1 Introduction

Several studies in literature related the intense and repeated climate anomalies, occurring during the last decades, with the existence of modifications in the observed climate. Trends in temperature probably represent the most known and evident effect of the changes occurring in the Earth’s climate system. According to Stocker (2013), over the period 1880–2012, a warming of 0.85 °C has been detected in the globally averaged combined land and ocean surface data as calculated by a linear trend. Several studies confirm this statement at local and continental extent: temperature increases have been recorded in the USA (Balling and Idso 1989; Lettenmaier et al. 1994), Canada (Zhang et al. 2000), Australia (Collins et al. 2000), Southeast Asia and the South Pacific (Manton et al. 2001), China (Zhai and Pan 2003), India (Kothyari and Singh 1996), South Africa (Kruger and Shongwe 2004), and Mediterranean area, (Piervitali and Colacino 1997; Sahsamanoglou and Makrogiannis 1992; Viola et al. 2014).

The largely documented global warming is supposed to change atmospheric circulation, which imply associated changes in winds and, in turn, waves and surface fluxes modification as well. Namely, because of the decrease of temperature difference between the poles and the equator, global air circulation will slow (particularly in the northern latitudes), but in specific areas, local wind speeds may increase due to local temperature gradients (Hazlett 2011).

Recently, the interest in analyzing historical wind data highly increased, since wind speed represents an essential forcing for hydrological processes. Surface wind is an important driver of oceanic and atmospheric circulation and wind trends of modest magnitude can deeply affect atmospheric and ocean dynamics, air-sea fluxes, and evapotranspirative fluxes. Moreover, also from a technical perspective, wind patterns are important, for instance, in wind farm design, because altered wind patterns could adversely influence, in the long term, the performance and the productivity of wind turbines (Bakker et al. 2013; Dale et al. 2004; Holt and Wang 2012; Pryor and Barthelmie 2005; Sailor and Smith 2008).

Currently, literature offers numerous studies on measured wind speed variations, carried out as regards to global or local scale. According to Ward and Hoskins (1996), the analysis of time series of local surface pressure gradients does not support the existence of any significant globally averaged trends in marine wind speeds, but reveals the existence of regional patterns of increasing trends in the tropical North Atlantic and extra-tropical North Pacific and decreasing trends in the equatorial Atlantic, tropical South Atlantic and subtropical North Pacific. Young et al. (2011) analyzed a 23-year database of calibrated and validated satellite altimeter measurements in order to detect changes in oceanic wind speed and wave height, finding a general global increasing trend in wind speed. Tokinaga and Xie (2011) found that the trade winds in the tropical Atlantic have weakened during the period 1950–2009. A global review of 148 studies showed that near-surface terrestrial wind speeds are declining in the Tropics and the mid-latitudes of both hemispheres at a rate of −0.14 m/s per decade (McVicar et al. 2012). Numerous recent studies have reported declines in near-surface observed wind speeds during the past 30–50 years over parts of North America (Klink 1999; Pryor and Ledolter 2010; Pryor et al. 2009; Tuller 2004), Southern and Western Canada (Wan et al. 2010), Australia (McVicar et al. 2008; Roderick et al. 2007), and regions of Europe (Brázdil et al. 2009; Cusack 2013; Earl et al. 2013; Hewston and Dorling 2011; Pirazzoli and Tomasin 2003). A great number of studies have documented the existence of wind speed trends in the last 50 years in China (Jiang et al. 2010; Lin et al. 2013; Xu et al. 2006; Zhang and Ren 2003). Among these, Zunya et al. (2004) and Guoyu et al. (2005) showed a decreasing trend in annual mean wind speed during 1954–2001 for mainland China, while Chen et al. (2013) quantified and compared the temporal trends in 10-m wind speeds over China during the period 1971–2007, detecting downward trends especially in the upper percentiles and during spring. Rahim et al. (2012) investigated the effects of global warming on the average wind speed field in Central Japan, detecting an increase of surface wind speed over the coastal and mountainous regions.

Finally, as regards to the northwestern European area, Atkinson et al. (2006) investigated the presence of trend in several windiness indices, identifying a downward trend in wind speed over the past 15 years. A study carried out by Siegismund and Schrum (2001) investigated the evolution of local wind climate in the North Sea area, finding a rising trend of ∼10 % during the last 40 years in annual mean wind speeds. In the Netherlands, Smits et al. (2005) observed that moderate wind events (occurring on average ten times per year) and strong wind events (occurring on average twice a year) had decreased between 5 and 10 %, over the period 1962–2002. With reference to Central Europe, Schiesser et al. (1997) reported a significant negative trend in the number of winter storms in Switzerland between 1864 and 1994.

