1. Introduction
a. Background
The Antarctic and Southern Ocean are important for understanding the implications of global climate change in general, and for evaluating and improving Earth system models (ESMs) in particular. ESMs presently give a poor representation of the pervasive low-level Southern Ocean cloud cover, which manifests as an underestimate in shortwave cloud radiative effects (Haynes et al. 2011; Bodas-Salcedo et al. 2019). This deficit has ramifications for numerous dynamical and thermodynamic fields throughout an ESM (e.g., Kay et al. 2016; Mechoso et al. 2016; Hyder et al. 2018; Xiang et al. 2018; Gettelman et al. 2020). Biases in satellite retrievals and the scarcity of in situ Southern Ocean cloud observations (e.g., Seinfeld et al. 2016) make this a particularly challenging problem. This deficiency results from inaccurately simulated cloud properties such as cloud droplet size and number concentration, and processes such as entrainment and precipitation (Mason et al. 2015; Sanchez et al. 2020). In recent years several field programs have made considerable progress toward a better understanding of Southern Ocean cloud cover and aerosol–cloud interactions. These programs have involved primarily aircraft (e.g., O’Shea et al. 2017; McFarquhar et al. 2021) and shipboard campaigns (e.g., Schmale et al. 2019 and references therein; McFarquhar et al. 2021) often deploying advanced aerosol sampling equipment for the first time. Over the Southern Ocean, there is often a trade-off between the sophistication and more advanced capabilities of the instrumentation and the duration and geographical coverage of the observations.
Over the Antarctic continent, and particularly over the regions where the Antarctic ice sheet and ice shelves are most at risk in a steadily warming climate (e.g., Scott et al. 2019), this observational trade-off problem can be even more acute. The West Antarctic Ice Sheet (WAIS) is now the second largest cryospheric contributor to global sea level rise (Mouginot et al. 2019; Rignot et al. 2019), after the Greenland Ice Sheet (GIS). The WAIS, being subject to marine ice sheet instability (Alley et al. 2015), has over the past two decades increasingly shown concerning evidence of rapid retreat past a “tipping point,” particularly in the vicinity of the Pine Island and Thwaites Glaciers (e.g., Joughin et al. 2014; Rignot et al. 2014). Most of the focus on West Antarctica emphasizes a steadily warming ocean, basal melting of the ice shelves, and retreat of the ice sheet grounding line toward deeper underlying topography that gives rise to more rapid glacier outflow (Pritchard et al. 2012; Paolo et al. 2015; Jenkins et al. 2018; Adusumilli et al. 2020). However, recent remote sensing and in situ observations have shown increasing episodes of surface melt during the Antarctic warm season, which inevitably adds stress to the ice shelf–ice sheet system (e.g., Kingslake et al. 2017; Bell et al. 2018). The ice shelves, which are the floating extensions of the grounded ice sheets, provide essential stability against rapid outflow through mechanical buttressing (Fürst et al. 2016). The ice shelves are attached to adjacent terrain and therefore substantially hold back ice sheet outflow at most key West Antarctic locations along the Amundsen Sea Embayment. When meltwater pools on an ice shelf during the warm season, it can filter through the entire ice shelf structure and weaken the structure through a process called hydrofracturing (Pollard et al. 2015; DeConto and Pollard 2016; Gilbert and Kittel 2021; Gilbert et al. 2022).
Several recent studies have emphasized the importance of cloud properties in modulating the Antarctic SEB and driving Antarctic surface melt events (Wille et al. 2019; Ghiz et al. 2021; Kittel et al. 2022), but current large-scale model simulations continue to show deficiencies in Antarctic cloud properties, even when using modern double-moment microphysical parameterizations (e.g., Hines et al. 2019). In one example, Lenaerts et al. (2018) have used the Regional Atmospheric Climate Model second version (RACMO2; van Wessem et al. 2018), a model frequently used for Antarctic simulations, to accurately simulate West Antarctic melt events between 1979 and 2015. In RACMO2, van Wessem et al. (2014) addressed the common cloud liquid water path deficiencies (e.g., Hines et al. 2019; Silber et al. 2019a) over Antarctica by altering the model cloud microphysics to allow for more extensive cloud liquid water transport. This was done primarily by making simple but defensible adjustments to the threshold for ice supersaturation (Tompkins et al. 2007), and the critical cloud content for efficient precipitation (Lenaerts et al. 2018). While these simple alterations allow for sufficient cloud liquid water to contribute radiatively to positive ME and surface melt onset, there is no guarantee that the cloud liquid water path values they produce are entirely accurate, and there is no published validation of them over Antarctica. Ideally, a large-scale model should produce good estimates of cloud amount and cloud liquid/ice water content from something closer to first principles than adjustments of thresholds.
