I	−2.36	−0.029 *	−1.86	−0.017	−0.75	−0.009	−2.25	−0.039 *	−0.04	−0.001	−0.54	−0.004
		II	−2.71	−0.021 **	−1.57	−0.012	−0.57	−0.007	−2.00	−0.036 *	−0.57	−0.008	−1.14	−0.012
		III	−2.50	−0.020 *	−1.25	−0.009	−0.96	−0.013	−1.32	−0.019	−0.61	−0.009	−0.61	−0.013
	1991–2019	Whole	2.91	0.036 **	2.46	0.032 *	3.28	0.059 **	2.12	0.030 *	2.76	0.031 **	0.84	0.009
		I	2.76	0.037 **	2.12	0.026 *	3.06	0.053 **	1.52	0.026	2.31	0.026 *	1.07	0.015
		II	2.87	0.035 **	2.27	0.028 *	3.10	0.062 **	1.97	0.02 9 *	2.08	0.026 *	0.73	0.011
		III	3.21	0.038 **	2.79	0.035 **	3.17	0.054 **	1.78	0.030	2.98	0.042 **	0.69	0.011
RH	1961–1990	Whole	1.78	0.0009	1.50	0.0008	0.14	0.0002	2.78	0.0012 **	0.36	0.0002	0.61	0.0008
		I	2.36	0.0016 *	1.86	0.0014	0.89	0.0011	2.93	0.0021 **	0.43	0.0005	0.54	0.0012
		II	1.32	0.0006	1.21	0.0007	−0.11	−0.0001	2.46	0.0012 *	0.21	0.0001	0.39	0.0004
		III	1.50	0.0005	1.00	0.0004	−0.18	−0.00004	1.46	0.0009	0.61	0.0004	0.89	0.0006
	1991–2019	Whole	−1.07	−0.0008	−1.37	−0.0009	−1.29	−0.0016	−1.29	−0.0009	0.66	0.0007	−0.06	−0.0001
		I	−0.88	−0.0005	−0.84	−0.0007	−0.47	−0.0009	−0.99	−0.0006	0.47	0.0007	−0.09	−0.0001
		II	−1.18	−0.0008	−1.18	−0.0007	−1.48	−0.0019	−0.88	−0.0007	1.03	0.0009	−0.36	−0.0003
		III	−0.88	−0.0005	−1.41	−0.0010	−1.67	−0.0016	−1.22	−0.0008	0.69	0.0006	0.43	0.0004
u2	1961–1990	Whole	−5.67	−0.021 ***	−5.28	−0.019 ***	−5.32	−0.024 ***	−4.78	−0.014 ***	−5.32	−0.023 ***	−5.03	−0.023 ***
		I	−5.53	−0.025 ***	−5.03	−0.023 ***	−4.92	−0.029 ***	−4.53	−0.020 ***	−5.25	−0.027 ***	−5.35	−0.031 ***
		II	−5.46	−0.019 ***	−4.78	−0.017 ***	−5.00	−0.023 ***	−4.32	−0.012 ***	−4.89	−0.023 ***	−4.82	−0.021 ***
		III	−5.71	−0.018 ***	−5.07	−0.016 ***	−5.46	−0.022 ***	−3.71	−0.012 ***	−5.07	−0.019 ***	−5.57	−0.020 ***
	1991–2019	Whole	−3.55	−0.008 ***	−3.25	−0.007 **	−4.60	−0.009 ***	−2.38	−0.005 *	−1.97	−0.005 *	−2.04	−0.004 *
		I	−3.43	−0.011 ***	−3.28	−0.010 **	−4.15	−0.014 ***	−2.91	−0.010 **	−2.91	−0.009 **	−1.82	−0.006
		II	−4.18	−0.011 ***	−3.62	−0.011 ***	−4.75	−0.013 ***	−3.36	−0.010 ***	−2.53	−0.009 *	−2.68	−0.007 **
		III	1.03	0.001	1.48	0.003	0.47	0.001	1.03	0.003	1.52	0.003	0.58	0.001
Rn	1961–1990	Whole	−3.00	−0.017 **	−2.36	−0.020 *	−0.71	−0.007	−2.57	−0.047 *	−1.14	−0.006	−2.53	−0.010 *
		I	−3.03	−0.016 **	−2.50	−0.021 *	−1.00	−0.010	−3.32	−0.053 ***	−0.75	−0.003	−3.07	−0.009 **
		II	−2.71	−0.018 **	−2.28	−0.021 *	−0.75	−0.008	−2.60	−0.048 **	−1.00	−0.007	−2.60	−0.010 **
		III	−2.78	−0.016 **	−2.32	−0.019 *	−0.61	−0.003	−1.96	−0.040 *	−1.75	−0.012	−2.28	−0.008 *
	1991–2019	Whole	−0.92	−0.005	−0.69	−0.008	0.84	0.008	−0.51	−0.010	−2.57	−0.017 *	−2.46	−0.008 *
		I	−0.66	−0.004	−0.92	−0.007	0.58	0.003	−0.62	−0.011	−1.89	−0.012	−2.57	−0.009 *
		II	−0.96	−0.006	−0.99	−0.011	0.66	0.006	−0.54	−0.009	−2.49	−0.017 *	−2.31	−0.008 *
		III	−0.32	−0.003	−0.06	−0.001	1.63	0.013	0.28	0.005	−2.49	−0.016 *	−1.67	−0.006