As regards to Italy, Pirazzoli and Tomasin (2003) analyzed trends in wind speed and direction during the second part of the last century in data from 17 Italian stations over the central Mediterranean and Adriatic areas. Results highlighted the presence of a general increasing trend after the mid-1970s that shows a positive correlation with temperature trends in the area of study.

Wind and temperature play a fundamental role on ecosystems and water cycle, since they strongly affect evapotranspiration rates. The process of evapotranspiration moves water vapor from ground or water surfaces to an adjacent shallow layer. Once a saturated layer moves, wind replaces it with a drier air layer, which can absorb water vapor. In literature, different studies have shown that evapotranspiration is widely conditioned by wind speed (Dinpashoh et al. 2011; Kousari and Ahani 2012; McVicar et al. 2012). According to Stocker (2013), on a global scale, evapotranspiration over land increased from the early 1980s up to late 1990; after 1998, a lack of soil moisture availability in the Southern Hemisphere land areas has acted as a constrain to further increase global evapotranspiration. In contrast with the expectation that global warming will produce an increase in evaporation, many recent studies have shown that pan evaporation and potential evapotranspiration had decreased over the past decades in different places of the world (Brutsaert and Parlange 1998; Chattopadhyay and Hulme 1997; Cohen et al. 2002; Elnesr and Alazba 2013; Golubev et al. 2001; Jhajharia et al. 2012; Lawrimore and Peterson 2000; Liu et al. 2014; Peterson 1995; Roderick and Farquhar 2002; Roderick and Farquhar 2004; Roderick et al. 2007; Talaee et al. 2013; Thomas 2000; Zongxing et al. 2014). Some of these studies have attempted to identify the meteorological variable that can be considered the driving force of these trends. In many cases, wind speed has been recognized to play a critical role in determining changes in evaporation and evapotranspiration. Rayner (2007) carried out a study in order to detect the main causes of pan evaporation trends observed in Australia. In the analyzed area, a pan evaporation rates decrease has been observed, even if air temperatures have increased (i.e., the so-called pan evaporation paradox, (Peterson 1995)). Trends in daily average wind speed were found to be an important cause of such trends in pan evaporation. In order to investigate changes in pan evaporation, Roderick et al. (2007) analyzed radiation, temperature, humidity, and wind speed data from 41 Australian sites for the period 1975–2004. Results showed that changes in temperature and humidity regimes were generally too small to impact pan evaporation rates, whose decrease was mostly due to decreasing wind speed trends. As regards to evapotranspiration, its assessment is a key topic for supporting agricultural planning and management, especially in Mediterranean managed ecosystems where rainfall is out of phase with temperature and consequently most of the water is provided through irrigation. Some studies focused on the effects of trends in the hydrometeorological variables that condition it, such as maximum and minimum temperature, dew point, wind speed, and net radiation (Irmak et al. 2012; Moratiel et al. 2011; Zuo et al. 2012). More recently, Liu and Zhang (2013) have focused just on two of these variables, investigating the roles of changing wind speed and surface air temperature on evapotranspiration trends in northwest China.

It is important to emphasize that pan evaporation and potential evapotranspiration are highly related, but a decreasing trend of pan evapotranspiration does not necessarily correspond to a decrease of potential evapotranspiration (Fu et al. 2009). In this regard, Cohen et al. (2002), carried out an analysis of evaporation measurements made between 1964 and 1998 at Bet Dagan in Israel’s central coastal plain, finding a slight increase of pan evaporation, but no changes in the total open water evaporation or reference crop evapotranspiration, calculated with Penman’s combined heat balance and aerodynamic equation. This result has been attributed to the combined effect of a decrease in the solar radiation balance term and an increase in the aerodynamic term.

The premises given above hint the role of wind speed and temperature trends on evapotranspiration and address the aim of this study to the detection of trend in such variables in Southern Italy and to a quantitative evaluation of the effects of their changes on reference evapotranspiration. Wind speed, direction, and temperature time series, measured over 10 stations during the period 1968–2004, have been investigated in order to detect statistical significant trend. Reference evapotranspiration has been calculated, using the FAO-56 Penman–Monteith equation, exploring a wide range of temperature and wind speed combinations arising from trend analysis. This approach is double-acting because from one side it quantifies the effects of local tendencies, while from the other side, it presents generally exportable results, allowing the identification of the dominant variable in determining evapotranspiration changes in the Mediterranean basin.