In response to the need to improve understanding of the Antarctic atmosphere, there have recently been several field campaigns and emerging permanent installations of advanced atmospheric science equipment. These include 1) Fourier transform infrared spectroradiometer measurements, micropulse lidar (MPL), and tethered balloon cloud microphysical observations at the South Pole (e.g., Walden et al. 2006; Town et al. 2007; Lawson and Gettelman 2014); 2) multisensor observations at Dome Concordia that include multispectral microwave radiometry (Ricaud et al. 2017); 3) research radars for cloud and precipitation at Dumont d’Urville Station in Adélie Land (Grazioli et al. 2017); 4) a climate observatory at Princess Elizabeth Base in Queen Maud Land, East Antarctica, that maintains a precipitation radar along with comprehensive meteorological measurements whose combined data enable studies of both cloud microphysics and surface mass balance (Gorodetskaya et al. 2015); 5) a French-Italian atmospheric observatory at Dome Concordia deploying a depolarization lidar, microwave radiometer, and broadband radiometer system, to discern the radiative impact of supercooled liquid water clouds (Ricaud et al. 2022); 6) a recently deployed X-band Doppler radar at Syowa Station, used to develop a snowfall climatology as influenced by synoptic-scale disturbances (Hirasawa et al. 2022); and 7) the British Antarctic Survey’s Rothera Station, in the southern Antarctic Peninsula region, which serves as a base for aircraft-based in situ cloud microphysical observations that have revealed details about warm-temperature secondary ice production in Antarctic clouds (Lachlan-Cope et al. 2016). In addition, aerosol, cloud, and radiometric measurements have been made from shipboard on resupply transects between Australia and the three permanent Australian Antarctic research stations (McFarquhar et al. 2021), and an atmospheric observation program is being established at New Zealand’s Scott Base, including an MPL (A. McDonald 2022, personal communication).
b. Scope of the present work
The work presented here is a sequel to the largest Antarctic atmospheric and climate science field program to date, in terms of the number and variety of state-of-the-art instruments deployed: the U.S. Department of Energy (DOE) Atmospheric Radiation Measurement (ARM) Facility West Antarctic Radiation Experiment (AWARE; Lubin et al. 2020). From December 2015 through January 2017, AWARE deployed an ARM Mobile Facility (Mather and Voyles 2013) to Ross Island, at a site near McMurdo Station, comprising cloud research radars, lidars, multiple broadband and spectral radiometers, an aerosol observation suite, and thorough meteorological sampling ranging from surface turbulent flux equipment to radiosondes. AWARE also deployed a smaller suite of instruments to the WAIS Divide Ice Camp during December 2015 and January 2016, comprising surface energy balance (SEB) measuring equipment, an MPL, a shortwave spectroradiometer, microwave radiometers, and sondes launched four times daily (Nicolas et al. 2017; Cadeddu et al. 2022; Riihimaki et al. 2021).
Even with these emerging atmospheric science programs, a recurring limitation in the Antarctic is the relatively small number of major research stations capable of supporting the most current atmospheric science equipment, either year-round or even consistently from one summer season to the next. Very few of these stations are located in the immediate vicinity of the ice sheet and ice shelf regions of greatest concern. For example, AWARE collected data at the summit of the WAIS, at 1851-m elevation. While the AWARE data at the WAIS Divide sampled that region’s synoptic meteorology, it did so far inland from the at-risk ice shelves, and where near-surface conditions were generally colder and drier. Following the overall success of AWARE, including use of AWARE data for large-scale model evaluation (Hines et al. 2019, 2021; Silber et al. 2019a, 2022) we were motivated to develop a compact and transportable instrument suite capable of key SEB measurements, and also some cloud remote sensing, that could be deployed at very remote locations more representative of regions of greatest interest for high-latitude climate change.
