Impacts of Climatic Change on Reference Crop Evapotranspiration across Different Climatic Zones of Ningxia at Multi-Time Scales from 1957 to 2018

,e impact of global climate change on agroecosystems is growing, affecting reference crop evapotranspiration (ET0) and subsequent agricultural water management. In this study, the climate factors temporal trends, the spatiotemporal variation, and the climate driving factors of ET0 at different time scales were evaluated across the Northern Yellow River Irrigation Area (NYR), Central Arid Zone (CAZ), and Southern Mountain Area (SMA) of Ningxia based on 20 climatic stations’ daily data from 1957 to 2018. ,e results showed that the Tmean (daily mean air temperature), Tmax (daily maximum air temperature), and Tmin (daily minimum air temperature) all had increased significantly over the past 62 years, whilst RH (relative humidity), U2 (wind speed at 2m height), and SD (sunshine duration) had significantly decreasing trends across all climatic zones. At monthly scale, the ET0 was mainly concentrated from April to September. And at annual and seasonal scales, the overall increasing trends were more pronounced in NX, NYR, and SMA, while CAZ was the opposite. For the spatial distribution, ET0 presented a trend of rising first and then falling at all time scales. ,e abrupt change point for climatic factors and ET0 series was obtained at approximately 1990 across all climatic zones, and the ET0 had a long period of 25a and a short period of 10a at annual scale, while it was 15a and 5a at seasonal scale. RH and Tmax were the most sensitive climatic factors at the annual and seasonal scales, while the largest contribution rates were Tmax and SD. ,is study not only is important for the understanding of ET0 changes but also provides the preliminary and elementary reference for agriculture water management in Ningxia.


Introduction
Evapotranspiration (ET) is not only an important component of the hydrological cycle but also essential for understanding land surface processes in climatology [1]. And the productivity is closely related to actual ET in agricultural research; thus, ET has important implications for improving local agricultural water management [2]. However, the data of ET are insufficient and limited due to lack of monitoring; therefore, many scholars use a variety of methods to study ET from different perspectives. e reference evapotranspiration (ET 0 ), which is defined as the potential ET of grass, can be used to prepare input data for hydrology models (e.g., SWAT model), schedule irrigation systems, and calculate the actual evapotranspiration (ET) in a basin or a region [3].
is is because the reference evapotranspiration (ET 0 ) has sufficient moisture, and it is not affected by soil factors. Linking ET 0 to a specific surface (grass) can provide a reference for ET on other surfaces [2]. In order to revise the reference evapotranspiration (ET 0 ) calculation criteria, the Penman-Monteith method has been recommended by Food and Agriculture Organization (FAO) Irrigation and Drainage Paper No. 56 [4]. And due to the easy calculation and ease of data access, the FAO56 has become one of the bestselling and most cited publications in the field of water resources management up to now [5].
Climate change has produced profound impact on society and environment. Many global issues are likely to be affected by climate change, such as extreme hydrological cycles, food security, biodiversity loss, and water scarcity [6,7]. ET 0 , which is the link between water balance and surface energy balance, is considered as a significant indicator for the climate change and the water cycle [8]. And as part of changing climate, ET 0 trends directly affect global and regional water resources. In order to assess agricultural demand, hydrological cycle, and ecological changes more effectively, a clear understanding of historical trends and future changes of ET 0 and its response to climate change are essential [9,10].
Climate change is intensifying and mainly characterized by a significant increase in temperature due to anthropogenic emissions of active greenhouse gases (such as CO 2 , CH 4 , and N 2 O) [11]. According to the report of IPCC, the average temperature of global land and sea surface showed a linear upward trend of rising by 0.85°C from 1880 to 2012, the average temperature increased by 0.78°C from 1850 to 1990, and the average temperature was projected to increase by 1.5°C for the end of 21st century [12]. Since the rising trend of air temperature has always existed, ET 0 has been increasing globally in the past few decades and reported in the Republic of Moldova [13], Greece [14], Iran [15], and France [16]. However, the ET 0 is declining in some areas at the same time, such as northwest China [17], New Zealand [18], United States [19], and southwestern China [20]. erefore, an increase in global temperature may not lead to the rise in ET 0 under any circumstances, which is called as the "evaporation paradox." To explain this problem, many scholars have done a lot of research by quantifying the influence of climate factors on ET 0 . ey concluded that changes in radiation and wind speed dominated the impact of ET 0 in some regions [21,22], while water vapor indicators and temperature were the main reasons for changes in other regions [23,24]. ese contradictory explanations and recent studies show that it is necessary to pay attention to the relationship between ET 0 and climate factors in the future research.
e Ningxia Hui Autonomous Region, one of the important parts of "the Belt and Road", is located in the temperate continental arid and semiarid area. e east, west, and north are surrounded by Maowusu, Tengger, and Ulanbu deserts, respectively. And the desertification land area accounts for 53.68% of the total area [25]. e amount of water resources is limited, and the demand for water has increased dramatically; thus, there is a sharp contradiction between economic development and limited water resources [26]. In the past decade, based on the Ningxia Water Resources Bulletin, the water consumption has increased by two times, and agriculture is the largest water-consuming sector, with more than 85% of water supplied for irrigation [25]. In addition, water resource is being directed from the agricultural sector to industry and other sectors; thus, food security and agricultural water security are facing severe pressure in Ningxia. ET 0 is a key component of the water cycle process and plays an important role in assessing water resources shortage [27]. Consequently, the analysis of ET 0 and its response to climate change can not only help to understand climate change but also improve water management.
In general, the main objectives of this study are as follows: (1) to discuss the interannual variation trend of climate factors; (2) to analyze monthly, seasonal, and annual ET 0 of spatiotemporal changes; (3) to explore the abrupt change and periodicity of ET 0 series; and (4) to quantify the impact of climate factors on ET 0 change and identify dominant factors. is research will help to raise awareness of climate change and provide valuable reference for researchers and policymakers to guide regional water management, agricultural production, and conservation of the environment.

