Abstract

It is essential to understand the changing patterns in reference evapotranspiration (ET0) and its relation to climate variables. In this study, meteorological data obtained from the Sanjiang Plain (SJP) between 1959 and 2013 are used to calculate ET0 via the Penman–Monteith method. This study analyses the spatial and temporal changes of ET0 and determines which meteorological variables have an impact on this. The Mann–Kendall test, moving t-test, sensitivity analysis and simulated results have been used to conduct these analyses. The results demonstrate the following. (1) Spatially, there is an increasing trend in the annual ET0 values in agricultural areas. However, significant decreasing trends (P < 0.05) can be found in mountainous regions. (2) Temporally, two abrupt changes can be detected in the early 1980s and the late 1990s for the entire SJP, leading to large inter-annual differences. (3) Sensitivity analysis shows that relative humidity (RH) is the most sensitive climate variable and has a negative influence on ET0, followed by temperature, sunshine duration and wind speed, all of which exert positive impacts. (4) The simulated result shows that ET0 is most sensitive to RH. However, significant reductions in wind speed can exert large influences on the ET0 values.

INTRODUCTION

Hydrological processes are influenced by both climate change and human activities, and the negative effects on such processes are becoming increasingly more critical. Since the 1950s, temperatures over land have increased on a global scale, which has made a huge impact on the water cycle (IPCC 2014). As a consequence, water supplies are in decline (Gao et al. 2012), and various water-saving techniques are now being implemented to ameliorate this effect (Zou et al. 2013).

Evapotranspiration is a major component of the water cycle. It involves both evaporation from soil and transpiration by plants, and accounts for 90% of the precipitation occurring in semi-arid and arid regions (Shweta & Krishna 2015). However, due to the lack of measured evapotranspiration data, reference evapotranspiration (ET0) is always used to estimate actual evapotranspiration, and it is widely utilized for predicting crop water requirements (Cruz-Blanco et al. 2014).

Although ET0 values have changed globally over the past several decades, the changing patterns vary across different climate zones. For example, a number of studies from Mediterranean countries and Central Asia have shown that ET0 values have been increased in these areas (Kitsara et al. 2012; Vicente-Serrano et al. 2014; Dadaser-Celik et al. 2016), whereas studies from other regions reported a decreasing trend (Xu et al. 2006; Irmak et al. 2012). In particular, Liu & Zhang (2011) indicated that ET0 has declined in China in most areas except for the Songhuajiang River Basin and the Yellow River Basin. In addition, Huo et al. (2013) reported a significant decreasing trend in ET0 in northwest China. ET0 shows multiple patterns in other areas of the world. For example, although Iran is categorized as having an arid and semi-arid climate, there are differing patterns in the trends of monthly and annual ET0 across the country (Dinpashoh et al. 2011), whereas studies in Argentina have found that most parts of the country have non-significant trends (de la Casa & Ovando 2016). Therefore, as regional differences in ET0 values are evident, it is important to analyse these changes on the Sanjiang Plain (SJP) in China, to inform decisions on agricultural management.

Research indicates that ET0 is affected by numerous factors, such as climate (Petkovic et al. 2015), agricultural development (Maeda et al. 2011), characteristics of vegetation (Anda et al. 2015) and environment. As the dominant factors differ from region to region, it is necessary to determine the main factors controlling variations in ET0. For example, previous research indicated that wind speed (WS) is the most sensitive variable in the north and northwest of China and that ET0 is less sensitive to relative humidity (RH) (Huo et al. 2013; Zhao et al. 2014). In contrast, Sharifi & Dinpashoh (2014) reported that ET0 is less sensitive to WS in Iran, particularly in summer. In addition, it has been widely reported that RH as well as temperature is one of the most dominant factors controlling ET0 (Vicente-Serrano et al. 2014; Petkovic et al. 2015; Zheng & Wang 2015). However, as it is currently unclear whether the reasons for these varying conclusions are related to climate, topography or latitude, detailed studies are thus required on the SJP to identify the dominant factors influencing changes in ET0 in this area.

The SJP is a vast alluvial floodplain that is home to a substantial area of wetlands. This area has attracted considerable attention mainly in relation to environmental protection, but a few studies have focused on evapotranspiration on the SJP. For example, Chen & Lu (1994) analysed the relation between evapotranspiration and water evaporation in marshland using evaporimeters, and Sun & Song (2008) calculated evapotranspiration using three methods and compared the differences with measured evapotranspiration; results showed that the Priestley–Taylor model was the most applicable method for use in modelling daily evapotranspiration from a wetland in the SJP. In addition, Guo & Sun (2012) determined that surface resistance and net radiation are the main factors affecting evapotranspiration. However, although both these studies were conducted on the wetlands of the SJP, data records were not of an adequate length to describe characteristics of evapotranspiration under climate change. Therefore, the objectives of this current paper are: (1) to analyse spatial and temporal variations in the potential influencing factors and ET0; (2) to determine the changes in ET0 in response to these factors; and (3) to quantify the impact from each factor.

