Impact of global dimming and brightening on reference evapotranspiration in Greece

Authors


Abstract

[1] In this study, the consistency of trends in radiation and temperature records and their implications for the hydrological cycle and especially the trends in reference evapotranspiration are examined during the period 1950–2001. The new reference evapotranspiration model for complex terrains (REMCT), with monthly time step, is used for estimating trends of reference evapotranspiration in Greece. REMCT is applied after developing a methodology for calibrating its parameter values with Penman-Monteith estimates. The calibrated REMCT estimates are independently validated against available pan evaporation measurements. The evolution of available sunshine duration anomalies measured in Athens during the period 1951–2001 are used for highlighting global dimming or brightening periods in Greece. The sign of trends in the modeled reference evapotranspiration and precipitation are examined according to the dimming or brightening periods 1950–1983 and 1958–1983 (for 16 and 22 stations, respectively) or 1983–2001 (for 16, 22, and 29 stations). The trends of REMCT estimates, precipitation, number of rainy days, and mean, maximum, and minimum air temperature for all sets of stations considered “as a whole” are evaluated during the same periods. The results show that the annual calibrated reference evapotranspiration trend shows a decline from 1950 until the early 1980s, followed by an upward trend until 2001, while the annual precipitation and rainy days indicate a downward trend during the whole period 1950–2001. The trends of mean, maximum, and minimum air temperature are found almost negligible during the dimming period and rather increased during the brightening period.

1. Introduction

[2] Many studies document the changes in surface temperature over the past decades and a few studies concentrate on the surface energy balance components, which are the processes that govern the evolution of surface temperature. Additionally, much evidence from the changes of precipitation, runoff, and soil moisture suggests that the hydrological cycle has intensified in many parts of the world during the past century [Allan et al., 2003; Huntington, 2006], along with the increasing global temperature. Changes in the climate, extended to the water cycle and not limited only to changes in the energy balance, have been investigated as a key research in some studies [Wild et al., 2004; Nazarenko and Menon, 2005; Wild et al., 2008; Teuling et al., 2009]. The connecting term between water and energy balance is the actual evapotranspiration, a parameter difficult to be measured directly [e.g., Brutsaert, 1982], but estimated by indirect approaches, as to first estimate potential evaporation or reference evapotranspiration. Consequently, several studies have examined the trends in potential or reference evapotranspiration, in order to investigate the changes to latent heat (the energy equivalent to evaporation) over land.

[3] Potential evaporation is commonly measured by means of pan evaporation or it is estimated from readily available climate variables using, i.e., the equation of Penman [1948] or simpler methods based on minimum data. Reference evapotranspiration can also be computed from meteorological data by different methods [Gong et al., 2006]. The Penman-Monteith formulation [Monteith, 1981; Allen et al., 1998] is regarded, as a good reference evapotranspiration estimator, for a wide variety of climatic conditions, but a major drawback to its application is the relatively high data demand, limited in many areas of the globe. Consequently, accurate approaches of estimating reference evapotranspiration, from minimum weather measurements, only available as monthly summaries, appropriate for climate trend studies [Xu et al., 2005], can be used for increasing spatial resolution in data-sparse areas. Among them, the new reference evapotranspiration model for complex terrains (REMCT), with monthly time step, developed by Diodato and Bellocchi [2007] based on air temperature and rainfall attributes (frequently available on a regional scale) has the advantage of skipping over limitations, observed with other simplified models as the ones by Hargreaves and Samani [1985] and Droogers and Allen [2002], in complex terrains. The model has been applied successfully to a range of terrains in Italy [Diodato and Bellocchi, 2007].

[4] Contrary to the expectation that a warmer climate will increase evaporation, many studies have reported decreasing pan evaporation over the past 50 years or so, for many places of the world with widely different climates. Declines are reported in the USA and across many parts of the former Soviet Union [Peterson et al., 1995; Brutsaert and Parlange, 1998; Lawrimore and Peterson, 2000; Golubev et al., 2001; Roderick and Farquhar, 2002; Groisman et al., 2004; Hobbins et al., 2004], India [Chattopadhyay and Hulme, 1997], Venezuela [Quintana-Gomez, 1998], Japan [Asanuma and Kamimera, 2004], China [Liu et al., 2004; Liu and Zeng, 2004; Chen et al., 2005], Australia and New Zealand [Roderick and Farquhar, 2004, 2005; Roderick et al., 2007; Jovanovic et al., 2008; Hobbins et al., 2008], Thailand [Tebakari et al., 2005], the Tibetan plateau [Shenbin et al., 2006; Zhang et al., 2007], Canada [Burn and Hesch, 2007] and Greece [Papaioannou et al., 2007]. These decreases in pan evaporation were not universal, and data from a few regions indicate increases, as for Israel [Cohen et al., 2002], Northeast Brazil [da Silva, 2004], the Tibetan plateau [Xu et al., 2005] and Australia [Kirono et al., 2008]. At individual sites, both decreases and increases have commonly been observed. Like the trends in pan evaporation, trends in reference evapotranspiration estimated from readily available climate variables, have been reported negative for some regions and positive for other ones. Thus, trends in Penman-Monteith estimates have been referred negative for Israel [Cohen et al., 2002], China [Thomas, 2000; Chen et al., 2005; Wu et al., 2006; Gao et al., 2007] and the Tibetan plateau [Zhang et al., 2007] and positive in the United Kingdom [Yang et al., 2005] and in Africa [Oguntunde et al., 2006]. There are also reports on increases in estimates based on other methods, as the method of Kondo and Xu [1997] for Eastern Asia by Xu et al. [2005], the method of Morton [1983] for Australia by Kirono et al. [2008], the method of Thornthwaite [1948] for Australia and New Zealand by Hobbins et al. [2008], or decreases in open water evaporation based on the equation of Penman [1948] for Israel by Cohen et al. [2002].

[5] The interpretation of the above mentioned negative trend has been controversial, and so far agreement on its significance, regarding terrestrial evaporation has remained elusive. At first it was thought as indication of a general negative trend in terrestrial evaporation in those areas, which seemed incompatible with the reported general increases in global precipitation in the same period [Dai et al., 1997; Karl and Knight, 1998]. According to Brutsaert and Parlange [1998], this “evaporation paradox” was suggested that may indicate increasing terrestrial evaporation, because of the complementary relationship between potential and actual evaporation. This explanation was generally consistent with other evidence by Lawrimore and Peterson [2000], Golubev et al. [2001], Hobbins et al. [2004] and Ramirez et al. [2005]. On the other hand in some studies [Cohen et al., 2002; Roderick and Farquhar, 2002; Ohmura and Wild, 2002; Liu et al., 2004; Wild et al., 2004] the decreasing pan evaporation was attributed in decreasing solar irradiance (global dimming) giving evidence of decreasing landscape evaporation. Brutsaert [2006] has shown that both effects (global dimming and the complementary relationship) can be combined, whereas, Zhang et al. [2007] have referred that actual evapotranspiration and pan evaporation exhibit a complementary relationship in the Tibetan plateau, but not Bouchet's complementary relationship [Bouchet, 1963]. Recently, Teuling et al. [2009] have shown that both scenarios of decreasing actual evapotranpiration with decreasing pan evaporation in regions with ample supply of water and of increasing evapotranspiration with decreasing pan evaporation can be encompassed.