Note: Z indicates the M–K test statistic; β refers to the estimated slope of meteorological factor, β > 0 (β < 0) denotes the increasing (decreasing) trend; *, ** and *** indicate significance level of 0.05, 0.01 and 0.001, respectively.

, T_a exhibited significant decreasing trends in the entire region and each subregion on an annual timescale and in Regions I and II in summer, while showing insignificant decreasing trends in other regions and timescales during 1961–1990. During 1991–2019, T_a showed increasing trends in all regions and study periods, with significant trends exhibited annually, during the growing season, spring, summer and autumn in most regions. RH showed increasing trends before 1990 in most regions, while it showed decreasing trends after 1990, except in autumn. u₂ presented significant decreasing trends in most regions during the two time periods, except in Region III after 1990, which exhibited slightly increasing trends. R_n decreased significantly annually, during the growing season, summer and winter, while insignificantly in other periods before 1990. Then, R_n decreased slowly in most regions and study periods, except in spring, which increased slowly after 1990.

3.3. Sensitivity of Meteorological Factors to ET₀

3.3.1. Temporal Variation Characteristics

The daily sensitivity coefficient of each meteorological factor to ET₀ is shown in Figure 5. As seen from Figure 5, the daily sensitivity coefficient of T_a, RH and R_n (i.e., S(T_a), S(RH) and S(R_n)) first increased and then decreased, while the daily sensitivity coefficient of u₂ (S(u₂)) showed a gradual changing trend, going downward first and then upward. The magnitude of these four meteorological factors was ranked as follows: S(RH) > S(R_n) > S(T_a) > S(u₂). In contrast to S(T_a), S(u₂) and S(R_n), S(RH) was negative.

Table 3. Sensitivities of ET₀ to meteorological factors during 1961–1990 and 1991–2019.