2 Data and methods

2.1 Datasets

The area of study, which has been selected for summarizing possible scenarios of wind and temperature changes, is the Southern Italy and is characterized by a Mediterranean climate with mild winters and hot and generally dry summers. Hot winds, blowing from the North African coast (i.e., Sirocco), are frequent during autumn and spring.

Datasets of three-hourly wind speed and daily mean temperature for the period 1968–2004 over 10 sites across Southern Italy (Fig. 1) have been provided by Servizio Meteorologico dell’Aeronautica Militare (Italian Air Force Meteorological Service). Wind speeds were measured at 10 m above the ground, while temperature at 2 m above the ground. Wind directions were recorded using 10° increments, calculated clockwise from the geographical north, corresponding to the average direction within the 10 min preceding the measurement time, whereas wind velocity corresponds to the average velocity within the same time period to which the direction is referred. After a preliminary inspection, data showing inconsistent wind direction (greater than 360°) or temperature (according to the Chauvenet’s criterion (Chavuenet 1871)) were identified and removed from the dataset. Although the number of missing observations is very low, missing data were replaced averaging the neighbor’s values; it has been verified that this procedure does not influence trend analysis. For each station and for each year of observation, mean annual and monthly wind speed and temperature have been calculated and then analyzed.

Fig. 1
figure 1

Location of wind speed and temperature gauging stations, and indication of significant trends for α = 0.1 in hourly and annual mean wind speed series

Full size image

For each wind gauging station, the wind rose plot has been evaluated in order to give an immediate representation of occurrence and mean wind speed for a given direction (Fig. 2). The wind rose plot is a diagram that provides the percentage of the time the wind blows from each direction during the observation period. The magnitude of the measured wind speeds (m/s) is plotted along the radius, which represents the frequency of occurrences, while the corresponding wind directions are given along the circumference; wind speed is also represented with color bands showing wind speed ranges. The analysis of Fig. 2 allows one to identify the most frequent wind directions throughout the considered period (1968–2004). As regards to most of the measurement stations (Capo Palinuro, Capri, Cozzo Spadaro, Gela, S. Maria di Leuca, Messina, Pantelleria, and Trapani Birgi), the main wind direction is north with most frequent speeds ranging from 0 to 10 m/s. In Capo Carbonara the prevailing direction is west and most frequent speeds range from 5 to over 10 m/s. The wind rose plot for Cozzo Spadaro shows that, in addition to north, west is another direction with a high frequency of occurrence. In about 10 % of data, wind speeds from west range from 5.0 to 10 m/s. In Marina di Ginosa, a considerable percentage of winds are from west-northwest and north-west. The basic wind statistics are coherent with those reported by Aeronautica (2009).

Fig. 2
figure 2

Wind rose plot for each measurement station

Full size image

2.2 Detection of trends

Several statistical procedures can be used for trends detection, in particular parametric and nonparametric tests. In this study, the nonparametric Mann–Kendall test for trend detection (Kendall 1962; Mann 1945) has been used. This test identifies the presence of a trend, without making an assumption about the distribution properties and, being nonparametric, is less influenced by the outliers’ presence. In a trend test, the null hypothesis H 0 is that there is no trend in the population from which the data are extracted, while the alternative hypothesis H 1 is that there is a trend in the records. The test statistic, Kendall’s S (Kendall 1962), is calculated as:

$$ S={\displaystyle \sum_{i=1}^{n-1}{\displaystyle \sum_{j=i+1}^n\mathrm{sign}}}\left({y}_j-{y}_i\right) $$
(1)

where y are the data values, namely wind speed and temperature, at times i and j, n is the length of the dataset and

$$ \mathrm{sign}\left({y}_j-{y}_i\right)=\kern0.36em \left\{\begin{array}{l}1\kern0.36em \mathrm{if}\;\left({y}_j-{y}_i\right)>0\\ {}\kern0.48em 0\kern0.36em \mathrm{if}\;\left({y}_j-{y}_i\right)=0\\ {}-1\kern0.36em \mathrm{if}\;\left({y}_j-{y}_i\right)<0\end{array}\right. $$
(2)

