Trend Analysis and Identification of the Meteorological Factors Influencing Reference Evapotranspiration

Abstract

Investigating the trends of reference evapotranspiration (ETo) is fundamental importance for water resource management in agriculture, climate variability analysis, and other hydroclimate-related projects. Moreover, it would be useful for understanding the sensitivity of such trends to basic meteorological variables, as the modifications of these variables due to climate change are more easily predictable. This study aims to analyze ETo trends and sensitivity in relation to different explanatory meteorological factors. The study used a 17 year-long dataset of meteorological variables from a station located in Piazza Armerina, Sicily, a region characterized by a Mediterranean climate. First, the FAO-Penman-Monteith method was applied for estimation of ETo. Next, the Mann-Kendall test with serial autocorrelation removal by Trend-free pre-whitening (TFPW) was applied to analyze ETo trends and the basic meteorological variables on which they depend. Sen’s slope was also used to examine the magnitude of the trend of monthly ETo and its related meteorological variables. According to the obtained results, ETo only showed a downward trend of 0.790 mm per year in November, while no trend is shown in other months or on seasonal and annual time scales. Solar radiation (November and Autumn) and rainfall (Autumn) showed a downward trend. The other meteorological variables (minimum temperature, maximum temperature, mean temperature, wind speed, and relative humidity) showed an upward trend both at monthly and seasonally scale in the study area. The highest and lowest sensitivity coefficients of ETo in the study area are obtained for specific humidity and wind speed, respectively. Specific humidity and wind speed give the highest (44.59%) and lowest (0.9%) contribution to ETo trends in the study area. These results contribute to understanding the potential and possible future footprint of climate change on evapotranspiration in the study area.

1. Introduction

Reference evapotranspiration (ETo) is a pivotal part of the hydrological cycle and of the most crucial physical processes in natural ecosystems and environmental systems on our planet [1]. The reference surface of ETo considered the hypothetical grass reference crop to have an assumed crop height of 0.12 m, a fixed surface resistance of 70 s m^−1, and an albedo of 0.23. ETo enables one to calculate the energy and water exchanges in vegetation [2,3], soil surface [4], land surface [5,6], and atmosphere [7,8,9]. Estimation and measurements of ETo contribute greatly to our understanding of earth’s energy budget, agricultural water management, water resource management, and climate change studies [6,10,11,12,13,14,15,16]. There are different methods and approaches for measuring and estimating ETo. These methods can be divided into two main groups: direct methods (field water balance approach and soil moisture depletion approach) and indirect methods (empirical/statistical methods, micrometeorological methods, and remote sensing methods [1,8,17,18,19,20,21,22]).

Moreover, it is of fundamental importance to analyze the trends in reference evapotranspiration in order to understand the potential impacts of climate change on ETo. Different methods for analyzing trends in hydroclimate variables over space and time are available in the literature. Both parametric and non-parametric methods were applied for hydroclimate time series analysis in different part of the world [17,23,24,25]. In parametric methods, linear regression is used for trend analysis for different meteorological variables including ETo [23,25,26]. These parametric methods are helpful for explaining the relationship between two or more variables using a linear relationship [26]. However, parametric methods require data to be independent and normally distributed, while non-parametric trend tests only require data to be independent and can tolerate outliers in the data [27]. One of the most commonly applied non-parametric tests for assessing trend significance is the Mann Kendhall test (MK-test) (Mann, 1945; Kendhall, 1975; [23,27,28,29,30,31]).

The MK-test allows one to determine whether or not the trend is monotonic, in terms of monthly, seasonal, and annual time series of evapotranspiration, as well as other hydro-climatological variables [28,32,33,34,35]. Additionally, to investigate trend magnitude, Sen’s slope is widely used as a non-parametric method, including in ETo analysis [23,26,36,37,38]. Meanwhile, sensitivity analysis is also essential for identifying the most influential factors in cases where a monotonic trend is present [36].

