On the Attribution of Changing Reference Evapotranspiration in a Coastal Area of China

Abstract

Reference evapotranspiration (ET₀) is a crucial parameter for hydrological modeling, land-atmospheric interaction investigations and agricultural irrigation management. This study investigated changes in ET₀ and attributed those changes to climate variations in a coastal area (Zhejiang province) of China by a numerical experiment method. The results indicated that annual ET₀ increased significantly (p < 0.05) at a rate of 1.58 mm·year⁻¹ from 1973 to 2013, which was mainly caused by an obvious increase in ET₀ in spring. Air temperature and water vapor pressure deficits increased significantly (p < 0.05) at rates of 0.04 °C·year⁻¹ and 0.005 kPa·year⁻¹, respectively, at an annual time scale during the study period, while wind speed and solar radiation decreased significantly (p < 0.05) at rates of −0.01 m/s·year⁻¹ and −3.94 MJ·m⁻²·day⁻¹·year⁻¹, respectively. The contributions of changes in air temperature, wind speed, water vapor pressure deficits and solar radiation to increases in ET₀ were 0.39, −0.56, 2.62 and −0.61 mm·year⁻¹, respectively. The decrease in wind speed and solar radiation negatively affected the increase in ET₀, which was offset by the positive effects of the air temperature and water vapor pressure deficits increase, where the water vapor pressure deficits was the dominant factor in increasing ET₀ in the coastal area. Moreover, the impact of topography on ET₀ was further discussed. ET₀ changes at plain stations were approximately 5.4 times those at hill stations, which may be due to the impact of a large water body and the augment of surface roughness from intense human activities in the well-developed plain area. The results are helpful for investigating spatial and temporal changes in the evaporative demand for well-developed regions under energy-limited conditions.

1. Introduction

Reference evapotranspiration (ET₀) is an important parameter for hydrological process simulations, land surface model investigations and agricultural irrigation management, especially in humid regions, which reflects a potential demand for evapotranspiration under a certain underlying surface [1,2,3,4]. Under a changing climate, significant decreases in ET₀ have been reported all over the world, which are associated with increasing air temperature, also known as an “evaporation paradox” [5,6,7]. However, recent studies indicated that ET₀ has shown an increasing trend, and the “evaporation paradox” has disappeared since the late 1990s [8,9,10].

Quantifying changes in ET₀ due to climate variations is an important issue in current studies. There are several methods for attributing ET₀ changes to climate variables, such as a regression analysis, sensitivity method, detrended method, differentiation approach and numerical modeling [11,12,13,14]. For example, Vicente-Serrano et al. [14] studied the changes in monthly ET₀ in Spain by a sensitivity analysis and found there were spatial gradients in ET₀ sensitivity. Xu et al. [12] used a detrended method to study ET₀ changes in the Yangtze River catchment and concluded that the decrease in total net radiation was the dominant factor for changes in ET₀. Liu and Sun [15] assessed atmospheric evaporation demands in 12 state-of-the-art global climate models against D20 pan evaporation observations by a differentiation approach. Sun et al. [16] attributed changes in ET₀ to climate variables in southwestern China using numerical experiments. These abundant investigations are beneficial for understanding the dynamic response of evaporative demand changes to complex land-atmospheric interactions [17,18,19].

Up to now, many studies on ET₀ changes are available for small catchments, continents and at the global scale [20,21,22]. Changes in ET₀ exhibited different trends, and controlling factors varied under different climate conditions at different time scales [3,23,24]. Wang and Dickinson [20] conducted an investigation on global atmospheric evaporative demands over land from 4250 globally distributed stations from 1973 to 2008 and found that the contribution of water vapor pressure deficits to the monthly variability of evaporative demands was much larger than that of solar radiation, wind speed and air temperature at the global scale. In addition, several studies have attempted to investigate the impact of natural factors (i.e., geography, water bodies, and climate conditions) and socio-economic factors on ET₀ (i.e., agriculture and household income) [10,25]. For example, Zhang et al. [10] indicated that the amplitude of ET₀ changes near a large freshwater lake was larger than that away from the lake, which was due to a larger sensitivity of ET₀ to climate variables near the lake. Han et al. [26] found that the decrease in ET₀ was accelerated in agricultural stations, with a more significant decrease in the aerodynamic term in arid/semi-arid regions and a more significant decrease in the radiation term in humid regions. Litvak et al. [25] found that evapotranspiration from the total land area was linearly correlated with the median household income in Los Angeles, California.