Under the null hypothesis that y i are independent and randomly ordered, the statistic S is approximately normally distributed when n ≥ 8, with zero mean and variance as follows:

$$ {\sigma}^2=\frac{n\left(n-1\right)\left(2n+5\right)}{18} $$
(3)

The standardized test statistic Z S is computed by:

$$ {Z}_{\mathrm{S}}=\left\{\begin{array}{l}\frac{S-1}{\sigma}\kern0.96em \mathrm{if}\;S>0\\ {}\kern0.6em 0\kern1.32em \mathrm{if}\;S=0\\ {}\frac{S+1}{\sigma}\kern0.96em \mathrm{if}\;S<0\end{array}\right. $$
(4)

and compared with a standard normal distribution at the required level of significance. If the significance level α is set equal to 0.05, the null hypothesis is verified when |Z s | < 1.96. A positive value of Z S indicates an increasing trend and vice-versa. Local significance levels (p values) for each trend test can be obtained from:

$$ p=2\cdot \left[1-\varPhi \left(\left|{Z}_S\right|\right)\right] $$
(5)

where Φ(⋅) denotes the cumulative distribution function of a standard normal variate.

The magnitude of trends was evaluated using a nonparametric robust estimate determined by Hirsch et al. (1982):

$$ \beta =\mathrm{Median}\left(\frac{x_j-{x}_l}{j-l}\right)\kern6em \forall \kern0.5em l<j $$
(6)

where x l is the lth observation.

2.3 Estimation of reference evapotranspiration

Evapotranspiration is the result of the combination of two separate processes, evaporation and plant transpiration, and plays a fundamental role in the water and energy balance on Earth’s surface. The process of evapotranspiration represents an important variable of hydrologic budgets at different spatial scales. For this reason, a great attention has been paid to this phenomenon and consequently a large number of mathematical models have been developed to estimate evapotranspiration. Potential evapotranspiration is defined as the amount of water that can be evaporated and transpired when soil water is sufficient to meet atmospheric demand (Allen et al. 1998). Potential evapotranspiration models can be classified into five categories (Xu and Singh 2002): mass transfer (e.g., Harbeck 1962), combination methods (e.g., Penman 1948), radiation-based (e.g., Priestley and Taylor 1972; Turc 1961) water budget (e.g., Guitjens 1982), and temperature-based (e.g., Blaney 1952; Hamon 1963; Hargreaves and Samani 1985; Thornthwaite 1948).

The Penman model (Penman 1948) can be considered as the first effort to combine both energy and atmospheric vapor transport components to estimate potential evaporation, since it refers to open water surface. The Penman–Monteith model (Monteith 1965) for the calculation of evapotranspiration is considered as an extension of the Penman model, but referring to vegetation. Starting from the original Penman–Monteith equation, the FAO-56 Penman–Monteith (Allen et al. 1998) method allows to estimate the so-called reference evapotranspiration, ET0 (mm/day), which has been derived parameterizing a hypothetical well-watered grass that is 0.12 m height, with a leaf area index of 4.8 [m2/m2], a bulk surface resistance of 70 s m−1, and an albedo of 0.23. Once the main physiology characteristics of the standard vegetation have been declared, then the Penman–Monteith equation becomes:

$$ {\mathrm{ET}}_0=\frac{0.408\varDelta \left({R}_{\mathrm{n}}-G\right)+\gamma \frac{900}{T+273}{u}_2\left({e}_{\mathrm{s}}-{e}_{\mathrm{a}}\right)}{\varDelta +\gamma \left(1+0.34{u}_2\right)} $$
(7)

where R n is the net radiation, G is the soil heat flux, γ is the psychrometric constant, Tis the mean daily air temperature (°C), u 2 is the wind speed at 2 m height (m/s), e ais the actual vapor pressure (kPa), e s is the saturation vapor pressure (kPa), and the difference (e s − e a) represents the vapor pressure deficit of the air, and Δ represents the slope of the saturation vapor pressure temperature relationship. Equation (7) has been used in order to assess, at a daily scale, reference evapotranspiration for 10 locations within the study area (Fig. 1) using daily climatic data provided by Servizio Meteorologico dell’Aeronautica Militare within the period 1968–2004.