In addition to estimation of ETo, there are studies concerning the trend and magnitude of ETo in different parts of the world. By applying the MK test and Sen’s Slope, ETo trends showed different configurations on a monthly, seasonal, and annual scale. The annual trend of ETo increases as the temperature increases in different studies [23,25,27,31,36,37]. There are also studies showing a downward trend of ETo as related to temperature. Such situations are called “Evapotranspiration Paradox” because this behavior contrasts with the upward trend of global temperature [25,36,38]. In seasonal trend analysis, ETo showed downward trends in summer in many parts of the world [23,27,36]. In monthly significant trend analysis, there are different pattern configurations in different studies.

With reference to the Mediterranean area, there are studies showing that climate variables and extreme events are changing [39]. There are also some studies conducted to assess and compare different estimation methods of ETo in Sicily [40,41,42,43,44,45]. There are also studies about the amount of ETo and its configuration for different crops in Sicily [46,47].

In general, ETo has shown both increasing [23,25,27,31,36,37] and downward trends across the world, as documented in several papers [25,36,38]. On the other hand, there are also different areas showing the absence of trends for ETo. ETo trends are subject to both spatial and temporal variations (monthly, seasonally, and annually).

Further studies analyzing and comparing trends of monthly, seasonal, and annual ETo are needed, especially with reference to the Mediterranean climate. Many previous studies in other regions do not investigate the main factors affecting ETo trends. Therefore, in this paper, apart from estimating ETo trends at multiple temporal scales, we provide some additional insights to provide a better understanding of the evapotranspiration dynamics by analyzing the sensitivity and contribution rate of each variable to evapotranspiration trends. To this aim, we refer to ETo estimations made by the Penman-Monteith formula. Analysis is carried out using the data collected in an experimental site located in Piazza Armerina, Sicily, Italy. In addition to these points, it may be also mentioned that, in many previous studies, the sensitivity of ETo trends to air humidity is carried out respective to relative humidity. However, this relative humidity represents a measure of the actual amount of water vapor in the air compared to the total amount of vapor that can exist in the air at its current temperature [48]. It implies that air will have a higher relative humidity if the air is cooler, and a lower relative humidity if the air is warmer. Hence, this study will examine the sensitivity and contribution rate of specific humidity for ETo rather than that for common relative humidity, because specific humidity is always considered a measure of the actual amount of water vapor (moisture) in the air, regardless of air temperature [49].

2. Data and Methods

2.1. The FAO-Penman-Monteith Method

The FAO-PM has been established as a standard for calculating reference evapotranspiration [50]. This method requires air temperature, relative humidity, solar radiation, and wind speed data input and is produced high quality output results of ETo compared to other empirical ETo estimation methods [51,52,53,54,55]. This method was also approved by the FAO and the American Society of Civil Engineers (ACSE) as the best and most comprehensive method, to be used when the necessary data inputs are available [52,56,57,58,59,60,61,62].

A simplified equation was recommended by the FAO [63] with the FAO-56 Penman-Monteith Equation, by assuming some constant parameters for a clipped grass reference crop. In particular, the reference crop was assumed to be a hypothetical crop with crop height of 0.12 m, a fixed surface resistance of 70 s m^−1, and an albedo value (i.e., portion of light reflected by the leaf surface) of 0.23.

This simplified equation is obtained by integrating the original Penman-Monteith equation and the equations of the aerodynamic and canopy resistance:

$$ETo=\frac{0.408\Delta \left(R_n-G\right)+\gamma \frac{900}{T+273}{U}_2\left(e_s-e_a\right)}{\Delta +\gamma \left(1+0.34 U_2\right)}$$

(1)

where: ETo is reference evapotranspiration (mm day⁻¹),(R_n) is the net radiation at the crop surface [MJ m⁻² day⁻¹], G is the soil heat flux density [MJ m⁻² day⁻¹], T is the air temperature at 2 m height [°C], $U_2$ is the wind speed at 2 m height [m s⁻¹], e_s is saturation vapor pressure [kPa], e_a is actual vapor pressure [kPa], e_s−e_a is the saturation vapor pressure deficit [kPa], $\Delta$ is the vapor pressure curve slope [kPa °C⁻¹], and $\gamma$ is the psychrometric constant [kPa °C⁻¹].

