First Measurements of the Infrared Sky Brightness at Dome C, Antarctica

The PAERI was used in this study to measure downwelling radiance from the atmosphere. The Polar AERI is a member of a group of AERI instruments developed by the Space Science and g Center at the University of Wisconsin‐Madison for the US Department of Energy's Atmospheric Radiation Measurement program (Knuteson et al. 2004a, 2004b).

The PAERI was operated at Dome C from 2003 January 13 to 29 and again from 2003 December 21 to 2004 January 25. The 17 days in 2003 January were unusually cloudy, with only about 4 or 5 days of clear‐sky conditions. The second field season had a higher percentage of clear skies, with about 14 to 15 days of clear‐sky conditions out of the 36 days of PAERI observations. The PAERI was operated on a tower and was 6 m above the snow surface in 2003 January and 24 m above the surface in 2003 December/2004 January.

The PAERI measures infrared radiance from 3 to 20 μm at a resolution of 1 cm⁻¹. All the PAERI observations used in this study are from the zenith direction, directly above the instrument. Various viewing schedules were used throughout both field seasons, but the PAERI typically viewed at zenith about every 10 to 12 minutes. (The frequency of zenith measurements depends on how many different angles the PAERI is viewing during a sequence and how long the instrument views each angle.) The PAERI field of view (full width at half‐angle) is 2 6 (Knuteson et al. 2004a).

The PAERI measures infrared radiance in two bands: a shortwave band from 3 to 5.5 μm and a longwave band from 5.5 to 20 μm. The instrument uses a "sandwich" indium antimonide (shortwave) and mercury cadmium telluride (longwave) detector that is cooled to either 68 or 78 K by a Stirling‐cycle cooler; a detector temperature of 78 K was used during the first field season, while 68 K was used during the second. The PAERI interferometer has a fine spectral resolution, with a resolving power between about 500 at 20 μm and 3000 at 3 μm.

Figure 1 shows examples of PAERI measurements of the spectral sky brightness under clear and cloudy sky conditions. The upper solid line shows the sky brightness under a thick cloud, which results in a radiation curve similar to a Planck function at the temperature of the cloud. The lower solid line shows the sky brightness under clear skies. The spectral regions that are most useful for astronomical observations have low radiance values and represent the portions of the atmosphere with the least emission. The long‐ and shortwave channels of the PAERI are both shown, separated by the vertical dashed line. The radiance values are extremely low, around 4.6, 8.3, 10.5, and 19 μm in the clear‐sky spectrum. These portions of the infrared sky brightness represent some of the clearest spectral regions for astronomical observations. The traditional spectral ranges of the M, N, and Q bands are shown along the top of Figure 1. The extremely cold and dry atmosphere over the Antarctic Plateau has semitransparent spectral regions in the Q band, which are clearly visible at wavelengths greater than 17 μm. Figure 1 shows that the PAERI bandwidth includes the clearest portions of the Q band; the sky brightness increases as the wavelength increases beyond about 19 μm, because of the increasing contribution from the water vapor rotational band, which becomes fairly opaque between 30 and 50 μm.

Zoom In Zoom Out Reset image size

Also shown in Figure 1 are estimates of the noise equivalent spectral radiance (NESR; dash‐dotted line) for both detector channels. The NESR is discussed in detail by Knuteson et al. (2004b) and is based on 2 minute observations of a stable, hot calibration source. Figure 1 shows that the NESR is about 0.3–0.4 Jy arcsec⁻² in the shortwave channel. In the longwave channel, the NESR is about 8–10 Jy arcsec⁻² throughout the N band and increases to about 30–100 Jy arcsec⁻² in the clearest portions of the Q band. The shape of the NESR in both channels is determined by the spectral characteristics of the detectors and the interferometer's optics. The noise is much lower in the shortwave channel than in the longwave, because the InSb detector is much more sensitive than the MCT detector. However, in both channels, the NESR is well below the measured signal of the sky brightness. The overall accuracy of the PAERI data is approximately twice that of the NESR when uncertainties in the temperature and emissivity of the calibration sources are accounted for (see Knuteson et al. 2004b for estimates of these uncertainties).

When measuring the infrared sky brightness over Antarctica, it is important to use an instrument that is calibrated accurately, because of the extremely low radiance values in the atmospheric window regions. The overall accuracy of AERI instruments, such as the PAERI, is discussed in detail by Knuteson et al. (2004b) and is stated as 1% (3 σ) of ambient radiance. The PAERI uses measurements of two infrared sources (one hot and one ambient) to calibrate a third measurement. The emissivity of the infrared sources is greater than 0.990, and the temperature of each source is controlled to better than ±0.03 K (Knuteson et al. 2004b). The temperature of each source is characterized by three thermistors that are traceable to temperature standards at the National Institute of Standards and Technology.