Materials and Methods
To analyze the changes in reference evapotranspiration (ET 0 ) over Ningxia at multi-time scales, an approach framework for ET 0 analysis pattern selection was proposed. e framework mainly included the climate trend analysis, ET 0 spatiotemporal analysis, abrupt change and periodicity analysis, and ET 0 influence factors analysis ( Figure 1).

Study Area and Climate Data.
e Ningxia Hui Autonomous Region (NX) (35°14′∼39°23′N, 104°17′∼107°39′E) is part of the Yellow River basin and is located among the Alxa Plateau, the North China Plateau, and the Qilian Mountains folds. e area consists of Northern Yellow River irrigation area (NYR), Central Arid Zone (CAZ), and Southern Mountain Area (SMA) and has a total area of 6.6 × 10 4 km 2 [28] (Figure 2). is region faces a serious shortage of water resources and a huge contradiction between water supply and demand. And it has also led to serious environmental problems due to excessive use of water resources, including soil erosion, water pollution, and deterioration of the ecological environment [29]. Moreover, the seasons are divided into climate standards, namely, spring from March to May, summer from June to August, autumn from September to November, and winter from December to February [30]. In addition, the statistical significance of the linear trend and correlation analysis in this study was labeled as significance levels p < 0.05 ( * ) and p < 0.01 ( * * ) unless otherwise stated.
In Ningxia, 20 climate stations with 62 years (from 1957 to 2018) of climate data were selected (Table 1). e climate data, including daily minimum air temperature (T min ,°C), daily mean air temperature (T mean ,°C), daily maximum air temperature (T max ,°C), sunshine duration (SD, h), relative humidity (RH, %), and wind speed at 2 m height (U 2 , m/s), were downloaded from the National Meteorological Information Center (http://data.cma.cn). e datasets are available because the missing data rate handled by quality control is less than 0.1%, and some missing points were interpolated based on the regression relationship with those of the adjacent stations.

Climate Trend Analyses.
Climate tendency rate is a widely used method for studying climate change, which is simple and effective. It is common practice to establish a linear regression equation: where x i is the climate factors with a sample size of n, t i is the time corresponding to x i , b is the regression coefficient, and a is the regression constant. b is the annual change rate of climate factors, while its symbol indicates the change direction. When b > 0, it indicates an increasing trend with time; when b < 0, it indicates a decreasing trend with time. b × 10 is known as the climate tendency rate, indicating the change rate of climate factors every 10 years.

ET 0 Estimation.
Analyzing the spatiotemporal characteristics of ET 0 is helpful to understand the distribution and management of agricultural water requirements. e Penman-Monteith (PM) method, which is recommended by the Food and Agriculture Organization (FAO), has been widely used for calculating ET 0 [31]. e model was used in this study to estimate daily ET 0 and cumulative monthly, seasonal, and annual ET 0 : where ET 0 is the daily reference crop evapotranspiration (mm · day −1 ), Δ is the slope of vapor pressure curve (kPa·°C −1 ), R n is the net radiation at crop surface (MJ·m −2 ·day −1 ), G is the soil heat flux density (MJ·m −2 ·day −1 ), c is the psychrometric constant (kPa·°C −1 ), T is the average daily air temperature (°C), u 2 is the wind speed at 2 m (ms −1 ), e s is the saturation vapor pressure (kPa), and e a is the actual vapor pressure (kPa).  Figure 1: Approach framework, as proposed for ET0 analysis.