2.2. Mann-Kendall Test

The Mann-Kendall test is the most effective method for supporting statistically significant trend tests for different hydro-climatological time series analysis and is widely applied in the literature [6,11,13,16,17,64]. The main advantage of the MK test is that it does not require the data to follow any statistical distribution and not sensitive to extreme values [17,36,65]. The test is based on two hypotheses: the null hypothesis (Ho) which supposes that the test is stationary and no trend exists, and the alternative hypothesis (H1), which rejects Ho and indicates the existence of a trend. Mann-Kendall’s statistical S is given by the following formula:

$$\mathrm{S}={\displaystyle \sum }_{\mathrm{k}=1}^{\mathrm{n}-1}{\displaystyle \sum }_{\mathrm{j}=\mathrm{k}+1}^{\mathrm{n}}\mathrm{Sgn}\left({\mathrm{X}}_{\mathrm{j}}-{\mathrm{X}}_{\mathrm{k}}\right)$$

(2)

where Xk and X_j are the values of the variable at time k and j, respectively, n is the length of the series and Sgn() is the sign function, defined as follows:

$$\mathrm{Sgn}\left({\mathrm{X}}_{\mathrm{j}}-{\mathrm{X}}_{\mathrm{k}}\right)=\{\begin{array}{c}1 if \left(X_j-X_k\right)>0\\ 0 if \left(X_j-X_k\right)=0\\ -1 if \left(X_j-X_k\right)<0\end{array}$$

(3)

It has been documented that when n ≥ 10, the statistic S is approximately normally distributed with the mean E(S) = 0, and its variance is:

$$Var \left(s\right)=\frac{n\left(n-1\right)\left(2n+5\right)-{{\displaystyle \sum }}_{i=1}^m{t}_i\left(t_i-1\right)\left(2t_i+5\right)}{18}$$

(4)

where n is the number of data points, m is the number of tied groups (a tied group is a set of sample data having the same value), and $t_i$ is the number of data points in the i^th group.

The standardized test statistic (Z) is computed as follows:

$$\mathrm{Z}=\{\begin{array}{c}\frac{S-1}{\sqrt{Var\left(S\right)}}, if S>0\\ 0, if S=0\\ \frac{S+1}{\sqrt{Var\left(S\right)}}, if S<0\end{array}$$

(5)

The null hypothesis, H₀, meaning that no significant trend is present, is accepted if the test statistic (Z) is not statistically significant, i.e., −Z_α/2 < Z < Z_α/2, where Z_α/2 is the standard normal deviate.

To overcome the limitation of the MK test related to autocorrelation of the original data that could affect the outcome of the test [31], a trend-free prewhitening (TFPW) algorithm was applied. This method enables removing serial dependence, which is one of the main problems in testing and interpreting time series data [28,31,66]. Trend-free prewhitening includes the following steps:

Calculate the first-order coefficient of autocorrelation (r):

$$r=\frac{{{\displaystyle \sum }}_{t=1}^{n-1}\left(X_t-{\overline{X}}_t\right)\left(X_{t+1}-{\overline{X}}_{t+1}\right)}{\sqrt{{{\displaystyle \sum }}_{t-1}^{n-1}{\left(X_t-{\overline{X}}_t\right)}^2}{{\displaystyle \sum }}_{t-1}^{n-1}{\left(X_{t+1}-{\overline{X}}_{t+1}\right)}^2}$$

(6)

Remove any trend items from the time series variables to form a sequence without trend items:

$$Y_t=X_t-{\beta }_t$$

(7)

Supplement the trend term β_t to obtain a new sequence without an autocorrelation effect:

$$\acute{Y_t=Y_t}-r_1{Y}_{t-1}+{\beta }_t$$

(8)

where:

$X_t$ is the value of the variable at time t of the time series, n is the length of the data, and ${\overline{X}}_t$ is the aver-age value. To assesses significance of the trend, the original MK test is applied to $Y_t$.