We were fortunate to be able to deploy this instrument suite during the austral summer of 2019/20 at Siple Dome Field Camp in West Antarctica (81°39′15″S, 149°00′18″W, 730-m elevation). Siple Dome Field Camp is a United States Antarctic Program (USAP) summer facility that is opened and operated intermittently as required by USAP continental science programs, either as a location for ice core, climatological, or cryospheric field work (Severinghaus et al. 2001; Ahn et al. 2004) or as a ground-transportation waypoint for glaciological or geological studies at even more remote locations such as the Kamb Ice Stream (e.g., Gades et al. 2000; Catania et al. 2005; Jacobel et al. 2009). Siple Dome has an ice core record showing increasing impact of summertime surface melting throughout the Holocene (Das and Alley 2008).
Figure 1 shows the location of Siple Dome, along with the station locations for the other measurement programs discussed above. Geographically, Siple Dome is an elliptical ice ridge approximately 100 km long and 75 km wide, with its major axis oriented northwest to southeast, located between the Kamb and Bindschadler Ice Streams. Its rise is gentle, with mean slopes ∼6 and ∼7 m km−1 on its southern and northern sides, respectively (Scambos et al. 1998). Satellite imagery and radar echo-sounding measurements reveal considerable spatial homogeneity in surface morphology and snow/ice accumulation patterns, compared with the adjacent ice streams (e.g., Nereson et al. 1998, 2000). Therefore on Siple Dome a single point measurement site at the surface should be fairly representative of a slightly larger area, for example, an aggregate of regional model grid cells.
Map of Antarctica with locations of the atmospheric research programs discussed in the text.
Citation: Journal of Climate 36, 19; 10.1175/JCLI-D-22-0731.1
As an example for model evaluation, in this study we compare the observations with the European Centre for Medium-Range Weather Forecasts (ECMWF) fifth-generation meteorological reanalysis data (ERA5; Hersbach et al. 2020). ERA5 provides reforecasts by assimilating myriad satellite, in situ, and surface meteorological observations with a global atmospheric circulation model. ERA5 has a prognostic cloud microphysical scheme whose limitations over Antarctica have already been substantially determined (Silber et al. 2019a). In addition to near-surface meteorology and meteorology on 37 pressure levels from 1000 to 1 hPa, ERA5 provides model estimates of total cloud cover, cloud base height, total cloud liquid and ice water content, and surface radiative fluxes, but not cloud optical depth. We present comparison of our observations with some of these variables to demonstrate how this type of dataset can clearly reveal a model’s strengths and weaknesses. We note that these surface data may be less useful for atmospheric properties above the surface, such as boundary layer height and inversion strength, which influence cloud formation.
With the successful data collection from this field program, we present a challenge to the climate modeling community: Can any current-generation models simulate atmospheric conditions and cloud properties with enough fidelity to continuously match the observed surface energy balance?
2. Field program
Observations at Siple Dome began with broadband radiometer data collection on 21 December 2019, followed the next day by turbulent flux measurements. The shortwave spectroradiometer and all-sky camera were both operational by 23 December, although the spectroradiometer was not operated during a brief period of bad weather between 24 and 26 December. Data collection concluded by 0600 UTC 19 January 2020. Figure 2 illustrates how the instruments were deployed.
Photographs of the equipment: (a) the broadband radiometer stand on 2-m truss towers, with the eddy covariance system approximately 100 m beyond, and the University of Wisconsin Automatic Weather Station in the distance (upper right of the photograph); (b) the Science Rac-Tent, with the shortwave spectoradiometer and all-sky camera housed at the top of the truss tower by the door.
Citation: Journal of Climate 36, 19; 10.1175/JCLI-D-22-0731.1
a. Equipment
We used a set of Kipp and Zonen (Inc.) instruments to measure the broadband radiative fluxes. The primary measurements of downwelling and upwelling shortwave flux were made by CMP22 and CMP10 pyranometers, respectively. The primary measurement of downwelling longwave flux was made by a CGR4 pyrgeometer. Backup measurements of these three SEB components, and the primary measurement of upwelling longwave flux, were made by a CNR4 net radiometer. Surface skin temperature was measured using a downward-looking Apogee Instruments (Inc.) SI-111 infrared radiometer supplied and calibrated by Campbell Scientific (Inc.). Data from these instruments were recorded by Campbell Scientific CR1000X and CDM-A108 dataloggers powered by a 20-W solar panel charging a 12-V 24-Ah (amp hour) battery. These instruments were mounted on a 2-m stand comprising an I-beam supported by two truss towers, configured with set screws and angle adjustments to provide rough two-axis leveling capability. Fine leveling was done with the stand adjustments on the individual Kipp and Zonen instruments. Measurement uncertainties involving pyrgeometers (generally of order ±2 W m−2 under cloudy conditions) are reviewed by Wang and Dickinson (2013). Measurement uncertainties involving pyranometers are discussed by Bush et al. (2000).