Advances in Meteorology
Organization and is widely used in climate trend analysis [32]: where UF K is a sequence of statistics calculated in time series x and UB K is a sequence of statistics calculated in reverse order of time series x. Uf i is a standard normal distribution; if a significance level α is given and |Uf i | > U (α/2) , it indicates a significant trend change in the sequence. And if there is an intersection point between the two curves of UF K and UB K , this is the moment when the abrupt changes starts.

Continuous Wavelet Analysis.
Wavelet analysis offers the possibility to better study time series problems, which can clearly reveal the multiple change cycles hidden in the time series, fully reflects the changing trend of the system in different time scales, and makes a qualitative estimation of the future development trend of the system. Although the continuous wavelet of the Morlet function was previously in the field of communication, it has been widely used in hydrological climate research with the interdisciplinary development [33]: where ψ(t) is the wavelet basic function, W f (a, b) is the wavelet variation coefficient, Var(a) is the wavelet variance, a is the scale factor of the wavelet period length, b is the time factor of time translation, t is time, f(t) is the time series number, and R is the real number field.

Sensitivity Analysis and Contribution Rate Assessment.
In order to quantitatively study the close relationship between ET 0 and different climatic factors, sensitive coefficient and contribution rate are the most commonly used methods by many scholars.
2.6.1. Sensitivity Analysis. Compared with other sensitivity analysis methods, the partial correlation coefficient method can analyze the complex nonlinear relationship of each factor by controlling the influence of other factors; thus, it is suitable for global sensitivity analysis, which is widely used in nonlinear dynamic problems [34]:

Advances in Meteorology
where S x is the sensitive coefficient; ΔET 0 and Δx are the change values of ET 0 and climate factors, respectively; and positive or negative of S x , respectively, indicates that ET 0 increases or decreases with the increase in climate factors.

Contribution Rate.
In order to determine the underlying causes of changes in ET 0 , sensitivity analysis needs to be combined with actual changes in climatic factors; thus, it is necessary to analyze the contribution rate of climatic factors to ET 0 [35]: where S x is the sensitive coefficient, C x is the contribution rate of climate factors to ET 0 changes, Rc x is the multiyear change rate of climate factors, Trend x is the annual climate tilt rate of climate factor x, x is the multiyear absolute average of climate factors, and n is the length of the time series.

Temporal Trends of Climate Factors.
From the climate data of each station from 1957 to 2018 (62 years), the average value of climate factors and the climate tendency rate was statistically calculated ( Figure 3 and Table 2). e climate in Ningxia had undergone significant changes in climate factors from the past 62 years. Spatially averaged T mean , T max , and T min all increased significantly (p < 0.01), and the change rates were 0.34°C/10a, 0.31°C/10a, and 0.51°C/10a, respectively.
Conversely, the RH and U 2 had a significant downward trend (p < 0.05), and the change rates were 0.42/10a and 0.10 m/s/ 10a, respectively. Although the SD was also showing a downward trend, the effect was not significant. Across the three different regions (NYR, CAZ, and SMA), T mean , T max , and T min also had significantly increased with similar variation. e increasing rates of T mean (0.38°C/10a), T max (0.36°C/ 10a), and T min (0.54°C/10a) were the largest in the NYR, while the CAZ was the smallest (0.31°C/10a, 0.24°C/10a, and 0.48°C/ 10a, respectively). Although the significant downward trends (p < 0.01) were found for RH in NYR and U 2 in CAZ, the RH, SD, and U 2 also showed a downward trend in the three different regions. e decreasing rates of RH (0.81%/10a) in the NYR and SD (0.40 h/10a) and U 2 (0.11 m/s/10a) in SMA were largest, while RH (0.06%/10a) and SD (0.04 h/10a) in CAZ and U 2 (0.08 m/s/10a) in NYR were smallest. Generally speaking, under the influence of global climate change in recent 62 years, the overall climate of Ningxia had been warming and drying, and the temperature has increased significantly. e T mean rise rate was 0.34°C/10a, which was lower than the rise rate of 0.37°C/10a in the northwest region of China, higher than the 0.23°C/10a of Chinese average [36,37], also higher than the 0.22°C/10a of global average [38,39]. And the rising rate was ranked as NYR > SMA > CAZ. Furthermore, the RH decreased in varying degrees, while the decreasing rate was also sorted by NYR > SMA > CAZ. e SD and U 2 had similar downward trend of change, and the decreasing rate was ranked as SMA � CAZ > NYR.
Additionally, the changes in climate factors during the year are basically consistent in all regions ( Figure 4). e temperature (T mean , T max , and T min ) showed a rising trend and then a decreasing trend, in which the maximum value appeared in July. And the temperature was ranked as CAZ > NYR > SMA. e RH performed a trend of falling first, then rising, and finally falling, in which the maximum value is September and the minimum value is April. Advances in Meteorology 5 And the RH was sorted by SMA > NYR > CAZ.
e SD and U 2 were relatively stable throughout the year without violent fluctuations, while they were ranked as CAZ > NYR > SMA.