2.3. Sen’s Slope Estimator

Sen’s slope is a method for estimating the magnitude of a trend in time series data [23,26,28,29,37,67] by evaluating the slope of the trend [68]. This study used a 0.05 significance level confidence. When |Z| > 1.96, the null hypothesis is rejected, and the trend is significant at 5%. If a trend is detected in the data series, its amount can be evaluated by the slope of the trend (β in the following). Hence, the magnitudes of the trends in ETo were studied using Sen’s slope estimator:

$$\beta =\mathit{Median}\left(\frac{X_i-X_j}{i-j}\right) \mathit{for} \mathit{all} i>j$$

(9)

where $X_i$ and $X_j$ are the data values at times i and j, respectively. While the value of β > 0, the time series of the ETo and other climatic factors are increasing and the vice versa.

2.4. Sensitivity Analysis

Sensitivity analysis enables one to calculate the influence of climatic variables on ETo [61,69,70,71,72]. The sensitivity coefficient is the rate of variation in ETo with respect to meteorological variables [36,72]. It is a quantitative parameter that represents the effect degree of change of ETo when one or several related meteorological factors are changed [15]. To precisely determine the sensitivity of ETo to humidity, it needs to differentiate the specific humidity from the relative humidity. The relative humidity does not show the humidity exactly; rather it consists of humidity and temperature on its partition [49]. Hence, this study used specific humidity (SHU) to precisely determine and examine the sensitivity and contribution of humidity to the ETo trend in this study.

The specific humidity is computed as follows:

$$e=6.112 \exp \frac{17.6 T_d}{T_d+243.5}$$

(10)

$$q=\frac{0.622 e}{p-\left(0.378 e\right)}$$

(11)

where e is vapor pressure in mbar, T_d the dew point in °C, p the surface pressure in mbar, and q the specific humidity in kg/kg.

The dew point temperature is also computed as follows:

$$e_s=6.112\ast \exp \frac{17.67T}{T+243.5}$$

(12)

Otherwise, for Equation (10) we can use the following:

$$e=e_s\frac{RH}{100}$$

(13)

$$T_d=\frac{\log \left(\frac{e}{6.112}\right)\ast 243.5}{17.67-\log \left(\frac{e}{6.112}\right)}$$

(14)

where T is temperature in °C, e_s is saturation vapor pressure in mbar, e is vapor pressure in mbar, and RH is relative humidity in percent.

T_max and T_min contribute differently to the ETo trend. The FAO PM equation (Equation (1)) includes the e_s saturation vapor pressure [kPa], e_a actual vapor pressure [kPa], and their difference (e_s−e_a saturation vapor pressure deficit [kPa]). These terms are computed using the T_max and T_min.

The sensitivity coefficient equation is:

$$S{v}_i=\underset{v_i\to 0}{\lim }(\frac{\Delta ETo/ETo}{\Delta v_i/v_i})$$

(15)

where $S{v}_i$ is the sensitivity coefficient of $v_i$, ΔETo is the variation in ETo, $v_i$ is the meteorological factor, and ∆$v_i$ is the variation in $v_i$.

The positive or negative sensitivity coefficient means that ETo increases or decreases with the increase or decrease of a climatic variable. The values of the sensitivity coefficient (SVI) for a particular climatic parameter show the magnitude of the sensitivity of ETo in variation in that parameter. The larger the absolute value of the sensitivity coefficient, the larger the effect of a given variable on ETo [61,66,71,72].

Moreover, [73] the range of variation of the sensitivity coefficient was divided into four levels, as shown in

Table 1. Classification of the sensitivity coefficient.