Zhejiang province is one of the most developed regions in China and is located in the southeastern coastal area of the country [27]. As a critical input for hydro-meteorological predictions, changes in ET₀ have been investigated by several researchers [28,29,30]. Xu et al. [28] studied future ET₀ changes from 2011 to 2014 by a global sensitivity analysis method that used one regional climate model in Zhejiang province and concluded that intra- and inter-annual changes in ET₀ were mainly due to changes in solar radiation, relative humidity and daily minimum temperature. Pan et al. [29] investigated the sensitivity of ET₀ to climate variables in the Shengsi Islands in Zhejiang province and found that ET₀ was most sensitive to solar radiation in the summer and most sensitive to relative humidity during the remaining seasons. These results are important for the improvement of hydrological modeling, urban flood-drought risk management and the planning of commercial crop irrigation. However, a comprehensive quantification of changes in ET₀ to climate variables in Zhejiang province has not been available until recently. Moreover, the influence of topography on ET₀ changes has not been clear in this coastal area.

The objectives of this study are (1) to analyze the changes in ET₀ and climate variables in Zhejiang province from 1973 to 2013; (2) to quantify the contribution of climate variations to ET₀ changes; and (3) to discuss the influence of topography on ET₀ changes.

2. Study Area and Data

2.1. Study Area

Zhejiang province is located at the eastern coast of China, south of the Yangtze River Delta (Figure 1), which is one of the most rapidly developed regions in China [31]. The area of this province is 1.02 × 10⁵ km², and the altitude ranges from −74 to 1922 m. Zhejiang province belongs to the typical Asian monsoon region. The average air temperature is 17.2 °C, the extreme maximum air temperatures is 43.2 °C and the extreme minimum air temperatures is −13.5 °C. The average annual rainfall is 1493.3 mm. The rainstorm days is 4.1 days and 50% of rainstorm occurs during the Mei-Yu period from June to July, which is caused by a persistent stationary front known as the Mei-Yu front [32]. This coastal province is affected by typhoon frequently, which mainly occurred during July to September [33,34]. Zhejiang province is a powerful economic province in China, and the per capita net income ranks the first in China for more than 20 years, which has reached the level of moderately developed countries [35].

2.2. Data

Daily meteorological records from 1973 to 2013 at 46 meteorological stations were collected from Zhejiang Meteorological Bureau and included sunshine duration (hour), wind speeds at 10 m (m/s), relative humidity (%), minimum air temperature (°C), and maximum air temperature (°C). Considering the continuity of meteorological data, missing data from the selected stations were less than 2%. All meteorological data were accumulated into monthly values. If the data missing for a station were more than 9 days in a certain month, then the value of this month was replaced by the average climate value. Climate variables and ET₀ for the whole province were averaged from all 46 stations.

The Digital Elevation Model (DEM) data for Zhejiang province was downloaded from the Global Land Cover Facility with a resolution of 90 m × 90 m [36], which was used to analyze the impact of topography on ET₀ changes.

3. Methods

3.1. Calculation of ET₀

The Penman-Monteith formula, recommended by the Food and Agriculture Organization of the United Nations, was used to estimate ET₀ (mm), which represents the evaporative capacity of a hypothetical reference crop with an assumed crop height of 0.12 m, a fixed surface resistance of 70 s·m⁻¹ and an albedo of 0.23 [37]:

$$E{T}_0=\frac{0.408\Delta (R_n-G)+\gamma \frac{900}{T+273}{U}_2·VPD}{\Delta +\gamma (1+0.34U_2)}$$

(1)

where R_n is the net radiation (MJ·m⁻²·day⁻¹), G is the soil heat flux density (MJ·m⁻²·day⁻¹), T is the mean air temperature (°C), VPD is the water vapor pressure deficits (kPa), U₂ is the wind speed at 2 m (m·s⁻¹), Δ is the slope of the vapor pressure curve versus temperature (kPa·°C⁻¹) and γ is the psychrometric constant (kPa·°C⁻¹).