Abrupt Change Test Analysis of Climate Factors.
Although there had been a trend of warming and drying in Ningxia in recent 62 years, when did this trend begin and what form did it take? e Mann-Kendall abrupt change test method was used to analyze the climatic factors in this study (Table 2). e results showed that there were different abrupt change points in each factor. Importantly, since the ET 0 in NX had an abrupt change point in 1990 ( Figure 5), the whole research period was divided into two periods (i.e., from 1957 to 1990 and from 1991 to 2018) for the convenience of analysis.
In general, the temperature trend changes were consistent in all regions.
e T mean and T min had increased significantly (p < 0.01) from two periods, while the T max decreased significantly from 1957 to 1990 and increased from 1991 to 2018. e T mean , T max , and T min changed by 0.094°C/10a, −0.044°C/10a, and 0.451°C/10a from 1957 to 1990 while changed by 0.451°C/10a, 0.446°C/10a, and 0.548°C/10a from 1991 to 2018 in NX, respectively. And this result showed that the extreme temperatures phenomenon in Ningxia was easing. e average value of T mean from two periods was both ranked as NYR > CAZ > SMA, while the largest value was 7.92°C/10a and 9.19°C/10a; the smallest value was 5.73°C/10a and 6.87°C/10a, respectively. And the average value trend of T min was similar to T mean . Moreover, the average value of T max was sorted by CAZ > NYR > SMA. e result showed that it is hotter in CAZ and colder in SMA. RH in NX changed significantly (p < 0.01) by 0.121%/10a and −1.598%/10a from two periods, respectively. It was also found across the three different climate regions, and this also implies the trend of drying. SD had an increased trend from 1957 to 1991 and a decreased trend from 1991 to 2018 in all regions, but the effect was not significant. e region with largest change was SMA, where increased by 0.1 h/10a from 1957 to 1991 and decreased by 0.175 h/10a from 1991 to 2018. U 2 had decreased significantly (p < 0.05) by 0.108 m/s/10a from 1957 to 1991 and 0.365 m/s/10a from 1991 to 2018 in NX, respectively. Although the three climate regions all had the similar trends, the effect was not significant in NYR and SMA.

Trend Analysis of ET 0 .
ere are significant differences in the temporal and spatial distribution of ET 0 due to the variability in the climate system and the complexity of the geographical environment. In this study, the time scale was divided into monthly, seasonal, and annual scales, and the spatial scale was divided into NYR, CAZ, and SMA.