STUDY AREA AND DATA

Study area

The SJP is situated in the northeast of Heilongjiang province, China (between 45°01′N–48°28′N and 130°13′E–135°05′E, Figure 1). It covers an area of 10.57 ×104 km2 and accounts for 23% of the entire area of Heilongjiang Province. It is bordered in the north and east by the Heilongjiang River and Wusuli River, respectively, and its boundaries in the west and south are the Lesser Khingan Mountains and Khanka Lake, respectively. The SJP is home to an area of flat plain at an elevation of 50–60 m above sea level, and a gentle slope (1/10,000) can be detected from the southwest to northeast. Mean annual precipitation in the study area is between 450 mm and 650 mm; however, there are large inter-annual and intra-annual variations. For example, the annual maximum precipitation at Fujin station was 829.6 mm in 1959, and the minimum value was 338.5 mm in 1977. Precipitation falling between May and August accounts for two-thirds of the annual value, and the mean annual temperature is between 1 °C and 3 °C (although it is relatively higher in the southern regions and on the plains). The frost period lasts from late September to the following April; melt begins in late April and thus the frost-free period is relatively short. In addition, frost depth can even reach up to 2 m below the soil's surface.

Figure 1

Location of meteorological stations, farmlands and elevation of SJP.

Figure 1

Location of meteorological stations, farmlands and elevation of SJP.

Soil types on the SJP are very suitable for cultivation and cultivated areas have increased rapidly since the 1950s. The SJP is now one of the most important agricultural areas in China and rice is the major crop. Rice is grown between May and August; May is the period for ponding, June and July are periods of rice development and the late season occurs in August. In this respect, this study mainly analyses the influences of meteorological variables on ET0 during the period from May to August.

Agricultural areas are mostly located on the plain area of the SJP (Figures 1 and 2). According to the future plan for the next 15 years, there will be an increase of 5,000 km2 in the area of paddy fields, leading to an increase in the need for agricultural water of more than 20 × 108 m3. It is thus necessary to investigate the changing pattern of ET0 and to understand its relation with various factors.

Figure 2

Land-use types of the SJP (obtained from remote sensing data, 2014). Figure obtained from remote sensing data provided by the Chinese Academy of Sciences.

Figure 2

Land-use types of the SJP (obtained from remote sensing data, 2014). Figure obtained from remote sensing data provided by the Chinese Academy of Sciences.

Data

Daily meteorological records from 10 stations (Figure 1) were collected for a 55-year period from January 1959 to December 2013. These data comprise daily precipitation, maximum/minimum/average temperature, sunshine duration (SD), WS and RH, as provided by the China Meteorological Data Sharing Service System (http://cdc.cma.gov.cn/cdc_en/home.dd). The mean annual and monthly data were then calculated from these records.

METHODOLOGY

Penman–Monteith (P–M) method

The reference crop is assumed to be a hypothetical crop with a height of 0.12 m and a surface resistance of 70 m/s (Allen et al. 1998). The FAO P–M method was developed based on the original P–M equation and is defined as follows:  
formula
(1)
where ET0 is the reference evapotranspiration (mm/d), Rn is the net radiation at the crop surface (MJ/(m2·d)), G is the soil heat flux density (MJ/(m2·d)), (es–ea) represents the saturation vapour pressure deficit (kPa), Δ is the slope vapour pressure curve (kPa/°C), γ is the psychrometric constant (kPa/°C), Tmean is the mean daily temperature (°C) and u2 is WS at 2 m height (m/s). The calculation for the parameters used in the FAO P–M method is provided in Chapter 3 of FAO-56 (Allen et al. 1998).

Data analysis

Several methods are used to analyse the long-term trend of ET0 and meteorological variables in this study, including the linear trend analysis, the moving average method and the non-parametric Mann–Kendall (M–K) test. The M–K test is a non-parametric method which is used to statistically assess whether a monotonic upward or downward trend of the variable exists over time.

As the M–K test only shows whether the trends of variables are significant, the moving t-test (MTT) is also used to detect abrupt changes in time series. This method can assess the magnitude of change between two subsequences of a sample. As the length of a subsequence can influence the result, we calculate the results by setting two subsequences at 5 and 10 years. A comparison between these results will then assist in identifying any abrupt changes occurring within a sample series.

Contributing rate

The rank correlation coefficient and sensitivity analysis are used in this paper to analyse the impact of each variable on ET0. The rank correlation coefficient commonly includes the Spearman, Kendall and Pearson method, which measures the degree of similarity between two rankings and can be used to assess the significance of the relation between them. The sensitivity analysis is used to assess uncertainty in the output of a mathematical model. In the study area, four meteorological factors (temperature, RH, WS and SD) are studied. The sensitivity of ET0 to these factors is then analysed by changing a single variable while holding the others constant. These factors are changed within a range between −20% and 20% at an interval of 1%. It is thus possible to study percentage change in ET0 in response to these variables.

RESULTS AND DISCUSSION

Evolution of monthly and annual ET0

Figure 3 shows the spatial distribution of monthly and annual ET0 from 1959 to 2013. Results indicate that ET0 values recorded in the agricultural area were higher than those in the mountainous areas, and monthly differences between the values were between 10 and 20 mm. This pattern can also be detected on an annual scale wherein the values at central stations exceeded those in the mountain areas by more than 100 mm.

Figure 3

Spatial distribution of monthly and annual ET0 at ten stations from 1959 to 2013 (mm).

Figure 3

Spatial distribution of monthly and annual ET0 at ten stations from 1959 to 2013 (mm).