Wind speed observed at 10 m was adjusted to 2 m by a logarithmic wind speed profile:

$$U_2=U_{10}\frac{4.87}{\ln (67.8\times 10-5.42)}$$

(2)

G is calculated by:

$$G=0.14(T_i-T_{i-1})$$

(3)

where T_i and T_i−1 are the air temperature at time i and i − 1.

Solar radiation (R_s) is calculated by:

$$R_s=(a_s+b_s\frac{n}{N})R_a$$

(4)

where a_s and b_s are coefficients, n is the sunshine duration and N is the maximum possible duration of sunshine. R_a is the extraterrestrial radiation calculated by the Angstrom formula [3].

R_n is calculated by:

$$R_n=(1-\alpha )R_s-\sigma \left[\frac{{T_{max,k}}^4+{T_{min,k}}^4}{2}\right]\left(0.34-0.11\sqrt{VP}\right)\left(1.35\frac{R_s}{R_{so}}-0.35\right)$$

(5)

where α (=0.23) is the albedo of the reference grassland. σ is Stefan-Boltzmann constant (4.903 × 10⁻⁹ MJ·K⁻⁴·m⁻²·day⁻¹), T_max,k is maximum daily temperature (K = °C + 273.16), T_min,k is minimum daily temperature (K = °C + 273.16), R_so is clear-sky radiation (MJ·m⁻²·day⁻¹).

3.2. Attribution of Changes in ET₀

A numerical experiment method proposed by Sun et al. [16] was employed to quantify the impact of climate variations on changes in ET₀ in this study. For ET₀, the combined effects of air temperature, solar radiation, wind speed and water vapor pressure deficits were designed in five experiments to attribute the influence of climate variations on ET₀ change, including one control experiment and four sensitivity experiments (

Table 1. Numerical experiment design for ET₀ contribution.

Experiment	Description
Control Test: ET0_CTR	air temperature, solar radiation, wind speed, actual vapor pressure from 1973–2013
Temperature Test: ET0_T	air temperature maintained in 1973, the others same as the control test
Solar radiation Test: ET0_Rs	solar radiation maintained in 1973, the others same as the control test
Wind speed Test: ET0_U2	wind speed maintained in 1973, the others same as the control test
Actual vapor pressure Test: ET0_VPD	water vapor pressure deficits maintained in 1973, the others same as the control test

). ET₀ calculated by the control experiment represented the combined influence of the four climate variables, whereas the ET₀ calculated by each sensitivity experiment represented the combined influence of the other three climate variables, excluding the non-changed variable. The impact of each climate variable on changes in ET₀ can then be estimated as:

$$C(x_i)=L(E{T}_0\_CTR)-L(E{T}_{0\_}{x}_i)$$

(6)

where L(ET₀_CTR) and L(ET₀_x_i) are the calculated long-term trends in ET₀ via a linear regression based on the control test and the sensitivity test, respectively. The variable x_i represents either air temperature, solar radiation, wind speed or the water vapor pressure deficits, and C(x_i) is the contribution of the corresponding climate variable to ET₀ changes. Errors in the experiment analyses can be calculated as:

$$err=L(E{T}_0\_CTR)-{\displaystyle \sum _{i=1}^{4}C(x_i)}$$

(7)

Then, the relative contribution R(x_i) of each climate variable to changes in ET₀ can be calculated as:

$$R(x_i)=\frac{C(x_i)}{L(E{T}_0\_CTR)}\times 100\%$$

(8)

4. Results and Discussion

4.1. Changes in ET₀

Figure 2 shows the spatial distribution of changes in annual ET₀ at 46 meteorological stations in Zhejiang province from 1973 to 2013. Annual ET₀ at 39 stations increased during the past 40 years and trends at 24 stations were significant at the 95% level. Highly increasing trends in ET₀ were generally located at northern and coastal stations, which are well-developed socio-economic regions. Annual ET₀ at 7 stations showed a decreasing trend, and only one station had a trend significant at the 95% level. Annual ET₀ increased significantly (p < 0.05) at a rate of 1.58 mm·year⁻¹, which was the provincial average (

Table 2. Trends in ET₀ and climate variables in the study area during the period of 1973–2013.