[6] However, the significant changes of solar radiation at the surface (a key determinant of surface temperature) on decadal time scales (“global dimming or brightening”) have been widely reported [Ohmura and Lang, 1989; Dutton et al., 1991; Stanhill, 1995; Stanhill and Cohen, 1997; Gilgen et al., 1998; Stanhill and Cohen, 2001; Liepert, 2002; Wild et al., 2004, 2005; Pinker et al., 2005; Wild et al., 2007; Makowski et al., 2008; Wild et al., 2008; Gilgen et al., 2009; Long et al., 2009; Liley, 2009; Makowski et al., 2009; Norris and Wild, 2009; Ohmura, 2009; Stanhill and Cohen, 2009; Wild, 2009; Wild et al., 2009; Chiacchio and Wild, 2010]. Especially, in the most recent of the referred studies, an emerging evidence for a widespread decline of solar dimming during the 1980s and reversal to a brightening thereafter is also reported. Trends in UV-A irradiances were also referred, as negative until the early 1990s, and positive during the last decade by Zerefos et al. [2009] for Thessaloniki, Greece. Furthermore, Wild et al. [2007] investigating the role of solar dimming and brightening in the context of recent global warming showed that the decadal changes of land mean, maximum and minimum surface temperature, as well as temperature range, are in line with the proposed transition in surface solar radiation from dimming to brightening during the 1980s and with the increasing greenhouse effect. The relationship between diurnal temperature range and surface solar radiation was also recently investigated for Europe [Makowski et al., 2008, 2009]. Both the global dimming and the recent brightening, still involve a remarkable degree of uncertainty in their description. Evidence of global dimming is thus far based mainly upon few measurements of solar radiation. Thus, a need for more accurate measurements and research extended on longer series of records has been urged [Stanhill and Cohen, 2005]. For this purpose, there have been analyses supported and extended with the use of other climatic variables, such as sunshine duration (SD), recorded for a longer time period and successfully used as a proxy for solar radiation over the past 80 years [Stanhill and Cohen, 2005]. A lot of studies, as for example summarized by Sanchez-Lorenzo et al. [2007], have analyzed SD variations in various regions. SD time evolution has been reported with a decrease since the 1950s until the mid-1980s, and a recovery thereafter [Sanchez-Lorenzo et al., 2007; Abakumova et al., 2008; Sanchez-Lorenzo et al., 2008; Stanhill and Cohen, 2008; Liley, 2009; Sanchez-Lorenzo et al., 2009], in line with studies investigating changes in surface solar radiation.

[7] Thus, in this study, the consistency of trends in radiation and temperature records and their implications for the hydrological cycle and especially the trends in reference evapotranspiration are examined, during the period 1950–2001. The new reference evapotranspiration model for complex terrains (REMCT) [Diodato and Bellocchi, 2007], with monthly time step, is used for estimating trends of reference evapotranspiration in Greece. REMCT is applied after developing a methodology for calibrating its parameter values with Penman-Monteith estimates. The calibrated REMCT estimates are independently validated against available pan evaporation measurements. The evolution of available sunshine duration anomalies measured in Athens during the period 1951–2001 is used for highlighting global dimming or brightening periods in Greece. Furthermore, the sign of trends in reference evapotranspiration based on REMCT are examined according to the dimming or brightening periods 1950–1983 and 1958–1983 (for 16 or 22 stations, respectively) or 1983–2001 (for 16 or 22 or 29 stations). In addition, the trends of REMCT estimates, precipitation, number of rainy days and mean, maximum and minimum air temperature for all sets of stations considered as a whole are evaluated, during the same periods.

2. Data and Methods

2.1. Data Sets

[8] Daily measurements of dry and wet bulb temperature, wind velocity and sunshine hours obtained from the National Meteorological Service for 29 stations in Greece, during the period 1974–2001, are used for estimating reference evapotranspiration by Penman-Monteith equation [Monteith, 1981]. The characteristics of the stations are shown in Table 1.

Table 1. Characteristics of the Stations and Yearly Averages of Mean (T) and Maximum (Tmax) Air Temperature, Number of Rainy Days (Rd), and Diodato-Bellocchi Evapotranspiration (ETDIOBEL) During the Period 1974–2001
StationLatitudeLongitudeAltitude (m)T (°C)Tmax (°C)Rd (month−1)ETDIOBEL (mm yr−1)
  • a

    Stations with available pan measurements.

Serresa41°05′23°34′34.015.220.57966.2
Florina40°48′21°25′4.012.117.47888.7
Mikraa40°31′22°58′4.015.820.491138.9
Kavala40°56′24°23′15.014.719.07906.6
Alexa/polia40°51′25°26′3.514.919.27913.9
Kozani40°17′21°47′625.013.117.87916.7
Kerkira39°37′19°55′2.017.621.9101119.9
Ioannina39°40′20°51′483.014.219.510940.5
Larissaa39°39′22°26′71.015.721.491181.7
Aghialosa39°13′22°48′12.216.320.881164.9
Mitilinia39°04′26°36′4.217.621.071209.4
Agrinioa38°37′21°33′24.017.023.081184.4
Lamiaa38°51′22°24′14.316.621.981194.9
Andravidaa37°55′21°17′11.117.221.881164.0
Argostoli38°11′20°29′21.218.221.381161.9
Chios38°21′26°09′4.017.520.941278.5
Pirgosa37°40′21°26′12.017.622.971139.8
Tripoli37°32′22°24′652.413.919.69979.2
Elliniko37°54′23°45′10.018.423.071332.2
Kalamataa37°04′22°06′222.017.422.591202.8
Naxos37°06′25°23′9.018.220.551316.3
Methoni36°50′21°42′53.017.921.181181.2
Milos36°44′24°26′164.017.520.761336.5
Kithira36°08′23°01′165.817.720.061295.8
Rhodes37°24′28°05′34.719.222.061278.8
Iraklioa35°20′25°11′37.018.721.781306.6
Ierapetraa35°00′25°44′10.019.723.351446.1
Sitia35°12′26°08′114.518.621.551359.9
Tymbakia35°00′24°46′5.018.823.051399.7

[9] Monthly values of mean (T) and maximum (Tmax) air temperature, precipitation (PR) and number of rainy days (Rd) available for the mentioned 29 stations, during the period 1974–2001 are used to obtain monthly reference evapotranspiration (ETDIOBEL) based on the reference evapotranspiration model for complex terrain, as developed by Diodato and Bellocchi [2007]. These meteorological parameters, available furthermore for 22 and 16 stations (evenly distributed across the country) during the longer periods 1958–2001 and 1950–2001, respectively, are also used for calculating ETDIOBEL. The selected 16 stations are shown in Table 3, while the group of 22 stations is formed, by adding Mikra, Aghialos, Andravida, Elliniko, Naxos and Tymbaki to the group of 16 stations.

