By continuing to browse this site you agree to us using cookies as described in About Cookies
Notice: Wiley Online Library will be unavailable on Saturday 7th Oct from 03.00 EDT / 08:00 BST / 12:30 IST / 15.00 SGT to 08.00 EDT / 13.00 BST / 17:30 IST / 20.00 SGT and Sunday 8th Oct from 03.00 EDT / 08:00 BST / 12:30 IST / 15.00 SGT to 06.00 EDT / 11.00 BST / 15:30 IST / 18.00 SGT for essential maintenance. Apologies for the inconvenience.
G. Stanhill, Department of Environmental Physics and Irrigation, Institute of Soil, Water and Environmental Sciences, Agricultural Research Organization, The Volcani Center, Bet Dagan, Israel. E-mail: Gerald@agri.gov.il
Radiative forcing (RF) at the tropopause is the metric adopted by the Intergovernmental Panel on Climate Change as the forcing factor determining changes in the mean global temperature at the Earth's surface, the level at which the many climate-dependent processes of interest to man to take place. At the Earth's surface, RF is subject to major regional, annual and seasonal variations due to changes in the concentration of water vapour in the troposphere, the gas most strongly influencing the downwelling long-wave irradiance. Additionally, changes in cloud cover and aerosol loads within the troposphere influence the short-wave irradiance reaching the Earth's surface.
For this reason, the committee of the US National Research Council charged with examining the relevance of the concept of RF to climate change studies stressed the need to consider the vertical as well as regional variations in this metric (NRC, 2005).
One such study of the influence of RF at the surface at a site in the temperate, marine climate of Northern Ireland, demonstrated that it accounted for three quarters of the temperature changes observed over a 120 year period, although a large and unexplained difference was found in the climate sensitivities of long as compared with short-wave forcing (Stanhill, 2011).
In this study, the effect of RF at the surface on temperature changes measured during the last 45 years was examined at a site in the very different, semi-arid climate of Israel's central coastal plain to explore the importance of local factors, including rainfall, in determining climate sensitivity.
2. Measurements and methods
Mean monthly values of global short-wave radiation Eg↓, maximum and minimum air temperatures T, relative humidity RH, atmospheric pressure Pa and rainfall measured at Bet Dagan [32°00′N, 34°49′E, 30 m MSL (mean sea level)], the Israel Meteorological Services Observatory in the central coastal plain during the 1964 to 2010 period were used in this study.
Values of RH were derived from eight daily measurements made at three hourly intervals. Those of Pa were taken from the NCEP reanalysis series adjusted for station height; comparison with 5 months of direct measurements showed agreement averaging 0.2 Pa with a maximum difference of 0.4 Pa.
The statistical significance of trends in the measured changes in Eg↓, T and specific humidity q, and rainfall was determined by the non-parametric Mann–Kendall test after pre-whitening to remove the effect of autocorrelation commonly occurring in climate time series (von Storch, 1999; Jhajharia and Singh, 2011). Standard parametric methods of linear regression were used to calculate the parameters of the trends.
Measured mean monthly values of Eg↓ were not available for 15 months; this missing data was replaced by estimates based on measured values of sunshine duration using the Angstrom–Prescott relationship derived from the 356 months when both sunshine duration and global radiation measurements were available. The linear relationship calculated was Eg↓/Eg0↓ = 0.617 SD/SD0 + 0.133, R2 = 0.668, P < 0.001, where Eg0↓ represents values of solar irradiance at the top of the atmosphere, SD sunshine duration and SD0 the potential sunshine duration, i.e. day length, for individual monthly values.
Direct measurements of atmospheric long-wave radiation El↓ were not available from Bet Dagan: in their absence values were calculated in two stages. In the first, the effect of the atmosphere's water content, at contemporary levels of the other radiatively active gases, was estimated using monthly mean values of specific humidity q, calculated from the mean air temperature, relative humidity and pressure at Bet Dagan. Values of q were used as a proxy to estimate El↓ on the basis of the relationship derived from 78 mean monthly values of q and El↓ measured between 2003 and 2010 at Sede Boqer (30°50′N, 34°46′E, 470 m MSL), the nearest Baseline Surface Radiation Network station, a desert site 130 km south east of Bet Dagan. The power function fitted to the data shown in Figure 1 was El↓ = 198.76q0.2539, R2 = 0.916, P < 0.001.