Two calibration checks were performed at Dome C during each field season. Each of these checks was performed with three calibration sources, and two of the calibration sources (hot and ambient) were used to calibrate a third source at an intermediate temperature. The calibrated radiance of the intermediate source is compared with the expected theoretical radiance, which is calculated using the measured temperature and emissivity of the intermediate source. One test was performed inside our research building, with the intermediate calibration source at 317 K, while the other test was performed outside, with the source at 253 K. The difference between the actual calibrated radiance and the expected theoretical radiance is less than 1% in all cases, as expected from Knuteson et al. (2004b).

In addition to the PAERI operations, over 100 Vaisala radiosondes (RS80‐15 GH and RS90) were launched in collaboration with the astrophysics group at the University of Nice. About 30 of these sondes were launched under clear‐sky conditions. Radiosondes measure profiles of temperature and relative humidity, in addition to wind speed and direction, as a function of atmospheric pressure. The sonde data are used to construct model atmospheres for use in radiative transfer modeling (discussed below). Vaisala reports that the temperatures are accurate to 0.1–0.2 K and that the relative humidity values with respect to water (RHw) are accurate to ±3%. The RHw values are somewhat uncertain, because of the long response times (minutes) of radiosonde humidity sensors at low temperatures. However, Hudson et al. (2004) have recently shown that Vaisala RS80 radiosondes can measure temperature, pressure, and humidity accurately in Antarctica if the sensors are properly equilibrated to ambient conditions before launch. (Vaisala RS‐80 GH and RS90 sondes use the same humidity sensor. However, the RS90 sondes use two sensors that are successively heated to prevent the accretion of frost.) Therefore, the sondes launched in 2003 January were equilibrated for at least 10 minutes outside in ambient conditions before being launched. In addition, the sondes were prepared in a tent structure that was near freezing and was, therefore, only 20–25 K above the ambient outside temperature; the absolute humidity inside the tent was probably quite similar to the humidity outside. The RS90 sondes launched during the second field season were stored and prepared for launch outside, at ambient temperatures.

Visual observations were made multiple times each day and when the sky conditions were changing. Clear‐sky ice crystal precipitation (diamond dust, ice haze) was only visually observed a few times during both field seasons. There were occasional observations of radiation fog just above the surface, forming in the late evening and early morning hours. These fogs were always composed of supercooled water droplets, as evidenced by riming on the tower (S. R. Hudson 2004, private communication). During the two field seasons, the radiation fog was only observed in January, not December, probably because of the lower surface temperatures in January. The percentage of time that radiation fog occurred in January is estimated to be about 8%; it occurred on eight occasions for about 6–8 hr each time. Aerosols from station pollution were not present over the field site where the infrared measurements were made, because the site was located about 1 km from the main station and was not downwind of the station's power generator.

The transmission calculations presented below were generated using the line‐by‐line radiative transfer model (LBLRTM) described by Clough & Iacono (1995). The input to this model is similar to the more widely used model MODTRAN. A model atmosphere is constructed, consisting of temperature and gaseous constituent profiles. The gases included here are carbon dioxide, water vapor, ozone, carbon monoxide, nitrous oxide, and chlorofluorocarbons (CFCs; CFC11 and CFC12). The surface values of carbon dioxide, carbon monoxide, nitrous oxide, and the CFCs come from measurements made by NOAA's Climate Monitoring and Diagnostics Laboratory at South Pole Station. The profiles of these gases are similar to those described by Walden et al. (1998).

Four spectral bands are defined in Table 1 that will be used in this paper. Each of these bands is considered a "dark band," because they encompass spectral regions that are relatively free of gaseous emission and absorption. The M^'', N^'', and Q^'' bands were chosen because they have the lowest values of sky brightness in the wider M, N, and Q bands traditionally used by astronomers. The N^' and N^'' bands are used here because these particular bands have been used in other studies of Antarctic sky brightness (Chamberlain et al. 2000). Table 1 also shows how many PAERI wavelengths are averaged within each of these bands.

