Surface albedo data from several Antarctic sites were compared to determine spatial and temporal variability in albedo. The highest degree of variability was observed at Hells Gate Station on the Ross Sea coast. The temperature close to the melting point and the reduced katabatic winds during summer allowed a strong metamorphism of the snow. At Neumayer, a coastal station by the Weddell Sea, snowfall and drifting snow were more frequent, and the surface albedo was constantly high. The albedo increased by an average of 0.07 from clear days to days with snowfall and overcast sky. Surprisingly, the hourly variation in albedo at Hells Gate Station showed a trend similar to the one observed at Neumayer Station and at Dome Concordia Station on the high plateau, when only those days with fresh snow at the surface were considered. The albedo steadily decreased during the day for solar zenith angles less than 80°. Snow metamorphism, sublimation during the day, and refreezing and/or crystal formation/precipitation during the night can explain the observed trend. To represent the daily trend in albedo over ice and fresh snow, we propose two parameterizations, which can be easily applied over other Arctic and Antarctic sites in summer. Small- and large-scale surface roughness elements can result in distortion in the measured albedo. The data at Reeves Névé Station show the effect produced on the albedo by changing slightly the sampling area immediately over a sastruga.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
 Ice and especially snow surfaces are poor absorbers of shortwave radiation, which is mostly scattered back to the atmosphere. The amount of incoming shortwave radiation is very large in summer over the Antarctic. Nevertheless, due to the high albedo the net shortwave radiation at the surface is small over the Antarctic, and it is very sensitive to small changes in the albedo: a decrease of 10% in the albedo can cause an increase of 50% in net shortwave radiation. The surface albedo is dependent on snow characteristics and other factors such as the solar zenith angle (z), atmospheric parameters, clouds, geometric pattern of the snow surface, and morphology of the area surrounding the measurement site [Warren, 1982]. In this paper we present albedo data collected in Antarctic stations at different altitudes and distances from the sea. The variety of snow conditions, determined by the different meteorological regimes, enables us to examine the spatial and temporal variations in surface albedo over Antarctica, and to describe and quantify the effect of some factors on albedo.
 During clear days, albedo undergoes variations due mostly to snow metamorphism and changes in z. Under cloudy conditions the incoming flux is mostly diffuse and isotropic, the effect of z on albedo is smaller and, if neither major snow metamorphism nor snowfall occur, the albedo is rather constant throughout the day. Occurrence of clouds at a certain location is dependent mostly on the distance from the coast and on the orography of the surrounding area. Clear skies are more common over central Antarctica than on the coast, since low- and often medium-level clouds do not penetrate into the high Antarctic Plateau [Rusin, 1964]. During summer on the coast the sky is usually cloudy due to the offshore persistence of low-pressure systems [King and Turner, 1997].
 In this paper, the effects of clouds and of major snow metamorphic changes on the surface albedo are examined through analysis of the daily mean albedo, obtained as the ratio between the daily mean outgoing and incoming shortwave flux. On clear days, the daily mean albedo is mostly an indicator of the surface reflectance during the central hours of the day, when most of the solar radiation is received. The daily means remove the variation in albedo due to variations in z, remaining good indicators of the surface properties.
 The second part of the paper addresses the hourly variation of the clear-sky albedo. Both on the coast and at the interior, clear-sky albedo is of fundamental importance not only for the correct estimation of the clear-sky radiation budget, but also for the calculation of the cloud radiative forcing, which is usually obtained from the difference between cloudy- and clear-sky radiation budget. Some parameterizations of incoming shortwave radiation over snow/ice use as an input parameter the clear-sky albedo value typical of a certain area [Konzelmann et al., 1994] or averaged over a specific surface [Shine and Henderson-Sellers, 1985], without taking into account the diurnal albedo cycle due to the dependence on z. Over sea ice the variations of clear-sky albedo due to the changes in z are of secondary importance compared with the variations due to snow metamorphism, but over the Antarctic continent albedo variations caused by snow metamorphism are very modest in most locations, and neglecting the albedo dependence on z may result in a poor estimation of the daily cycle of radiation budget and in an erroneous evaluation of the cloud radiative forcing. Microphysical modeling of the albedo dependence on z includes serious difficulties: the scattering of light from snow crystals depends on the snow crystal size, shape, and orientation, especially at the low Sun elevations typical of high latitudes. The geometry of the snow crystals is often unknown, and a sophisticated treatment of the snow grain geometry and scattering is complex and would require extensive computing time. An easier solution can result from suitable parameterizations based on measurements. Through the analysis of clear-sky data, we study the combined effects of variation in z and daily cyclic snow metamorphism, and we summarize our results into two empirical albedo parameterizations.
 To clarify the various aspects we will consider during the data analysis, we present in section 2 a review of the dependence of the clear-sky albedo on several surface and atmospheric parameters. In addition to physical factors, there are geometric aspects that must be considered to avoid misinterpretation of observations. Slopes, shadows, as well as small- and large-scale surface roughness elements can cause distortion in the measured albedo; these effects are discussed in section 3. In section 4 we described the measurement sites and the data, together with the experimental errors. Section 5 contains the data analysis and discussion, which are summarized in section 6.
2. Factors Affecting the Clear-Sky Albedo
 Surface albedo is the ratio of outgoing irradiance to incoming irradiance above the surface at a particular wavelength, according to the definition given by Warren . Albedo is a characteristic of the material at the surface and is dependent on its optical properties. Snow albedo is very high for the visible band of the spectrum, but is lower in the near-infrared, decreasing with increasing wavelength. The data analyzed in this paper are obtained from broadband pyranometers, which provide an albedo integrated over the wavelength range 0.3–3 μm (hereafter simply albedo).
 In comparing the clear-sky albedo at various stations, the interpretation of albedo variation, trend, and differences among sites is complicated by the overlapping of several factors affecting the albedo. Here we discuss the most relevant of these factors, excluding snow impurity, which we believe is so low that its effect on albedo is negligible throughout Antarctica, even nearby manned stations [Warren and Wiscombe, 1980; Warren and Clarke, 1990; Wolff and Cachier, 1998].
2.1. Albedo Dependence on Snow Grain Size and Metamorphism
 Model studies [e.g., Wiscombe and Warren, 1980] and experimental results [e.g., Nakamura et al., 2001] have showed that albedo decreases when snow ages and the surface-to-volume ratio decreases (destructive metamorphism), so that the grains become more rounded and increase in size. This snow metamorphism is one of the main factors affecting albedo in coastal Antarctic regions, where the warm summer temperatures accelerate the metamorphism and snowmelt [Rusin, 1964].