Figure 4 shows a box plot of average ET0 values obtained at ten stations on the SJP for every month from 1959 to 2013. Significant seasonal differences are noted in relation to changes in climate. Values of ET0 were relatively small (<50 mm) from November to March, and differences among stations were small. However, during the frost-free period, the averages of ET0 were higher than 120 mm and variations among stations were comparatively large in May, June and July. There were large differences between the maximum value (ET0 was 199 mm in May 1963 in Jixi) and the minimum value (ET0 was 65 mm in June 1983 in Hulin). It should also be mentioned that 60% of total annual ET0 was recorded during the rice growing period.

Figure 4

Box plot of monthly ET0 at ten stations. The plot shows the median, interquartile range (box) and whiskers and * indicates outliers.

Figure 4

Box plot of monthly ET0 at ten stations. The plot shows the median, interquartile range (box) and whiskers and * indicates outliers.

The inter-annual variation of monthly and annual ET0 from 1959 to 2013 can be seen in Figure 5. During the frost-free period, the linear regression shows an increasing trend in all five monthly time series, whereas the opposite trend is shown in the frost period. Significant variations (P < 0.05) of monthly ET0 can be detected in April, May, June, August, September and December. Therefore, there were strong inter-annual variations of monthly ET0 and the range of these variations was at a maximum in May, June and July (at more than 60 mm). The largest positive slope is detected in June (2.30 mm/10a), whereas the largest decrease in ET0 is in April (−1.42 mm/10a). However, there are no obvious changes in the annual time series (−0.44 mm/10a). Nevertheless, the decadal variation shows differences in the changes in annual average ET0 on the SJP during different periods, and further analysis indicates that there were several abrupt changes in the amounts of annual ET0.

Figure 5

Temporal changes in monthly and annual ET0 on the SJP from 1959 to 2013.

Figure 5

Temporal changes in monthly and annual ET0 on the SJP from 1959 to 2013.

Figure 6 provides the MTT results of annual ET0 on the SJP. Three abrupt changes took place in 1980, 1991 and 1995 when the two subsequences were both 5 years long. However, when the two subsequences were both 10 years long, two abrupt changes occurred in 1983 and 1996. Although it therefore appears that abrupt changes are sensitive to the choice of subsequence length, both the results reflect an increasing change in the early 1980s and a decreasing change in the late 1990s. Moreover, these changes are confirmed by the results provided in Table 1, wherein it can be seen that similar patterns occurred in the frost-free period.

Table 1

Temporal changes in monthly and annual ET0 during different periods (mm/10a)

Periods Jan. Feb. Mar. Apr. May Jun. Jul. Aug. Sep. Oct. Nov. Dec. Annual 
1959–1980 −0.04 −1.10 2.00 −0.84 10.65 4.42 12.17 11.57 8.03 3.14 0.38 −0.54 49.83 
1980–1995 0.04 2.59 4.05 −2.42 −6.15 −3.59 −4.42 −2.20 −3.70 1.47 4.51 −0.20 −10.02 
1995–2013 −1.31 −2.91 −5.36 −10.17 −5.44 −1.38 −4.60 6.50 0.21 −0.18 −4.78 − 2.16 − 36.22 
Periods Jan. Feb. Mar. Apr. May Jun. Jul. Aug. Sep. Oct. Nov. Dec. Annual 
1959–1980 −0.04 −1.10 2.00 −0.84 10.65 4.42 12.17 11.57 8.03 3.14 0.38 −0.54 49.83 
1980–1995 0.04 2.59 4.05 −2.42 −6.15 −3.59 −4.42 −2.20 −3.70 1.47 4.51 −0.20 −10.02 
1995–2013 −1.31 −2.91 −5.36 −10.17 −5.44 −1.38 −4.60 6.50 0.21 −0.18 −4.78 − 2.16 − 36.22 

Note: Bold numbers indicate significant changes (P < 0.05), as determined by the M–K tau test.

Figure 6

Moving t-test for annual ET0 on the SJP from 1959 to 2013.

Figure 6

Moving t-test for annual ET0 on the SJP from 1959 to 2013.

Figures 7 and 8 show the spatial and temporal changes of ET0 at each station; furthermore, spatial and temporal patterns can be observed. Most of these stations show an increasing trend during the frost-free period, and the maximum increases are mostly located in the agricultural area of the SJP. In contrast, the ET0 values are observed to decrease during the frost period, with the exception of the month of February. There is a large difference between the two figures: fluctuations in ET0 are relatively small in Figure 7, whereas the per cent changes seen in Figure 8 are comparatively large. This is because the ET0 values were very small during the frost period, and small variations can cause a large percentage change. A similar reason can explain the difference between the figures on an annual scale. Values of annual ET0 are much higher than those of monthly ET0; therefore, the percentage change would be small, even if there was a large fluctuation. However, these patterns indicate strong correlations between ET0 and climate variables. For instance, higher increases in ET0 may occur in areas with a large reduction in the amount of precipitation, as cloud cover would be reduced and thus a larger amount of solar radiation would reach the Earth's surface. It is evident that further analyses are required to identify the dominant driving force of ET0. As previously discussed in the section ‘Study area and data’, the rice growing period is from May to August on the SJP, and the ET0 values are relatively small during the frost period. Therefore, this study only analyses the sensitivity of ET0 to the different meteorological variables during the rice growing period.

Figure 7

Slopes in ET0 values on the SJP on monthly and annual scales.

Figure 7

Slopes in ET0 values on the SJP on monthly and annual scales.

Figure 8

Percentage changes in ET0 on the SJP on monthly and annual scales.

Figure 8

Percentage changes in ET0 on the SJP on monthly and annual scales.