3 Results and discussion

3.1 Wind speeds trend analysis

The trend analysis has been carried out using the Mann–Kendall nonparametric test, while the magnitude of significant trends has been estimated using Eq. (6). Different temporal scales, namely hourly, monthly, and annual, and different significance levels α equal to 0.1, 0.05, and 0.01 have been considered.

Results obtained from the analysis of hourly and annual averaged data are summarized in Table 1, where for each station, the presence of trend, their magnitude, and the p value have been reported. Considering the hourly series, all the stations showed a statistically significant trend; in particular in seven out of ten stations (Capo Carbonara, Capo Palinuro, Gela, Marina di Ginosa, Messina, Pantelleria, S. Maria di Leuca) there has been an increase in wind speed over the period 1968–2004 (Fig. 1), while the remaining three stations (Capri, Cozzo Spadaro, Trapani Birgi) showed a negative trend. The increases range from 0.01 to 0.11 m/s per year, while decreases are between–0.05 to–0.03 m/s per year. As regards to mean annual data (Fig. 1), six stations (Capo Carbonara, Capo Palinuro, Marina di Ginosa, Messina, Pantelleria, S. Maria di Leuca) showed an increase, three stations (Cozzo Spadaro, Gela, Trapani Birgi) have not been of interest by a statistically significant trend, while a negative trend has been detected just in a station (Capri). The magnitudes of the positive trends range from 0.011 to 0.073 m/s per year, while the unique negative trend has a magnitude of −0.02 m/s per year. Figure 3 shows the mean annual wind speeds and the relative trends for the stations in which the positive and negative trends with the highest magnitudes β (m/s per year) have been detected (Capo Carbonara and Capri, with β equal to 0.00034 m/s per year and−0.000017 m/s per year, respectively). When wind speed has been averaged at monthly time scale (Table 2), generally positive trends have been detected. Monthly wind speed data from Capo Carbonara and Messina have been interested by positive significant trends in every month, with the greatest increases (α = 0.01) in September (0.106 m/s per year) and October (0.053 m/s per year) respectively. In Capo Palinuro, monthly mean wind speed showed a significant trend from March to November with significance level from 0.05 to 0.01. For the station of Gela, positive trends have been detected for the months from July to October, with the greatest increase in September (0.036 m/s per year). In Pantelleria, which is in the south border of the considered domain, the months belonging to the growing season, namely from April to November, showed upward trends with magnitude ranging from 0.029 (October) to 0.047 m/s per year (June). As regards to the station of S. Maria di Leuca, again within the growing season, from May to September, significant increases have been detected, ranging from 0.023 (August) to 0.052 m/s per year (June).

Table 1 Location of measurement stations, and hourly and mean annual wind speed trends and their significance levels, magnitudes and p values
Full size table
Fig. 3
figure 3

Mean annual wind speeds and relative trends for the stations in which the positive and negative trends with the highest magnitudes β (m/s per year) have been detected (Capo Carbonara and Capri respectively)

Full size image
Table 2 Monthly mean wind speed trends and their significance levels, magnitudes and p values
Full size table

Negative trends have been detected in monthly wind speeds time series recorded in Capri, Cozzo Spadaro, and Trapani Birgi. Monthly data from the station of Capri are of interest by decreases from January to August and in December. The greatest (lowest) magnitude of these negative trends occurs in February with −0.047 m/s per year (in July with −0.014 m/s per year). In Cozzo Spadaro, downward trends have been detected from January to May and in October and December and the decreases range from −0.047 (February) to −0.026 m/s per year (May). For the station of Trapani Birgi, significant trends have been detected only in February (−0.068 m/s per year) and March (−0.033 m/s per year) with significance level 0.05 and 0.1, respectively.

Results of trend analysis on monthly averaged wind speeds are summarized in Fig. 4 where, for each station, the long term monthly mean and the likely shifts are depicted. While this figure shows the absence of marked seasonality in monthly averaged wind speeds, it illustrates the potential increase/decrease in a 50-year temporal horizon just where a positive/negative trend has been detected (with a significance level α equal to 0.1). This representation provides the opportunity of a general overview of wind tendencies in the analyzed area. It could be observed how monthly averaged wind speed generally increases, with the exception of three cases. Focusing the attention on the Mediterranean growing season, different extents of this tendency have been observed.