Turbulent fluxes were measured with a Campbell Scientific IRGASON eddy covariance system comprising an integrated sonic anemometer and a CO2/H2O open path gas analyzer. This system was installed 50 m farther from the station than the radiometer stand, with the sonic anemometer sensor oriented into the prevailing wind direction when conditions do not involve bad weather. The system was powered by two 90-W solar panels charging a 12-V 84-Ah battery. Once the system was started on 22 December 2019, it ran continuously throughout the field program.
The shortwave spectroradiometer system consisted of two StellarNet (Inc.) spectrometer modules connected by fiber optic cables to a single radiometric cosine receptor, an upward-looking Teflon diffuser sampling shortwave irradiance (flux) from the full 180° field of view. The visible wavelength range (350–1150 nm) is covered by a miniature BLUE-Wave spectrometer with a 2048 element Si detector array and 50-μm slit, for a spectral resolving resolution of 1.6 nm. This spectrometer was powered directly by the datalogging computer’s USB port. The near-infrared wavelength range 900–1700 nm is covered by a miniature DWARF-Star spectrometer with a 512-element InGaAs photodiode array and 25-μm slit for a resolving resolution of 2.5 nm. A thermoelectric cooler maintains the detector temperature at −10°C. Both spectrometers have 16-bit digitization and integration times as fast as 0.1 ms.
Digital all-sky photographs were taken using an ALCOR System ALPHEA 6CW color camera with 2.7 megapixels and a 180° field of view fisheye lens. Exposure time was controlled by software, and would typically vary between 32 and 320 μs. The all-sky camera and the spectroradiometer system were installed together in a ruggedized L-com (Inc.) 45 cm × 40 cm × 20 cm enclosure with a small space heater and temperature controller maintaining an interior temperature of 0°C. This enclosure was mounted on a truss tower approximately 2.5 m above ground to clear the roof of the adjacent science Rac-Tent. This heated enclosure, the all-sky camera, the InGaAs detector’s thermoelectric cooler, and datalogging computer operating both the camera and spectroradiometer, were powered by the camp’s solar photovoltaic system.
b. Interpreting the all-sky photographs
The ALPHEA 6CW system is a color CCD camera viewing a fixed f/1.6 lens with no auto-iris capability. Scene illumination varies widely throughout a 24-h diurnal cycle, and also throughout the 180° field of view particularly when the direct solar beam is present. While these digital images are excellent for showing the texture of cloud cover, and hence for estimating sky coverage, the color balance is inconsistent due to the highly variable illumination. Therefore different shades of blue that appear throughout a cloud-free day should not be interpreted in terms of surface visibility conditions or tropospheric aerosol content. Even under overcast skies, the CCD sometimes records a very pale blue rather than a shade of gray.
Figure 3 shows example all-sky photographs to assist with interpretation. These photographs were taken every five minutes throughout the campaign. Figures 3a–c are examples that can be manually interpreted as clear sky (CLR), few or scattered clouds (SCT, sky coverage < ∼50%), and broken clouds (BKN, sky coverage between ∼50% and 90%), respectively. Figure 3d shows a bright patch off-center to the left, which is the direct solar beam partly visible through an overcast cloud deck, which can be classified as thin overcast (TN OVC). Figure 3e shows a solid overcast layer (OVC), and the relative brightness in the center of the image is consistent with the well-documented zenith brightening in the illumination under overcast skies (e.g., Kittler and Valko 1993; Grant and Heisler 1997); note that the sun is never directly overhead in Antarctica. Figure 3f is also from overcast sky that obscures the direct solar beam, but with an annular pale blue region around the bright center, which is an example of the color balance problem emerging under a relatively bright overcast sky. When confusion arises as to the appearance of pale blue, one can check the downwelling longwave fluxes to make additional discrimination between broken and overcast. With the above considerations in mind, it may be possible to adapt automated image processing methods to derive quantitative cloud fractions from these photographs (e.g., Long et al. 2006).