Changing patterns of meteorological variables

To analyse the influence of each meteorological variable on ET0 in the study area, we analysed monthly changes of the six meteorological variables used to calculate ET0 (Figures 9 and 10). Increases in temperature were relatively high during May and June, and changes in minimum temperature were the largest (in contrast to changes in maximum and average temperature). Changes in WS were the largest among these variables, and most of the ten stations (except Jixi station) show a significant decreasing trend. The increase in RH was greater than 1%/10a during May. However, decreasing trends were detected during June to August. These changes in RH could therefore be important, particularly because ET0 values decreased during May but increased during the other months.

Figure 9

Box plot of meteorological variables at ten stations: plot shows median (centre circle), interquartile range (box) and whiskers; * indicates outliers. Results of each variable are divided into six blocks separated by the vertical dashed lines. The four bars in each block represent changes at ten stations for a single variable during the rice growing period.

Figure 9

Box plot of meteorological variables at ten stations: plot shows median (centre circle), interquartile range (box) and whiskers; * indicates outliers. Results of each variable are divided into six blocks separated by the vertical dashed lines. The four bars in each block represent changes at ten stations for a single variable during the rice growing period.

Figure 10

Percentage changes in meteorological variables on SJP.

Figure 10

Percentage changes in meteorological variables on SJP.

Throughout the entire study area, the results of trend analyses and the moving t-test reveal that inter-annual changes of meteorological variables follow a similar pattern to that of the values of annual ET0, which changes abruptly several times (Table 1). For example, temperature increases significantly (P < 0.01) at a rate of 0.28 °C/10a. As an abrupt change point can be detected in 1993, the temperature series was classified into two stages. Compared with the variation (0.25 °C/10a) of temperature during 1959 to 1993, there was a sharp increase in temperature after 1993 (0.44 °C/10a) in comparison with the preceding years. However, other variables declined over the 55-year period. Precipitation and RH show a slight downtrend (−6.32 mm/10a and −0.27%/10a), but an abrupt change in both series can be detected. Both trends are significant (P < 0.01) at a rate of −74.9 mm/10a and −3.1%/10a before 1980, after which the trends increase slightly until 2013. The change in the WS trend is statistically significant (P < 0.01) at a rate of −0.15 (m·s−1)/10a, and this can also be classified into two periods: it decreased at a rate of −0.09 (m·s−1)/10a before 1993, but the decrease accelerated to −0.19 (m·s−1)/10a from 1993 to 2013. SD fluctuated, with no significant trend, at a rate of −0.04 h/10a. However, the trend was statistically significant (P < 0.01) before 1980 at a rate of 0.47 h/10a.

Influence of meteorological factors on ET0

Rank correlation coefficients

To determine key factors impacting ET0 on the SJP, rank correlation coefficients between ET0 and precipitation were analysed. Results of the Kolmogorov–Smirnov test showed that the data series of ET0 and related factors do not have a standard normal distribution. Thus, the Spearman's rank correlation coefficient was introduced to measure the statistical dependence between meteorological variables and changes of ET0 on a monthly and annual scale (Table 2). On an annual scale, changes in ET0 have positive correlations with temperature (minimum and average), WS and SD, whereas correlations with RH and precipitation have an opposite pattern. These results help to explain the changes in ET0 over different months and periods. For instance, correlations between ET0 changes and several meteorological variables are relatively strong during June to September, leading to a rising trend of ET0 in these months, in accordance with climate change. However, there are significant changes in most of the meteorological variables from 1959 to 1980, which could explain temporal patterns in ET0 mentioned in Table 1.

Table 2

Spearman's rank correlation coefficients between changes in ET0 and meteorological variables