 Falling snow is usually in the form of fine grains, therefore the surface albedo is usually higher immediately after snowfall [Yamanouchi, 1983; Wendler and Kelley, 1988; Grenfell et al., 1994]. Exceptions were observed by Liljequist : at Maudhein Station in Queen Maud Land, the new-fallen snow consisted of large three-dimensional crystals that produced a porous snow cover with albedo lower than on average.
 In the interior of the continent, metamorphism progresses more slowly due to the lower surface temperatures (Ts). On the high Antarctic plateau, frost deposition and ice-crystal precipitation (“diamond-dust”) may contribute to change the texture of ice crystals at the surface [Brandt et al., 1991]. Snow drifting is common in many places, both on the coast and in the interior. In the suspended drifting snow the smallest grains are the last to fall at the surface, therefore the surface albedo increases after snow drifting, as observed, e.g., by Liljequist , Kuhn et al. , and Grenfell et al. .
2.2. Albedo Dependence on Solar Zenith Angle
 When the snow surface is smooth, uniform and horizontal, the surface albedo increases when z increases, since light incident at grazing angles has a larger probability of escaping from the snow grains without being absorbed, while light incident at lower values of z penetrates deeper into the snowpack and is more likely trapped. Choudhury and Chang  argued that at low values of z albedo is dependent primarily on the grain size, but when z increases the shape of the grains becomes more important, and albedo is indeed higher for more faceted grains. Yamanouchi  confirmed with his observations that the albedo of a renewed snow surface has a weak dependence on z, in agreement with model calculations, while hard eroded snow has a stronger dependence.
2.3. Albedo Dependence on Atmospheric Parameters
 With their multiple scattering radiative transfer model, Aoki et al.  compared three modeled atmospheres and showed that the variation of the broadband albedo with z depends also on atmospheric conditions. For example, an increase in water vapor in the atmosphere will decrease the fraction of light in the near-infrared region, for which albedo is low, with a resulting increase in the spectrally integrated surface albedo. Based on the results of Aoki et al. , we should expect the albedo to be 1% lower on the Antarctic inner continent than on the coast.
 Over a surface with a high albedo the diffuse radiation is large, due to the multiple reflections between the snow surface and atmospheric molecules. With model simulations it is easy to show that on a surface with an albedo of 0.8 the amount of diffuse radiation generated by Rayleigh scattering is twice as large as over a surface with an albedo of 0.0.
 Diffuse radiation is richer in the visible and ultraviolet region compared to direct radiation. An increase in air density will increase the visible diffuse radiation and thus the albedo, and will also shorten the path of photons in the air before being scattered, with a decreased probability that photons from distant areas will reach the instrument. At Palmer Station on the Antarctic coast, Ricchiazzi and Gautier  showed that the snow/ocean albedo contrast affects the irradiance at the snow surface up to 7 km inland from the coastline.
 The effects of air density and humidity on the surface albedo are certainly present on our data, but they cannot be investigated on the basis of the little information we have, and they are less significant than the effects of snow metamorphism and albedo dependence on z.
2.4. Effect of Surface Patterns and Large-Scale Morphological Features on Albedo
 In nature, almost no snow surface is perfectly uniform and horizontal. The wind effects on transport of snow and erosion of the surface may generate more or less regular microscale patterns, larger striations parallel to the wind direction (sastrugi), and macroscale dunes with axis perpendicular to the prevailing wind. Formation of these features is dependent on the surface morphology, snow density, wind speed, and wind directional constancy [Frezzotti et al., 2002].
 Over the Antarctic slopes and in areas with strong katabatic winds, sastrugi are common features at the surface. Warren et al.  reviewed the effect of sastrugi on the albedo. The sides of sastrugi facing the Sun receive more irradiance than a flat surface because the incidence angle of light (measured with respect to the normal to the surface) is lower. The sides opposite to the Sun and the areas in shadow receive less incident irradiance because the incidence angle is higher than z or, in the case of shadow, the irradiance is only diffuse. Over a sastrugi field, the average of the incidence angles, weighted with the amount of irradiance, is therefore lower than z, hence the albedo is lower. Another effect is “light-trapping,” for which the radiation reflected from the side of a single sastruga may reach and be absorbed by the facing side of a neighboring sastruga, thus decreasing the upward diffuse radiation. Both effects (change in effective z and light-trapping) reduce the albedo, especially at high z and when the Sun's azimuth is perpendicular to the long sastrugi axes [Kuhn, 1974; Carrol and Fitch, 1981; Yamanouchi, 1983]. Nevertheless this reduction is small; Kuhn concluded that the daily mean decrease in albedo due to sastrugi is about 2%. Carrol and Fitch  observed a 2–3% decrease in albedo when z was less than 72° and an increasing albedo reduction for higher values of z. From observations at the South Pole, Warren et al.  concluded that the reduction in albedo is extremely small (0–1%) at those values of z from which most of the incoming radiation came.
3. “True” and “Apparent” Albedo Over Sloping and Nonhomogeneous Surfaces
 When carrying out albedo measurements, particular attention should be focused on selecting the location at which to place the instrumentation. Three-dimensional obstacles in the field of view of the sensors, surface slopes, or the nearby presence of surfaces with markedly different reflectance compared with the area under study, may affect the radiance received by the instruments, especially in the case of highly reflective snow surfaces. The albedo measured under particular geometric conditions may be quite different from the “true” albedo, which corresponds to the actual reflectance of the surface.
 If the surface is partly shadowed but the instruments are not, the downward facing pyranometer experiences a lower reflected flux, while the upward facing one measures the flux that the surface would have received without shadows. The resulting “apparent” albedo is therefore lower than the true snow albedo, depending on the extension of the shaded area.
 If the snow surface is tilted, the angles of incidence of the direct radiation on the surface and on the horizontally aligned upward facing sensor are different. When the Sun is in the downhill direction, the amount of sunlight incident on the tilted surface is greater than the amount incident on the upward facing pyranometer. A greater incident flux increases the intensity of the upward scattered radiation from the surface, causing an increase in the ratio between the fluxes measured by the downward and upward facing pyranometers, i.e., an increase in the apparent albedo. The reverse situation occurs when the Sun is in the uphill direction.
 To avoid this bias between the true and apparent albedo, Brock et al.  placed the radiation sensors in a surface-parallel plane to measure the albedo over a tilted snow surface. For sloping surfaces, Mannstein  proposed an equation that relates the apparent albedo, measured by horizontal instruments, to the true albedo. He, however, used as true albedo an average value obtained under overcast conditions and constant throughout the day, but under clear skies this value is obviously not realistic.