[10] Daily evaporation measurements from class-A pans (EPAN), available for 14 stations (footnoted entries in Table 1) during the period 1979–1999 are used to validate the calibrated REMCT estimates. The stations are chosen on the basis of the records, being available for as long as possible.

[11] In order to contribute to the knowledge of global dimming or brightening in Greece, daily SD measured in the National Observatory of Athens (37°58′N, 23°43′E), by Campbell-Stokes heliographs, are used as a proxy to explore changes in solar radiation conditions in Greece, during the period 1951–2001.

2.2. Methods

2.2.1. Reference Evapotranspiration

[12] The reference evapotranspiration model for complex terrains, based on monthly rainfall occurrence and air temperature [Diodato and Bellocchi, 2007], is selected for estimating trends of reference evapotranspiration in Greece, (a country characterized by a complex terrain), since temperature and rainfall attributes are available for more sites or for longer time periods. REMCT includes also, the effect of climate and site elevation and incorporates a monthly function of minimum air temperature, conditioned by the precipitation status to account for seasonal shifts. Monthly values of reference evapotranspiration (ETDIOBEL) based on REMCT, are estimated by the following nonlinear equation:

equation image

where, Rso is the mean monthly extraterrestrial solar radiation (MJ m−2 d−1), λ is the latent heat of vaporization (MJ kg−1), T is the mean monthly air temperature (°C), Tmax is the mean monthly maximum air temperature (°C), Rd is the number of rainy days in a month (month−1), ref(Tmin) is a reference minimum air temperature (°C), f(clim) and f(w) are semiempirical functions (month), modulating the rate at which the energetic forcing, involved in the evapotranspirative processes (principally the term under square root in equation (1)), is approached as air temperatures and rainy days change. The reference minimum air temperature is determined as:

equation image

Where, m is the month, a, α, b and c are empirical parameters equal to: −2.64 (°C), 1.79, 6.5 and 12.0, respectively. The two semiempirical functions are determined as:

equation image

where T(a)max is the long-term yearly mean daily maximum temperature (°C), E is the site's elevation, (m), P is the long-term mean annual precipitation (m), d, e, g, h, k and q are empirical parameters equal to 0.00247, 0.000000831 (m−1), 2000 (m), 0.1 (m−1), 0.0397 (°C0.5 month−1) and 0.000079228 (°C−3). The parameter g [in f(clim) function], set equal to 2000 m, is the upper elevation for REMCT to be applied.

[13] Penman-Monteith estimates (ETPENMON) are calculated for comparisons with the obtained REMCT estimates. The daily ETPENMON used to estimate potential evapotranspiration from hypothetical reference grass with an assumed height of 0.12 m, a fixed surface resistance of 70 s m−1 and an albedo of 0.23 [Allen et al., 1998] is expressed as:

equation image

Where, Δ is the slope of saturated vapor pressure in relation to air temperature (kPa °C−1), γ is the psychrometric constant (kPa °C−1), Rn is the net radiation (MJ m−2 d−1), G is the soil heat flux density (MJ m−2 d−1), (es–ea) is the vapor pressure deficit (kPa), T is the mean daily air temperature at 2 m above ground level (°C), u2 is the wind speed at 2 m above ground level (m s−1). Daily values of G or Rn (based on sunshine hours) are estimated according to Allen et al. [1998].

[14] The obtained monthly ETDIOBEL values, based on the original parameters of REMCT, are compared to monthly ETPENMON estimates (calculated from their daily values), for each of the 29 stations for the period 1974–2001. The results from their linear regressions (determination coefficient R2 and slope a) and “difference measures” (root mean square error (RMSE), mean bias error (MBE) and index of agreement (IA)) are used for the evaluation of the model.

[15] For improving REMCT estimates, local calibration of the model parameters is employed in various ways. The best results (as compared to ETPENMON estimates) are achieved, when the new REMCT parameters are determined independently for groups of stations, and not uniquely for the whole country. Thus, the final calibration procedure is as follows. As a first step, the 29 stations of the study are grouped, according to their monthly values of T, Tmax and Rd, following cluster analysis. For each group, some stations are selected as calibration set and the rest ones (footnoted entries in Table 2) are used as validation set. As a second step, the parameter values of REMCT are basically recalculated from comparisons against mean monthly ETPENMON for the 15 sites of the calibration set. Keeping constant the parameters of ref (Tmin), the following calibration process is employed for equation (1): First, the values are determined for f(clim) and second for f(w), fitting the estimates of the depended upon Rso, λ, T, Tmax, ref(Tmin), Rd terms of equation (1), against the Penman-Monteith estimates. Next, the parameters of the functions f (clim) and f (w) are determined by multiple regression analysis.

Table 2. Results of the Linear Regressions (Determination Coefficient (R2) and Slope (a)) and Difference Measures (Root Mean Square Error (RMSE), Mean Bias Error (MBE), and Index of Agreement (IA)) Between Monthly Uncalibrated or Calibrated Estimates of Diodato-Bellocchi Evapotranspiration and Penman-Monteith Evapotranspiration for the Period 1974–2001
StationUncalibratedCalibrated
R2aRMSEMBEIAR2aRMSEMBEIA
  • a

    Stations belonging to the validation set.