  Jan. Feb. Mar. Apr. May Jun. Jul. Aug. Sep. Oct. Nov. Dec. Annual 
Slope 
 Tmax 0.24 0.64 −0.09 −0.36 0.09 0.24 0.48 0.40 0.45 0.24 0.37 0.17 0.25 
 Tmin −0.16 0.42 −0.16 0.07 0.42 0.72 0.71 0.56 0.39 −0.03 0.27 0.19 0.64 
 Tmean −0.08 0.52 −0.17 −0.25 0.31 0.53 0.67 0.57 0.51 0.10 0.29 0.13 0.61 
 RH −0.12 0.69 −0.18 0.59 0.35 −0.25 −0.50 −0.10 −0.48 −0.31 −0.40 −0.42 −0.17 
 WS −0.11 0.13 −0.34 0.06 0.58 0.63 0.47 0.24 0.32 0.15 0.20 0.05 0.42 
 SD 0.04 0.07 0.29 0.14 0.03 0.16 0.38 0.26 0.79 0.74 0.61 0.21 0.57 
 P −0.18 −0.01 0.04 0.47 −0.53 − 0.86 −0.07 −0.47 −0.59 −0.18 0.12 0.10 − 0.66 
Percentage change 
 Tmax 0.24 0.58 −0.04 −0.25 0.18 0.20 0.52 0.37 0.43 0.26 0.45 0.36 0.27 
 Tmin −0.15 0.19 −0.14 0.24 0.51 0.70 0.73 0.55 0.39 0.06 0.43 0.45 0.69 
 Tmean −0.07 0.32 −0.14 −0.08 0.42 0.50 0.71 0.55 0.51 0.18 0.43 0.39 0.66 
 RH −0.16 −0.53 −0.24 0.52 0.27 −0.19 −0.52 −0.06 −0.45 −0.37 −0.53 −0.62 −0.19 
 WS −0.06 −0.01 −0.35 0.18 0.62 0.60 0.47 0.23 0.31 0.23 0.33 0.27 0.45 
 SD 0.02 −0.07 0.27 0.19 0.10 0.11 0.40 0.22 0.78 0.80 0.72 0.43 0.57 
 P −0.28 −0.09 0.01 0.44 −0.57 − 0.83 −0.05 −0.47 −0.59 −0.19 0.07 −0.18 0.57 
  Jan. Feb. Mar. Apr. May Jun. Jul. Aug. Sep. Oct. Nov. Dec. Annual 
Slope 
 Tmax 0.24 0.64 −0.09 −0.36 0.09 0.24 0.48 0.40 0.45 0.24 0.37 0.17 0.25 
 Tmin −0.16 0.42 −0.16 0.07 0.42 0.72 0.71 0.56 0.39 −0.03 0.27 0.19 0.64 
 Tmean −0.08 0.52 −0.17 −0.25 0.31 0.53 0.67 0.57 0.51 0.10 0.29 0.13 0.61 
 RH −0.12 0.69 −0.18 0.59 0.35 −0.25 −0.50 −0.10 −0.48 −0.31 −0.40 −0.42 −0.17 
 WS −0.11 0.13 −0.34 0.06 0.58 0.63 0.47 0.24 0.32 0.15 0.20 0.05 0.42 
 SD 0.04 0.07 0.29 0.14 0.03 0.16 0.38 0.26 0.79 0.74 0.61 0.21 0.57 
 P −0.18 −0.01 0.04 0.47 −0.53 − 0.86 −0.07 −0.47 −0.59 −0.18 0.12 0.10 − 0.66 
Percentage change 
 Tmax 0.24 0.58 −0.04 −0.25 0.18 0.20 0.52 0.37 0.43 0.26 0.45 0.36 0.27 
 Tmin −0.15 0.19 −0.14 0.24 0.51 0.70 0.73 0.55 0.39 0.06 0.43 0.45 0.69 
 Tmean −0.07 0.32 −0.14 −0.08 0.42 0.50 0.71 0.55 0.51 0.18 0.43 0.39 0.66 
 RH −0.16 −0.53 −0.24 0.52 0.27 −0.19 −0.52 −0.06 −0.45 −0.37 −0.53 −0.62 −0.19 
 WS −0.06 −0.01 −0.35 0.18 0.62 0.60 0.47 0.23 0.31 0.23 0.33 0.27 0.45 
 SD 0.02 −0.07 0.27 0.19 0.10 0.11 0.40 0.22 0.78 0.80 0.72 0.43 0.57 
 P −0.28 −0.09 0.01 0.44 −0.57 − 0.83 −0.05 −0.47 −0.59 −0.19 0.07 −0.18 0.57 

Note: Bold numbers indicate significant changes (P < 0.05).

Sensitivity analysis

Sensitivity analysis can be used to quantitatively analyse changes in ET0. The analysis is achieved by changing one variable while holding others constant. Figure 11 shows the curves determined by sensitivity analysis for ET0 over the entire region; the curves are near linear. It is also evident that different meteorological variables exert different influences on ET0. On an annual scale, the most sensitive variable on the SJP is RH, which has a noticeable negative correlation with ET0. In response to the changes of RH at rates of 20% and −20%, the percentage changes in ET0 are −15.8% and 17.2%, respectively. The other variables have positive correlations with ET0. However, temperature also has a relatively large impact on ET0, 20% increase in temperature can cause a percentage change in ET0 of 12.4%. In contrast, SD and WS have small effects on ET0, 20% increases in SD and WS can cause percentage changes in ET0 of 4.0% and 3.4%, respectively.

Figure 11

Percentage changes in ET0 with respect to changes in meteorological variables.

Figure 11

Percentage changes in ET0 with respect to changes in meteorological variables.

It can be seen that the influence of each variable during the rice growing period is similar to that shown in the annual results. The influence of RH is again higher than the other variables, particularly in July and August. However, the sensitivity of ET0 to each variable varies monthly. For example, the sensitivity of ET0 to WS decreases from May to August, and the values in May are twice as large as during the other months. However, the opposite pattern is seen for other variables. This could, therefore, be one of the reasons why ET0 was significantly reduced in May.

DISCUSSION

According to the FAO-56 P–M method, this study analyses the trends of ET0 and different meteorological variables that are related to, or used for, calculating ET0 on annual and monthly scales. Results show that the annual averages of ET0 were slightly decreased over the whole SJP (−0.44 mm/10a), but that regional differences were significant. Previous studies have reported decreases in ET0 in the past 50 years throughout most regions of China, and decreasing trends have been found to be significant in northwest China (Liu & Zhang 2011; Huo et al. 2013; Zhao et al. 2014). These results are consistent with changing patterns in the mountainous areas of the SJP, and differences in agricultural areas can be explained by changes in meteorological variables.