Grenfell et al.  developed a parameterization of the apparent spectral albedo as a function of the true albedo, using data collected at South Pole Station over a snow slope of 2° and at a fixed z of 70°. The instruments over the slope were horizontally aligned, and the true albedo was obtained by averaging measurements taken over horizontal surfaces. Their data showed that in the visible spectral region (0.4–1.4 μm) the apparent albedo is about 10% higher than the true albedo when the Sun is downhill or 10% lower when the Sun is uphill.
 Most of the radiation measurements are carried out by placing horizontally aligned sensors 1–1.5 m above the snow, but some authors do not specify if the measurement area is tilted or not [Wendler and Kelley, 1988; Bintanja, 2000]. The difference between apparent and true albedo is larger for slopes oriented in the north-south direction (if the station is not at the South Pole) and for increasing tilting angle. For slopes oriented east-west, the error occurring during the morning is opposite in sign to that occurring in the afternoon; therefore the daily mean albedo is unaffected by the tilt if the surface is azimuthally homogeneous.
 When the horizontally leveled pyranometers are placed near large three-dimensional features such as sastrugi, the amount of reflected radiation reaching the downward facing pyranometer depends on the separate contribution of the areas in shadow, tilted toward the Sun or opposite to it, and also on the spatial distribution of the sastrugi. The weight of each single contribution is determined by the position of the sastrugi with respect to the sensor, the height of the sensor above the surface, and the shape, size, and orientation of the sastrugi.
 When the radiation instruments are near the surface in a sastrugi field, the sampling bias could be avoided by collecting a large number of samples at various places, or a long record of measurements if the sastrugi change over time [Carrol and Fitch, 1981]. An alternative strategy is to increase the height of the sensor above the surface, so that the irradiance measured will also be more representative of the effective reflectance of the sastrugi field [Warren et al., 1998].
4. Locations, Instrumentation, and Data Description
 The albedo data were selected from the data sets of three Italian Antarctic campaigns and from the German Neumayer Station (NM), which belongs to the Baseline Surface Radiation Network (BSRN). The Italian stations were located approximately along a 1200-km-long line at 75°S latitude. At one end of the line was Dome Concordia Station (DC), situated over a large flat dome on the Antarctic Plateau (the average slope is 0.004% [see Capra et al., 2000]). At the other end was Hells Gate Station (HG), situated on a flat area 5 × 10 km wide, on the ice shelf near the Ross Sea coast at Terra Nova Bay. The Reeves Névé Station (RN) was in between, over a 2–3° down slope toward the sea (oriented southeastward, ∼120°) on the Reeves Glacier, about 70 km inland from the coast. At 75°S the Sun is continuously above the horizon from 5 November to 9 February. NM station is situated on the Ekström Ice Shelf on the opposite coast of Antarctica at 70.7°S latitude. The surface is flat and homogenous, about 15 km south of the open ocean, and 5 km from the ice-covered Atka Bay on the east. At this latitude the Sun remains permanently above the horizon from 19 November to 24 January. The geographic coordinates and altitudes of the stations are reported in Table 1, and the locations of the stations are marked in Figure 1.
Table 1. Clear Days at the Four Antarctic Stations During the Measurement Campaignsa
Temperature, K (Mean; Max.; Min.)
For each day the mean albedo is given, together with the daily mean, maximum, and minimum temperature at the surface. The surface temperature at Reeves Névé was not available, and the 3-m air temperature is shown in parentheses.
Neumayer (70°39′S, 8°15′W) ∼20 m a.s.l.
20 Jan. 1994
261.3; 269.9; 249.5
17 Nov. 1994
259.5; 268.5; 249.8
24 Nov. 1998
261.8; 267.4; 256.3
29 Nov. 1998
261.5; 267.9; 256.4
Hells Gate (74°51′S, 163°48′E) ∼20 m a.s.l.
13 Nov. 1997
267.6; 270.3; 264.9
28 Nov. 1997
269.1; 272.2; 261.8
1 Dec. 1997
266.1; 271.3; 257.8
2 Dec. 1997
266.2; 269.2; 260.4
3 Dec. 1997
265.2; 270.7; 259.2
12 Dec. 1997
266.3; 269.2; 262.5
16 Dec. 1997
268.0; 273.0; 261.6
22 Dec. 1997
269.2; 273.4; 265.4
15 Jan. 1998
266.3; 272.1; 259.0
Reeves Névé (74°39′S, 161°35′E) ∼1200 m a.s.l.
28 Nov. 1994
29 Nov. 1994
(256.2; 261.0; 251.7)
30 Nov. 1994
(256.4; 261.0; 252.3)
2 Dec. 1994
(256.4; 260.7; 251.3)
10 Dec. 1994
30 Dec. 1994
(261.0; 265.6; 257.5)
5 Jan. 1995
(261.4; 263.6; 258.5)
Dome Concordia (75°09′S, 123°06′E) ∼3232 m a.s.l.
26 Jan. 1997
236.0; 244.7; 226.3
29 Jan. 1997
235.4; 244.6; 224.8
 For the analysis of the clear-sky albedo, we selected days with no clouds or small amount of cloud occurring for a very short time during the day. Nine clear days occurred during the three-month campaign at HG, while we had to look through the data collected during ten summers to find four clear days at NM. The difference between HG and NM in the occurrence of clear days is related to differences in the local katabatic winds and the synoptic-scale circulation in the Ross Sea and Weddell Sea regions [King and Turner, 1997]. The campaign at RN lasted for 42 days, during which seven clear days were selected. At DC the campaign lasted only ten days: although clear sky prevailed in almost all that period, only 2 days were regarded as completely clear. The clear days selected at the various stations are shown in Table 1.
 RN was characterized by a strong and persistent wind [Bromwich, 1985], channeled by orography along the steep Reeves Glacier toward the Ross Sea. The high directional constancy of the wind created a persistent wavy sastrugi field at the surface, with the axis oriented along the direction of the wind. In contrast, DC experienced rather weak winds, being the station located at high altitude on a large dome in central Antarctica, far from the major cyclone tracks.
 The incoming shortwave radiation (Swin) and the outgoing shortwave radiation (Swout) were measured with a Schenk dual pyranometer (Type 8104) at RN and HG, and with downward and upward facing Kipp and Zonen pyranometers CM3 and CM11 at DC and NM, respectively. The instruments were horizontally leveled above the surface. The flux densities received by the upper and lower pyranometers correspond to the fluxes incident and reflected at the surface, if the surface is smooth and horizontal. The viewing angle of the downward facing pyranometer is 180°, but most of the Swout received by the instrument comes from the area immediately below it. If the Swout is isotropically distributed, a total of 50%, 90%, and 99% of the flux intercepted by the pyranometer comes from circular areas centered on the vertical projection of the sensor and radius respectively long 1, 3, and 10 times the height of the instrument above the surface [Schwerdtfeger, 1976]. For many practical purposes (cleaning and checking of the instruments, stability of the instrument mast under strong wind conditions) the height of the downward facing pyranometer is usually 1–1.5 m above the surface, and this was also the case for the measurements at HG, NM, and RN. At DC the instruments were only about 50 cm from the snow.