Sensitivity Coefficient	Sensitivity Level
0.00 ≤ |$\mathit{S}{\mathit{v}}_{\mathit{i}}$| < 0.05	Negligible
0.05 ≤ |$\mathit{S}{\mathit{v}}_{\mathit{i}}$| < 0.2	Moderate
0.2 ≤ |$\mathit{S}{\mathit{v}}_{\mathit{i}}$| < 1	High
1.00 ≤ |$\mathit{S}{\mathit{v}}_{\mathit{i}}$|	Very high

.

2.5. Contribution Rate

This is computed by multiplying the sensitivity coefficient of a single meteorological factor by its relative change rate [15]. If the contribution rate results > 0, then the change of the factor means ETo is increasing, which means that the factor had a positive contribution to the variation of ETo. If the contribution rate < 0, then the change in the factor means that ETo is decreasing, and that the factor had a negative contribution [15].

$$Con{v}_i=S{v}_i\ast RC{v}_{i }$$

(16)

$$RC{v}_i=100\frac{n\ast Trend{v}_i}{\left|a{v}_i\right|}100$$

(17)

where, $\mathrm{Con}{v}_i$ is the contribution rate of $v_i$, $\mathrm{RC}{v}_i$ is the relative change rate in $v_i$, n is the number of years, $\mathrm{a}{v}_i$ is the mean value of $v_i$, and Trend $v_i$ is the annual trend in $v_i$.

2.6. Study Area and Data

The proposed approach was applied to an experimental site located in Piazza Armerina (Sicily, Italy). The climate in this area is typically Mediterranean, with hot but not torrid summers, mild and short winters, and moderate annual rainfall mainly occurring in the period from October to March [74]. The annual average temperature along the coast is between 17 and 18.7 °C, with July being the hottest month [75]. The maximum (T_max) and minimum temperature (T_min), relative humidity (RH), wind speed (WS), and solar radiation (SR) data were obtained from Piazza Armerina meteorological station installed and managed by the Sicilian Agro-meteorological informative service (Servizio Informativo Agrometeorologico Siciliano—SIAS, http://www.sias.regione.sicilia.it/, accessed on 20 July 2022). The astronomical location of the meteorological site is 37.382171° N and 14.3666704° E, and its elevation is 697 m a.s.l. (Figure 1). The dataset consisted of 17.25 years of daily data covering the period from 1 December 2003 to 28 February 2021, for all variables.

3. Results

The Penman Monteith was applied for estimation of ETo in the 17-year timeframe of analysis. The result showed that maximum daily ETo was 14.47 mm per day on June 24, 2007 (Figure 2A), and the minimum daily ETo was 0.24 mm per day on 4 January 2016.

3.1. MK-Test Trends of Meteorological Factors and ETo

Table 2. MK trend test result with 95% significance level.

	Jan	Feb	Mar	Apr	May	Jun	Jul	Aug	Sep	Oct	Nov	Dec	Win	Spr	Sum	Aut	Annual
Tmax	1.1	1.19	2.27	1.44	0.91	1.77	1.61	1.69	2.43	0.78	1.69	1.86	1.77	2.14	1.98	1.49	1.31
Tmin	0.08	0.95	0.99	0.62	−0.49	−0.21	1.19	2.18	2.14	0.01	1.36	−0.37	0.12	0.33	1.03	1.4	1.19
Tmean	0.12	1.03	1.94	1.36	0.29	0.45	1.69	1.77	2.35	0.29	1.77	0.62	0.95	2.51	2.18	1.49	1.58
SR	−0.95	−0.54	−0.7	0.62	−1.65	−0.33	−0.78	−1.2	−1.07	−1.44	−2.02	0.95	−0.12	−0.87	−0.91	−2.6	−1.44
WS	2.68	0.99	1.03	1.28	2.76	2.18	2.97	1.73	0.77	0.95	2.39	2.23	2.76	2.02	0.95	0.86	0.95
HU	0.41	0.5	2.27	2.27	2.12	2.84	3.17	2.39	2.12	2.02	1.22	2.02	0.86	2.21	2.03	1.31	1.31
ETo	−1.85	−0.95	−1.11	0.45	−1.28	−0.62	0.54	0.21	0.45	0.04	−2.51	−0.95	−0.54	−0.78	−0.21	−0.54	−0.78
RF	−0.04	0.45	−0.62	−1.19	0.29	−0.49	0.46	0.64	−0.21	0.33	0.95	0.5	−1.28	−0.29	−0.04	−2.6	−1.61

