Journal of Geophysical Research: Atmospheres

Significance of multidimensional radiative transfer effects measured in surface fluxes at an Antarctic coastline

Authors


Abstract

[1] At a coastal high-latitude site, multiple reflection of photons between the high albedo surface and an overlying cloud can enhance the downwelling shortwave flux out over the adjacent open water to a distance of several kilometers. This coastal albedo effect has been predicted by theoretical radiative transfer studies and has also been measured under ideal conditions. In this study, three multispectral solar ultraviolet radiometers were deployed in the vicinity of Palmer Station, Antarctica (64° 46′S, 64° 04′W) to determine the prevalence of the coastal albedo effect under the region's natural variability in cloud cover. One radiometer was deployed near the base of a glacier, and the other two radiometers were deployed on Janus Island and Outcast Island, islets ∼2.8 km (1.5 nautical miles) and 5.6 km (3 nautical miles) distant from Palmer Station, respectively. The radiometers were operated simultaneously for 16 days during late December 1999 and January 2000. Under all cloudy sky conditions sampled by this experiment the coastal albedo effect is seen in the data 60% of the time, in the form of a decreasing gradient in surface flux from Palmer Station through Janus and Outcast Islands. During the other 40% of the cloudy sky measurements, local cloud inhomogeneity obscured the coastal albedo effect. The effect is more apparent under overcast layers that appear spatially uniform and occurs 86% of the time under the low overcast decks sampled. The presence of stratus fractus of bad weather, under higher overcast layers, obscures the coastal albedo effect such that it occurs only 43% of the time. A wavelength dependence is noted in the data under optically thin cloud cover: the ratio of a flux measured at an islet to that measured at the station increases with wavelength. This wavelength dependence can be explained by plane-parallel radiative transfer theory.

1. Introduction

[2] This study addresses the spatial distribution in surface solar flux at conservative scattering wavelengths, in a region where three-dimensional radiative transfer effects are expected to govern the radiation field. Coastal regions at high latitudes offer a unique location to test the results of multidimensional radiative transfer modeling. Near the Antarctic Peninsula, and in similar regions, it has been shown that multiple reflection of photons between a high albedo surface (glacier or snow-covered sea ice) and a cloud layer enhances the downwelling shortwave surface flux by a factor of 2 or more relative to a low albedo surface such as open water [Gardiner, 1987]. Theoretical studies have suggested that photons reflected between a high albedo surface and a cloud should enhance the downwelling shortwave flux not only at the coastal location but also out over the open ocean to a distance of several kilometers [Degünther et al., 1998; Podgorny and Lubin, 1998; Ricchiazzi and Gautier, 1998]. The existence of this phenomenon, which we herein refer to as the “coastal albedo effect,” has been measured under near-ideal conditions by Smolskaia et al. [1999] at Davis Station, Antarctica, using broadband ultraviolet (UV) radiometers mounted on both a sledge to move inland and in a boat to make open water transects perpendicular to the high albedo coastline.

[3] Here we are interested in the prevalence of this effect at Palmer Station, Antarctica (64° 46′S, 64° 04′W), for two reasons. First, Palmer Station is the central facility for an extensive Long Term Ecological Research (LTER) program [Smith et al., 1995, 1999], in which many of the physical forcing mechanisms acting on the base of the marine food web, primary production by phytoplankton, are monitored. The coastal albedo effect may be important with respect to the light levels experienced by photosynthetic organisms in the upper water column near the coast. To facilitate marine ecological studies involving primary production throughout the Southern Ocean, satellite-based methods have been proposed to map the spatial distribution of Antarctic UV and photosynthetically active radiation (PAR, 400–800 nm), using plane-parallel radiative transfer theory and an independent pixel approximation [Lubin et al., 1994; Nunez et al., 1997; Krotkov et al., 2001]. If the coastal albedo effect prevails throughout the region, some sort of parameterized correction to the independent pixel approximation may be necessary in coastal waters. Second, an empirical determination of the significance of multidimensional radiative transfer effects is of general interest to atmospheric science for validation and refinement of ongoing theoretical work.