This paper also conducts a detailed study of the sensitivity of ET0 to different meteorological variables. In comparison with previous studies, it is possible to ascertain whether the results are consistent and reasonable. In our study, RH and temperature are found to be the dominant driving forces in changes of ET0. Similar patterns are not exclusive to the SJP, for example, Zheng & Wang (2015) performed a wide study of the sensitivity of ET­0 to climate variables from 1960 to 2009, using data from 668 stations in China; this showed that RH and temperature were the most sensitive parameters in northeast and northern China. In addition, Zhao et al. (2015) reported that RH was the most sensitive meteorological factor in the Heihe River Basin of China, followed by WS and temperature. However, ET0 was found to be less sensitive to SD on the SJP, and this could be explained by the high latitude. In general, solar radiation is closely related to SD, and Zheng & Wang (2015) reported that stations at low latitudes are more sensitive to solar radiation than others. It is also relevant that WS is an important factor likely to affect ET0 to a large degree in northwest and northern China, and some studies have reported that ET0 is extremely sensitive to WS (Li et al. 2013; Zhao et al. 2014; Xu et al. 2015). These results can be contributed to the higher WS in these areas (Huo et al. 2013), and methods adopted in conducting sensitivity analyses could also influence results. Although WS is the least sensitive factor on the SJP, it has comparatively large fluctuations compared to the other variables. Therefore, the contribution of WS to annual or monthly ET­0 should not only be determined by the sensitivity analysis, but the trends of meteorological variables should also be considered. With this in mind, in order to further analyse the influences of meteorological variables on ET0, we generated six simulated series of ET0 for different meteorological variables. These series were calculated from the P–M method, which assumes that one variable could reach its highest value in history while the others remained at actual values. We compared the simulated series with actual ET0 to find the influence of each climatic variable on the evolution of ET0 (Figure 12). The results show that changes in RH could explain most of the observed changes of ET0 between the years 1959 to 2013. Although both mean and minimum temperature were increased significantly, their contributions to ET0 were found to be relatively small. These results can possibly explain the different trends of ET0, because changing patterns in meteorological variables have not previously been considered in sensitivity analyses.

Figure 12

Comparison between simulated series and actual ET0. The variable of interest was constant (we used the historical daily maximum value), while other variables remained at actual values. Simulated results are then obtained using the P–M method.

Figure 12

Comparison between simulated series and actual ET0. The variable of interest was constant (we used the historical daily maximum value), while other variables remained at actual values. Simulated results are then obtained using the P–M method.

This study only considers meteorological factors and ignores human activities, such as population and the underlying surface, and some of these could be the root causes of climate change. Previous studies have indicated that the SJP has suffered major land cover changes, especially in cultivated land, which increased in area from 8.2% to 57.97% during 1954 to 2010 (Wu et al. 2015). The urban area has also gradually increased over the past 50 years. Such changes could affect changing patterns in certain meteorological variables. For instance, WS decreased significantly during 1959 to 2013, which could be ascribed to the effect of urban development (Jeong & Kim 2014). Similar phenomena have been reported in most parts of China (Jiang et al. 2010). In addition, previous studies have also indicated that near-surface WS increases as land-surface elevation increases (McVicar et al. 2010), and this could help to explain the increasing trend of WS at Jixi station, which has a comparatively higher elevation (280 m) than others (<100 m). Other meteorological variables influence each other, and complex mechanisms exist among them, suggesting further future analysis is required.

CONCLUSION

By utilizing the FAO-56 Penman–Monteith method, we analysed the changing patterns of annual and monthly ET0 on the SJP during 1959 to 2013. Evolution of related meteorological variables and the sensitivity of ET­0 to different variables was also analysed. The main results are as follows:

  1. ET0 changed slightly during the 50-year period, at a rate of −0.44 mm/10a; however, there were large spatial and intra-annual differences. The annual averages of ET0 were increased in agricultural areas, while significant decreasing trends were found in mountainous regions.

  2. In most cases, there were decreasing trends in WS, SD and RH, while temperature increased continually. The changing trends of WS and temperature were statistically significant (P < 0.05).

  3. According to the sensitivity analysis, ET0 is most sensitive to RH, followed by the temperature, SD and WS.

  4. The simulated result showed that RH is the predominant factor influencing changes in ET0, but WS can also affect changes in ET0 to a large degree (Figure 12).

Although it is thus acknowledged that changes in one meteorological variable can affect others, this effect is not considered in this paper. Further studies should thus be conducted to develop a high-precision model with the aim of conducting a superior analysis of this effect.

ACKNOWLEDGEMENTS

The authors wish to acknowledge the two anonymous reviewers and Editor Dr Quinn for their detailed and helpful comments to the original manuscript. This study was supported by the National Natural Science Foundation of China (No. 51679252 and No. 51409246), the National Basic Research Program of China (973 Program) (No. 2013 CB036406 and No. 2015CB452701), the Open Research Fund of State Key Laboratory of Simulation and Regulation of Water Cycle in River Basin (China Institute of Water Resources and Hydropower Research) (No. IWHR-SKL-201603) and by the Jiangsu Province Natural Science Foundation (BK20130178).