Region	Season	1961–1990				1991–2019
		S(Ta)	S(RH)	S(u2)	S(Rn)	S(Ta)	S(RH)	S(u2)	S(Rn)
Whole	Annual	0.489	−1.033	0.074	0.709	0.492	−0.886	0.082	0.731
	Growing season	0.660	−0.824	0.049	0.778	0.641	−0.683	0.060	0.791
	Spring	0.499	−0.971	0.049	0.724	0.507	−0.753	0.066	0.740
	Summer	0.728	−0.691	0.037	0.817	0.692	−0.628	0.044	0.830
	Autumn	0.578	−1.078	0.090	0.707	0.575	−0.906	0.098	0.726
	Winter	0.152	−1.398	0.119	0.588	0.194	−1.253	0.121	0.628
I	Annual	0.474	−1.024	0.102	0.652	0.477	−0.871	0.104	0.690
	Growing season	0.666	−0.817	0.070	0.737	0.643	−0.664	0.072	0.770
	Spring	0.505	−0.903	0.079	0.658	0.511	−0.713	0.086	0.698
	Summer	0.736	−0.713	0.054	0.783	0.696	−0.618	0.053	0.814
	Autumn	0.565	−1.095	0.115	0.660	0.559	−0.904	0.119	0.692
	Winter	0.090	−1.393	0.159	0.507	0.144	−1.243	0.160	0.556
II	Annual	0.499	−1.080	0.075	0.701	0.501	−0.920	0.086	0.720
	Growing season	0.669	−0.861	0.051	0.771	0.650	−0.713	0.064	0.780
	Spring	0.507	−1.019	0.048	0.718	0.516	−0.777	0.071	0.728
	Summer	0.737	−0.720	0.038	0.813	0.701	−0.657	0.046	0.823
	Autumn	0.589	−1.123	0.095	0.694	0.584	−0.943	0.104	0.713
	Winter	0.164	−1.464	0.120	0.580	0.203	−1.297	0.125	0.616
III	Annual	0.483	−0.954	0.052	0.763	0.487	−0.838	0.060	0.777
	Growing season	0.640	−0.758	0.033	0.820	0.626	−0.644	0.045	0.823
	Spring	0.482	−0.930	0.029	0.778	0.491	−0.736	0.046	0.785
	Summer	0.707	−0.619	0.025	0.849	0.676	−0.581	0.033	0.854
	Autumn	0.568	−0.979	0.065	0.763	0.570	−0.846	0.074	0.773
	Winter	0.176	−1.288	0.090	0.660	0.211	−1.185	0.088	0.696

Note: The bold font represents the most sensitive factor.

presents the seasonal variation of the sensitivities of meteorological factors to ET₀. The general change trends of sensitivity coefficients in the seasonal timescale were consistent with those in the daily timescale. During 1961–1990, the ET₀ was most sensitive to RH annually, during the growing season, spring, autumn and winter, and the sensitivity magnitude of four meteorological factors was ranked as follows: S(RH) > S(R_n) > S(T_a) > S(u₂). However, the ET₀ was most sensitive to Rn in summer, followed by T_a, RH and u₂ in all regions. Furthermore, the ET₀ was most sensitive to R_n in the growing season, followed by RH, T_a and u₂ in Region III. In contrast to that during 1961–1990, the ET₀ was most sensitive to R_n in the growing season, followed by RH, T_a and u₂ during 1991–2019. Moreover, R_n was also the most sensitive factor in spring in Region III. The differences between these two time periods were not distinct.

Figure 6 shows the annual sensitivity coefficients of meteorological factors to ET₀. S(T_a) exhibited a slow downward trend during the whole study period, decreased before 1990 and then increased. S(RH) and S(R_n) both presented evident increasing trends in the past 59 years, especially during 1961–1990. Because the S(RH) was negative, as far as the magnitude was concerned, the sensitivity of RH decreased generally. S(u₂) was similar to S(T_a), while it increased more obviously after 1990.

3.3.2. Spatial Variation Characteristics

In order to illuminate the sensitivity coefficients of ET₀ to meteorological factors, the spatial distribution of annual mean sensitivity coefficients is displayed in Figure 7. As shown in Figure 7, S(T_a) was higher in Region II, especially along the Yangtze River, than in the north of Region I, west of Region II and south of Region III. The spatial distribution of S(RH) was similar to that of S(T_a), while the highest value was located in the central Region III (namely, the Huangshan station). S(u₂) was comparatively higher in the north and lower in the south. The S(u₂) in the west of Region II was lower than in surrounding regions. Contrary to the spatial distribution of S(T_a) and S(RH), the higher values of S(R_n) were mainly located in the west of Region II and most areas of Region III.

3.4. Contribution of Meteorological Factors to ET₀

3.4.1. Validation of Differential Method

To verify the validity of the differential method, we compared the ET₀ trends calculated using a differential method (C_ET₀) and with those, the Sen slope estimator (E_ET₀) in three time periods (during 1961–2019, 1961–1990 and 1991–2019). As shown in Figure 8, the fitting results from the differential method and the Sen slope estimator were all concentrated in a 1:1 line, with an R² value generally greater than 0.90. Thus, all the above phenomena indicated that the four selected meteorological factors in this research were reasonable and could be well explained by the contributions to ET₀ trends by employing the differential method.