Group 1
Serresa0.971.0823.118.540.970.990.9713.40−0.480.99
Florina0.981.1728.1017.940.960.990.9812.670.010.99
Kavala0.991.0515.683.830.990.990.9113.29−7.890.99
Alexa/polia0.990.9810.38−2.990.990.990.8519.53−14.050.97
Kozania0.981.1726.7916.990.970.990.9511.85−3.580.99
Tripoli0.990.9415.29−7.310.990.980.9615.91−0.950.98
Ioannina0.970.9220.68−8.650.980.990.8718.22−11.130.98
 
Group 2
Mikra0.980.9117.88−11.280.980.990.9610.62−2.590.99
Larissa0.980.9718.61−4.580.980.991.0612.899.650.99
Kerkira0.960.7633.18−29.390.920.991.018.921.860.99
Agrinioa0.950.8628.15−18.380.950.991.1117.6314.190.98
Ahgialosa0.980.9515.39−7.420.990.991.029.012.970.99
Lamiaa0.970.8921.29−14.100.970.991.009.431.860.99
Andravidaa0.970.8721.86−17.440.960.991.027.993.220.99
Kalamata0.960.8426.61−21.760.950.991.008.732.060.99
 
Group 3
Mitilini0.980.8521.97−18.730.960.990.9412.09−6.860.99
Argostolia0.980.8423.01−19.760.950.990.9810.18−3.590.99
Chiosa0.990.9215.47−8.810.980.990.9712.40−1.540.99
Ellinikoa0.980.8721.08−17.300.970.990.9312.27−8.060.99
Naxosa0.980.9914.56−5.150.980.991.0211.070.080.99
Methonia0.970.9020.78−16.060.960.991.0311.43−0.560.99
Milosa0.980.9714.72−6.480.980.991.0211.140.360.99
Kithira0.990.9414.06−9.370.980.991.009.71−1.010.99
Sitia0.980.9614.13−7.510.980.991.029.761.350.99
Rhodes0.980.8423.01−19.520.950.990.9411.91−6.080.99
Iraklio0.960.8524.40−20.700.940.990.9313.81−10.200.98
Ierapetraa0.980.7927.38−21.910.930.980.8322.53−14.660.95
Tymbaki0.980.9118.49−12.140.970.990.9712.57−2.980.99
Pirgos0.970.8325.30−21.580.950.991.0110.46−0.070.99

[16] These new parameters are used in the REMCT for each group, for estimating monthly ETDIOBEL (calibrated) values. The improvement in the calibrated ETDIOBEL is tested on the sites of the validation set, which have not been used in the calibration procedure, by using linear regressions and difference measures with ETPENMON estimates. Since pan measurements have been reported usually as direct measurement of reference evaporation, monthly values of calibrated ETDIOBEL are furthermore independently validated, by the results of linear regressions with EPAN available for limited number of stations and time period.

2.2.2. Time Evolution and Trends in Sunshine Duration, Reference Evapotranspiration, Precipitation, Rainy Days, and Temperature

[17] In order to examine the consistency of trends in radiation and reference evapotranspiration, according to dimming/brightening periods over Greece, the trends of calibrated REMCT estimates and SD measurements (as a proxy of solar radiation) in Athens are mainly examined for the longest available periods. For this purpose, the anomalies of either the annual calibrated REMCT estimates for all stations considered as a whole, or the annual values of SD in Athens, are calculated as relative deviations from their corresponding 1971–2000 mean [Sanchez-Lorenzo et al., 2007], during the periods 1950–2001 or 1951–2001, respectively. The analysis is furthermore accomplished by applying the best fits, from first order to third order, to the yearly time series of the mentioned anomalies, following the technique referred by Chiacchio and Wild [2010]. The latest is also applied on the yearly Precipitation, ETDIOBEL and limited EPAN data.

[18] The trends in all considered time series are examined as statistically significant, by the nonparametric Mann-Kendall test, recommended by the World Meteorological Organization [Sneyers, 1990]. Having established their significance, the magnitude of the time trends for the highlighted dimming or brightening periods in Greece is determined, using standard parametric regression methods.

[19] According to the available data, the trends and relative trends of ETDIOBEL and PR are calculated during the periods 1950–1983 and 1958–1983 (for 16 or 22 stations, respectively) or 1983–2001 (for 16 or 22 or 29 stations) or 1950–2001 and 1958–2001 (for 16 or 22 stations, respectively). Trends of ETDIOBEL, PR, Rd, T, Tmax and Tmin for all stations considered as a whole are also evaluated during the same periods.

3. Results

3.1. Calibration and Evaluation of Reference Evapotranspiration by REMCT

[20] The results of the linear regressions (R2, a) and difference measures (root mean square error (RMSE), mean bias error (MBE) and index of agreement (IA)) between the monthly ETDIOBEL (uncalibrated) values, obtained from the original parameters of REMCT and the monthly ETPENMON estimates (calculated from their daily values), are presented in Table 2, for each of the 29 stations for the period 1974–2001. It is evident that the MBEs are greater than 10% in a lot of stations and the RMSEs are quite high, although the indices of agreement are excellent (ranging from 0.93 to 0.99). Thus, the model is calibrated, employing the steps of the described calibration process. The results from the grouping of the stations are apparent in Table 2. Considering only data from the calibration stations of each group, the recalculated REMCT parameters d, e, h, k and q, are: (d = 1.342E-03, e = 4.770E-08 m−1, h = 2.761 m−1, k = 5.620E-03°C0,5 month−1, q = 3.160E-06°C−3) or (d = 1.766E-03, e = 2.000E-07 m−1, h = 1.512E-01 m−1, k = 1.770E-02°C0,5 month−1, q = 1.360E-05°C−3) and (d = 2.158E-03, e = 4.600E-07 m−1, h = 4.236E-01 m−1, k = 2.619E-02°C0,5 month−1, q = 3.880E-05°C−3) for group 1, group 2 or group 3, respectively. These new parameters are used in the REMCT for each group, for estimating monthly ETDIOBEL (calibrated) values. The results of the linear regressions and difference measures between the mean monthly ETDIOBEL (calibrated) values and ETPENMON estimates, are presented in Table 2 (calibrated). It is evident that the excellent indices of agreement (already referred for the uncalibrated ETDIOBEL values) are improved (ranging from 0.95 to 0.99), the MBEs become smaller than 10% in 25 stations and the RMSEs are mainly reduced ranging from 9% to 14% for 24 stations.

[21] The improvement in the calibrated ETDIOBEL estimates is evaluated, when considering especially the sites kept as validation set (footnoted entries in Table 2). All results are generally improved with MBEs being smaller than 10% for 11 sites and only up to 14% for 3 sites. RMSEs are also reduced remaining up to 17–22% only for 3 stations. This is also shown from the linear regressions between ETDIOBEL (uncalibrated) or ETDIOBEL (calibrated) and ETPENMON, apparent in Figures 1a and 1b, respectively, for the stations kept as validation set. Figure 1b shows a good agreement between ETDIOBEL (calibrated) and ETPENMON values, as most points tend to line up around the 1:1 line. It must be noticed though, that a spread is apparent at high values of ETPENMON > 7 mm d−1, mainly caused by REMCT underestimations for Ierapetra, a site characterized by high aerodynamic trends.

Figure 1.