Variable	Annual	Spring	Summer	Autumn	Winter
ET0 (mm·year−1)	1.58 *	1.12 *	0.23	0.34	−0.17
T (°C·year−1)	0.04 *	0.04 *	0.03 *	0.03 *	0.03 *
U2 (m/s·year−1)	−0.01 *	−0.01 *	−0.01 *	−0.01 *	−0.01 *
VPD (kPa·year−1)	0.005 *	0.006 *	0.009 *	0.005 *	0.001 *
Rs (MJ·m−2·day−1·year−1)	−3.94 *	1.37	−2.43 *	−1.05	−1.84 *

Note: * means the slope is significant at the level of 0.05 by t-test.

). For seasonal distribution, the changes in ET₀ clearly varied (Figure 3). It is worth noting that ET₀ decreased at 35 stations in winter, and the trends at 8 stations were significant at the 95% level. For the provincial average, ET₀ increased at rates of 1.12, 0.23 and 0.34 mm·year⁻¹ during spring, summer and autumn, respectively, whereas ET₀ decreased at a rate of −0.17 mm·year⁻¹ in winter.

4.2. Changes in Climate Variables

Figure 4 shows the changes in climate variables related to ET₀ in Zhejiang province during the period of 1973–2013. The annual mean air temperature at the 46 stations presented an upward trend, and all trends passed the 0.05 significance level by the t-test (

Table 3. Number of stations (out of 46 stations) with ET₀ and climate variables changes at annual and seasonal time scales.

Time	Trend	ET0	T	U2	VPD	Rs
Annual	Upward	39(24)	46(46)	11(4)	46(45)	12(0)
	Downward	7(1)	0	35(29)	0	34(26)
Spring	Upward	46(45)	46(45)	11(3)	46(46)	42(14)
	Downward	0	0	35(27)	0	4(0)
Summer	Upward	28(5)	46(45)	14(4)	46(45)	3(0)
	Downward	18(1)	0	32(23)	0	43(24)
Autumn	Upward	35(16)	46(45)	14(5)	46(44)	9(0)
	Downward	11(0)	0	32(23)	0	37(13)
Winter	Upward	11(0)	46(27)	10(3)	43(35)	0
	Downward	35(8)	0	36(30)	3(0)	46(33)

Note: the number in the brackets indicates the number of stations where its trend is significant at the level of 0.05 by t-test.

). For the provincial average, the annual mean air temperature increased significantly (p < 0.05) at a rate of 0.04 °C·year⁻¹. Changing rates in air temperature were 0.04, 0.03, 0.03 and 0.03 °C·year⁻¹ during the four seasons, respectively.

Wind speed showed declining trends at 35 stations, and the declining trend was significant (p < 0.05) at 29 stations. Wind speed at the other 11 stations showed rising tendencies, and the increasing trend was significant (p < 0.05) at 4 stations. For the regional average, the decreasing rate of wind speed for the whole province was −0.01 m/s·year⁻¹ (p < 0.05) from 1973 to 2013 at both annual and seasonal time scales.

The water vapor pressure deficits at all 46 stations showed an increasing trend, and the trends were significant at the 0.05 level at 45 stations. Water vapor pressure deficits increased significantly (p < 0.05) by 0.005 kPa·year⁻¹ for the regional average. For the four seasons, the water vapor pressure deficits increased at rates of 0.006, 0.009, 0.005 and 0.001 kPa·year⁻¹, respectively.