Analysis at Monthly Scale.
Monthly averaged ET 0 at each region from 1957 to 2018 is listed in Table 3 and Figure 6. e average monthly ET 0 was sorted as CAZ (108.06 mm) > NYR (105.70 mm) > NX (101.87 mm) > SMA (83.02 mm). e ET 0 ranged from 33.49 mm·month −1 to 168.95 mm·month −1 with a total value of 1222.41 mm over the NX, while the extremum ratio was 5.04. In the three climate regions, both the ET 0 of NYR and CAZ were larger than NX, but the extremum ratio of NYR (5.80) exceeded CAZ (4.48). e ET 0 and extremum ration of SMA was the smallest among NYR, CAZ, and NX, while its values were 996.19 mm and 4.43, respectively. In general, ET 0 showed a trend of rising first and then falling during the year across all the regions. And the time of the monthly ET 0 extreme was almost the same in different regions, the maximum value was in May or June, and the minimum value was in December. In addition, the ET 0 from May to July was larger during the whole year in all regions, accounting for more than 40% of the total ET 0 , respectively.
In order to compare the spatial distribution of ET 0 in different seasons, the typical months were used for analysis ( Figure 7). Because autumn in Ningxia is too short to be representative, this study chose the typical months of spring (April), summer (August), and winter (December) for analysis. Climate and topography changed from north to south; ET 0 showed a trend of first falling, then rising, and finally falling in the typical months. And the ET 0 was the largest in the north and middle, while the smallest in the south. In April, ET 0 was the highest in Weizhou, Mahuangshan, Xingren, and Zhongning, and Huinong and Shitanjing were also higher. e overall change ranged from 90 mm to 160 mm, while it is in line with the climate characteristics of Ningxia's fast warming and strong wind. In August, the semiarid climate in Ningxia was characterized by less rainfall and longer sunshine, and ET 0 continued to increase with the range of variation from 100 mm to 165 mm. In December, ET 0 was significantly reduced compared to April and August due to the severe cold, and   rate was 0.80 mm/10a. Compared with the spring, the trends in summer, autumn, and winter were similar except for the proportion of ET 0 in NYR and CAZ, and they were all showed as CAZ > NYR > SMA (Table 4). In order to better understand the changing trend of ET 0 , this study was divided into two time periods (from 1958 to 190 and from 1991 to 2018), which is consistent with the analysis of climatic factors (Table 5). ET 0 tendency rate mainly showed two trends: the downward trend in the first stage and the upward trend in the second stage in NYR and SMA, while the CAZ had always shown a downward trend. Additionally, this study found that the ordering of ET 0 tendency rate was consistent in the two periods of three regions, which was showed as spring > summer > winter > autumn.
In terms of spatial distribution, ET 0 varies in different regions and seasons. But in general, it showed a trend of increase first and then decreased from north to south, and     gradually increased from east to west (Figure 9). e high value area of ET 0 presented two distinct distribution regions, namely, the Shitanjing and Huinong in the north and the Tongxin and Xingren in the middle. e low value area was mainly distributed in the south of Liupanshan, Jingyuan, and Longde. It was worth noting that the size of ET 0 in Lingwu, Qingtongxia, and Wuzhong has always been in the middle position. e ET 0 in Yanchi and Mahuangshan was special, where the spring and autumn were relatively large and the summer and winter were relatively small. In addition, the ET 0 in spring was significantly higher than the autumn, and the ET 0 in summer was significantly higher than other seasons. Overall, it was sorted as summer > spring > autumn > winter, which was consistent with the local climate characteristics of fast spring, short summer heat, cool autumn early, and long winter and cold.

Analysis at Annual Scale.
e ET 0 in each climate zone had similar trend in the past 62 years ( Figure 10 and Table 6). NX, NYR, and SMA (p < 0.01) showed significant growth trends of 5.19 mm/10a, 12.12 mm/10a, and 5.09 mm/ 10a, respectively, while the CAZ decreased at a trend rate of 4.52 mm/10a. And the change rate of different regions was ranked as NYR > NX > SMA > CAZ. In particular, although the temperature of CAZ was rising, ET 0 showed a downward trend, indicating that there was "evaporation paradox" in parts of Ningxia. Overall, the annual ET 0 ranking was CAZ (1296.67 mm) > NYR (1268.34 mm) > NX (1222.41 mm) > SMA (996.19 mm), while the extreme ratio was ranked as NYR (1.30) > SMA (1.29)> CAZ (1.28) > NX (1.27). In addition, it was similar to the monthly scale for the maximum and minimum of annual ET 0 , with the trend of both NYR and CAZ being greater than NX and CAZ less than NX. e spatial distribution characteristics of ET 0 have significant regional ( Figure 11). As the terrain changes from north to south, ET 0 showed a trend of decline first, then rise, and finally decline. Located at the southern of Ningxia, the SMA is cool and humid, and the T mean was 2.25°C and 2.08°C lower than NYR and CAZ, respectively, while the RH larger by 8.18% and 8.02%, respectively. us, ET 0 in SMA was smaller than the NYR and CAZ. CAZ, located in the middletemperate and semiarid zone, is a typical continental climate with strong drought and heavy evaporation. Compared with before the abrupt change point of 1990, the T mean rose by 1.3°C and the rainfall dropped by 4.2%; thus, the ET 0 of CAZ was larger than the NYR and SMA. Particularly, the ET 0 of NYR was larger in the north climate stations, while the