Monthly (a) uncalibrated and (b) calibrated Diodato-Bellocchi evaportranspiration (ETDIOBEL) as a function of monthly Penman-Monteith evaportranspiration (ETPENMON) for the validation set of stations (marked with two asterisks in Table 2) for the period 1974–2001. The solid line is the 1:1 line. (c) Monthly calibrated ETDIOBEL as a function of monthly Pan evaporation (EPAN) for the stations marked with an asterisk in Table 1 for the period 1979–1999.

[22] In addition, an independent validation of the calibrated REMCT estimates is achieved by their correlation with available pan measurements (EPAN). The correlation between all monthly ETDIOBEL (calibrated) estimates and EPAN measurements is presented in Figure 1c for 14 stations, during the period 1979–1999. The results of their linear regression show a very good agreement with a high determination coefficient (R2 = 0.865) (at 99% significant level). The slope is equal to 0.611 ± 0.004.

[23] The average yearly ETDIOBEL (calibrated), T, Tmax and Rd calculated for 29 stations, during the period 1974–2001, are presented in Table 1.

3.2. Time Evolution and Trends in Sunshine Duration, Reference Evapotranspiration, Precipitation, Rainy Days, and Temperature

[24] In order to investigate the reference evapotranspiration time evolution over Greece, the calibrated ETDIOBEL is furthermore calculated for 16 stations (shown in Table 3), during the longest period 1950–2001. The annual time evolution of ETDIOBEL and PR for the 16 sites (considered as a whole) is apparent in Figure 2 for the period 1950–2001. The solid lines represent a third order or a linear trend line as the best fits to ETDIOBEL or PR data (R2 = 0.32 or R2 = 0.40), respectively. The third-order fit shows a decline in ETDIOBEL from 1950 until the early 1980s. This is followed by an upward trend through the rest of the time series until 2001. On the contrary, annual precipitation over Greece exhibits a downward trend during the whole period.

Figure 2.

Annual time evolution of ETDIOBEL and precipitation (PR) for 16 stations considered “as a whole” during the 1950–2001 period. The solid lines are the best fit trend lines.

Table 3. Trends and Relative Trends in Precipitation (PR) and Diodato-Bellocchi Evapotranspiration (ETDIOBEL) During the 1950–1983 and 1983–2001 Periods for 16 and 29 Stationsa
StationPR Trend (1950–1983)ETDIOBEL Trend (1950–1983)PR Trend (1983–2001)ETDIOBEL Trend (1983–2001)
mm/yr2%/yrmm/yr2%/yrmm/yr2%/yrmm/yr2%/yr
  • a

    Statistically significant trends are noted in bold.

Florina−1.87 ± 2.30−0.27−0.70 ± 0.39−0.08−0.52 ± 6.84−0.091.19 ± 1.530.21
Kavala−8.79 ± 4.54−1.093.55 ± 1.03−0.380.62 ± 4.170.153.19 ± 0.980.35
Alex/poli−2.39 ± 1.90−0.401.04 ± 0.36−0.112.27 ± 4.900.492.81 ± 1.050.30
Kozani0.82 ± 2.000.151.25 ± 0.45−0.14−4.50 ± 3.68−1.113.70 ± 1.460.40
Kerkira9.70 ± 3.65−0.80−0.69 ± 0.43−0.065.75 ± 7.650.611.46 ± 1.300.13
Ioannina−6.01 ± 3.31−0.521.25 ± 0.41−0.13−5.83 ± 8.41−0.621.40 ± 1.120.15
Larissa−2.51 ± 2.25−0.562.45 ± 0.59−0.21−1.60 ± 3.20−0.414.74 ± 1.610.40
Mitilini1.31 ± 2.190.19−0.47 ± 0.42−0.047.88 ± 7.001.364.45 ± 1.740.36
Agrinio0.50 ± 3.960.052.39 ± 0.50−0.20−4.66 ± 8.54−0.594.04 ± 1.330.34
Argostoli10.08 ± 3.47−1.021.64 ± 0.63−0.151.05 ± 7.980.152.06 ± 1.400.17
Tripoli−0.17 ± 2.82−0.021.94 ± 0.44−0.20−8.67 ± 6.81−1.24−0.69 ± 1.75−0.07
Methoni6.11 ± 2.43−0.83−0.70 ± 0.53−0.06−4.77 ± 6.69−0.731.81 ± 1.310.15
Milos−1.12 ± 1.88−0.261.24 ± 0.60−0.101.10 ± 3.880.286.85 ± 1.650.50
Kithira−2.71 ± 2.04−0.50−0.49 ± 0.53−0.04−6.53 ± 6.31−1.303.70 ± 1.460.28
Rhodes8.36 ± 3.14−1.101.61 ± 0.53−0.132.92 ± 8.800.452.06 ± 1.450.16
Iraklio1.94 ± 2.180.39−1.05 ± 0.61−0.080.91 ± 4.860.203.97 ± 1.550.30
Mikra    −2.83 ± 4.20−0.694.74 ± 1.450.41
Aghialos    −1.55 ± 3.88−0.364.18 ± 1.300.36
Andravida    −5.66 ± 6.97−0.802.03 ± 1.170.17
Elliniko    −2.85 ± 4.50−0.807.25 ± 1.770.54
Naxos    −0.40 ± 3.58−0.123.43 ± 1.140.26
Tymbaki    −8.28 ± 5.44−1.922.83 ± 1.490.20
Serres    2.38 ± 3.980.553.28 ± 1.100.34
Lamia    3.32 ± 5.380.614.34 ± 1.250.36
Chios    1.68 ± 6.200.342.43 ± 1.710.19
Pirgos    −0.68 ± 7.57−0.083.23 ± 1.360.28
Kalamata    0.43 ± 8.520.062.87 ± 1.100.24
Ierapetra    −10.12 ± 5.11−2.555.54 ± 1.560.38
Sitia    −10.27 ± 5.20−2.231.46 ± 1.600.11

[25] Annual time evolution of anomalies in ETDIOBEL (considered as a whole for 16 sites) and SD in Athens is shown in Figures 3 and 4 for the periods 1950–2001 and 1951–2001, respectively. Both series, calculated as relative deviations from their corresponding 1971–2000 mean are best fitted to a third-order equation (R2 = 0.31 and R2 = 0.24, respectively). A negative trend lasting some decades, up to minimum of the anomalies in either ETDIOBEL or SD at the early up to the middle 1980s and then a partial recovery until the end of the analyzed period are shown, following the globally detected solar dimming/brightening periods.

Figure 3.

Annual time evolution of anomalies of ETDIOBEL for 16 stations considered as a whole during the 1950–2001 period. Anomalies are expressed as relative deviations (%) from the 1971–2001 mean. The solid line is the best fit trend line.

Figure 4.