2. Experiment

[4] We deployed three radiometer systems at Palmer Station, one at the station itself and two on neighboring islands. These were Biospherical Instruments (Inc.) model GUV-531 multispectral photodiode radiometer systems. The GUV-531 measures downwelling flux in four UV wavelength bands of ∼10 nm full width half maximum, centered on 305, 320, 340, and 380 nm. These UV wavelength bands are established by interference filters, and their filter functions are approximately Gaussian. A fifth channel measures downwelling PAR flux, and the filter function of this channel is designed to be flat in the quantum (photon-counting) domain. Downwelling flux is measured by means of a Teflon diffuser empirically adjusted for correct cosine response, followed by a collimating lens, and then by five individual Si photodiodes each with its own bandpass filter. Absolute calibration was established by prefield and postfield season calibration at Biospherical Instruments with reference to NIST-traceable lamps. Radiometric stability in the field is established by maintaining the input optics and the detectors at 40°C (well above the ambient outdoor temperature) using a temperature controller. The Si photodiode voltages are digitized within the temperature-stabilized optical unit before being sent along the data cables to the data logger. This eliminates uncertainties related to long signal transmission lines. Because of the high stability achieved by temperature regulation, the radiometric precision of each channel is only limited by the precision of the analog-digital converter. This is approximately one part in 12 bits over a voltage range 625 mV to 1 V, which is effectively ±0.15%.

[5] In the vicinity of Palmer Station there are numerous islets that are home to penguin, cormorant, and petrel rookeries and seal breeding grounds and which are the sites for much biological research. Our original intention was to deploy one radiometer at the station, a second on as far distant an islet as we could reach, and a third in a boat that could motor along transects between the station and the islet. This would have been similar to the experiment of Smolskaia et al. [1999]. However, local waters proved too rough during most days, and on the very light and nimble boats available at Palmer Station (zodiacs) we could not stabilize the radiometer in a gimbal satisfactorily for downwelling flux measurements. We therefore deployed the three radiometers at fixed locations as shown in Figure 1, between 29 December 1999 and 22 January 2000. One radiometer was mounted atop the T-5 laboratory building, approximately 100 m inland. This building was also approximately 500 m from the edge of a large glacier behind Palmer Station that extends several kilometers inland on Anvers Island. During most of the measurement period this glacier's upper surface was mostly bare ice and hard-packed snow, and therefore most likely exhibited an albedo less than that of fresh Antarctic snow [Warren, 1982]. The second radiometer was deployed on Janus Island, an islet ∼2.8 km (1.5 nautical miles) from Palmer Station. The third radiometer was deployed on Outcast Island, an islet ∼5.6 (3 nautical miles) from Palmer Station. We therefore covered a geographical distance that is relevant to the predicted and measured two-dimensional radiative transfer effects reported in previous studies [Podgorny and Lubin, 1998; Smolskaia et al., 1999]. The three sites lie as closely along a straight line as we could achieve over this distance. By periodically resupplying the two islet radiometer installations with freshly charged marine batteries as weather permitted (the temperature controllers and heaters were a significant current drain), we achieved continuous radiation measurement throughout much of January 2000. While power was available at the islet sites, flux measurements were sampled from the detectors once per second and were recorded by the dataloggers as 2-min averages.

Figure 1.

Map of the region encompassing Palmer Station, Antarctica, showing the locations of the three GUV radiometers.

[6] These continuous measurements gave us a representative sample of how the downwelling radiation field varies under Palmer Station's changing meteorological conditions. Throughout the measurement period, cloud coverage was observed visually and recorded at 6-hour intervals by station science support personnel. These observations were made according to World Meteorological Organization (WMO) standards and used the associated numeric codes for different cloud types.

[7] Because we are interested in subtle differences in downwelling flux induced by nonuniform surface albedo and related cloud multiple reflection effects, we could not rely solely on the preseason and postseason radiometer calibration. Absolute radiometric calibration could only be ascertained to within ±2.5%. Therefore at the end of the field program all three radiometers were installed at T-5 and were operated simultaneously for 2 hours for intercomparison. Offsets between the radiometers were determined for all channels, and the fluxes recorded by the radiometers at the islets were normalized to that of the T-5 instrument on the basis of this intercomparison. The required normalization factors were small (all within 3%), as shown in Table 1. The one exception is the large offset (7%) for the 305-nm channel in the radiometer deployed at Outcast Island. Further examination of this channel's data indicated unreliable detector performance, and the 305-nm measurements from the Outcast Island radiometer are not analyzed in this study.