Mann-Kendall Test and Wavelet Analysis of ET 0
3.3.1. Analysis at Annual Scale. In the past 62 years, ET 0 had shown a trend of first drop and then rise in NX, NYR, CAZ, and SMA, while the abrupt change points were 1990, 1990 (the same to NX), 1991, and 1989, respectively ( Figure 5). NX and NYR had a clear upward trend in the 21st century (Z > 1.96), and CAZ showed an obvious downward trend in the 1970s (Z < 1.96), while the SMA did not have a significant trend of upward and downward. Furthermore, the solid line is a positive value, indicating that ET 0 is more than the multiyear average value. And the dashed line is negative, indicating that ET 0 is less than the multiyear average value. e junction of the solid line and the dotted line indicates the change point of ET 0 . e long and short periods of ET 0 in NX and NYR were 25a and 10a, respectively, with the 10a being the most significant ( Figure 12). Similarly, the periods of ET 0 in SMA were 10a and 5a. Nevertheless, the CAZ was special with one period of 15a.

Analysis at Seasonal Scale.
From the seasonal scale of Ningxia, the performance of ET 0 could be divided into two stages: the alternate change of the early rise and decline to the stable rise in the later period ( Figure 13). e abrupt change point in spring, summer, and autumn was the same (1990), while the abrupt change point in winter was earlier (1980). ere was an obvious upward trend in the 21st century (z > 1.96) in summer, autumn, and winter. But in spring, the trend was not significant. Seasonal periods all showed the similar trends. ET 0 had a long period of 15a and a short period of 5a over the entire time domain in spring, summer, autumn, and winter ( Figure 14). Take the 15a period in spring as an example; the time point of larger ET 0 was the early 1960s and 1980s, while the smaller ET 0 was the late 1960s, 1970s, and 1990s. (5), the sensitive coefficients between ET 0 and other six climate factors were obtained ( Table 7). All regions had the similar trends, T mean , T max , T min , SD, and U 2 were positive values, while the RH was negative. is suggested that the increase in RH caused the decline of ET 0 , and other climatic factors were the opposite. Moreover, among the 20 climate stations, the average minimum and maximum climate factors for the sensitive coefficient of ET 0 were T min (0.08) and RH (0.46), respectively. In general, the sensitive coefficient of Ningxia was ranked as RH > T max > U 2 > T mean > SD > T min . Additionally, there were two trends of the sensitive coefficient in spatial distribution, the T max , T mean , and RH increased first from north to south and then decreased, while the SD, T min , and U 2 were the opposite. us, the CAZ was special, where the sensitive coefficient of T max , T mean , and RH was the largest, and other climate factors were the smallest.     As depicted in Figure 15, although the contribution rates of T max , T mean , T min , and RH were all positive, the reason for the change in ET 0 was exactly the opposite. Positive sensitive coefficient and positive climate trend rate of T max , T mean , and T min led to the positive contribution rate, while the sensitive coefficient and climate trend rate of RH both were negative. In particular, due to the positive sensitive coefficient and negative climate trend rate, the contribution rates of SD and U 2 were both negative, indicating that they led to the decline of ET 0 . In general, the most contribution rates in NX, NYR, CAZ, and SMA were T min , T min , T max , and U 2 , respectively, while the SD was the smallest in all regions.

Analysis at Seasonal Scale.
As for the seasonal scale, the sensitive coefficients of six climate factors were consistent with the annual scale. All showed that T mean , T max , T min , SD, U 2 , and RH were positive values, while the RH was negative value (Table 8). In the spring and summer of NYR, CAZ, and NX, the most sensitive coefficient of main climate factors was T max , while the RH (absolute value) was the largest in autumn and winter. e SMA was different, and its most sensitive coefficient was always RH (absolute value) in four seasons. In addition, there were also two cases for the smallest sensitive coefficient of main climate factors, and it was T min in four seasons of CAZ and NX. Moreover, the NYR and SMA were T min in spring, summer, and autumn while SD in winter. Additionally, the sensitive coefficients of the main climate factors were also varying in the four seasons. e T mean , T max , and SD showed an increase in trend first and then decrease, while the U 2 was the opposite. As for T min and RH, they were special, and T min appeared as an alternating change in rise and decline, while RH was always showing an increase in trend.
As for the contribution rate depicted in Figure 16. In spring, SD and U 2 were negative values, whilst the T mean , T max , T min , and RH were positive values. And the largest contribution rate in NYR and CAZ was RH, while the U 2 was largest in SMA and NX. Moreover, there was no regularity in summer and autumn. In summer, the largest contribution rates in NYR, CAZ, SMA, and NX were U 2 , T max , T min , and U 2 , respectively. And in autumn, the largest contribution rates were T max , T max , T mean , and U 2 , respectively. In winter, it was extremely consistent in different regions, in which the T max was always the dominant factor. In addition, the climate factors varied from spring to winter, but it could be roughly divided into two situations. e T mean , T max , and T min were relatively regular, which showed a trend of decreasing first and then increasing. On the contrary, the SD, U 2 , and RH did not show obvious regularity.