The calculated values of El↓ were then corrected for the changing concentrations of the radiatively active gases other than water vapour, i.e. carbon dioxide, methane, nitrous oxide, etc. The values of RF at the troposphere since 1750 cited by Gohar and Shine (2007) were used for this correction which was updated on the basis of the increasing anthropogenic gas concentrations listed at (http://data.giss.nasa.gov/modelforce/ghgases). The value at 2006, the midyear of the Sede Boqer series, was taken as the reference; for earlier years, the differences due to changes in anthropogenic gas concentrations were subtracted and for later years, the differences were added.
3.1. Climate trends
Annual values of global irradiance, mean daily air temperature, specific humidity and rainfall at Bet Dagan are shown as normalized anomalies in Figure 2. Temperature shows marked trends confined to the last 20 years when the mean annual temperatures increased. Temporal changes in both specific humidity and global irradiance were more complex: In the case of specific humidity, a small increase in the first half of the record was followed by a more marked increase during the last 20 years of the record. Global irradiance showed marked decreases and then recoveries in the mid 1980s and late 1990s against a background of an overall decrease over the entire record. Annual rainfall displayed large interannual variation without any significant trend.
Analysis with the non-parametric Mann–Kendall test of the pre-whitened time series indicated the trends in annual values of Eg↓, Ta, q and rainfall were significant (P < 0.05). Changes in the trends of the four climate parameters, as indicated by crossovers of the forward and backward ut and ut1 values at significant levels, were not found.
Parametric analysis by linear regression of annual temperatures on year of measurement showed increases in maximum, minimum and mean temperature which, respectively, averaged 0.14, 0.74 and 0.42 °C per decade: all increases were statistically significant, at P < 0.05 for the maximum and at P < 0.001 for the minimum and mean temperature. The linear decrease in global irradiance at Bet Dagan, averaging − 3.65 W m−2 per decade, was statistically highly significant (P < 0.01), as was the increase in specific humidity which averaged 0.662 g kg−1 per decade (P < 0.001). This increase in q was less than that expected from the measured increase in mean air temperature as this was mitigated by the decrease in relative humidity: this decrease was minor between 1964 and 1998, 72 to 70%, but steep in the last decade of the record when relative humidity fell to 65%.
3.2. Changes in short- and long-wave irradiances
Interannual variations in the long-wave radiation reaching the surface at Bet Dagan, both those resulting from changes in concentrations of water vapour E 1 ↓ H 2O, and those including changes in the other, anthropogenic gases E 1 ↓ H 2O + CO2, are analysed in Table 1 which summarizes the large and irregular variations due to changes in q as well as the smaller, monotonic increases due to the rising concentrations of the anthropogenic radiatively active gases. The magnitude of the mean annual radiation fluxes, their interannual variability and linear time trends presented include similar data for the means of the three consecutive midwinter months, December, January and February, when 85% of the annual rainfall occurs (Cohen et al., 2002), and the three rainless midsummer months, June, July and August. In both absolute and relative terms, the interannual variation in annual values of El↓ was much less than that in Eg↓. The interannual variation in the total of these two fluxes, i.e. the total all-wave incident radiance Et↓ is less than that of the short-wave irradiance but greater than that of the long-wave irradiance as values of the two fluxes were not significantly correlated for annual, summer or winter season values.
Table 1. Atmospheric, global and total all-wave radiation, 1964–2009
Linear time trends in annual values of El↓, Eg↓ and Et↓ were significant at P < 0.01; the increase in long-wave radiation averaged 0.7 W m−2 per decade; the decrease in short-wave radiation averaged 3.6 W m−2 per decade resulting in a net decrease in total irradiation, that is a negative RF, of 2.9 W m−2 per decade. In the summer season, the increase in long-wave radiation was greater, 1.6 W m−2 per decade and statistically more significant, while the decrease in short-wave radiation was similar to that of the annual values. In contrast, during the midwinter season, there was no significant time trend in El↓ while the decrease in Eg↓ was similar to that during the summer and the yearly values (Table 1).