Annual time evolution of anomalies of sunshine duration (SD) in Athens, during the 1951–2001 period. Anomalies are expressed as relative deviations (%) from the 1971–2001 mean. The solid line is the best fit trend line.

[26] According to the nonparametric Mann-Kendall test, the ETDIOBEL and ETDIOBEL anomaly trends are found significantly either decreasing (at 99% level), for the period 1950–1983 or increasing (at 95% level), for the period 1983–2001. SD anomalies also show statistically significant decrease (at 99% level) or increase (at 90% level), during the period 1951–1983 or 1983–2001, respectively. It must be noticed that trends in pan measurements or their anomaly, investigated for the 14 stations considered as a whole, are found decreasing or increasing for the periods 1979–1983 or 1983–1999 but not significantly.

[27] Furthermore, the signs of trends in the modeled ETDIOBEL and PR are estimated for every available station, according to the indicated global dimming or brightening periods. Table 3 shows the trends and relative trends in PR and ETDIOBEL, during the 1950–1983 and 1983–2001 periods, for 16 or 29 stations, respectively. Statistically significant trends are noted as bold. The results from the analysis applied to time series 1950–1983 (16 stations) highlight a dominant decrease in ETDIOBEL. The most significant trends of ETDIOBEL are detected in Kavala and Larissa with decreases of the order −0.38% and −0.21% per year, respectively. PR trends for the same period show mainly decreases (significant for four sites) and only four (not significant) increases. When the 1983–2001 period is considered, the available stations (29) exhibit positive ETDIOBEL trends in almost all stations (except of Tripoli), ranging from 0.11% in Sitia to 0.54% in Elliniko, per year. These trends are statistically significant for most stations and verify (for individual sites) a reversal (from decrease to increase) of the ETDIOBEL trend, which is detected in the early 1980s. PR trends for 6 out of the 16 stations, already examined during the period 1950–1983, seem to change their direction to positive (but not significantly and with great standard errors noticed in their obtained slopes). Totally, PR trends are negative in most stations (17 out of 29), but none of them is significant.

[28] Table 4 shows the number of stations with positive or negative slopes of the trends in ETDIOBEL (bETDIOBEL) and Precipitation (bPR) for the periods 1950–1983 or 1958–1983 (16 or 22 stations, respectively), or 1983–2001 (16 or 22 or 29 stations) and 1950–2001 or 1958–2001 (16 or 22 stations, respectively). The number of stations with statistically significant trends is shown in parenthesis. It is obvious that almost all stations (16 out of 16, or 20 out of 22) exhibit decreasing evapotranspiration trends, during the dimming period 1950–1983 or 1958–1983 and increasing (15 out of 16, 21 out of 22 or 28 out of 29) during the brightening period (1983–2001). The trends are significant, in more than half of the considered stations either during the dimming or the brightening period. It must be noticed, that during the longest periods 1950–2001 or 1958–2001, a tendency for positive evapotranspiration trends is indicated by most of the stations (10 out of 16, or 17 out of 22, respectively), whereas precipitation trends are decreasing in almost all stations (16 out of 16 and 21 out of 22). The number of stations with negative trends in precipitation is reduced to 12 or 15, during the period 1950–1983 or 1958–1983, respectively and is furthermore reduced to 8 or 14, during the period 1983–2001.

Table 4. Number of Stations With Positive or Negative Slopes of the Trends of Diodato-Bellocchi Evapotranspiration (bETDIOBEL) and Precipitation (bPR) for 16, 22, and 29 stations for the Periods Used in This Studya
PeriodbETDIOBELbPR
++
  • a

    The number of stations with statistically significant trends is shown in parenthesis.

16 Stations
1950–1983016 (10)4 (0)12 (4)
1983–200115 (9)1 (0)8 (0)8 (0)
1950–200110 (4)6 (1)016 (12)
 
22 Stations
1958–19832 (0)20 (15)7 (1)15 (4)
1983–200121 (13)1 (0)8 (0)14 (0)
1958–200117 (7)5 (1)1 (0)21 (11)
 
29 Stations
1983–200128 (18)1 (0)12 (0)17 (0)

[29] Finally, the trends of ETDIOBEL, PR, Rd, T, Tmax, Tmin are shown in Table 5, when considering the stations as a whole, for every period. The slopes of these trends (bETDIOBEL, bPR, bRd, bT, bTmax, bTmin) are evaluated during the periods 1950–1983, 1983–2001, 1950–2001 (16 stations) or 1958–2001, 1983–2001, 1958–2001 (22 stations) or 1983–2001 (29 stations). The statistically significant trends are shown as bold. The slopes of ETDIOBEL trends are statistically significant either negative during the periods 1950–1983 (or 1958–1983) for 16 (or 22) stations, respectively or positive during the period 1983–2001 for both analyses (16 or 22 stations). ETDIOBEL trend for 29 stations considered as a whole is equal to 3.48 (or 4.86) mm yr−2 during 1983–2001 (or 1992–2000) and 2.94 mm yr−2 during 1986–2000 (after discarding the Pinatubo-affected years 1991–1993). These results indicate evidence for the effect of global dimming/brightening periods in reference evapotranspiration, reassuring the results already obtained from the analysis referred to each station (Table 4). Precipitation and rainy days exhibit (statistically significant) negative trends, during the longest periods 1950–2001 or 1958–2001. PR and Rd trends are negative during the dimming period (statistically significant during 1950–1983, but not significant during 1958–1983). As the latest period 1983–2001 is considered, PR and Rd trends remain negative (but not significant), with their slopes rather weakening, for all sets of stations (16 or 22 or 29). On the contrary, bT, bTmax and bTmin are almost zero, during the periods 1950–1983 (or 1958–1983), for 16 (or 22) stations. They are significantly positive, during the period 1983–2001 for 16 or 22 or 29 stations. When considering 16 (or 22) stations as a whole for the longest periods 1950–2001 (or 1958–2001), bT, bTmax and bTmin are almost zero (except of bTmin being weakly positive during 1958–2001).

Table 5. Slopes of the Trends of Diodato-Bellocchi Evapotranspiration (bETDIOBEL), Precipitation (bPR), Number of Rainy Days (bRd), and Mean (bT), Maximum (bTmax), and Minimum (bTmin) Air Temperature for 16, 22, and 29 Stations Considered as a Whole, for the Periods Used in This Studya
PeriodbETDIOBEL (mm/yr2)bPR (mm/yr2)bRd (month yr)−1bT (°C/yr)bTmax (°C/yr)bTmin (°C/yr)
  • a

    The statistically significant trends are shown in bold.