Figure 2 shows time series of the sky brightness in the N^' band throughout the two field seasons in 2003 January and 2003 December/2004 January. Higher values of sky brightness are indicative of cloud cover, while the lowest values denote clear skies (see Fig. 1). Figure 2 shows that the second half of 2003 January was quite cloudy, with few time periods of completely clear skies, while 2003 December and 2004 January had significant periods of clear‐sky conditions. Figure 3 shows the time period between 2004 January 12 and 17 in more detail. The values of sky brightness for this band are between about 40 and 60 Jy arcsec⁻² under clear‐sky conditions. In addition, a diurnal cycle is present, with peak values of sky brightness occurring around noon (local time). This spectral band (N^') contains several strong water vapor emission lines (visible in the fine‐resolution PAERI spectra), the strength of which are modulated by fluctuations in the near‐surface temperature and humidity. The temperature structure near the surface changes significantly throughout the day as the zenith angle of the Sun changes. The solar zenith angle during this time period varied from about 53° at noon to 84° at midnight (local time), with corresponding changes in the surface temperatures (under clear skies) of roughly −20°C to −40°C. It is much more difficult to measure changes in the actual humidity of the near‐surface air, but the existence of radiation fog at times indicates that the relative humidity increases when the solar zenith angle is large. It is possible that water vapor is sublimated from the surface to the atmosphere during the part of the day when the solar zenith angle is near its minimum.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Figure 4 shows histograms of the sky brightness in each of the four spectral bands. Each histogram has a single peak at low values of sky brightness, with a long tail toward higher brightness values. These distributions from Dome C are not bimodal, as are those reported by Chamberlain et al. (2000) for the South Pole. This indicates that during the summer at Dome C, a wide range of cloud optical depths can exist, and a continuum of atmospheric opacities can occur, from very clear skies to thick, overcast conditions.

Zoom In Zoom Out Reset image size

The histograms also give estimates of the sky brightness in each of the bands under clear skies. Table 2 shows the mode (most probable value), median, and minimum values of the distribution of sky brightness for each band. The distributions shown here, and thus the values shown in Table 2, do not necessarily represent climatological conditions at Dome C, but rather only the conditions during our two field seasons.

As expected from the shape of the Planck curve, the M^'' band has the lowest brightness levels. Values of around 1 Jy arcsec⁻² are comparable to those measured at South Pole Station in winter by Phillips et al. (1999). The values in the N^' and N^'' bands are slightly higher than those reported by Chamberlain et al. (2000), but range from about 30 to 50 Jy arcsec⁻² for clear‐sky conditions. The values in the Q^'' band vary from about 200 to 300 Jy arcsec⁻² under clear skies. These values are much lower than those reported by Smith & Harper (1998), because their results were centered on a band around 17 μm, which is much more opaque than the band reported here (18.69–18.94 μm). In fact, the Q^'' band as defined here appears to be an optimal spectral region for astronomical observations, because it lies between the emission band of carbon dioxide (centered at 15 μm) and the emergence of the water vapor rotational band at wavelengths longer than 20 μm. It is also worth noting that the important fine‐structure line of [S iii] at 18.71 μm lies in a portion of this window that is particularly clear at Dome C.

In addition to knowing how clear the sky can be, it is also important to know how often the sky is clear. The determination of "clear sky" conditions is difficult over the Antarctic Plateau, because of precipitation that can occur in an otherwise "clear sky." Chamberlain et al. (2000) have reported that the N‐band sky brightness in winter is limited not by emission from water vapor, but rather by atmospheric aerosols. Although we show evidence here of a slight diurnal cycle in sky brightness, the presence of any ice crystals or supercooled liquid water droplets (fog) can greatly increase the atmospheric opacity.

By examining PAERI data during the few visual observations of clear‐sky ice crystal precipitation, it was determined that when a human observer begins to visually detect ice crystals in the near‐surface air, the N^'' band sky brightness is about 50 to 70 Jy arcsec⁻². This result was confirmed by calculating "clear sky" brightness thresholds, using the radiance‐ratio method described by Town et al. (2005). This method computes the ratio of the integrated sky brightness (over the bandwidth of the PAERI longwave detector), divided by the integrated Planck function of the near‐surface air temperature (as determined from the brightness temperature between 15.6 and 15.9 μm, a region that contains strongly emitting carbon dioxide lines). The ratio is low (perhaps 0.3 to 0.4) when the sky is clear and when the sky brightness is small compared to a Planck function of the near‐surface air temperature. The ratio is close to unity when the sky is cloudy and the sky brightness is nearly equal to that of the Planck function.

Here we determine three different thresholds for a "clear sky." First, we calculate the sky brightness at zenith, using only the average temperature and humidity profiles from radiosondes launched at Dome C under exclusively clear‐sky conditions. The radiance‐ratio method then provides a threshold. This "clear‐sky threshold" represents our most stringent criterion for truly clear‐sky conditions. The second threshold again calculates the zenith sky brightness, but uses the average temperature and humidity profiles from all of the radiosondes launched at Dome C. This is called the "all‐sky threshold." The clear‐sky and all‐sky temperature and humidity profiles are shown in Figure 5. The primary difference between the two cases is near the surface, where the clear‐sky temperature profile has a lower temperature because of the presence of a stronger temperature inversion under clear skies. In addition, the relative humidity (with respect to ice) is nearly saturated near the surface for the clear‐sky case.