3.3. Effects of short- and long-wave irradiances on air temperatures
Relationships between mean annual values of maximum, minimum, mean air temperatures and the diurnal temperature range (DTR) and the mean annual values of short-wave, long-wave and total downwelling radiation are presented in Table 2 together with their statistical significance. The slopes of the linear relationships of temperature to radiation fluxes, that is the climatic sensitivities, are in units of °C per W m−2. The same parameters derived from multilinear regressions of temperatures on both short- and long-wave radiation are also tabulated. Seasonal values of these relationships are presented for the means of the three midwinter and midsummer months in Table 2.
Table 2. Relationships between air temperatures °C, and surface radiation fluxes W m−2, Bet Dagan, 1964–2009
All four indices of annual temperature were significantly related to long-wave radiation but only one, DTR, was significantly correlated with short-wave radiation. Total radiation fluxes were not significantly related to annual temperatures but all multilinear relationships between radiation and temperature were because of the very different climate sensitivities for long- and short-wave radiation: Averaged for the annual maximum, minimum and mean temperatures, the climate sensitivity was 0.253 °C per W m−2 for El↓, but only 0.007 °C per W m−2 for Eg↓. During midsummer, the temperature dependencies on radiation were similar to, but more marked than those for annual periods.
Interannual variations in annual mean temperatures not accounted for by changes in long- and short-wave radiation; amounting to 26% for the rainless midsummer, 42% for annual values and 40% for the wet, midwinter months, was partially attributable to the effect of the large interannual variation in rainfall at Bet Dagan, which has a coefficient of variation of 34% (Cohen et al., 2002). The relationships between rainfall and the residuals of the multilinear regressions on mean air temperatures shown in Figure 3 were significant for both the annual and midwinter periods (P < 0.05), the relationship with number of rain days was statistically more significant (P < 0.01) for both periods.
The only significant relationship between short-wave radiation and air temperature found was that for maximum temperature in midwinter.
4.1. Accuracy of estimates of radiation fluxes
The pyranometer used to measure Eg↓ was calibrated annually so that an error of 3% can be assumed for the measured short-wave irradiances (WMO, 2008): for the small proportion (3%) of Eg↓ values estimated from sunshine duration measurements, the error was larger, the root mean square error (RMSE) of individual monthly values using the Angstrom–Prescott relationship described in Section 2.1 was 15.9 W m−2 or 4.3%.
Monthly values of E 1 ↓ H 2O calculated with the power equation shown in Figure 1 had a RMSE of 7.5 W m−2 or 2.2% of the mean. The use of specific humidity as a proxy to estimate long-wave radiation at Bet Dagan is supported theoretically by the fact that q is a combined measure of the atmosphere's radiating temperature and emissivity. Empirically, the use of the Sede Boqer relationship is sustained by its similarity to the relationships found at the seven other sites covering a global range of climates shown in Figure 1, noting that much of the difference between the five relationships illustrated could be attributed to the 5% uncertainty in pyrgeometer output (WMO, 2008). The uncertainty in the small correction to values of E l ↓ H 2O for the effect of changing concentrations of anthropogenic gases has been estimated at 10% (Ramaswamy et al., 2001).
4.2. Climate trends at Bet Dagan, magnitudes and causes
The statistically significant increase in mean air temperature, 0.42 °C per decade, measured at Bet Dagan exceeds the average for the Northern Hemisphere over the comparable period but is within the range reported for the central coastal region of the Eastern Mediterranean. The fivefold larger increase in minimum (0.74 °C per decade) as compared with maximum temperatures (0.14 °C per decade) measured at Bet Dagan is greater than the ratio reported for this region (Trenbarth et al., 2007).
The decrease in solar radiation measured at Bet Dagan, averaging 3.6 W m−2 per decade, is among the highest of the widespread negative trends tabulated in a recent review (Cohen, 2009), and approaches the maximum value listed, 4.1 W m−2 per decade, measured between the mid 1950s and 1990s at sites with populations exceeding 100 000 (Alpert and Kischka, 2008).