16 Stations
1950–19831.40 ± 0.373.45 ± 1.580.42 ± 0.140.00 ± 0.010.00 ± 0.010.00 ± 0.01
1983–20012.94 ± 1.23−0.91 ± 3.59−0.09 ± 0.420.05 ± 0.020.04 ± 0.020.05 ± 0.01
1950–20010.13 ± 0.274.80 ± 0.830.47 ± 0.080.00 ± 0.010.00 ± 0.010.00 ± 0.01
 
22 Stations
1958–19831.91 ± 0.51−2.38 ± 1.94−0.38 ± 0.180.00 ± 0.010.00 ± 0.010.00 ± 0.01
1983–20013.21 ± 1.27−1.36 ± 3.46−0.07 ± 0.420.05 ± 0.020.05 ± 0.020.05 ± 0.01
1958–20010.56 ± 0.354.59 ± 0.940.48 ± 0.100.00 ± 0.010.00 ± 0.010.01 ± 0.01
 
29 Stations
1983–20013.48 ± 1.18−1.82 ± 3.64−0.10 ± 0.410.05 ± 0.020.05 ± 0.020.05 ± 0.01

4. Discussion and Conclusions

4.1. Calibration and Evaluation of Reference Evapotranspiration by REMCT

[30] The reference evapotranspiration model for complex terrains has been applied successfully to estimate reference evapotranspiration in Greece. The obtained uncalibrated ETDIOBEL values (based on the original REMCT parameters) compared to ETPENMON values present indices of agreement very high but MBEs greater than 10% (in a lot of stations) and RMSEs quite high. This may be due to the fact that specific geographical locations in Greece may have particular characteristics that cause the parameters to deviate from reference values (determined as generic optimized parameters estimated over multiple years of data from sites in Italy) and they require local calibration. After calibrating locally the parameters of the model, with monthly Penman-Monteith estimates in the calibration set of sites (according to the developed methodology), the accuracy of the predictions of the model is improved. The indices of agreement become even better, the MBEs are reduced (becoming smaller than 10% in most sites) and the RMSEs become smaller (ranging from 8 to 14% in most stations). It must be mentioned though, that a high RMSE remains apparent in Ierapetra, a site characterized by very high values of the aerodynamic term, which rather could not be approximated by the model. Generally, the performance of the calibrated REMCT model over Greece seems to be very high taking into account the results from the validation set of sites which have not been used in the calibration procedure. This indicates that rainy days number in a month (Rd in equation (1)) is used rather successfully as an air humidity proxy in the model, modulated by a function of the long-term maximum air temperature (f(w) in equation (1)). In addition, the function f(clim) helps matching the reference evapotranspiration fluctuations for different combinations of precipitation (proxy of air humidity) and elevation (proxy of solar transmittance). The dependence of reference evapotranspiration on air temperature, seems also successfully limited in the model, by daytime air temperatures and thus roughly matched by maximum air temperatures (Tmax), since air temperature lapse rate is usually greater for Tmax than for Tmin, and as a result, smaller values of estimated radiation transmittance and reference evapotranspiration at higher elevations are obtained [Thornton et al., 1997; Diodato and Bellocchi, 2007].

[31] The locally calibrated REMCT estimates are independently validated from measured values of water loss. The correlation between the monthly values of calibrated ETDIOBEL and the available pan measurements is found very high (R2 = 0.865). Thus, REMCT based on minimum data, is not limited only to Italian sites [Diodato and Bellocchi, 2007], but is also shown as suitable to estimate accurately the monthly mean reference evapotranspiration on heterogeneous surfaces in Greece and hence to investigate its time evolution for longer time periods over the country.

4.2. Time Evolution and Trends in Sunshine Duration, Reference Evapotranspiration, Precipitation, Rainy Days, and Temperature

[32] The analysis of the temporal evolution of sunshine duration anomalies over Athens (Figure 4) indicates a decrease since the 1950s until the early 1980s, followed by a positive trend, up to the end of the twentieth century. These results match generally the globally detected global dimming/brightening periods and they are in agreement with SD trends reported for the Iberian Peninsula [Sanchez-Lorenzo et al., 2007] and the Western Europe [Sanchez-Lorenzo et al., 2008]. These SD evolutions are in accordance with the studies investigating surface solar radiation changes over the Western Europe [Sanchez-Lorenzo et al., 2008] or Europe [Norris and Wild, 2007; Stjern et al., 2009; Philipona et al., 2009; Chiacchio and Wild, 2010]. Decreasing SD until the mid-1980s and recovering thereafter, reflecting the concurrent decadal surface solar radiation changes measured at the same time, was as well referred for Moscow [Abakumova et al., 2008]. Decreasing surface solar radiation variation was also determined in eastern Mediterranean and especially in Turkey between 1960 and 1994 [Aksoy, 1997], in Israel between 1954 and 1994 [Stanhill and Ianetz, 1997], in Egypt between 1968 and 1994 [Omran, 2000].

[33] The potential causes for the annual mean evolution of SD trends for Greece during 1951–2001 may be generally related to the changes in the transmissivity of the atmosphere or modifications to the aerosol loadings in Europe over this period [Norris and Wild, 2007]. A possible connection between the decreasing surface solar radiation trends and changes in aerosols has first pointed by Stanhill and Moreshet [1992]. Recently, an increasing number of studies suggest that changes in aerosol concentrations have contributed to the observed dimming and the subsequent brightening around the globe [Wild et al., 2005; Streets et al., 2006; Ohmura, 2006; Ruckstuhl et al., 2008; Ohvril et al., 2009; Vautard et al., 2009; Hatzianastassiou et al., 2009]. Especially in Europe, a significant part of the brightening is reported as related to the reduction of local aerosol amounts due to air quality control measures and stringent measures taken all over Europe and the deindustrialization in eastern European countries. Norris and Wild [2007] found that the reversal (from decrease to increase) of the trend in surface solar radiation in Europe (after removing the effect of cloud cover changes) was more likely due to changes in aerosol direct and indirect effects, in line with the trend in aerosol emissions reaching its maximum in the 1980s. Sanchez-Lorenzo et al. [2009] as well, using long-term SD data for the Iberian Peninsula found the dimming and subsequent brightening periods even in records that only contained SD of cloud-free days, indicating the contribution of changes in aerosol concentrations to these trends.