Win = winter; Spr = spring; Sum = summer; Aut = Autumn. Red color text = showed decreasing/upward trend.

shows the results concerning trend significance, as obtained from the MK-Test. The T_max exhibits positive trends for March and September and spring and summer, while no significant trends were obtained for other time series. The T_min presents positive monthly trends in August and September, and no other significant trends were observed. For Tmean, there was no significant trend except for September and spring and summer. Conversely, SR presents a negative trend in November and in Autumn. As regards WS, positive trends were observed on both seasonal and monthly scales. On the monthly scale, January, May, June, July, November, and December showed positive trends; for the remaining months no trend is evidenced. The HU also has a positive significant trend in March, April, May, June, July, August, September, October, and December, as well as in spring and summer. Rainfall showed a negative trend only in Autumn. The trend of ETo is negative only in November, and non-significant in the other cases.

3.2. Sen’s Slope (Magnitude of the Trend)

Climatic variables and ETo showed both negative and positive trends in seasonal and monthly scale analysis. The ETo showed a negative downward trend in November of 0.790 mm per year. The T_max increased in March (0.10 °C) and September (0.14 °C) (monthly trend analysis), and spring (0.10 °C) and summer (0.09 °C) (seasonal trend analysis). The T_min also increased in August (0.09 °C) and September (0.07 °C). The T_mean trend also increased in September (0.10 °C) in monthly analysis and in spring (0.07 °C), and summer (0.06 °C) in the seasonal analysis. The SR showed a downward trend in November (0.09 MJ/m²), at the monthly scale, and in Autumn (0.076 MJ/m²), at the seasonal scale. The WS also showed an upward trend in January (0.054 m/s), May (0.038 m/s), June (0.021 m/s), July (0.043 m/s), November (0.040 m/s), and December (0.042 m/s), as well as in winter (0.037 m/s) and spring (0.029 m/s). The HU trend also increased in March (0.711%), April (0.543%), May (1.169%), June (0.741%), July (1.012%), August (0.824%), September (0.816%), October (0.614%), and December (0.412%), as well as in spring (0.840%) and summer (0.942%). The RF also decreased in Autumn by 14.019 mm in the seasonal trend analysis.

3.3. Sensitivity of ETo to Climatic Factors

As the study stated above, for ETo estimation, T_max, T_mean, T_min, SR, WS, and SHU were used as input climatological variables. ETo showed different levels of sensitivity to these climatological variables. In particular, the result showed that SHU, T_mean, and T_max have a very high sensitivity level, with sensitivity coefficients of 2.68, 1.46, and 1.35 respectively. The RS and T_min also showed a high sensitivity level, with the sensitivity coefficient of 0.53 and 0.28 respectively. In contrast, the sensitivity level of wind speed was negligible, with a value of 0.02 for the sensitivity coefficient (

Table 3. Sensitivity coefficient of climatic factor for ETo.

Climatological Element	Sensitivity Coefficient |x|	Sensitivity Level
Net solar radiation	|0.53|	High
Maximum temperature	|1.35|	Very high
Minimum temperature	|−0.28|	High
Mean temperature	|1.46|	Very high
Specific humidity	|−2.68|	Very high
Wind speed	|0.02|	Negligible

).