3.4.2. Contribution of Meteorological Factors to ET₀

The contribution of meteorological factors to the ET₀ trend during the two study periods is shown in Figure 9. During 1961–1990, all meteorological factors offered negative contributions to ET₀ trends, which ultimately led to decreasing trends for almost all regions and periods (Table 1). Specifically, R_n was the leading factor annually, during the growing season and summer, while u₂ played leading roles in spring, autumn and winter. However, RH was the leading factor in Region I annually and R_n was the leading factor in Region III in autumn. During 1991–2019, T_a and RH showed positive contributions to ET₀ trends for most regions and periods, except for RH in autumn and winter. In contrast, u₂ and R_n devoted negative contributions to ET₀ trends, except for R_n in spring and summer and u₂ in Region III. Concretely, the main reason for changes in the ET₀ was T_a annually, during the growing season, spring and summer, and then R_n in autumn and winter for most regions. Moreover, notably, the contribution magnitude of u₂ to ET₀ trends dropped sharply in Region III for all seasons. Overall, the negative contributions from u₂ and R_n could not offset the positive contributions from T_a and RH, which led to the slow upward ET₀ trends in the entire region, eventually, while the upward trend was higher in Region III.

Table 4. Contributions of meteorological factors to ET₀ trends during 1961–1990 and 1991–2019.

Region	Season	1961–1990				1991–2019
		C(Ta)	C(RH)	C(u2)	C(Rn)	C(Ta)	C(RH)	C(u2)	C(Rn)
Whole	Annual	−0.746	−0.772	−0.842	−1.223	0.993	0.499	−0.459	−0.267
	Growing season	−0.317	−0.533	−0.470	−1.063	0.612	0.304	−0.274	−0.225
	Spring	−0.088	−0.049	−0.221	−0.085	0.486	0.309	−0.160	0.146
	Summer	−0.368	−0.424	−0.150	−1.158	0.265	0.159	−0.094	−0.023
	Autumn	−0.041	−0.094	−0.268	−0.136	0.201	−0.133	−0.105	−0.300
	Winter	−0.047	−0.077	−0.196	−0.106	0.047	−0.018	−0.061	−0.095
I	Annual	−0.959	−1.560	−1.277	−1.204	0.910	0.454	−0.785	−0.181
	Growing season	−0.438	−1.101	−0.738	−1.118	0.483	0.122	−0.490	−0.068
	Spring	−0.115	−0.311	−0.355	−0.119	0.455	0.111	−0.272	0.087
	Summer	−0.471	−0.693	−0.257	−1.290	0.215	0.140	−0.180	0.062
	Autumn	0.000	−0.164	−0.374	−0.062	0.161	−0.152	−0.197	−0.233
	Winter	−0.021	−0.145	−0.282	−0.096	0.051	−0.030	−0.113	−0.093
II	Annual	−0.713	−0.528	−0.781	−1.304	0.980	0.572	−0.715	−0.333
	Growing season	−0.294	−0.381	−0.424	−1.110	0.592	0.307	−0.461	−0.306
	Spring	−0.057	0.050	−0.207	−0.114	0.523	0.399	−0.238	0.126
	Summer	−0.402	−0.378	−0.116	−1.223	0.265	0.146	−0.170	−0.085
	Autumn	−0.044	−0.048	−0.268	−0.118	0.170	−0.170	−0.179	−0.304
	Winter	−0.055	−0.034	−0.187	−0.117	0.042	0.015	−0.094	−0.103
III	Annual	−0.611	−0.387	−0.505	−1.161	1.043	0.212	0.009	−0.098
	Growing season	−0.216	−0.244	−0.273	−0.981	0.676	0.260	0.052	−0.095
	Spring	−0.093	0.037	−0.123	−0.052	0.442	0.277	−0.017	0.260
	Summer	−0.227	−0.240	−0.093	−0.977	0.264	0.116	0.035	0.094
	Autumn	−0.064	−0.065	−0.156	−0.222	0.260	−0.117	0.033	−0.320
	Winter	−0.052	−0.060	−0.129	−0.102	0.048	−0.074	0.001	−0.079

Note: The bold font represents the most dominant factor.

and Figures S1 and S2 in the Supplementary Materials provide detailed information and the spatial distribution of dominant factors.