[34] ETDIOBEL anomalies also indicate a negative trend, especially after late 1960s until a minimum at the early 1980s and then a partial recovery until the end of the analyzed period (Figure 3). There is roughly a general correspondence with the annual time evolution of SD anomalies (Figure 4). It must be referred that the curvilinear annual time evolution of global radiation during the last half century may provide an explanation for the superior fit of a higher order than linear trend line to time trends in evaporation [Stanhill and Möller, 2008]. This may be attributed to the changes in the available radiative energy at the surface, during the dimming and brightening periods. Thus, during the dimming period, the greenhouse gas induced enhancement of the downward thermal radiation has not exceeded the decreasing surface solar radiation, resulting in a reduction of surface net radiation [Wild et al., 2004]. This reduction has to be balanced predominantly by the changes in the turbulent fluxes of the surface energy balance (changes in ground heat flux and melt are small). The majority of the turbulent energy transport over land is due to latent heat (the energy equivalent to evaporation) and as a consequence, reduced evaporative cooling at the land surface was observed, consistent with reported decreasing pan evaporation for many regions throughout the world with widely different climates. In contrast, during 1986–2000, increased latent heat flux was estimated in accordance with increased surface net radiation caused by the concurrent increased downward solar radiation (because of a more transparent atmosphere) and increased thermal greenhouse effect [Wild et al., 2008]. This increasing trend is also supported by pan evaporation measurements especially in energy-limited environments, which partially indicate, after decades of decrease [Roderick and Farquhar, 2002; Ohmura and Wild, 2002], a recent recovery, in line with changes in surface solar radiation [Liu et al., 2004].

[35] The analysis for the sign of reference evapotranspiration trend for every station (Tables 3 and 4) shows that almost all stations exhibit negative trends during the dimming period 1950–1983 (or 1958–1983) and positive trends during the brightening period 1983–2001. Furthermore, the ETDIOBEL trends for 16 (or 22) stations considered as a whole are also significantly negative equal to −1.40 (or −1.91) mm yr−2, during the dimming period and significantly positive equal to 2.94 (or 3.21) mm yr−2, during the brightening period (Table 5). The trend in annual evapotranspiration for 29 stations as a whole is 3.48 mm yr−2, for the period 1983–2001. The trend estimated as 2.94 mm yr−2, when the period 1986–2000 is considered, after discarding the Pinatubo-affected years 1991–1993, and equal to 4.86 mm yr−2 for the period 1992–2000 seems to be reasonable as compared with an increase in annual global evapotranspiration by 2.4 and 6.6 mm yr−1 reported by Wild et al. [2008], for the same periods. The overall ETDIOBEL trend for the longest period 1950–2001 (or 1958–2001) is positive in most of the stations (Table 3 and 4) or for 16 or 22 stations, considered as a whole (Table 5). This is generally in agreement with positive trends in reference evapotranspiration reported by Yang et al. [2005] and Oguntunde et al. [2006] or Xu et al. [2005] or Kirono et al. [2008] or Hobbins et al. [2008], based on Penman-Monteith or Kondo and Xu or Morton or Thornthwaite method, respectively, and in disagreement with reported negative trends in Penman-Monteith estimates [Thomas, 2000; Cohen et al., 2002; Chen et al., 2005; Wu et al., 2006]. The tendency for even more positive trends is indicated, when the solar brightening period is considered and this increase is consistent with the increased surface radiative heating [Wild et al., 2004, 2007, 2009].

[36] Although annual calibrated ETDIOBEL trend indicates a decline from 1950 until the early 1980s, followed by an upward trend until 2001, the annual precipitation over Greece shows a downward trend during the whole period 1950–2001 (Figure 2). This is in accordance with similar findings for precipitation in Greece indicating a significant negative trend in winter for the period 1958–2000 [Maheras et al., 2004] or a downward trend in winter and annual precipitation for the period 1955–2001 [Feidas et al., 2007]. Precipitation across Greece was also reported as changing toward drier conditions during the last two decades of the period 1951–1990 [Kutiel et al., 1996] or decreasing during winter [Xoplaki et al., 2000] with similar results obtained for the neighboring Mediterranean coast of Turkey [Kadioğlou, 2000]. Precipitation trends in different Mediterranean regions were attributed to temporal evolution of particular large-scale circulation modes [Dünkeloh and Jacobeit, 2003]. Furthermore, the precipitation decline over the Mediterranean region during the last decades of the past century was combined with the positive trend in the East Atlantic Western Russia pattern, which was induced by the positive trend of the North Atlantic Oscillation [Krichak and Alpert, 2005]. Statistically significant negative trend of winter cloudiness was also found for the Mediterranean region, associated with the corresponding positive trends in North Atlantic Oscillation and European blocking activity [Lolis, 2009]. Maheras and Kutiel [1999] interpreted the spatial distribution of the positive and negative precipitation anomalies in terms of mean circulation over the Mediterranean. The relations between atmospheric circulation and precipitation conditions were also investigated for the eastern Mediterranean [Kutiel et al., 1996] and Greece [Metaxas et al., 1993; Xoplaki et al., 2000; Maheras et al., 2004]. The downward trend in winter and annual precipitation in Greece was also linked to the rising trend in the hemispheric circulation modes of North Atlantic Oscillation [Feidas et al., 2007].

[37] During the longer period 1950–2001 (or 1958–2001), precipitation seems decreasing in almost all independent stations (Table 4) and additionally, precipitation and rainy days exhibit significant negative trends, when the stations are considered as a whole (Table 5). This is in accordance with general decrease of rainy days and negative trend of rainfall during the period 1958–2000, reported as consistent with the negative trend of the cyclonic circulations types in Greece, [Maheras et al., 2004]. It must be noticed though, that during the brightening period the number of stations with negative trends is less (Table 4) and precipitation and rainy days for the stations considered as a whole, exhibit (insignificant) negative trends, with their slopes being weaker (Table 5). These results are in disagreement to the referred increase in global land precipitation over the period 1986–2000 [Wild et al., 2008].

[38] The trends of mean, maximum and minimum air temperature are found almost negligible during the dimming period and rather increased (0.05 ± 0.02, 0.05 ± 0.02, and 0.05 ± 0.01°C yr−1, respectively), during the brightening period, considering the stations as a whole (Table 5). These findings for the consistency of trends in radiation and temperature records and their implications for the reference evapotranspiration, are consistent with the ones reported by Wild et al. [2007] for a rapid mean or maximum or minimum temperature rise over land surfaces since the mid-1980s (0.38, 0.37, and 0.40°C/decade, respectively). They suggested that solar dimming was effective in masking greenhouse warming, but only up to the 1980s, when dimming gradually transformed into brightening and since then, the uncovered greenhouse effect is apparent with a rapid temperature rise over land since mid-1980s.

Acknowledgments

[39] We thank the editor and the Associate Editor M. Wild and the two anonymous reviewers for their critical comments that helped to improve the paper. We also thank N. Diodato for his advice. We would also wish to express our thanks to the National Meteorological Service of Greece and the National Observatory of Athens for providing the data. Gianna Kitsara is awarded a scholarship from the operational program Education and lifelong learning investing in knowledge society (IRAKLITOS II) which is cofunded by the European Union and Greece.

Advertisement