Table 1. Intercomparison of the three Biospherical Instruments (Inc.) GUV-531 Radiometer Systems as Determined by Operating the Radiometers Simultaneously Atop the T-5 Laboratory Building at Palmer Stationa
Channel, nmJanus Island to T-5Outcast Island to T-5
RatioσRatioσ
  • a

    For each channel, the table gives the ratio of the downwelling flux measured by the radiometer deployed on one of the islets to that measured by the instrument at the T-5 laboratory building. The standard deviation in this ratio is also given.

PAR0.97380.01000.99230.0072
3801.00200.00850.99950.0060
3400.97790.00880.99760.0057
3201.00200.00800.99650.0059
3050.97830.00950.93000.0185

3. Results

[8] Table 2 gives all 69 weather observations that pertain to the time periods when all three radiometer systems were operating simultaneously. Clear sky or scattered cloud conditions were rare throughout January 2000, as is typical of the region [Warren et al., 1986, 1988]. The most common sky conditions involved some combination of broken or overcast stratocumulus or altocumulus. Stratus fractus of bad weather (WMO code L7), a layer of patchy broken cloudiness ∼100 m above the surface and below higher overcast decks, also occurred frequently. This type of cloud is commonly known as “scud.” Despite the persistence of extensive cloud cover, there were few occasions when the clouds resembled ideal horizontally homogeneous “plane parallel” scattering layers. Considerable spatial inhomogeneity was evident nearly all the time for clouds at all altitudes. Throughout most of the experiment the water between the station and the islands was ice-free, although occasionally brash and pancake ice drifted in with the prevailing winds.

Table 2. Surface Weather Observations from Palmer Station, Antarctica, During the Island Radiometer Deployments of December 1999 and January 2000a
Observation NumberDay and Time, UTCSky Conditions
  • a

    Observation numbers and dates match those in Figure 2. Sky coverage is given in octas, for each type of cloud observed. Cloud type was recorded according to World Meteorological Organization codes, which are listed with translation. Estimates of cloud base height are given in feet.