 In considering the measurement errors, one major problem is that the pyranometers deviate from the ideal cosine response. When z is less than 80°, the error produced by the Kipp and Zonen CM3, Schenk, and Kipp and Zonen CM11 radiometers in the incoming radiation is not larger than ±3%, ±3%, and ±1%, respectively, but it increases for higher values of z. For this reason, albedo is calculated only for z < 80°. The horizontal alignment of the sensors was periodically checked. The frequency of this check varied from one day to one week, depending on weather conditions and accessibility of the station, but the eventual correction applied to the inclination of the instruments was not recorded. Therefore we also check the horizontal alignment by carefully comparing the daily cycles of Swin among each other. The application of this method resulted in the rejection of some data at RN. As reported by the manufacturers, the expected calibration accuracy for daily cumulative shortwave radiation is ±10% for the pyranometer at DC, ±5% at HG and ±2% at NM. Nevertheless, due to the high albedo over snow, most of the errors associated to the absolute sensitivity of the instruments occurred rather similarly in both, global and reflected solar radiation. Thus, in calculating the albedo, many of the errors are compensated for. We subjectively estimate that the percentage of compensated errors is 75%, and that the remaining 25% is attributable partly to random errors occurring only in the diffuse or direct component and partly to the difference in response efficiency of the upward and downward facing instruments. Thus the errors associated with daily mean albedo become ±5% at DC, ±2.5% at HG, and ±1% at NM.
 In considering the sensitivity errors associated with instantaneous measurements and the possible errors due to zero-offset (when the radiation is zero the instrument continues to measure a certain noise), we assume that the accuracy of the pyranometers is between 10 and 20 Wm−2. A rough estimation of the percentage error associated with instantaneous albedo (ERI) can be obtained from
where ERswin and ERswout are the relative errors associated with the calibration accuracy of the pyranometers (the upward and downward facing, respectively), and CR is the relative error associated with the deviation from a perfect cosine response. In formula 1 we also assumed that 75% of the errors in the accuracy are compensated when calculating the ratio of Swout and Swin. In Figure 2 we plotted our estimate of the errors in the instantaneous albedo, depending on z: the two curves represent the minimum and maximum errors, if accuracy is between 10 and 20 Wm−2 and maximum CR for z < 80° is between ±1% and ±3%. Based on Figure 2, we assume that the uncertainties in clear-sky albedo measurements for z < 80° are about ±2% for the CM11 pyranometer, ±5% for the CM3 pyranometer, and between these values for the Schenk pyranometer. However, we should bear in mind that our error estimations are based on some general assumptions and therefore are somewhat uncertain, but we believe they are useful to an understanding of at least the order of magnitude of the errors.
 Another possible source of error in the data is the deposition of frost or snow on the dome of the pyranometer. For that reason, routine checks of the domes were performed and domes were cleaned after snowfalls.
 The snow surface below the radiometers at DC from 0:30 to 7:40 solar time (ST) was in the shadow produced by the data logger box and the support system of the instrumentation. To calculate the daily mean Swout, we replaced the Swout data collected during the period of shadow with values calculated through trigonometric fits of the form:
where a, b, and c are coefficients, slightly different for each day, obtained by fitting the equations to all the data of a single day, excluding the period of shadow. The hour angle is represented by θ and Swin* is the incoming shortwave radiation at noon. Since at the time of shadow the amount of shortwave radiation is low (z is between 78° and 86°), we estimated that the error introduced in the daily mean by this interpolation was minimal and would not increase the ±5% error mentioned. During the rest of the day, the effect of shadow on the measured albedo quickly decreased and became insignificant, due to the proximity of the instruments to the surface. The shadow of the detectors, on the other hand, becomes more important the closer the instrument is to the surface. However, the sizes of the instruments were such that the effect of the shadow was in the noise level of the other uncertainties.
5. Data Analysis and Discussion
5.1. Daily Mean Albedo
5.1.1. Comparison Among Stations
 The daily mean albedo is defined as the ratio of the daily averaged reflected shortwave radiation to the daily average incoming shortwave radiation , as
The variation of daily mean albedo on clear days (αcl) was 0.24 and 0.04 at the two coastal stations HG and NM, respectively (Table 1). The αcl variability at NM was close to the αcl variability at the two inland stations (0.02 at RN and 0.00 at DC). To better explain these results, in Figure 3 we compared α and Ts at the coastal stations. For NM the summer of 1998–1999 was selected, since in that period we found 2 of the 4 clear days examined. The clear days are marked with squares (HG) and triangles (NM). The large difference in α between the two sites did not correspond to a significant difference in Ts, which indeed was very similar at the two stations during the summer. Only in November was Ts significantly lower at NM, since at that latitude the summer season, with the Sun always above the horizon, begins later than at HG. Although Ts varied at NM between the first summer month and the rest of the season, a corresponding difference in α is not apparent. We can therefore argue that in summer Ts had a minor importance in driving the variations of α at NM. Further, Ts was not of primary importance for the difference in αcl between the two stations, although in the clear days examined Ts was lower at NM than at HG (Table 1). Snowfall at NM occurred between 1 and 5 days before the clear days, and drifting snow was observed almost every day. Indeed, snowfall and especially drifting snow are the predominant all-year weather conditions at NM, as reported by König-Langlo and Herber . The continuous renewal of snow at the surface with the fine grains deposited after snow-drifting episodes explains the high albedo with very little variation during the summer: αcl retained values between 0.79 and 0.83 (see Table 1). The highest albedo values (up to 0.93) were observed in conjunction with snowfall and overcast sky (Figure 4). In overcast days without snowfall, α was on average 0.04 higher than in clear days, while on overcast days with snowfall the increase of α was 0.07 (Table 2). Clouds increase the surface albedo because they shift toward the visible the spectral distribution of irradiance. During snowfall the albedo is further increased by the highly scattering crystals of newly fallen snow.