4. Discussion

In this study, T_a exhibited significant decreasing and increasing trends before and after 1990, respectively, which was similar to that in the whole of China [52] and also surrounding regions, such as the Huai River Basin [53] and Jiangsu province [2]. Ding et al. [52] pointed out that the sum of positive radiative forcing generated by greenhouse gases was the cause of climate warming, and that the surface temperature is likely to continue to rise. On the contrary, the RH first increased and then decreased, which may also be explained by the climate warming in this region (namely, the change trends in T_a). Furthermore, the larger vapor pressure deficit (VPD) caused by climate change from 1983 to 2008 [54] could explain the RH changing trends. Here, the changing trend of u₂ was similar to that in mainland China [55]. However, because Region III is mainly a mountainous terrain zone, it was greatly affected by the narrow tube effect, which resulted in a higher wind speed here. These phenomena might be responsible for the relatively lower decreasing trends of u₂ during 1961–1990 and slightly increasing trends during 1991–2019. Moreover, Tao et al. [56] reported that urbanization also had an impact on the annual mean wind speed decline in Anhui province after 2000 and its contribution rate reached 40%, particularly in spring. Meanwhile, the attenuation factor u₂ might suppress the diffusion of aerosols and strengthen the influence of aerosol emissions on solar dimming [57]. In addition, Qian et al. [58] indicated that a fog-like haze caused by increasing air pollution might absorb or reflect the radiation, resulting in a reduction in surface solar radiation. Similar results were also shown by Ma et al. [59], Qi et al. [60] and Tao et al. [61].

This research also revealed that the ET₀ decreased significantly before 1990 and then increased slowly. Similar phenomena occurred in a desertification-prone region of China [62], the Yellow River Basin [31], the Jing-Jin-Ji region of North China [35], Xinjiang province [63] and even mainland China [14,64]. All these changing trends in ET₀ were attributed to changes in meteorological factors before and after the change point in specific regions. Generally, in this research, ET₀ was most sensitive to RH, followed by R_n, T_a and u₂, while the most sensitive factor shifted to R_n in summer, followed by T_a, RH and u₂ for most regions. Similar results could be found in the Huai River Basin [1] and Jiangsu province [2]. However, in other regions of China, Xu et al. [27] pointed out that T_max was, generally, the most sensitive factor for ET₀, followed by RH, R_s, T_min and u₂ in the Jing River Basin of Northwest China. Wang et al. [65] reported that the ET₀ was more sensitive to T_max and SD than RH, u₂ and T_min in the three-river headwaters region of the Qinghai–Tibetan Plateau. Li et al. [29] discovered that the RH had the highest sensitivity, followed by T_max, u₂, SD and T_min. She et al. [31] and Yang et al. [66] both indicated that the ET₀ was the most sensitive to R_s, followed by RH and T_a in parts of the Yellow River Basin. From the previous research above, the difference in sensitivity factors of ET₀ in different regions of China may have been mainly caused by the lower water vapor carried by the wind in dry regions and the higher humidity of the wind flow in humid regions [64]. However, in summer, and even in the growing season of Anhui province, T_a and R_n increased, while the air pressure and RH decreased, which could explain the transition of the most sensitive factor of ET₀ from RH to R_n.