Variation in Climate Factors.
In the past thousand years, the temperature had showed an unprecedented upward trend, indicating the global warming as an indisputable fact.
is study found that the Ningxia was suffering from climate warming (i.e., 0.34°C/10a in T mean , 0.31°C/10a in T max , and 0.51°C/10a in T min ( Table 2)), which is consistent with most research in the past few decades [40] and almost triple the global temperature increasing rate shown in the IPCC Fifth Assessment Report (0.12°C/10a). e rise in temperature may be due to two reasons: one is the increase in solar radiation caused by the destruction of the ozone layer, and the other is the large amount of greenhouse gas emissions caused by economic development and population growth [41].
e U 2 showed a downward trend in Ningxia based on this study (0.10 m/s/10a, Table 2), which was reported in some relevant studies [42]. e natural reason, internal role of the natural system, is the main reason for the decrease in wind speed, including the weakening of the East Asian winter monsoon and summer monsoon in recent decades [43]. Additionally, the SD also showed a decreasing trend in Ningxia (0.03 h/10a, Table 2), which is generally consistent      [44]. ere are different views on the reduction of sunshine hours, and decreased visibility due to rising aerosol content in the troposphere may be the main cause [45]. Similar to U 2 and SD, the RH showed a downward trend from 1957 to 2018 (0.42%/10a, Table 2). And this result is coincident to the other studies [46]. e impact of climate change on RH is a complex issue, involving temperature, surface runoff, vegetation types, and even the underground water [47]. erefore, it is difficult to determine the fundamental causes of RH change. Furthermore, this study found that the 1990s seemed to be a time code of abrupt change, in which the six climate factors were beginning to change before and after it. e temporal trend of the six climate factors in Ningxia is consistent with most of the findings, while there are minor differences in spatial distribution, which is mainly due to the form of climate data and the area size.

ET 0 Trends and "Evaporation Paradox".
As the rise in T and the decrease in U 2 , RH, and SD, the ET 0 is generally considered to gradually increase. However, the performance of different regions in Ningxia is inconsistent. e annual and seasonal ET 0 showed an increasing trend in NX (5.79 mm/10a), NYR (12.24 mm/10a), and SMA (5.09 mm/ 10a), while it was decreasing in CAZ (4.52 mm/10a). is implied that there was "evaporation paradox" in the CAZ of Ningxia, and this estimate for ET 0 was higher than the entire China (3.5 mm/10a) [34] and the Northern Loess Plateau of China (3.3 mm/10a) [9]. For decades, many scholars studied the causes of the "evaporation paradox" in different regions and have reached different conclusions [48,49]. It is generally believed that there are three main reasons for the decline of ET 0 . e decrease in solar radiation was caused by the increase in atmospheric cloud amount (aerosol), and the decrease in water vapor pressure was caused by the increase in air humidity and the decrease in wind speed. In order to obtain the cause of the evaporation paradox in the CAZ of Ningxia, the MODIS satellite data (MOD08) from 2000 to 2018 were selected to analyze the distribution and variation in cloud optical thickness and aerosol optical thickness. As depicted in Figure 17, the CAZ's annual  average cloud optical thickness is 24 in 2018, and the upward trend is very significant compared to 16 in 2000. Generally, the increase in the thickness of the cloud optics reduces the total amount of radiation reaching the ground. is study analyzed the total solar radiation variation in CAZ since 1957. e total radiation of CAZ had a significant downward trend from 1957 to 2018 (−13.418 MJ/m 2 /a). In general, the increase in cloud optical thickness in CAZ led to a decrease in total solar radiation, which is the main reason for the decline in ET 0 .