1363.758/8 L6 (stratus in a sheet or layer)
23648/8 L7 (stratus fractus of bad weather)
3364.253/8 L5 (stratocumulus) at 1500′
  5/8 M7 (altocumulus)
  5/8 H1 (cirrus filaments, not expanding)
4364.58/8 L7 (stratus fractus of bad weather)
5364.758/8 L7 (stratus fractus of bad weather)
63651/8 L7 (stratus fractus of bad weather)
  6/8 M5 (altocumulus)
72.757/8 L5 (stratocumulus) at 1000′
838/8 L7 (stratus fractus of bad weather) at 800′
93.258/8 L7 (stratus fractus of bad weather) at 2000′
103.56/8 L8 (cumulus and stratocumulus, bases at different levels
  8/8 M1 (altostratus, semitransparent)
113.758/8 L6 (stratus in sheet or layer) at 2000′
1248/8 L7 (status fractus of bad weather) at 1000′
134.258/8 L7 (stratus fractus of bad weather) at 1000′
144.58/8 L6 (stratus in a sheet or layer)
154.758/8 L7 (stratus fractus of bad weather) at 800′
1658/8 L7 (stratus fractus of bad weather) at 1000′
175.256/8 L5 (stratocumulus) at 600′
  8/8 M7 (altocumulus, mainly opaque)
185.54/8 L8 (cumulus and stratocumulus, bases at different levels
  7/8 M6 (altocumulus from the spreading of cumulus)
195.757/8 L6 (stratus in a sheet or layer) at 2000′
2061/8 L4 (stratocumulus from the spreading of cumulus) at 1200′
  8/8 M7 (altocumulus, not expanding)
216.258/8 L7 (stratus fractus of bad weather)
228.758/8 L7 (stratus fractus of bad weather) at 1200′
2397/8 M3 (altocumulus, semitransparent) at 6000′
249.258/8 L7 (stratus fractus of bad weather)
259.56/8 L8 (cumulus and stratocumulus, bases at different levels)
269.751/8 L1 (cumulus with little vertical extent) at 1200′
  1/8 M3 (altocumulus, semitransparent) at 5500′
27101/8 L1 (cumulus with little vertical extent) at 3000′
  1/8 M3 (altocumulus, semitransparent) at 5500′
2810.252/8 L4 (stratocumulus from the spreading of cumulus)
  7/8 M3 (altocumulus, semitransparent)
3010.755/8 L4 (stratocumulus from the spreading of cumulus) at 2000′
31111/8 L1 (cumulus with little vertical extent) at 1200′
  1/8 M2 (altostratus, dense) at 5500′
3211.254/8 M3 (altocumulus, semitransparent) at 8000′
3311.53/8 M3 (altocumulus, semitransparent) at 8000′
  2/8 H1 (cirrus filaments, not expanding)
3411.753/8 L4 (stratocumulus from the spreading of cumulus) at 2000′
  1/8 H1 (cirrus filaments, not expanding)
35124/8 L4 (stratocumulus from the spreading of cumulus) at 2000′
  3/8 M7 (altocumulus, not expanding) at 5000′
3612.258/8 L7 (stratus fractus of bad weather)
3712.57/8 L8 (cumulus at stratocumulus, bases at different levels) at 2000′
3812.758/8 L6 (stratus in a sheet or layer) at 2000′
3913.753/8 L5 (stratocumulus) at 3000′
40141/8 L4 (stratocumulus from the spreading of cumulus) at 1000′
  1/8 M4 (altocumulus patches) at 6000′
  7/8 M2 (altostratus, dense) at 8000′
4114.258/8 L6 (stratus in a sheet or layer) at 4000′
4214.58/8 L7 (stratus fractus of bad weather)
4314.752/8 L1 (cumulus with little vertical extent) at 1400′
  1/8 L5 (stratocumulus)
44151/8 L1 (cumulus with little vertical extent) at 800′
  6/8 L5 (stratocumulus)
4515.258/8 L6 (stratus in a sheet or layer)
4615.758/8 L5 (stratocumulus) at 2000′
47161/8 L6 (stratus in a sheet or layer) at 600′
  7/8 L5 (stratocumulus) at 2000′
4816.258/8 L6 (stratus in a sheet or layer) at 1000′
4916.54/8 L6 (stratus in a sheet or layer)
  6/8 M1 (altostratus, semitransparent)
5016.757/8 L5 (stratocumulus) at 2000′
51174/8 L4 (stratocumulus from the spreading of cumulus) at 2000′
  4/8 M1 (altostratus, semitransparent)
5217.254/8 L5 (stratocumulus) at 3000′
  7/8 M2 (altostratus, dense)
5317.51/8 H8 (cirrostratus, not increasing)
5418.757/8 L5 (stratocumulus) at 4000′
55197/8 M3 (altocumulus, semitransparent) at 6000′
  1/8 H5 (cirrostratus, increasing)
5619.257/8 M7 (altocumulus, not expanding) at 5000′
5719.57/8 M7 (altocumulus, not expanding)
5819.757/8 L8 (cumulus and stratocumulus, bases at different levels) at 5000′
59204/8 L4 (stratocumulus from the spreading of cumulus) at 1800′
  8/8 M2 (altostratus, dense)
6020.258/8 L7 (stratus fractus of bad weather)
6120.58/8 L7 (stratus fractus of bad weather)
6220.755/8 L7 (stratus fractus of bad weather) at 1500′
  8/8 M1 (altostratus, semitransparent)
63214/8 L4 (stratocumulus from the spreading of cumulus) at 2000′
  8/8 M2 (altostratus, dense)
6421.254/8 L4 (stratocumulus from the spreading of cumulus) at 2000′
  8/8 M2 (altostratus, dense)
6521.55/8 L7 (stratus fractus of bad weather)
  8/8 M2 (altostratus, dense)
6621.758/8 L7 (stratus fractus of bad weather)
67227/8 L5 (stratocumulus) at 2000′
  1/8 M3 (altocumulus, semitransparent)
6822.258/8 L7 (stratus fractus of bad weather)
6922.58/8 L7 (stratus fractus of bad weather)