Table 2. Daily Mean, Maximum, and Minimum Albedo Under All Sky Conditions, and Daily Mean Albedo Under Clear Sky () and Overcast Sky () During the Measurement Campaigns at HG, RN, and DC and During the Summer of 1998–1999 at NMa
 The snow conditions were different at HG. The high variation in albedo was well described by Casacchia et al. , who analyzed the spectral albedo of various snow surface samples in that area. They identified three different types of surface, with differing albedo: (1) drifted snow surface, (2) ice surface with snow inclusions, and (3) bare ice surfaces. During summer at HG the frequency of katabatic wind decreases compared with the situation in winter and spring [Bromwich, 1985]. This leads to a decrease in the seasonal mean value of wind speed and to a decrease in the frequency of drifting snow events. Rusin  observed that during the summer in coastal zones subject to katabatic wind, the relatively rare and weak drifting-snow events are only associated with cyclonic winds. The high amount of solar radiation, Ts close to the melting point, and the relatively calm wind favor large metamorphism in the snow cover and successive transformation into ice. Casacchia et al.  observed the three different surface types in different places over the HG area, but the measurements presented here show that these three surface characteristics also occurred at a fixed place during the summer. Throughout the field campaign, α varied from 0.54 to 0.89; the highest α was observed on days when snowfall occurred, while the lowest corresponded to the bare ice surface. During strong winds the smoothness of the bare ice prevented the drifting snow crystals from accumulating. In contrast, low wind and Ts near the melting point favored snow crystal embedding into the ice during snowfall events and subsequent snow accumulation. Near the summer solstice, snow metamorphism, sublimation, and melting were enhanced by the large amount of Swin. As the snow layer became thinner, the underlying ice increasingly affected the surface, thus increasing the surface absorption of solar radiation. The large albedo changes caused by the variation in surface characteristics predominate over the cloud effect on albedo; thus no difference is apparent between the mean αcl and the mean overcast-sky albedo (αov) throughout the campaign (Table 2), although on most overcast days snowfall occurred.
 At RN, as at NM, the range in variation of α was modest: the lowest values occurred during clear days, and α increased by 0.07 in overcast days (Table 2). Snowfall occurred on all overcast days except one. Special care should be taken at RN in interpreting the albedo data, since the surface down sloped southeastward (about 2–3°) and a sastruga underlaid the instruments. From the discussion in section 3, we suggest that such an orientation of the slope does not affect the daily mean albedo, but sampling bias should be expected due to the presence of the sastruga. Details of the bias are discussed in section 5.2.
 The campaign at DC was very short, weather conditions were in general good, and αcl was 0.80 during the two clear days. Despite the shortness of the measurement period, we believe that the albedo observations reflect those situations typical of the high plateau during the summer. Indeed, snow precipitation is scarce and the extremely low temperatures inhibit rapid snow metamorphism. Nevertheless, as discussed in section 2.1, some metamorphic changes of the snow occur in summer even at very low temperatures, due to the high amount of Swin reaching the surface. This could explain the slightly lower αcl observed on the high plateau site compared with the values at NM. Indeed, even if both surfaces were snow-covered, there was continuously renewed fine drifting snow at NM, while at DC the snow was at least 7 days old (we do not know the precise age, since before the beginning of the campaign nobody was there). In addition, the fraction of incoming light in the near-infrared is higher at DC, since the amount of atmospheric water vapor is low compared to NM (see section 2.3). Therefore, even with the same snow conditions at the surface, following Aoki et al.  we would expect that at DC αcl would be 1% lower than on the coast.
 The robustness of the albedo comparison among the stations is clearly limited by the instrumental errors. The pyranometers used at the various sites showed differing accuracies, and in the worst case at DC the uncertainty due to experimental errors is probably larger than the difference between α there and at the other stations. Nevertheless, we considered it useful to also include the less accurate measurements in the comparison, although final confirmation of certain results may require further observations.
5.1.2. Parameterization for Daily Mean Albedo at Hells Gate
 From the monthly mean values of albedo reported by Rusin  at Mirny, a station on the East Antarctic coast, we can infer that the surface conditions are similar to those at HG in the summer: the monthly mean albedo decreased from 0.80 in October 1956 to 0.69 in November and to 0.61 in January 1957. During various Antarctic campaigns, Bintanja and van den Broeke , Bintanja et al. , and Bintanja  obtained a 42-day-mean albedo of 0.56, a 432-day-mean albedo of 0.55 and a 37-day-mean albedo of 0.58, 0.65, and 0.68 at different sites over blue-ice areas in Dronning Maud Land, between 1150 and 1310 m a.s.l. The lower mean albedos obtained by the authors, compared to the 0.72 observed at HG, were most probably due to the blue-ice surface with less frequent snow precipitation (the region is about 300 km from the coast). In West Greenland (1155 m a.s.l.) during the summer, Greuell and Konzelmann  showed variation in α similar to that observed at HG. An analogous range of values was reported by Brock et al. , based on observations made over old snow at Haut Glacier d'Arolla, Switzerland, during the summer.
 In places where the surface undergoes large changes such as these, it is of paramount importance for radiation models to reproduce the variation in α, due to its dramatic impact on the surface radiation budget. Several albedo parameterizations have been proposed for the use in sea-ice and snowpack models. Over sea ice, the albedo in summer is highly variable: due to snow/ice melting, snowfall, melt pond formation and refreezing, the albedo can vary over a range even larger than that observed at HG [Grenfell and Maykut, 1977; Perovich et al., 2002]. In the sea-ice albedo parameterizations, albedo is expressed as a function of one or more of the following variables: air temperature, Ts, snow/ice type, cloud cover fraction, snow depth, and sea-ice thickness [Curry et al., 2001]. Among these parameters, the only one available at HG is Ts. In Figure 5, α is plotted versus the daily mean Ts: days with clear sky are marked with squares, days with snowfall with triangles, and the other days with points. From the large scatter in the data in Figure 5 it is clear that Ts alone can explain only a modest amount of the variance in α. Nevertheless, we observed that α was in general high when snowfall occurred, mostly above 0.7, and Ts was in most cases only a few degrees below the melting point. On all other days, α appeared to vary in two different modes, depending on temperature. For Ts < 269 K, the average of α was 0.76, while for 269 K < Ts < 273 K, the average α was 0.61. The albedo dependence on Ts is related to the fact that when Ts approaches 273 K, the growth of snow grains, snowmelt, and liquid water content at the surface strongly increase, thus decreasing the snow albedo. When the daily mean Ts was above 269 K, in most cases snowmelt occurred at the warmest hours of the day, otherwise Ts reached values above 271 K, allowing the occurrence of subsurface melting [Liston et al., 1999]. We believe that the sudden decrease in α for Ts > 269 K is mostly due to the onset of surface/sub-surface melting.