Fig. 4
figure 4

Mean monthly wind speeds and their increase or decrease in 50 years when a significant trend (α = 0.1) has been detected

Full size image

The obtained results, at hourly, monthly, and annual time scales can be summarized in a general increase of wind speed, with some local exceptions, which confirm the outcomes of previous studies (Pirazzoli and Tomasin 2003), whose explanation can be shared. In this cited work, the authors tried to relate Mediterranean wind speed trends with the North Atlantic Oscillation, observing a negligible correlation, while the temperature increase has been demonstrated to be better correlated with wind tendencies.

As regards to wind direction, the number of wind events blowing from a given direction per year has been analyzed and results summarized in Table 3. North is the direction with the greatest number of trends, nine out of ten stations, and the greatest magnitudes β, ranging from −40 to −11 events/year. The number of wind events from north-northwest decreased for four stations (Capo Carbonara, Gela, Marina di Ginosa, and Pantelleria), with β ranging from −4 to −1 events/year, and increased for three stations (Capo Palinuro, Cozzo Spadaro and Messina), with β equal to 2 and 7 events/year. Similar results have been obtained considering wind events blowing from West. For the remaining directions, the β of the statistically significant trends ranges from 1 to 13 events/year and from −9 to −1 events/year for positive and negative trends respectively.

Table 3 Wind direction trends and their significance levels, magnitudes β [no of events/year] and p values
Full size table

3.2 Temperature trend analysis

Trend analysis has been carried out using daily temperatures, aggregated at annual and monthly scales. According to trend analysis results, and to previous local assessments (Viola et al. 2014), the analyzed sites are affected by statistically significant positive trends. Significance levels, magnitude of these trends, and relative changes with respect to monthly averages are reported in Table 4. Results showed that, in the area of study, trends are more frequent in summer months (June, July, and August), but no seasonality emerges in relative changes. June is the month with the greatest number of trend (nine stations), with minimum magnitude in Cozzo Spadaro (0.01 °C/year) and maximum magnitude in Marina di Ginosa, Capri, and Pantelleria (0.07 °C/year). Instead, September is the month with the smallest number of trend (only Gela, 0.02 °C/year).

Table 4 Monthly mean temperature trends and their significance levels, magnitudes and p values
Full size table

Mean monthly temperatures from the Sicilian stations Cozzo Spadaro, Gela, and Trapani are affected by trend in most of the months; in particular, temperatures measured in Gela are of interest by a trend in every month, ranging from 0.04 °C/year (October) to 0.02 °C/year (September).

The lowest number of trend has been detected in Capo Carbonara. Mean monthly temperature from this station are affected by positive trend only in March and June, with magnitude 0.1 °C/year and 0.06 °C/year, respectively.

3.3 Evaluation of wind speed and temperature trend effects on reference evapotranspiration

As previously mentioned, several studies tried to relate changes in reference or potential evapotranspiration ET0 taking into account the trends of temperature, relative humidity, solar radiation, and wind speed. In this study, the influence of wind speed and temperature trends on future ET0 have been analyzed. In order to evaluate and quantify the effects of wind speed and temperature trends on reference evapotranspiration, different scenarios have been created, analyzed, and then compared with the current climate conditions. First, daily potential evapotranspiration series have been evaluated by the FAO-56 Penman–Monteith equation for the current climate condition, named hereafter as NT (“No trend”) scenario. In particular, this reference scenario consists of daily values of ET0 calculated with Eq. (7), in which the climatic variables have been obtained as the historical means calculated in each day of the year within the period 1968–2004.

Afterwards, three pairs of scenarios have been defined. Each pair invokes climate projections to the next 50 and 100 years based on the extrapolation into the future of statistically significant detected trends. The first pair, identified with the letter W, takes into account the effects of the observed significant wind speed trends; the second pair, identified with the letter T, takes into account the role of increasing temperature as detected in the trend analysis, whereas the third pair, identified with the letters WT, considers the coupled effects of wind speed and temperature changes. As before, these scenarios refer to daily ET0 calculated with Eq. (7) in which climatic variables are the daily historical means, with the exception of wind speed and temperature. In fact, for these variables statistically significant monthly trends (α = 0.1) have been superimposed to the daily averages in order to quantify the increase/decrease of these variable in the next 50 and 100 years. The method used for the scenarios definition implicitly assumes the stationarity of relative humidity and solar radiation while their seasonal variability has been reproduced miming historical time series. It is also assumed, for sake of simplicity, that a constant CO2 concentration equal to 350 ppm., notwithstanding it is widely recognized its role in modifying water use efficiency and consequently potential evapotranspiration (Eamus 1991; Drake et al. 1997).