The probable cause of the anomalously high increase in minimum temperature and decrease in solar irradiance at Bet Dagan is the growth of Tel-Aviv, the urban conurbation and road transport network whose eastern boundary lies 4 km upwind of the Israel Meteorological Services observatory. Between 1964 and 2009, the population of the city increased from 0.794 to 1.277 million; over the same period, the total national number of motor vehicles increased from 134 000 to 2.459 million (http://www.cbs.gov.il).
Although urban effects enhancing increases in minimum temperature are well established (Mishra and Lettenmaier, 2011) local studies of the magnitude of Tel-Aviv's heat island effect have yielded diverse results (Goldreich, 2003). By contrast, a study of urban influences on the decrease in solar radiation at Bet Dagan, based on an analysis of areal and temporal variations of Eg↓ measured in the surrounding area, yielded unambiguous results: a decrease of 7% attributed to urban effects, particularly motor vehicle emissions (Stanhill and Cohen, 2009).
4.3. The paradox of negative RF and positive temperature trends
The net trend in RF at Bet Dagan was negative as the decrease in Eg↓ exceeded the increase in El↓. The resulting low and even negative values of temperature sensitivity to solar forcing reported in Table 2, although not statistically significant, are surprising in view of the strong, positive diurnal, seasonal and latitudinal relationships which exist.
In the last decade, after the reality of widespread and significant negative RF in short-wave irradiance was recognized, many attempts were made to calculate its effect on global warming using climate models in which the warming effect resulting from anthropogenic emissions of CO2 and other gases radiatively active in the long-wave bands was modified by the cooling effect of reductions in Eg↓ calculated to result from anthropogenic emissions of aerosols (Liepert et al., 2004; Ma et al., 2004; Nazarenko and Menon, 2005; Romanou et al., 2007; Jones et al., 2011; Wild and Schmuki, 2011). These and other studies resolved the paradox by assigning a minor role to short wave RF without, however, providing a physical explanation for the low climate sensitivity to solar irradiance changes at the Earth's surface.
4.4. Interactions with rainfall
The role of the large interannual variations in rainfall in supplementing RF effects on temperature changes shown in Figure 3, agrees with previously described inverse relationships between annual rainfall and temperature in the region. An analysis of 100 years of rainfall and temperature measurements at Jerusalem, 50 km SE of Bet Dagan, yielded a decrease of 105 mm rainfall per °C increase in temperature (Streim, 1974), in exact agreement with the sensitivity reported here. It also agrees with the inverse relationship between mean air temperatures and the annual volumes of rain falling on the entire region west of the Jordan River over the 78 year period between 1931 and 2008 (Stanhill and Rosa, 2009/2010).
The probable explanation of this interaction, the association of rainfall with incursions of cold moist air masses, stresses the role of changes in atmospheric circulation patterns in controlling the interannual variations of temperature at Bet Dagan. An analysis of such changes in the Eastern Mediterranean Region (30° to 40°E, 28° to 37°N) between 1948 and 2000 showed that significant changes had occurred, notably a doubling in the number of days with the mostly dry Red Sea Trough systems had occurred since the 1960s (Alpert et al., 2004). Another analysis showed that circulation changes accounting for half the increases in the summer temperatures over the Mediterranean region occurring during the second half of the 20th century (Xoplaki et al., 2003). An additional explanation of the rainfall–temperature interaction, emphasizing local influences, is that more frequent wetting of the land surfaces in wet years will lead to reductions in the convective heat flux through a changed partitioning of radiant into sensible and latent heat. The existence of such a local feedback mechanism at Bet Dagan is supported by the significant inverse relationship found between monthly values of evaporation and rainfall measured at this site (Moller and Stanhill, 2007).
4.5. Comparison with Armagh results
The key parameters of climate change and sensitivity at the two sites presented in Table 3 are based on the 1964–2000 period common to the two measurement series. The major common feature is the ability of the multilinear regression equations to explain a significant fraction of the interannual variation in temperature on the basis of changes in long- and short-wave irradiances; at both sites, the climate sensitivity to atmospheric long-wave irradiance greatly exceeded that to solar short-wave irradiance.