 The result obtained at HG presents interesting analogies with the model proposed by Ross and Walsh  for representing the fluctuations of Arctic summertime snow albedo (Figure 5). Their threshold Ts was 268 K instead of 269 K, and their albedo values for Ts lower than the threshold and at 273 K were 0.80 and 0.65, respectively, i.e., 5% higher than in the present case. They also observed a linear albedo decrease between 268 K and 273 K, not clearly evident in the HG data, which anyhow are too few to exclude a linear relationship. The Ross and Walsh parameterization was also confirmed by observations of snow albedo measured at four Russian stations during the period 1978–1983 [Roesch et al., 1999, Figure 6]. The albedo data hovered around 0.8 for Ts < 268 K and showed an approximately linear decrease over the range 268 K < Ts < 273 K.
5.2. Daily Cycle of Albedo
5.2.1. Effects of Solar Zenith Angle and Snow Metamorphism
 The daily cycle of clear-sky albedo at HG was strongly dependent on the surface conditions. In Figure 6 the clear-sky albedo curves for surfaces with fresh or drifting snow and for bare ice are presented. These two cases represent the extremes; all other curves, observed at various stages of snow metamorphism, fell in between. Albedo values are plotted only for z < 80°, since for higher values of z the measurement errors were higher, and the amount of Swin is extremely low compared with that during the rest of the day. On the left side of the figure are the morning values and on the right the afternoon values. Three of the four clear days at HG were preceded by snowfall 1–2 days before. The albedo of fresh snow during these three days was high and varied only slightly, gradually decreasing during the day.
 The albedo curves result from the superimposition of the albedo dependence on z and on snow grain size and shape. Previously, a minimum albedo in the afternoon was observed by Liljequist  at Maudheim, by Dirmhirn and Eaton  over various glaciers, and by McGuffie and Henderson-Sellers  at Resolute, Canada. As these authors suggested, the albedo decrease during the course of the day may have been due to the snow grain metamorphism caused by heating of the surface. Under clear skies when the Sun approaches the horizon and the temperature decreases, crystals can form on the snow by sublimation, and eventually ice crystals may condense in the air and fall at the surface. The crystals increase the surface albedo, until z decreases again and with increasing Swin the metamorphism recycles. On 28 November at HG the maximum Ts hovered around the melting point in the afternoon, while on 1 and 2 December the maximum Ts were 271 K and 269 K, respectively. In all three cases there were the conditions for a rather pronounced snow metamorphism, which evidently predominated over the albedo dependence on z in the afternoon (Figure 6). On 15 January snow drifting occurred, as it appeared from the strong wind (between 10 and 15 m/s) recorded by an automatic weather station (“Sofia”) located about 10 km upwind from HG. Due to the high reflectivity of the small crystals deposited at the surface when the wind set down, the albedo was high and similar to the albedo of fresh snow. Again, albedo decreased during the day, with a rate slightly faster than on days with freshly fallen snow.
 The daily curves of bare ice albedo show a much steeper decrease during the morning, from 0.8 to 0.54, and a constant albedo in the afternoon (for z < 80°). Other studies (see section 2.2) have already shown that the albedo dependence on z is steep for hard erosional snow and weak for fine-grained snow. Although we cannot totally exclude the presence of a thin layer of frost on the dome of the pyranometers, the total absence of symmetry between morning and afternoon is probably again attributable to the daily cycle of melting and refreezing/frost crystal formation at the surface. McGuffie and Henderson-Sellers  observed a similar albedo curve at Resolute on 27–29 May 1970. For z < 80°, albedo decreased from 0.7 in the early morning to about 0.5 at noon, and remained approximately constant throughout the afternoon. Although they did not specify the surface temperature and the snow conditions during those days, they attributed the albedo variation to the diurnal deposition and evaporation of a hoar-frost coating on the snow surface.
 At HG the minimum Ts values during the early morning were 265 K and 266 K, and maximum Ts in the afternoon were 270 K and 273 K on the two days with bare ice surface. The bare ice albedo curves are very similar to each other, although the two days were in early and midsummer. This leads us to suggest that the observed curves are quite typical for the bare ice of the Antarctic Ice Sheet, when temperatures are near the melting point. We therefore propose the following albedo parameterization:
where αi,m and αi,a refer to the morning and afternoon ice albedo. The simple formula is valid for z < 80°, and for such conditions when Ts approaches the melting point during the day. The root-mean-square error (RMS) associated with formula 4 for z < 70°, i.e., when the surface receives about 80% of the daily amount of Swin, is ±0.017 in the morning and ±0.007 in the afternoon (corresponding to relative errors of ±2.5% and ±1%, respectively). By modeling the surface albedo with a constant value equal to 0.59, which corresponds to the average daily mean value over bare ice, the RMS for z < 70° would be ±0.035 in the morning and ±0.040 in the afternoon (corresponding to a relative error of ±6–7%). Compared with the proposed parameterization, the use of a constant value for albedo would result in significant overestimation of the observed values in the afternoon, with an error of about 30 Wm−2 in the afternoon shortwave budget at the surface.
 To examine the daily albedo curves over highly reflective snow, we compare in Figure 7 the averaged daily cycle of clear-sky albedo at HG, DC, and NM when z < 80°. At HG the mean curve is calculated only for those four clear days with fresh or drifting snow at the surface. For each station, the mean daily cycle was obtained by interpolating the albedo values of the single days at fixed values of z at 1° interval. The number of days used in the calculation of the average is marked in brackets. Standard deviation of the averaged clear-sky albedo for z < 70° was practically zero at DC, lower than 0.03 at NM and reached the maximum value of 0.05 at HG in the late afternoon, as can be deduced from Figure 6.
 The trend in albedo already observed at HG is also present at NM and DC. On each clear day studied and in the daily average for each station, albedo decreased gently from early morning to late afternoon, approximately at the same rate for all stations. During the clear days examined here, maximum Ts were 245 K and 270 K at DC and NM, respectively.
 To investigate the reasons for the observed albedo curves, at first we considered the factors that may have caused the albedo measured to be different from the true surface albedo. In section 4 we mentioned that shadows compromised the albedo measurements until 7:40 ST (at elevation 22°), but, after this time, we estimated that the effect of shadows was in the noise level of the other uncertainties. The shadows at NM and HG were much smaller, and their effect on the measured albedo was considered insignificant.
 As discussed in section 4, other geometric factors that can affect the daily albedo cycle are surface tilting and the presence of three-dimensional features near the measurement stations. However, we know that all three stations were located over flat and rather uniform snow far from obstacles. If the stations were over small slopes, we would conclude from the curves in Figure 6 that the direction and magnitude of the inclinations were the same for all three sites, which is highly improbable.