Solar radiation decreased at 34 stations, and trends were significant at 26 stations. Solar radiation at the other 12 stations showed an increasing trend. For the provincial average, solar radiation decreased significantly at a rate of −3.94 MJ·m⁻²·day⁻¹·year⁻¹ from 1973 to 2013. It is worth noting that solar radiation increased at a rate of 1.37 MJ·m⁻²·day⁻¹·year⁻¹ in the spring.

4.3. Quantification of ET₀ Changes

Figure 5 shows there are significant relationships between trends in ET₀ from the control test and the accumulated trends in ET₀ from the four sensitivity tests at annual and seasonal time scales. The R² is 0.98, and the relative error is −16.5% at the annual time scale for the provincial average (

Table 4. Contributions of climate variations to ET₀ changes in Zhejiang province during the period from 1973 to 2013.

	Variable	Annual	Spring	Summer	Autumn	Winter
Contribution (mm·year−1)	ET0	1.58	1.12	0.29	0.34	−0.17
	T	0.39	0.08	0.11	0.10	0.11
	U2	−0.56	−0.12	−0.17	−0.16	−0.12
	VPD	2.62	0.82	0.95	0.61	0.22
	Rs	−0.61	0.41	−0.55	−0.17	−0.35
	Error	−0.26	−0.07	−0.05	−0.04	−0.03
Relative contribution (%)	T	24.7	7.1	37.9	29.4	−64.7
	U2	−35.4	−10.7	−58.6	−47.1	70.6
	VPD	165.8	73.2	327.6	179.4	−129.4
	Rs	−38.6	36.6	−189.7	−50.0	205.9
	Error	−16.5	−6.2	−17.2	−11.8	17.6

). R² ranges from 0.97 to 0.99, and the relative error ranges from −17.2 to 17.6% for the four seasons. This result indicates that the method used in this study is effective for quantifying the impact of climate variations on ET₀ changes.

During the period from 1973 to 2013, changes in air temperature, wind speed, water vapor pressure deficits and solar radiation led to changes in ET₀ of 0.39, −0.56, 2.62 and −0.61 mm·year⁻¹, respectively, at the annual time scale. The positive effects of the increase in air temperature and water vapor pressure deficits were offset by the negative effects of the decrease in wind speed and solar radiation. The combined effects of the four climate variables led to an increase in ET₀ of 1.84 mm·year⁻¹. The increase in the water vapor pressure deficits was the dominant factor for the increase in ET₀. For the 46 stations, air temperature, wind speed, water vapor pressure deficits and solar radiation were the dominant factors for changes in ET₀ at 0, 1, 43, and 2 stations, respectively (Figure 6).

Seasonally, air temperature and water vapor pressure deficits have positive effects on the increase in ET₀ in spring, summer and autumn and negative effects on the decrease in ET₀ in winter. The decrease in wind speed has a negative effect on the increase in ET₀ in spring, summer and autumn and a negative effect on the decrease in ET₀ in winter. The decrease in solar radiation has a negative effect on the increase in ET₀ in summer and autumn, but it has a positive effect on ET₀ changes in spring and winter. The increase in water vapor pressure deficits is the controlling factor for the increase in ET₀ in spring, summer and autumn, while a rapid decrease in solar radiation is the controlling factor for the decrease in ET₀ in winter (Figure 7).

4.4. Influence of Topography on ET₀ Changes

It seems plausible that the changes in ET₀ at plain stations are larger than those at hill stations (Figure 2 and Figure 3), which reflects the impacts of topography on ET₀ changes. Therefore, all 46 stations are classified into two categories based on the degree of topographic relief: stations in the plain area, where the degree of topographic relief is lower than 100 m, which reflects the influence of a large water body, and stations in the hill area (25 stations), where the degree of topographic relief is larger than 100 m, which reflects the influence of topography [38,39,40,41].