Zoom In Zoom Out Reset image size

The third threshold uses the calculation from the all‐sky threshold, but adds radiance to account for atmospheric hydrometeors, such as ice haze and radiation fog. We approximated this thin cloud as having a temperature of 250 K and an emissivity of 0.05, which was justified by fitting an actual PAERI measurement observed during ice‐haze conditions. This threshold is termed the "ice‐haze threshold"; this is the same clear‐sky threshold used by Town et al. (2005).

Figure 6 shows the thresholds (vertical dashed lines) and the radiance ratios for the entire data set of about 5500 PAERI measurements. The bottom panel is an enlargement of the main peak of the distribution shown in the top panel. (The bin size is different in the two panels.) The clear‐sky and all‐sky thresholds only distinguish the smallest values of the radiance ratios. The ice‐haze threshold encompasses the primary peak of the distribution.

Zoom In Zoom Out Reset image size

Because there is a one‐to‐one correspondence between each radiance‐ratio value and the PAERI spectrum from which it was derived, the maximum value of the sky brightness below each threshold can be determined. Table 3 summarizes these results. The values for the ice‐haze threshold are somewhat arbitrary, since they depend on the definition of the thin cloud used to calculate the threshold. However, the value of 65 Jy arcsec⁻² for the N^'' band agrees quite well with the value deduced by looking at the actual PAERI data during times when ice haze was reported in the visual sky observations.

It is interesting to note that the distributions of the clearest sky data (those below the clear‐sky and all‐sky thresholds) are bimodal, again suggesting the influence of the summertime diurnal cycle. (This is not observed in Fig. 4, because the bin size of the histograms is too coarse.) This is most prevalent in the N^' band (shown as a time series in Fig. 3), but is also somewhat apparent in the other bands as well.

By simply dividing the number of ratios below a given threshold by the total number of observations, an estimate can be made of the fraction of time that exceptionally clear conditions will occur. Again, these values are not necessarily representative of the true climatology of Dome C. Using the results listed in Table 3, one can deduce, for instance, that the sky brightness in the dark portions of the N band will be less than about 50–60 Jy arcsec⁻² for about 10% of the time, and less than about 75 Jy arcsec⁻² for close to 50% of the time, at Dome C in the summertime if the conditions encountered during our two field seasons are representative of climate conditions at this site.

One interesting caveat to these results is the potential effect of radiation fog on astronomical observations. Even though radiation fog occurred about 8% of the time in January, it was almost always confined to a very thin layer above the surface. The observers noticed the fog as they walked up the tower. By the time they reached the height of the PAERI instrument, they could see the top of the fog layer. Therefore, to maximize the amount of time of clear‐sky observing during the summer, infrared telescopes should be placed high enough above the surface to avoid this very shallow, low‐lying fog. Future studies should be conducted at other times of year to determine the frequency of fog and its height above the surface.

The stability of the sky brightness under clear‐sky conditions is also of primary interest. Aristidi et al. (2005) have recently reported on the wind conditions at Dome C, which are very important for atmospheric stability. Because the PAERI data were rarely obtained on a consistent schedule over a long period of time (days), it was difficult to accurately perform spectral analysis (e.g., Smith & Harper 1998). In other words, the longest period of measurements at a fixed sampling rate under clear‐sky conditions was less than 12 hr. As mentioned above, the sampling rate of PAERI measurements was quite long (about 10 minutes), compared with past experiments (Smith & Harper 1998; Chamberlain et al. 2000).

Because of these sampling issues, we simply report the standard deviation of the sky brightness measurements for the longest period of clear‐sky observations during the two field seasons. This was a 5 day period from noon on 2004 January 11 at 12:00 UTC until noon on 2004 January 16, which is basically the data shown in Figure 3. Two hours of data were removed around 00:00 UTC on 2004 January 14, because of the presence of thin clouds. The average sky brightness for the remainder of this time period was 47.7 Jy arcsec⁻² and varied by only 4.4 Jy arcsec⁻² (1 σ). Therefore, except for a brief period of a couple of hours, the sky was extraordinarily stable for several days.

This work was supported by a grant from the National Aeronautics and Space Administration (NAG5‐11112). Vital logistical support for our field work was provided by the Programma Nazionale di Ricerche in Antartide (PNRA), the Institut Polaire Francais–Emile Victor, and the Office of Polar Programs at the National Science Foundation. We thank Karim Agabi (Univ. Nice), Tony Travouillon (Univ. New South Wales), and W. Lance Roth (Univ. Idaho) for launching radiosondes at Dome C. We also thank Stephen Hudson (Univ. Washington) for providing his observations of radiation fog for 2003 December and 2004 January. We acknowledge the Climate Monitoring and Diagnostics Laboratory of the National Oceanic and Atmospheric Administration for providing data on concentrations of trace gases from South Pole Station.

We dedicate this paper to Dr. Mario Zucchelli, late director of the PNRA. This project would not have been possible without his extraordinary vision and generosity.