Once the seven scenarios (NT, W50, W100, T50, T100, WT50, and WT100) have been defined, the relative daily reference evapotranspiration series have been assessed. Results have been aggregated at monthly and annual time scales (mm/month and mm/year respectively) in order to smooth the daily stochasticity.

In Table 5, for each station, the monthly ET0 values have been reported and, when a statistically significant wind speed or temperature trend has been detected, the percentage variations δ of ET0 in each scenario compared with the NT scenario; when both significant wind speed and temperature trends have been observed, the percentage variations relative to the WT scenarios are reported.

Table 5 Monthly evapotranspiration ET0 [mm] calculated for the NT scenario and the percentage of increases/decreases δ for the W, T, and WT scenarios
Full size table

As regards to W scenarios, increases and decreases of ET0 have been observed, as a natural consequence of positive and negative wind speed trends. The highest percentage increases of ET0 have been observed for Capo Carbonara in November (+24.9 % for W50, and +36.6 % for W100) while the highest decreases have been observed for Capri in February (−39.6 % for W50 and −59.7 % for W100). In the T scenarios, the positive temperature trends resulted in increases of ET0 for all the stations (at least for 2 months a year) ranging from +1.7 to +27 % in the T50, and from +3.3 to +58.5 % in the T100.

In the WT scenarios, the combined effects of wind speed and temperature trends produced an increase of total annual ET0 for eight stations and a decrease for two stations (Capri and Cozzo Spadaro). At the monthly scale, it is interesting to observe how the coupled effects of wind speed and temperature positive trends produced nonlinear effects: in fact the WT′ outputs are always higher than the sum of the ones obtained in W and T scenarios. The highest increases have been detected for Capo Palinuro in November (+31.6 % for WT50 and 65.1 % for WT100). In Trapani Birgi, which could be considered as a paradigm case, the downward wind speed trends in February and March counterbalances the effect of the positive temperature trends, and then the decrease of ET0 in the WT scenarios is lower than the one relative to the W scenarios.

In Fig. 5 the monthly values of ET0 for W100, T100 and WT100 scenarios are reported and compared for the ten sites. Regarding the seasonal absolute variations of ET0 two behaviors emerge: the first showing almost constant shifts between the NT and future scenarios and the second in which absolute changes are concentrated during the growing season. At the same time, from the analysis of Fig. 5, it is clear that the effect of the significant wind speed rise on ET0 has been amplified by the increase of temperature.

Fig. 5
figure 5

Monthly ET0 [mm] in NT, W100, T100, and WT100 scenarios

Full size image

The sensitivity of ET0 to wind speed and temperature changes has been investigated, in order to further explore the obtained results. Figure 6a shows the relative changes of ET0 against the relative changes of wind speed or temperature at the monthly scale provided by the scenarios W50, W100, T50, and T100. It is clear that both wind speed and temperature have a positive correlation with potential evapotranspiration, as expected and arguable from Eq. 7. Temperature is the most sensitive variable, because its contribution to the increasing of ET0 is four times greater than that from wind speed, for fixed relative variation, as one can argue comparing the interpolating lines’ slopes. The analysis of Fig. 6a reveals some dispersion around the interpolating lines, which are wider when wind speed is allowed to change. This could be explained, taking into account the seasonality of wind speed and temperature. Significant wind speed changes trigger high relative variations in ET0 if they occur during the growing season (i.e., summer months) when temperatures are higher too. If instead these wind variations occur during the dormant season, small relative variations of ET0 are expected. This explains how a given wind speed change could result in different relative variations of ET0. From the other side, a given significant temperature increase must be coupled with seasonal windiness, which is not marked as already observed (cfr. Fig. 4), thus producing less pronounced relative changes in ET0.

Fig. 6
figure 6

Relative changes of ET0 as a function of wind speed or temperature relative changes at the monthly scale in the scenarios W and T (a) and considering also the WT (b) scenarios

Full size image

The arguments depicted in Fig. 6a have been extended in Fig. 6b where the joint effects of relative changes in temperature and wind speed, as forecasted in the scenarios WT50, WT100, on relative changes in ET0 have been added. The surface has been obtained interpolating the values of relative changes in ET0 for all the considered scenarios (namely W50, W100, T50, T100, WT50 and WT100, blue dots in Fig. 6b). The plotted points include significant wind speed change cases (43 %), temperature change cases (36 %) and simultaneous variations in wind speed and temperature cases (20 %). The interpolated surface shows a maximum in correspondence of the maximum relative changes in temperature and wind speed. At the same time, the minimum has been obtained for the case of no variation in temperature and maximum negative change in wind speed.