Although ET₀ was most sensitive to RH for most regions, the change rate of RH was relatively small compared to other factors. Before 1990, R_n was the leading factor of ET₀ trends annually, during the growing season and summer, while u₂ was the leading factor in spring, autumn and winter. However, the high contribution of RH to the ET₀ trend in Region I in the annual timescale could be interpreted reasonably by its significant increasing trend (Table 2). During 1991–2019, the leading factor of ET₀ trends changed to T_a annually, during the growing season, spring and summer, then to R_n in autumn and winter for most regions. Similarly, Han et al. [35] reported that a decreasing sunshine duration and wind speed offset the impact of increasing air temperature, resulting in a decrease in ET₀ between 1961 and 1991, while T_a was the most dominant factor contributing to an increase in ET₀ in the Jing-Jin-Ji region between 1992 and 2015. Wang et al. [14] reported that the ET₀ presented a significant increasing trend after the 1990s in China due to the increasing T_a. In this study, we also demonstrated an interesting phenomenon that the significance of climate variables was proportional to their dominance in ET₀ trends. This finding was supported by a similar finding in our previous study on pan evaporation in the Huai River Basin [3]. As shown in Table 2, during 1961–1990, R_n represented significant decreasing trends annually, during the growing season and summer, which may explain its dominant role in the corresponding seasons. Although u₂ always showed significant downward trends in these periods, the magnitude of u₂ trends in spring, autumn and winter was larger than those in other seasons, which might have been responsible for its dominance in corresponding seasons. However, during 1991–2019, T_a presented prominently increasing trends for most regions and seasons except winter, which could account for its leading role in corresponding seasons. R_n only showed significant trends in autumn and winter for most regions, which corresponded to its dominant position in these two seasons. Furthermore, the insignificant trends of u₂ in Region III for all seasons might also decipher its small contributions here. As shown in Figure 1, Region III is mainly a mountainous area with high elevation and the land use types are mainly forest and grassland. Meanwhile, Tao et al. [56] also found that the decreasing trend of u₂ of urban stations was significantly greater than that of rural stations in Anhui province, which could give a possible explanation for the insignificant upward trends of u₂ in Region III.

Although the effect of meteorological factors on ET₀ was well quantified in this study, some uncertainties still existed in this aspect. Firstly, the differential equation method was adopted in this study to determine the contribution of each meteorological factor to ET₀ trends. This method assumes that each major climate variable is independent and has been proven to be equivalent to the performance of the detrending method in a previous study [53]. However, each meteorological factor is not completely independent and may interact with one another, and the differential equation method adopts the averaged partial derivatives of each variable to attribute to the ET₀ changes, which may also introduce uncertainty in the ultimate results [67]. In addition, considering the complexity of the underlying surface, although the most complete observational data of national meteorological stations were employed in this study, the density of current meteorological stations was still sparse, which was not enough to fully reflect ET₀ changes and their causes at the spatial scale. Moreover, the changes in climate factors caused by human activities are likely to eventually lead to ET₀ changes. Therefore, human activities, especially land use and cover changes and urbanization, increase the errors and uncertainties of ET₀ calculation and attribution [31,68], which needs further research.

5. Conclusions

In this paper, we found that the ET₀ decreased significantly before 1990 (−3.89 mm·a⁻²) and then increased slowly (0.62 mm·a⁻²) throughout the Anhui province. Among the meteorological factors affecting ET₀ changes, T_a decreased significantly before 1990 and then increased significantly, with RH performing the opposite, while u₂ and R_n both declined significantly before 1990 and then slowed down. T_a, RH and R_n daily sensitivity coefficients to ET₀ increased first and then decreased, whereas u₂ showed a gradual change trend on the opposite. Generally, RH was the most sensitive factor except in summer, when R_n was the most sensitive factor. The four selected meteorological factors were reasonable and could explain well the contributions to ET₀ trends by employing the differential method. During 1961–1990, all meteorological factors provided negative contributions to ET₀ trends, which ultimately led to decreasing trends for almost all regions and periods. R_n was the leading factor annually, during the growing season and summer, while u₂ played a dominant role in other seasons. During 1991–2019, the leading factor of ET₀ trends changed to T_a annually, during the growing season, spring and summer, then to R_n in other seasons for most regions. T_a and RH exhibited positive contributions to ET₀ trends for most regions and periods, while u₂ and R_n showed negative contributions. Overall, the negative contributions from u₂ and R_n could not offset the positive contributions from T_a and RH, which, eventually, led to the slow upward ET₀ trends. Furthermore, the slightly increasing trends of u₂ and its extremely small contributions to the ET₀ trends in Region III deserve more attention. The outcomes obtained from this research should help in the understanding of the changing climate and provide a theoretical basis for the agricultural crop production and sustainable management of water resources in similar world regions.