3.1. Natural Variability in Downwelling Fluxes

[9] For this experiment to demonstrate the coastal albedo effect reported by Podgorny and Lubin [1998], Ricchiazzi and Gautier [1998], and Smolskaia et al. [1999], the results should show the downwelling flux measured at Janus Island to be consistently lower than that measured at T-5, and at the same time the flux measured at Outcast Island should be consistently lower than that measured at Janus Island. This would indicate the presence of a decreasing gradient in downwelling flux extending out over the open water, induced by multiple reflection of photons between the high albedo coastal surface and the cloud bases. Figures 2a–2i show the entire time series of downwelling flux measurements, expressed in terms of the ratios of the 380 nm flux measured at each of the islet sites to the flux measured at T-5. We denote these ratios as RJ and RO for Janus Island and Outcast Island, respectively. For much of the time we notice that RJ < 1, and at the same time RO < RJ. However, there are numerous instances where both ratios are larger than unity, and there are some instances where RO > RJ. Referring to Table 2, these conditions appear to occur often during the presence of stratus fractus of bad weather. During the two brief time periods when the sky was mostly clear (weather observations 26, 27, 31, and 32 in Table 2), RJRO ∼ 1, indicating the lack of a measurable multiple reflection effect in cloud-free conditions. The behavior of these flux ratios shown in Figure 2 for 380 nm was similar for the other four spectral channels.

Figure 2.

Time history of the ratios of downwelling 380-nm flux measured at the islets to that measured at T-5 (on Palmer Station). These ratios, RJ and RO, for Janus Island and Outcast Island are shown by the solid and dotted curves, respectively. Bold integers at the bottom of each panel, within the frame, are the weather observation numbers of Table 2. The solar zenith angle is given by the two-digit numbers at the top of each plot.

[10] The time series of Figure 2 suggests that the coastal albedo effect does occur but that there are numerous instances where the local cloud inhomogeneity dominates the radiation field and obscures the gradient owing to multiple reflection. To determine the fraction of time the coastal albedo effect prevails, under the region's natural cloud cover, we examine the histogram of the ratio RJ, shown in Figure 3. Figure 3a shows that RJ < 1 for 85% of the cloudy-sky observations when the solar zenith angle θo is <85° and that RJ < 1 for 89% of the cloudy-sky observations when the solar zenith angle is between 85–94°. Thus there is no appreciable difference in the frequency distribution of these flux ratios related to Sun elevation. We next select flux measurements that could be identified with broken or overcast low cloud decks (base altitudes lower than 600 m), broken or overcast midlevel cloud decks (base altitudes higher than 600 m), and stratus fractus of bad weather. To sort the data into these categories, we used only the flux measurements within ±15 min of a weather observation (Table 2) indicating the desired sky condition. These histograms are shown in Figures 3b and 3c. The frequencies of occurrence of RJ < 1 for midlevel clouds, low-level clouds, and stratus fractus of bad weather are 83%, 91%, and 78%, respectively. For all measurements in which RJ < 1 we then determine the number of occurrences of RO/RJ < 1. These histograms are shown in Figures 4a and 4b, and they show that RO < RJ for all cloudy observations 70% of the time. The corresponding frequencies of occurrence for midlevel clouds, low-level clouds, and stratus fractus of bad weather are 70%, 94%, and 55%, respectively. We therefore determine from this data set that the gradient in flux out over the open ocean, established by multiple reflection of photons between the high albedo coastal surface and the cloud base, appears 59% of the time for all cloudy observations with θo < 85° and 62% of the time for all cloudy observations with θo > 85°. The gradient appears 58% of the time for the midlevel cloud decks sampled here, 86% of the time for the low-level cloud decks, and only 43% of the time under stratus fractus of bad weather. The smaller fraction of occurrence for the midlevel clouds is somewhat surprising, as Podgorny and Lubin [1998] predict that the gradient should be more pronounced for clouds with higher base altitudes. Examination of Table 2 suggests a possible explanation. Most of the midlevel cloud observations also have scattered low clouds associated with them, and this might provide a source of horizontal cloud inhomogeneity that partially obscures the gradient.