Finally, in Fig. 7 the annual aggregated values of ET0 [mm/year] for each scenario are reported. As regards to W scenarios, increases and decreases of ET0 have been observed, since both positive and negative wind speed trends have been detected. Increases of ET0 range from +2.9 to +23.8 % for the W50 scenarios and from +5.5 to +39 % for the W100 scenarios. Negative wind speed trends observed in Capri, Cozzo Spadaro resulted in decreases of ET0, ranging from −3.0 to −11.2 % in the W50 and from −3.2 to −22.9 % in the W100. In the T scenarios ET0, increases range from +1.4 to +10.0 % for the T50, and from +2.5 to +21.0 % for the T100. As regards to WT scenarios, ET0 increases range from +1.3 to +30.0 % for the WT50, and from +5.5 to +54.2 % for the WT100, while the decreases, observed for Cozzo Spadaro and Capri, are −8.2 and −5.9 %, respectively, for the WT50 and −9.5 and −14.1 % for the WT100.

Fig. 7
figure 7

Total annual ET0 [mm] in each scenario

Full size image

It has to be remarked that this study provides results on the effects of wind speed and air temperature changes on ET0, neglecting the other variables involved in the Equation (7). Among these, solar radiation represents a very important factor for the evaluation of ET0, therefore the existence of a trend on this variable could significantly affect it (e.g., McKenney and Rosenberg 1993; Donohue et al. 2010). However, recently, Jiménez-Muñoz et al. (2012) analyzed trends in net radiation at global scale during the period 1980–2010, distinguishing between land and ocean and between different land cover classes, finding a net radiation reduction of about −1 W/m2 during the period 1980–2010 on world land, but not significant over sea. On the basis of the previous cited study, in this analysis, the investigation of solar radiation trend on ET0 has not been carried out, because of its slight magnitude. Moreover, the analyzed sites are all located in the coastal area, that according to Jiménez-Muñoz et al. (2012), can be considered as a “boundary zone” between land and sea, in terms of solar radiation trend, separating a negative trend zone (land) from a no trend zone (sea). For these reasons, with regard to the case of study here presented, neglecting solar radiation variations on ET0 did not lead to unreliable results.

4 Conclusion

This work has been aimed to investigate the impact of wind speed and temperature trends on reference evapotranspiration in the southern Mediterranean area

Significant positive wind speed trends have been detected at different time scales, in accordance with a previous literature study (Pirazzoli and Tomasin 2003); but also some negative significant tendencies have been observed. As regards to wind direction, the trend analysis showed statistically significant decreasing trends in north winds occurrences at the annual scale in 9 out of 10 measurement stations without any compensation in other directions. Temperature trends for the area of study have been detected and quantified as well. Results showed a general increase of mean temperature, in particular during the summer months, in accordance with global observations.

Once climatic trends have been evaluated, the effects of these trends on reference evapotranspiration ET0, calculated using the FAO-56 Penman–Monteith equation, have been investigated. This method allowed for the investigation on the sensitivity of ET0 to change in two different variables (wind speed and temperature).

The result showed that temperature is a key factor in determining ET0, since trends in this meteorological variable widely affect its time variation, much more than wind speed variation. Since the latter is another key variable driving ET0, the combined effect of wind speed and temperature trends on ET0 has been investigated. In order to carry out this analysis, some future scenarios have been defined starting from the magnitude of the statistically significant trends detected for the area of study.

In summary, where upward wind speed trends occurred, the effect of the wind speed rise on ET0 has been clearly amplified by the temperature increase. On the contrary, the positive trends of temperature has been counterbalanced by negative wind speed trends but still causing ET0 increase at all the analyzed time scales.

In order to have a complete view of the effects of climatic variables change on ET0, future developments of this work could involve the analysis of sensitivity of ET0 to changes in the other crucial variables, i.e., the relative humidity, the shortwave radiation and the CO2 concentration. Indeed, a full understanding and quantification of the effects of climate change on ET0 can be obtained by analyzing the contribution of all the variables involved in its evaluation.