REFERENCES

REFERENCES
Allen
,
R. G.
,
Pereira
,
L. S.
,
Raes
,
D.
&
Smith
,
M.
1998
Crop Evapotranspiration Guidelines for Computing Crop Water Requirements
.
FAO Irrigation and Drainage Paper 56
,
Rome
,
Italy
.
Anda
,
A.
,
Soos
,
G.
,
Teixeira da Silva
,
J. A.
&
Kozma-Bognar
,
V.
2015
Regional evapotranspiration from a wetland in Central Europe, in a 16-year period without human intervention
.
Agricultural and Forest Meteorology
205
,
60
72
.
http://dx.doi.org/10.1016/j.agrformet.2015.02.010
.
Chen
,
G. Q.
&
Lu
,
X. G.
1994
A study on marsh evapotranspiration in the Sanjiang Plain
.
Chinese Geographical Science
4
(
2
),
159
167
.
Cruz-Blanco
,
M.
,
Lorite
,
I. J.
&
Santos
,
C.
2014
An innovative remote sensing based reference evapotranspiration method to support irrigation water management under semi-arid conditions
.
Agricultural Water Management
131
,
135
145
.
http://dx.doi.org/10.1016/j.agwat.2013.09.017
.
Dadaser-Celik
,
F.
,
Cengiz
,
E.
&
Guzel
,
O.
2016
Trends in reference evapotranspiration in Turkey: 1975–2006
.
International Journal of Climatology
36
,
1733
1743
.
http://dx.doi.org/10.1002/joc.4455
.
de la Casa
,
A. C.
&
Ovando
,
G. G.
2016
Variation of reference evapotranspiration in the central region of Argentina between 1941 and 2010
.
Journal of Hydrology: Regional Studies
5
,
66
79
.
http://dx.doi.org/10.1016/j.ejrh.2015.11.009
.
Dinpashoh
,
Y.
,
Jhajharia
,
D.
,
Fakheri-Fard
,
A. P.
,
Singh
,
V.
&
Kahya
,
E.
2011
Trends in reference crop evapotranspiration over Iran
.
Journal of Hydrology
399
,
422
433
.
http://dx.doi.org/10.1016/j.jhydrol.2011.01.021
.
Gao
,
G.
,
Xu
,
C. Y.
&
Chen
,
D. L.
2012
Spatial and temporal characteristics of actual evapotranspiration over Haihe River basin in China
.
Stochastic Environmental Research and Risk Assessment
26
,
655
669
.
http://dx.doi.org/10.1007/s00477-011-0525-1
.
Guo
,
Y. D.
&
Sun
,
L.
2012
Surface energy fluxes and control of evapotranspiration from a Carex lasiocarpa mire in the Sanjiang Plain, Northeast China
.
International Journal of Biometeorology
56
,
221
232
.
http://dx.doi.org/10.1007/s00484-011-0415-8
.
Huo
,
Z. L.
,
Dai
,
X. Q.
,
Feng
,
S. Y.
,
Kang
,
S. Z.
&
Huang
,
G. H.
2013
Effect of climate change on reference evapotranspiration and aridity index in arid region of China
.
Journal of Hydrology
492
,
24
34
.
http://dx.doi.org/10.1016/j.jhydrol.2013.04.011
.
IPCC
2014
Climate Change 2014: Synthesis Report
. In:
Contribution of Working Groups I, II and III to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change
(
Core Writing Team
,
Pachauri
,
R. K.
&
Meyer
,
L. A.
, eds).
IPCC
,
Geneva
,
Switzerland
, pp.
151
.
Irmak
,
S.
,
Kabenge
,
I.
,
Skaggs
,
K. E.
&
Mutiibwa
,
D.
2012
Trend and magnitude of changes in climate variables and reference evapotranspiration over 116-yr period in the Platte River Basin, central Nebraska–USA
.
Journal of Hydrology
420–421
,
228
244
.
http://dx.doi.org/10.1016/j.jhydrol.2011.12.006
.
Jeong
,
J. H.
&
Kim
,
Y. K.
2014
Modeling of urbanization effect on the simulation of wind fields in urban building environments and its application to air quality studies
. In:
Systems, Signals and Image Processing, International Conference on IEEE
,
Dubrovnik, Croatia
, pp.
35
38
.
Jiang
,
Y.
,
Luo
,
Y.
,
Zhao
,
Z.
&
Tao
,
S.
2010
Changes in wind speed over China during 1956–2004
.
Theoretical & Applied Climatology
99
(
3–4
),
421
430
.
http://dx.doi.org/10.1007/s00704-009-0152-7
.
Kitsara
,
G.
,
Papaioannou
,
G.
&
Kerkides
,
P.
2013
Changes of Pan Evaporation Measurements and Reference Evapotranspiration in Greece. Advances in Meteorology, Climatology and Atmospheric Physics. Springer Atmospheric Sciences, Springer, Berlin, Heidelberg, pp. 527–533. Available from: http://dx.doi.org/10.1007/978-3-642-29172-2_75
.
Li
,
Z. L.
,
Li
,
Z. J.
,
Xu
,
Z. X.
&
Zhou
,
X.
2013
Temporal variations of reference evapotranspiration in Heihe River basin of China
.
Hydrology Research
44
(
5
),
904
916
.
http://dx.doi.org/10.2166/nh.2012.125
.
Liu
,
C. M.
&
Zhang
,
D.
2011
Temporal and spatial change analysis of the sensitivity of potential evapotranspiration to meteorological influencing factors in China
.
Acta Geographica Sinica
66
(
5
),
579
588
(
in Chinese
).