Figure 8 shows the comparison of ET₀ changes and the contributions of the two categories. The annual trends in ET₀ for the plain category and hill category are 2.67 and 0.50 mm·year⁻¹, respectively. The former increase is 5.4 times the latter one. This might be caused by the location of plain stations in well-developed cites, and intense human activities must influence underlying surface characterizations [42,43]. For example, the increase in air temperature and water vapor pressure deficits for the coastal plain category is 1.5 and is 1.6 times that of the hill category (Figure 9). It is worth noting that the decreases in solar radiation and wind speed for the plain category are smaller than those for the hill category, which has a negative effect on the increase in ET₀. The contributions of solar radiation, air temperature, wind speed and water vapor pressure deficits to the increase in ET₀ are −12%, 17.92%, −19.3% and 124.2%, respectively, for the plain category, and the increase in water vapor pressure deficits plays a dominant role in ET₀ increase. For the hill category, the contributions of solar radiation, air temperature, wind speed and water vapor pressure deficits to the increase in ET₀ are −181.1%, 59%, −122.5% and 387.5%, respectively. Although the water vapor pressure deficits is still the dominant factor for ET₀ increase, influences from solar radiation and wind speed cannot be ignored either. Moreover, it can be found in Figure 8 that the increase in ET₀ in these two categories is due to the rapid increase in ET₀ in spring.

Generally, humidity in this coastal province is very high, and the water vapor pressure deficits is the dominant factor affecting changes in ET₀. The other two important factors affecting increases in ET₀ are wind speed and solar radiation. The coastal Zhejiang province has a strong sea-land breeze due to land-atmosphere interactions [44,45]. Therefore, the decrease in wind speed in the plain category is smaller than that of the hill category, although the increase in surface roughness from intense human activities is much more obvious for the plain category. In addition, the plain category also has a smaller decrease in solar radiation than the hill category. Lin et al. [46] indicated that solar radiation considerably decreases with wind stilling for weakened wind, which amplified the impact of aerosol emissions on solar dimming. The contribution of air temperature to the increase in ET₀ is the smallest of the four factors. However, numerous studies have found that air temperature is the dominant factor affecting the increase in ET₀ in past decades [13,20,22]. This might be due to the coastal area being characterized by the energy-limited condition, where the impact of air temperature is limited on ET₀ increase. In addition, this study considers the water vapor pressure deficits as an influential factor, rather than water vapor pressure, which leads to a different contribution result from previous studies.

5. Conclusions

This study investigated the spatial and temporal changes in ET₀ along the coast of Zhejiang province during the period of 1973 to 2013, and a numerical experiment method was used to quantify the contribution of climate variations to changes in ET₀. The results indicated that the annual ET₀ increased significantly (p < 0.05), with a slope of 1.58 mm·year⁻¹ during the past 40 years. At the seasonal time scale, ET₀ increased in spring, summer and autumn but decreased in winter. Air temperature, water vapor pressure deficits, wind speed and solar radiation changed significantly (p < 0.05) at rates of 0.04 °C·year⁻¹, 0.005 kPa·year⁻¹, −0.01 m/s·year⁻¹ and −3.94 MJ·m⁻²·day⁻¹·year⁻¹, respectively.

The results of the five numerical experiments manifested that changes in air temperature, wind speed, vapor pressure deficits and solar radiation led to annual changes in ET₀ by 24.7%, −35.4%, 165.8% and −38.6%, respectively, and the relative error of the numerical experiment method was −16.5%. An increase in the water vapor pressure deficits was the controlling factor for increasing ET₀ along the coastal province. The impacts of topography on ET₀ changes were further analyzed. The trend in ET₀ for the plain category was approximately 5.4 times that of the hill category. This might be caused by the influence of a large water body, as well as the augment of surface roughness by intense human activities in the well-developed plain area.

This study provides a comprehensive understanding of ET₀ changes in a well-developed coastal province and highlights the important role of water vapor pressure deficits to increases in ET₀. Further investigation is required to properly quantify the influence of human activities on evaporative demand, such as the urban heat island effect and the augment of surface roughness. Our results are helpful for hydrological modeling, land-atmospheric interaction simulations and irrigation of eco-agriculture planning.