3.4. Contribution Rate of Climatic Factors for the Variation of ETo

The above-mentioned climatic factors have different contribution rates for the ETo trends at different temporal scales. Figure 3 shows that the contribution rate of SHU, SR and T_min are negative, with values of 91.73, 2.48, and 1.06, respectively. On the other hand, T_max, T_mean, and WS contribute positively with an 11.2, 9.74, and 0.9 contribution rate. This result shows that SHU has the highest contribution to the decrease of ETo and maximum temperature has the highest contribution to the increase of ETo in the study area.

4. Discussion

4.1. Trend of Climatic Factors and ETo

Globally, studies conducted by many authors showed that there is an upward trend of mean, maximum, and minimum temperature [76,77,78,79,80,81]. Our study also revealed that there were monthly and seasonal temperature upward trends at the experimental site in Sicily (Figure 4B). This is in line with similar research results showing an upward trend in temperature in Sicily [77,82]. The studies cited also showed a decreasing rainfall trend in autumn in Sicily. In particular, from 1921 to 2012 there was a downward trend of rainfall throughout the island in autumn [83] and a downward trend at the annual scale [39]. Moreover, a seasonal decrease in rainfall has been shown and documented in Sicily and Calabria [34]. Our study revealed that there was an increase in monthly and seasonal minimum temperature. Globally, in about the 37% of the landmasses from 1951–1990 the minimum temperature and maximum temperature increased by 0.84 °C and 0.28 °C respectively [84]. Similar results are also documented in several studies for different parts of Italy. From 1865 to 1996 southern Italy showed an upward trend of the minimum and maximum temperature, especially in southern Italy [85]. Moreover, from 1952 to 1990, Bologna showed an increase of 0.7 °C in 48 years in annual mean temperature and an increase in higher minimum and maximum temperature [86]. In Calabria (southern Italy), trend analysis results showed an increase in maximum and the minimum temperature specifically in the summer and spring seasons (from 1951 to 2010) [87]. This result is in line with our study since the maximum temperature showed a positive trend in summer and spring (Table 2). The maximum temperature also had a positive trend in Sardinia from 1982 to 2011 [88]. This study confirmed that there is an upward trend in mean temperature in summer, spring, and in the month of September (Table 2). The mean temperature of Sicily from 1924 to 2013 also showed an upward trend in summer and spring [82].

This study showed that there was increasing monthly and seasonal trend of relative humidity. In Ravenna, Italy, for the period 1989 to 2008, relative humidity had an upward trend [89]. Moreover, by considering the past and projected future data from 1971 to 2050, relative humidity also had an upward trend for Sicily [90]. Similar results showing an increase in relative humidity were confirmed in different part of the world [91,92,93]. Unlike other main meteorological factors, solar radiation showed a downward trend in November and in autumn (Table 2). Southern Italy showed a decrease in solar radiation after the mid-1980s [94]. In China from the 1960s to 2010s, solar radiation showed a downward trend [95,96]; and globally also showed a downward trend in solar radiation after the 1980s as well as the trend known as “global dimming” [97].

Like the most climatological elements, wind speed also showed an upward trend both in monthly and seasonal analysis (Table 2). Comprehensive studies that analyzed several meteorological sites showed both upward and downward trends in wind speed. Meanwhile, these studies also showed that there were increases in wind speed in monthly as well as seasonal timescales [98,99,100,101]. However, the study which analyzed 24 meteorological sites seasonal, monthly, and annually in Iran from 1975 to 2005 showed a downward trend in wind speed in most of its stations [92].

Table 2 showed that there was a downward trend in monthly reference evapotranspiration in November. This result is known as the “evaporation paradox”. There are many studies discussing a monthly, seasonal, and annual downward trend in ETo. For instance, studies conducted in China [25,66], India [38,69], and Senegal [36] revealed a downward trend in ETo seasonally, monthly, and annually. In Iran, a downward trend in seasonal and monthly trend was also found [27], and specifically in November [102]. In general, this study did not show seasonal or annual trends other than for November.