Table 3. Comparison of Bet Dagan and Armagh, 1964–2000
A mechanism decoupling changes in Ta from those in Eg↓ on an annual scale could be that although absorbed solar radiation is a major component of the energy balance, its influence on local Ta is modulated both temporally and absolutely by factors, such as rainfall, which partition the energy balance between latent and sensible heat. In the areal dimension, the influence of strong advection will favour decoupling. Thus, in the case of Bet Dagan and Armagh, which are near the coasts of the Mediterranean Sea and Irish Sea, respectively, long-term trends in Eg↓, which have been found to occur at the scale of a few 100 km (Stanhill and Cohen, 2001; Cohen, 2009; Wild, 2009), may only have a minor influence on the temperature of the water bodies whose advective influence on local climate is large.
This explanation is similar to the findings of a model study into the causes of the large variability of winter precipitation in the Mediterranean region during the 1960–1999 period which showed that the major cause was multidecadal variability in circulation with trends in RF playing a minor role (Kelley et al., 2011).
Large interannual changes in specific humidity and a significant long-term increase, averaging 0.622 g kg−1 per decade (P < 0.001), were measured at Bet Dagan giving rise to large variations in the flux of atmospheric radiation reaching the surface. The net increase in El↓ attributable to a greater water content of the lower atmosphere was 0.4 W m−2 per decade, constituting 57% of the total increase during the period of study: together with the effect of the increased concentrations of carbon dioxide and other well-mixed anthropogenic gases active in the long wavebands, the total long-wave RF averaged 0.7 W m−2 per decade. During midsummer, the increase in El↓ was double the annual increase and the contribution of the increase in specific humidity correspondingly larger.
Interannual variations in the global short-wave radiation reaching the surface at Bet Dagan were much greater than those in atmospheric long-wave radiation, the coefficients of variation were, respectively, 4.7 and 0.7%. The long- and short-wave fluxes were not significantly correlated. The long-term decrease in Eg↓ averaging 3.6 W m−2 per decade was five times greater than the increase in El↓ resulting in a negative all-wave forcing of 2.9 W m−2 per decade. Solar dimming was similar in midwinter and midsummer and was highly significant (P < 0.01) in both seasons. Measurements from a small pyranometer network in the central coastal plain indicated that the decrease of 7% in Eg↓ was caused by pollution from Tel Aviv, the large conurbation upwind of Bet Dagan.
The long-term increases in maximum, minimum and mean annual temperatures measured at Bet Dagan, 0.14, 0.74 and 0.42 °C per decade, respectively, were significant as was the 0.6 °C per decade decrease in the diurnal temperature range. The increase in mean air temperature was comparable to the regional value but the increase in minimum temperature and decrease in DTR were much larger and was attributed to urban effects of Tel Aviv.
The sensitivity of mean annual temperature to RF in the long-wave, 0.252 °C per W m−2, greatly exceeded that for short-wave forcing, 0.009 °C per W m−2. Similar differences were found when sensitivities were derived from multilinear regressions or from midwinter and midsummer relationships. The dependence of annual and seasonal temperatures on irradiances were highly significant (P < 0.001), accounting for 74% of the interannual variation in midsummer temperatures, 60% of the midwinter variation and 58% of the mean annual year to year variation.
A significant fraction of the residual annual and midwinter temperature variation was explained by the large interannual variation in rainfall at Bet Dagan, each 1 °C decrease in annual mean temperature was associated with a 104 mm increase in annual rainfall.
The results of this study validate the use of surface RF to account for the local interannual variation in temperature in the semi-arid Mediterranean climate. The greater climate sensitivity to long-wave irradiance than to solar irradiance previously noted in the temperate climate of Northern Ireland was confirmed and explained the paradox of a negative all-wave RF coinciding with the marked rise in air temperature measured at Bet Dagan. The physical mechanism effectively decoupling interannual changes in solar radiation from those in air temperature requires further study.
The help of A Furshpan and Isabella Osetinsky-Tzidaki of the Israel Meteorological Service for providing advice as well as much of the data used in this study is gratefully acknowledged. Data from Sede Boqer was supplied by the courtesy of G. Konig-Langlo, Director of the WMO Baseline Surface Radiation Network.