Figure 3.

Histogram of the ratio RJ at 380 nm (a) for all cloudy observations, (b) for data obtained under both midlevel and low-level cloud decks, and (c) for data obtained under stratus fractus of bad weather.

Figure 4.

Histogram of the ratio RO/RJ at 380 nm for all cloudy sky measurements having RJ < 1 (a) for all cloudy observations, (b) for data obtained under both midlevel and low-level cloud decks, and (c) for data obtained under stratus fractus of bad weather.

3.2. Wavelength Dependence

[11] Podgorny and Lubin [1998] predicted that there should be a wavelength dependence in the flux enhancements due to the high albedo coastline. Over the open water their calculations show a flux enhancement that is larger for 500 nm than for 308 nm for cloud optical depths larger than 10–20. For smaller cloud optical depths, flux enhancements are larger for the shorter wavelength. According to radiative transfer calculations and analysis of Palmer Station UV spectroradiometer data by Frederick and Erlick [1997], there is a wavelength dependence in the effective attenuation of UV and visible radiation by conservative scattering clouds which results from differential extinction due to Rayleigh scattering. At shorter wavelengths the atmospheric transmission through the troposphere is reduced owing to greater scattering optical depth contributed by Rayleigh scattering. Thus a cloud-scattering layer with fixed optical depth will appear to attenuate a larger fraction of the insolation. This mechanism does not depend on the horizontal movement of photons due to surface albedo inhomogeneities.

[12] Figure 5 shows some examples of wavelength-dependent RJ from our data. A wavelength dependence appears for the sample data under low-level stratocumulus clouds (Figure 5a) and under midlevel altocumulus clouds (Figure 5b) but not under stratus fractus of bad weather (Figure 5c). Furthermore, in Figures 5a and 5b the wavelength dependence is barely perceptible during the first half of each time series.

Figure 5.

Examples of the wavelength dependence in RJ (a) under a stratocumulus layer, (b) under an altocumulus layer, and (c) under stratus fractus of bad weather.

[13] We can explain this temporal variability in the wavelength dependence of RJ using plane-parallel radiative transfer theory alone. We first applied a delta-Eddington radiative transfer model [Joseph et al., 1976] to the Outcast Island radiometer data collected simultaneously with the measurements of RJ from Janus Island and T-5 to estimate the effective cloud optical depth that prevailed over the region for the examples of Figure 5. There are two approximations here. First, we are neglecting the two-dimensional multiple reflection effect due to the high albedo coastline, which is expected to be less important at the more distant islet. Second, we are assuming that a plane-parallel cloud optical depth derived from Outcast Island is relevant to the downwelling flux at all three locations. The results of the previous section show that neither assumption is entirely valid: both two-dimensional radiative transfer effects and small-scale cloud horizontal inhomogeneity are important. However, plane-parallel optical depth estimates from the Outcast Island radiometer data can serve to illustrate order of magnitude variability in cloud attenuation. In Figure 6 we show the cloud optical depths derived from all two-minute measurements in the examples of Figure 5. The time series begins near weather observations 46, 39, and 15 (Table 2) for the stratocumulus, altocumulus, and stratus fractus cases, respectively, and solar zenith angle increases with time in each case. Under the stratus fractus of bad weather the effective cloud optical depth is consistently larger than 20. In the other two examples the effective cloud optical depths are of the order of 20 at the beginning of the time series then happen to decrease throughout much of the second half of the time series.

Figure 6.

Estimates of the effective cloud optical depth derived from 380 nm flux measurements at Outcast Island, corresponding to the three examples of Figure 5.