4.2. Sensitivity of ETo and Contribution Rate of Climatological Elements

The sensitivity analysis of ETo for climatic factors showed different magnitudes of sensitivity. The main and very high-level sensitivity in ETo trends was observed for specific humidity, mean temperature, and maximum temperature. Specific humidity was the most sensitive climatic factor for ETo trends in the study area. On the other hand, wind speed had a negligible sensitivity climatic factor for ETo trend. Likewise, relative humidity was the highest sensitivity climatic factor in the Loess Plateau of Northern Shaanxi, China [15]; in the Tao’er River Basin, China [71]; and in the Yellow River Basin, China [25]. Mean temperature was also one of the most sensitive factors for ETo in Iran [61]; in the Tarim River basin, in Central Asia [66] and in the Yellow River Basin, China [25].

Similarly maximum temperature was also the another prime sensitivity climatic factor for ETo in Senegal [36]; the Himalayan region of Sikkim, India [72]; the Tarim River basin, Central Asia [66], and in the Tons River Basin in Central India [69]. Wind speed also showed the lowest sensitivity in Santa Barbara [70]; the Himalayan region of Sikkim, India [72] and in the Tarim River basin, Central Asia [66].

This study showed that specific humidity, solar radiation, and minimum temperature contributed negatively to the trend of ETo; whereas maximum temperature, mean temperature, and wind speed contributed positively to ETo in the study area (Figure 3). The contribution values showed that the absolute value of the climatic factors showed the contribution magnitude. Hence specific humidity is the climatological factor that contributed the most to monthly decrease in ETo in November. On the other hand, wind speed made the least contribution to ETo in the study area. Maximum temperature was prominent in ETo in the Loess Plateau of Northern Shaanxi, in China and the Tarim River basin, in Central Asia; and wind speed was the factor that contributed less to ETo in the Loess Plateau of Northern Shaanxi, China [15].

5. Conclusions

In this study, referring to the FAO Penman-Monteith method for estimating reference evapotranspiration, we investigated the trends and sensitivity of ETo to meteorological variables. This analysis is important as there is a lack of similar studies. Also, in a climate change context it is key to understand the main climatological factors controlling such an important process for water resources management as evapotranspiration [103]. The analysis considered the data collected at a Mediterranean climate site in Sicily, Italy (Piazza Armerina). The results show that there was no annual trend for all the analyzed variables. Trends are present only at the sub-annual scale (monthly or seasonally). Significant trends at the monthly and seasonal time scales were exhibited for all variables other than minimum temperature and ETo. ETo showed a significant trend only monthly, for November (ETo), and minimum temperature only monthly, for August and September.

The sensitivity analysis also showed that specific humidity, mean temperature, and maximum temperature are those factors that have a greater influence on ETo. Sensitivity to wind speed is negligible. In terms of the contribution of the climatic factors for ETo trends, specific humidity, solar radiation, and minimum temperature contribute negatively to ETo. On the other hand, maximum temperature, mean temperature, and wind speed contribute positively. These results contribute to understanding the potential and possible future footprints of climate change on evapotranspiration in the study area. On the one hand, the fact that no significant trends are exhibited by ETo seems to imply that the impacts of climate change have not left a relevant footprint on this hydrological variable. On the other hand, given the high sensitivity of ETo to temperature, it must be expected that ETo will be highly impacted by climate change in the future; as temperature is expected to increase by 2–3 °C in Sicily, depending on emission scenario [104,105]. Moreover, the study in Pertouli and Taxiarchis in Greece also exhibited an increase in the potential evapotranspiration (PET) from 1974 to 2016 [106]. Further development of this study will consider more meteorological stations in Sicily, as well as other sources of data, to extend the length of the series and thus to improve the significance of trend assessments. Moreover, investigation could be extended, where data is available, to actual evapotranspiration.