 The fact that all the instruments at the four stations observe the same daily albedo cycle, although they have different accuracy and operate in different environments, allows considerations on the reliability of the measurements. We noted in Figure 2 that ERI is dependent on z. Nevertheless, since the daily albedo cycle observed is not correlated with the daily change in z, the shape of the albedo curves cannot be caused by the daily variation of ERI. Therefore we believe that the errors described in Figure 2 mainly give an idea of the confidence we have in the absolute values of albedo, while the hourly albedo variations and differences among the albedo curves at each single station are less affected by measurement errors.
Liljequist  undertook a detailed study of the clear-sky albedo at Maudheim, a Norwegian base located on the ice shelf, about 115 km to the west of NM and about 5 km from the coastline. The meteorological and surface conditions were very similar to those of NM. Liljequist observed that in November the diurnal cycle of clear-sky albedo was symmetric around noon, with albedo decreasing with decreasing z, while in December and January albedo was rather asymmetrical, as shown in Figure 7. The method used by Liljequist to calculate the averaged albedo curve differed from the one adopted in the present study. He concluded that, even if no melting was ever observed, metamorphism into larger snow grains occurred during the central hours of the day. When z increased and temperature decreased, refreezing and/or crystal formation by sublimation formed a layer of highly scattering grains at the surface. This explanation appears realistic also for the albedo curve observed at NM, where the maximum Ts during conditions of clear sky was below 270 K and melting is seldom observed throughout summer [König-Langlo and Herber, 1996].
 The extremely low Ts at DC clearly allowed only a slow metamorphism to occur in the snow, even if the daily temperature excursions were very large (Table 1). During the coldest hours, strong temperature and humidity inversions developed at the surface layer. Rusin  observed rapid development of frost crystals during the coldest hours on the Antarctic Plateau, followed by sublimation of the rime due to the radiative heating during the warmest hours. This may explain the daily albedo cycle observed at DC. Yamanouchi  showed a decreasing albedo during the day at Mizuho Station (2230 m a.s.l.), on East Antarctica, even after flattening the surface to remove the sastrugi. He argued that the albedo curve could be caused by the sastrugi that remained around the flattened area. However, the effect observed by Yamanouchi could probably be due to the daily metamorphism of the snow surface, as also suggested by McGuffie and Henderson-Sellers . During December, surface conditions at Mizuho Stations were between those at NM and DC: the daily mean albedo was never lower than 0.8 [Yamanouchi, 1983], and the daily mean surface temperature was between −263 K and −248 K [Yamanouchi and Kawaguchi, 1984].
 We have shown that snow melting/refreezing and crystal formation/sublimation could explain the common daily trend in albedo on the coast and on the inner high plateau. This type of trend may be approximated with the following parametric equation:
where αs refers to the snow albedo, αn is the mean albedo at noon, and zmin is the averaged minimum solar zenith angle (in degrees) at each station. The values for αn were 0.81, 0.83, and 0.79, and for zmin 53°, 60°, and 58° for HG, NM, and DC, respectively. By fitting equation (5) to the data, the best value for c was 0.003 for the morning and −0.003 for the afternoon. The linear equation proposed represents the clear-sky albedo variation for z < 80°. In Figure 6 the parameterization 5 is plotted against the albedo observations at HG. During the central hours of the day (z < 70°), when Swin is 80% of the total daily amount, the daily RMS errors associated with formula 5 at HG, NM and DC, calculated with respect to the averaged daily cycles, are ±0.006, ±0.008, and ±0.008, respectively (corresponding to a relative error of ±1%). By comparison, using a constant daily mean albedo throughout the day, the RMS errors are ±0.035, ±0.034 and ±0.027 for the three stations, respectively (corresponding to a relative error of ±4–5%). A constant albedo at HG would result in a relative error of ±17% in the morning shortwave budget and ±6% in the afternoon (for z < 70°), while with formula 5 the relative error would be ±3% and ±1% in the morning and the afternoon, respectively. The equation proposed represents a simple and practical tool for expressing the albedo variation due to superimposition of the effects of variation in z, snow metamorphism, and crystal formation/sublimation. Since measurements of snow/ice grain size or density were not available, in equations (4) and (5) we implicitly included the albedo dependence on crystal metamorphism in the albedo dependence on z. We stress that the robustness of the parameterizations is not compromised by the experimental errors described in Figure 2, since these errors affect mostly the absolute value of the albedo and not to the albedo variation during the day.
 In the study done by Carrol and Fitch , the daily albedo cycle has been presented as a function of z, thus averaging over morning and afternoon values. Presumably, it is also because of this that the albedo variation during the central hours of the day was obscured, and albedo varied only for z > 70°. Rusin  averaged all the clear-sky albedo observations as a function of z, without morning/afternoon distinction and without discriminating between the various surface conditions. He obtained a curve in which albedo decreased with decreasing z, from 0.83 at z = 80° to 0.74 at z = 45°. The changing rate of albedo was intermediate between the two morning slopes obtained at HG for the extreme situations of fresh snow and ice. At Terre Adelie, Wendler and Kelley  presented a clear-day albedo curve that gently decreased with decreasing z, probably calculated by averaging morning and afternoon values, and they mentioned no asymmetry between morning and afternoon. However, in that location it is possible that crystal formation during the coldest hours and successive sublimation in the warmest time of the day did not occur, due to the continuously blowing strong katabatic wind (the station was on a slope at an altitude of 1560 m). The observations of Kuhn et al.  at Plateau Station show that the monthly averaged daily variation in albedo was symmetric around noon, when (and only when) the surface irregularities were leveled by drifting or falling snow. Since their albedo trends resulted from monthly averages and most probably included both clear and cloudy days, it is not possible to make a direct comparison with our data.
 The measurement errors at DC were probably larger than at the other stations. Nevertheless, we believe that the daily albedo curves presented here are reliable and physically explainable.
5.2.2. Sastrugi Effect at Reeves Névé
Figure 8 shows the effect of sastrugi on the daily albedo cycle at RN. Each of the seven lines plotted in the figure corresponds to a clear day. A difference can be seen between the curves measured on days before and after 1 December 1994, when the structure supporting the radiometers was rotated 77° counterclockwise.
 The difference between the two groups of curves in Figure 8 arose from new positioning of the albedometer with respect to a sastruga, about 20 cm high and 1 m long, present on the snow surface just below the instrument. There were no other large sastrugi in the 2- to 3-m2 field of view below the albedometer, which was at about 1 m above the surface. Since during most of the campaign the wind was from the same north-nortwest direction, the sastrugi field did not undergo significant alterations. This was also confirmed by the fact that the daily albedo trends in the days before rotation of the structure were very similar to each other, as were the albedo trends after rotation (Figure 8).