http://dx.doi.org/10.11821/xb201105001
.
Maeda
,
E. E.
,
Pellikka
,
P. K. E.
,
Clark
,
B. J. F.
&
Siljander
,
M.
2011
Prospective changes in irrigation water requirements caused by agricultural expansion and climate changes in the eastern arc mountains of Kenya
.
Journal of Environmental Management
92
,
982
993
.
http://dx.doi.org/10.1016/j.jenvman.2010.11.005
.
McVicar
,
T. R.
,
Van Niel
,
T. G.
,
Roderick
,
M. L.
,
Li
,
L. T.
,
Mo
,
X. G.
,
Zimmermann
,
N. E.
&
Schmatz
,
D. R.
2010
Observational evidence from two mountainous regions that nearsurface wind speeds are declining more rapidly at higher elevations than lower elevations: 1960–2006
.
Geophysical Research Letters
37
(
6
),
460
472
.
http://dx.doi.org/10.1029/2009GL042255
.
Petkovic
,
D.
,
Gocic
,
M.
,
Trajkovic
,
S.
,
Shamshirband
,
S.
,
Motamedi
,
S.
,
Hashim
,
R.
&
Bonakdari
,
H.
2015
Determination of the most influential weather parameters on reference ET by adaptive neuro-fuzzy methodology
.
Computers and Electronics in Agriculture
114
,
277
284
.
http://dx.doi.org/10.1016/j.compag.2015.04.012
.
Sharifi
,
A.
&
Dinpashoh
,
Y.
2014
Sensitivity analysis of the Penman-Monteith reference crop evapotranspiration to climatic variables in Iran
.
Water Resource Management
28
,
5465
5476
.
http://dx.doi.org/10.1007/s11269-014-0813-x
.
Shweta
,
A. P.
2015
Selection of the best method of ETo estimation other than Penman–Monteith and their application for the humid subtropical region
.
Agricultural Research
4
(
2
),
215
219
.
http://dx.doi.org/10.1007/s40003-015-0153-y
.
Sun
,
L.
&
Song
,
C. C.
2008
Evapotranspiration from a freshwater marsh in the Sanjiang Plain, Northeast China
.
Journal of Hydrology
352
,
202
210
.
http://dx.doi.org/10.1016/j.jhydrol.2008.01.010
.
Vicente-Serrano
,
S. M.
,
Azorin-Molina
,
C.
,
Sanchez-Lorenzo
,
A.
,
Revuelto
,
J.
,
Moran-Tejeda
,
E.
,
Lopez-Moreno
,
J. I.
&
Espejo
,
F.
2014
Sensitivity of reference evapotranspiration to changes in meteorological parameters in Spain (1961–2011)
.
Water Resources Research
50
,
8458
8480
.
http://dx.doi.org/10.1002/2014WR015427
.
Wu
,
D.
,
Hou
,
W.
,
Zhang
,
S. W.
,
Bu
,
K.
,
Xiang
,
B.
,
Wang
,
Y.
&
Li
,
Y.
2015
Processes and prediction of land use/land cover changes (LUCC) driven by farm construction: the case of Naoli River Basin in Sanjiang Plain
.
Environmental Earth Sciences
73
,
4841
4851
.
http://dx.doi.org/10.1007/s12665-014-3765-9
.
Xu
,
C. Y.
,
Gong
,
L. B.
,
Jiang
,
T.
,
Chen
,
D. L.
&
Singh
,
V. P.
2006
Analysis of spatial distribution and temporal trend of reference evapotranspiration and pan evaporation in changjiang (Yangtze River) catchment
.
Journal of Hydrology
327
,
81
93
.
http://dx.doi.org/10.1016/j.jhydrol.2005.11.029
.
Xu
,
L. H.
,
Shi
,
Z. J.
,
Wang
,
Y. H.
,
Zhang
,
S. L.
,
Chu
,
X. Z.
,
Yu
,
P. T.
,
Xiong
,
W.
,
Zuo
,
H. J.
&
Wang
,
Y. N.
2015
Spatiotemporal variation and driving forces of reference evapotranspiration in Jing River Basin, Northwest China
.
Hydrological Processes
29
,
4846
4862
.
http://dx.doi.org/10.1002/hyp.10541
.
Zhao
,
Y. F.
,
Zou
,
X. Q.
,
Zhang
,
J. X.
,
Cao
,
L. G.
,
Xu
,
X. W. H.
,
Zhang
,
K. X.
&
Chen
,
Y. Y.
2014
Spatio-temporal variation of reference evapotranspiration and aridity index in the Loess Plateau Region of China, during 1961-2012
.
Quaternary International
349
,
196
206
.
http://dx.doi.org/10.1016/j.quaint.2014.06.050
.
Zhao
,
J.
,
Xu
,
Z. X.
,
Zuo
,
D. P.
&
Wang
,
X. M.
2015
Temporal variations of reference evapotranspiration and its sensitivity to meteorological factors in Heihe River Basin, China
.
Water Science and Engineering
8
(
1
),
1
8
.
http://dx.doi.org/10.1016/j.wse.2015.01.004
.
Zheng
,
C. L.
&
Wang
,
Q.
2015
Spatiotemporal pattern of the global sensitivity of the reference evapotranspiration to climatic variables in recent five decades over China
.
Stochastic Environment Research and Risk Assessment
29
,
1937
1947
.
http://dx.doi.org/10.1007/s00477-015-1120-7
.
Zou
,
X. X.
,
Li
,
Y. E.
,
Cremades
,
R.
,
Gao
,
Q. Z.
,
Wan
,
Y. F.
&
Qin
,
X. B.
2013
Cost-effectiveness analysis of water-saving irrigation technologies based on climate change response: a case study of China
.
Agricultural Water Management
129
,
9
20
.
http://dx.doi.org/10.1016/j.agwat.2013.07.004
.