e Impacts of Climate Change on the Variation in ET 0 .
e impact of climate factors on the contribution of ET 0 changes is not only related to the sensitivity of ET 0 but also to the changes in climate factors themselves. e sensitivity analysis at annual scale in this study showed that the ET 0 in Ningxia was most sensitive to RH (−0.46), followed by T max (0.35), U 2 (0.28), T mean (0.24), SD (0.15), and T min (0.08),  which is consistent with some relevant studies [41,50]. Simultaneously, the positive changes in ET 0 in Ningxia due to the relative increase in T max was larger than other climate factors. And the relative contribution rate of RH was small although it was the most sensitive. Additionally, another explanation for what is known as "global dimming" in previous studies was the reduction in SD, which was the most important controlling factor leading to the reduction in ET 0 [51]. Nevertheless, this study found that the SD had less contribution rate to ET 0 in Ningxia compared to other climate factors. And this result is in accordance with Qi [30], who indicated that the SD had less influence on ET 0 in Northeast China. Compared to the annual scale, the sensitivity analysis in this study illustrated that the ET 0 in Ningxia at seasonal scale had no obvious regularity. e T max was the most sensitive in spring (0.42) and summer (0.46), while the RH was the largest in autumn (−0.54) and winter (−0.55), and this is different from Wang [41], who found that T max was the most sensitive to ET 0 in four seasons. As for the contribution rate of Ningxia, it was similar to the sensitivity. And this is inconsistent with omas [52], who emphasized that the U 2 changes ET 0 in water-limited areas of west China. e possible reason for this is that the climate difference among the three regions (NYR, CAZ, and SMA) of Ningxia was relatively large [53].

e Future Study of ET 0 .
ET 0 is significant for the water and energy balance of terrestrial ecosystems, and reasonable prediction of future ET 0 is not only beneficial to water resources management but also important for guiding agricultural production [54]. Although Ningxia is deeply inland in the northwest of China, the ET 0 is significant affected by the climate. us, we can study the ET 0 from two perspectives in the future.
First, there are many atmospheric teleconnection patterns (i.e., Atlantic Oscillation (AO), the Indian Ocean Dipole (IOD), Pacific Decadal Oscillation (PDO), and El Niño-Southern Oscillation indices (ENSO)) in the global climate system and the effects of these atmospheric circulations have broken through time and space constraints.
at is to say, the influence can occur simultaneously or sequentially in time, and the change can be far apart in space. erefore, these atmospheric teleconnection models can change ET 0 in annual and seasonal scales by affecting climatic factors. With a deep understanding of the atmospheric circulation system in the future, it is possible to take advantage of the atmospheric teleconnection models to predict future ET 0 [23].
Second, the air temperature had increased over the past 50 years and is reported in Xinjiang [55], Gansu, [56] and Mongolia [57]. And McVicar observed a drying trend due to the reduced rainfall in northwest China [58]. e future study is necessarily motivated by the abovementioned considerations to assess sensitivity of the evapotranspiration due to ±20% change in several climatic factors. Specially, the IPCC report for the 21 st century can be considered in the future study, which can describe and quantify the impacts of climatic factors on seasonal and annual ET 0 based on the climate change [59].

Conclusions
e climate factors temporal trends, the spatiotemporal variation of ET 0 at different time scales, and its climatic driving factors across different climatic zones of Ningxia were investigated with the climate tendency rate, Mann-Kendall test, continuous wavelet analysis, sensitivity analysis, and contribution rate assessment based on daily data of 20 climatic stations from 1957 to 2018. e main conclusions of this study are as follows: (1) T mean , T max , and T min all have increased significantly over the past 62 years in Ningxia, whilst RH, U 2 , and SD have significantly decreasing trends. And this trend has become more pronounced with 1990 as the abrupt change point. (2) e ET 0 is mainly concentrated from April to September in a year. In NX, NYR, and SMA, the ET 0 series has a significant increase in both annual and seasonal scales, while CAZ is the opposite. In terms of spatial distribution at monthly, seasonal, and annual scales, there is a trend of increasing first and then decreasing from north to south. (3) An abrupt change point in annual ET 0 is detected around the year of 1990, and the annual ET 0 decreased significantly from 1957 to 1990, while it increased significantly from 1991 to 2018. e ET 0 has a long period of 25a and a short period of 10a at annual scale, while it is 15a and 5a at seasonal scale. (4) At the annual and seasonal scales, the most sensitive climatic factors are RH and T max , while the largest contribution rates are T max and SD.
e results of this study can not only help to guide the agricultural water management in Ningxia but also contribute to agricultural production and environmental protection. In the future work, the relationship between atmospheric circulation and ET 0 can be analyzed for ET 0 prediction, which is most significant for the researchers and decision makers.
Data Availability e data used in this paper are provided by the National Meteorological Information Center (http://data.cma.cn).

Conflicts of Interest
e authors declare that they have no conflicts of interest.

Authors' Contributions
Ziyang Zhao contributed substantially to conceptualization, methodology, validation, data curation, data interpretation, and writing. All authors participated in drafting the article or revising it critically and gave final approval of the version to be submitted.