[14] In Figure 7 we have repeated the type of radiative transfer calculation done by Frederick and Erlick [1997], using the delta-Eddington model, for the approximate conditions of our region during summer. We calculate the cloud transmission TC(λ) = F(λ)/FC(λ), from the cloudy and clear sky downwelling surface fluxes F(λ) and FC(λ), respectively, over surface albedos 0.05 (sea surface) and 0.7 (land covered by glacier). These calculations were done first with an optically thick cloud (optical depth 27) and then with an optically thin cloud (optical depth 5). As expected from Frederick and Erlick [1997], TC(λ) decreases with increasing wavelength for most UV-B, UV-A, and visible wavelengths. The exception occurs for the optically thick cloud over high surface albedo, under which cloud attenuation appears stronger in the UV-B and then is constant with wavelength throughout the UV-A and visible. If we evaluate the ratio of the downwelling flux at sea to that over land, our simulated RJ(λ), we find that this ratio increases with wavelength in clear skies. Under a large cloud optical depth (27), the simulated RJ (λ) is constant in wavelength. Under a small cloud optical depth (5) the simulated RJ (λ) increases with wavelength. This is what we notice in the measurements of Figure 5.

Figure 7.

Delta-Eddington radiative transfer simulations of the cloud transmission TC(λ), over a high albedo surface (solid curve) and over a sea surface albedo of 0.05 (dotted curve) (a) for cloud optical depth of 27 and (b) for cloud optical depth of 5. The upper and lower dashed curves denote the ratio in flux over a sea surface to that over a high albedo surface under a clear sky and under the cloud optical depth, respectively.

[15] To determine if this wavelength dependence occurs throughout the data, we selected all measurements of RJ(λ) for which a there was a matching cloud effective optical depth derived from Outcast Island data. For each measurement we obtained the slope of a linear regression on RJ(λ)/RJ (305 nm). These slopes were averaged over 20-min intervals and sorted into bins of corresponding cloud optical depth. The resulting histogram is shown in Figure 8. There are large standard deviations in these slopes because this is a subtle measurement approaching the radiometers' performance limit. However, for cloud optical depths of 15 or less we see a tendency for the slopes in RJ (λ)/RJ (305 nm) to be larger than those for larger cloud optical depths. This suggests that the wavelength dependence illustrated in the examples of Figure 5 is manifest throughout the data set.

Figure 8.

Histogram of the mean slopes R(λ)/R(305 nm), averaged over Outcast Island cloud optical depth bins of width 5. The vertical bars give 1 σ standard deviations in these slopes.

4. Conclusions

[16] Under the natural cloud cover variability in the Antarctic Peninsula region the enhancement of downwelling shortwave surface flux over the open ocean due to high coastal albedo appears ∼60% of the time. This effect is more pronounced when cloud cover is overcast and appears in distinct stratiform layers. Even under such stratiform conditions sampled here, the low cloud decks from our weather observations, horizontal cloud inhomogeneity was significant enough to obscure the coastal albedo effect 14% of the time. Low-pressure systems frequently bring to the region multiple layers of stratus and stratocumulus clouds, with underlying patches of stratus fractus. Although the sky is overcast in these cases, the presence of the stratus fractus of bad weather signifies high cloud optical depth with large horizontal inhomogeneity. Under these conditions the coastal albedo effect appears in our data less than half the time. When we examine the wavelength dependence of the flux ratios between the islet and coastal radiometer installations, we find that the one consistent effect that appears in our data, increasing values of the ratio at longer wavelengths under smaller cloud opacity, can be explained by plane-parallel radiative transfer theory.

[17] Although these data demonstrate the existence of the coastal albedo effect, they also show that the maritime Antarctic's meteorology is dynamic enough that cloud inhomogeneity over small temporal and spatial scales is an equally important factor governing the surface radiation field in coastal waters. One motivation for this work was the possible need to correct satellite radiation budget mapping algorithms over coastal waters for horizontal radiative transfer effects that would not be treated by the independent pixel approximation. These results suggest that such corrections should include parameterizations for both the coastal albedo effect and small-scale cloud inhomogeneity.

Acknowledgments

[18] This research was supported by the National Science Foundation Office of Polar Programs under NSF OPP-9725403. We thank Palmer Station's science support staff for facilitating regular small boat trips to the islets to keep the experiment running. We thank J. Ehramjian of BSI for engineering support.

Ancillary