 In addition to the presence of sastrugi, the surface at RN showed a 2–3° down slope toward the southeast. Measuring the albedo from horizontally leveled sensors, the tilting alone would produce an apparent albedo higher when the Sun is downhill (during morning) and lower when the Sun is uphill. The data, however, showed the opposite trend: evidently the sastruga effect dominated over the slope effect. Prior to rotation of the mast, the daily mean clear-sky albedo was 0.01 higher than after rotation (Table 1). The RN data demonstrate how sensitive albedo measurements can be to surface inhomogeneities.
 The average daily variation in albedo presented by Kuhn et al.  over several months displayed quite asymmetric trends, with albedo minima occurring sometimes in the morning or afternoon, depending on the geometrical shape of the features formed at the surface by the action of wind and drifting snow. The authors concluded that the effect of large three-dimensional surface features on the monthly mean albedo was of the order of 1%.
6. Summary and Conclusions
 Data on surface albedo over four Antarctic sites during the austral summer were compared to examine the spatial and temporal variability in the albedo. The highest albedo variability was observed at HG, on the Ross Sea coast. Temperatures near the melting point and the decreased intensity and frequency of katabatic wind during the summer allowed a strong metamorphism of the snow. The albedo variability associated with snow/ice metamorphism was larger than the variability due to changes in sky conditions: even among clear days, the daily mean albedo varied between 0.58 and 0.82. To explain this variability, we examined the relation between daily mean albedo and surface temperature. Although the scattering of data was rather large, we observed a trend comparable to that obtained by Ross and Walsh  for the snow-covered Arctic sea ice. A similar variability in albedo has been observed also in Greenland [Greuell and Konzelmann, 1994] and the Alps [Brock et al., 2000]. Thus the alternation of snow and bare ice characterizes those glacier and sea ice surfaces that are subject to both snowfall and strong melting during the summer.
 Measurements of snow depth variation and snow grain size would be fundamental to better explain this large albedo variation and to develop more accurate parameterizations. Nevertheless, precipitation and optical properties of snow grains at their various stages of metamorphism are poorly represented in most of present-day climate models, while the simple formula derived by Ross and Walsh  and substantially confirmed here is easy to apply. It also yields much better results than using seasonally averaged albedo values from climatology. Simple albedo parameterizations, which only include the albedo dependence on temperature, can give reasonably good results when used in sea-ice models [Curry et al., 2001], allowing a realistic representation of many processes affected by albedo. On the other hand, it is not possible to correctly simulate the snow/ice-albedo feedback and the impact of albedo changes in future climate scenarios without taking into account the albedo dependencies on snow/ice depth and metamorphism.
 At NM and RN there was little variation in the daily means of clear-sky albedo (the maximum change was 0.04), and the occurrence of overcast sky together with snowfall increased the daily mean albedo by 0.07. At DC, on the high plateau, only few data were available and the daily mean clear-sky albedo was constant. Although NM station was also located at the coast, its surface characteristics showed greater similarity with those at DC on the plateau than at HG. The temperature at NM was about the same as at HG, but the occurrence of cyclones and easterly winds produced frequent snowfall and almost continuous drifting snow, which supplied the surface with small and highly reflective snow grains.
 The hourly variation of clear-sky albedo showed, surprisingly, a similar trend among HG, NM and DC, when at HG only those days with fresh snow on the surface were considered. Albedo steadily decreased during the day for z < 80°. Snow metamorphism or sublimation during the day and refreezing and/or crystal formation/precipitation during the night can explain the observed trends. In the case of bare ice at HG, albedo decreased rapidly in the morning and maintained a constant low value during the afternoon. In the case of snow the albedo decrease was weaker, but constant throughout the day. Some studies have reported a similar daily trend in albedo [Liljequist, 1956; Dirmhirn and Eaton, 1975; Yamanouchi, 1983; McGuffie and Henderson-Sellers, 1985], but other authors have not shown any asymmetry in the albedo before and after noon, probably because they have examined the albedo dependence on z by averaging morning and afternoon values.
 Two parameterizations were proposed (equations (4) and (5)) to represent the daily albedo cycle over ice and fresh snow. Both formulae express the changes in albedo due to the superimposition of two main effects: the direct effect of variation in z and the associated metamorphism at the surface. The use of these parameterizations instead of a daily constant albedo results in an improvement in the albedo estimation of about 3–4% over fresh snow and 2–5% over bare ice. Equation (4) is based on two days at HG with z varying between 51° and 80° and with temperatures approaching the melting point. We think that the formula may also be applied to other bare-ice surfaces at near-melting temperatures at polar latitudes, i.e., over glaciers at the Antarctic coast, in Greenland, in the Arctic Arcipelago and over thick sea ice. Equation (5) is based on data from three stations, located at latitudes from 70°S to 75°S, and characterized by surface temperatures ranging from 225 K to 273 K and z varying from 41° to 80°. We believe that the formula is applicable in summer over those snow surfaces that do not experience large metamorphic changes. This is the case for surfaces with fresh/drifting snow or temperature much lower than the melting point. These surfaces can be found over most of the Antarctic Continent and probably most of Greenland and Arctic archipelago. A comparison with the albedo curves shown by Dirmhirn and Eaton [1975, Figure 1] may suggest that equation (5) is applicable also to lower-latitude glaciers, although more data should be analyzed to confirm the validity of the equation at lower z and with shorter daylight duration.
 The largest uncertainty in the results is perhaps represented by the instrumental errors associated with the measurements. Nevertheless, we believe that the daily albedo cycle observed, with similar characteristics at all stations, is less affected by errors than the instantaneous values themselves. Analyses of albedo data from other field campaigns would be very useful in confirming or not the results we obtained. Particular attention should be paid to the possible presence of slopes, shadows, and other macroscopic striations on the surface, which may complicate the interpretation of the clear-sky albedo data. The RN data represent a good example of albedo alteration due to decimeter-scale surface irregularities, showing the effect produced on the albedo by merely changing slightly the sampling area immediately over a sastruga.
 This work was funded by the Academy of Finland (project 178457). I wish to thank the Italian National Agency for New Technologies, Energy, and the Environment (ENEA) for logistic and technical assistance during the three Italian expeditions at Hells Gate, Reeves Névé, and Dome Concordia. T. Georgiadis and G. Trivellone, from the Italian National Research Council (CNR), provided the radiation data from the three Italian stations and the description of the measurement sites. A. Pellegrini and P. Grigioni from ENEA provided the synoptic data from the Automatic Weather Station “Sofia.” The Alfred Wegener Institute (AWI) from Germany is acknowledged for making available in its web page the radiation and synoptic data from the Neumayer Station. I am grateful to T. Vihma, H. Savijarvi and T. C. Grenfell for useful discussions and suggestions.