Examples of all-sky images recorded by the ALCOR System camera at Siple Dome: (a) clear skies (CLR) at 1203 UTC 3 Jan 2020; (b) scattered clouds (SCT) at 0832 UTC 4 Jan 2020; (c) broken cloud cover (BKN) at 1635 UTC 4 Jan 2020; (d) thin overcast (TN OVC) at 0451 UTC 5 Jan 2020; (e) overcast (OVC) at 1800 UTC 11 Jan 2020; and (f) overcast, with an annular color balance issue, at 2035 UTC 12 Jan 2020.
Citation: Journal of Climate 36, 19; 10.1175/JCLI-D-22-0731.1
c. Data
Data from the broadband radiometer stand were recorded in one-minute averages and reduced using manufacturer calibration coefficients dated in February 2019, when these new Kipp and Zonen instruments were shipped to the lead authors’ laboratory. Sensible and latent fluxes were reduced from the IRGASON raw data using the Campbell Scientific Easy-Flux software resident in the instrument’s datalogger. This software also derived wind speed and direction from the sonic anemometer. These four data streams were recorded in 30-min averages. Shortwave spectroradiometer data were recorded in approximately 1-min averages. The spectroradiometer system was not fully automated, and had to be manually started to record data in packets of typically 60 spectra. Each packet would cover slightly longer than one hour, due to the accumulating detector integration time (between 125 and 400 ms for the infrared spectrometer, depending on sun elevation). The spectroradiometer measurements therefore covered approximately 24% of the campaign, with measurements being made during hours of highest sun elevation.
Radiometric calibration of the spectrometer-cosine receptor system was provided by the manufacturer using the standard protocol (Parr 1996) with a National Institute for Standards and Technology (NIST) traceable 1-kW tungsten-halogen lamp mounted 50 cm above the cosine receptor. While in the field, preseason manufacturer radiometric calibration of the visible-wavelength spectrometer showed discrepancies compared with clear-sky radiative transfer calculations, so both spectrometers were sent back to the manufacturer in April 2020 for recalibration. These postseason calibration constants proved satisfactory and were used for data reduction. All-sky photographs were taken approximately every five minutes continuously throughout the campaign.
For this type of commercial spectroradiometer system, a good general estimate of radiometric uncertainty relative to the NIST traceable calibration lamp is on the order of 2% (e.g., Meywerk and Ramanathan 1999), but absolute calibration discrepancies between spectroradiometers can sometimes be as large as 10% (e.g., Flynn 2016). Errors in cosine response of the diffusing optical collector can introduce additional uncertainties on the order of 10% (e.g., Wilson et al. 2018), which become most pronounced under clear skies with direct solar beam illumination. In this work we are considering spectral variability in irradiance measured under overcast skies (i.e., almost entirely diffuse irradiance), for which these absolute radiometric calibration uncertainties have a relatively small effect on interpretation.
All data from this field campaign have been archived at the United States Antarctic Program Data Center based at Lamont-Doherty Earth Observatory (Lubin and Ghiz 2022). The radiometer stand data are provided in three formats: 1) the original sensor data in 1-min averages; 2) a composite 1-min dataset constructed by choosing the most reliable shortwave data under clear skies; and 3) the composite dataset averaged to 30 min, along with the turbulent fluxes and wind speed and direction. The intent of the composite dataset is to avoid some of the pyranometer measurement uncertainties discussed below. The composite dataset was assembled by comparing both upwelling and downwelling sensor data under cloud-free or nearly cloud-free conditions with the diurnal cycle simulated by a clear-sky radiative transfer model (Atwater and Ball 1981), and either the primary or backup (net radiometer) measurement was chosen if it was the best match. Under cloudy skies only the primary sensor data are included in the composite.
3. Results
a. Surface albedo
We generally expect broadband albedo of a snow surface to be several percent higher under cloud cover than under clear skies (e.g., Key et al. 2001). The spectral albedo of snow is larger at visible than near-infrared wavelengths, and cloud cover attenuates solar irradiance at near-infrared wavelengths more strongly than does the clear sky; hence there is a larger contribution to the broadband reflectance from shorter wavelengths under cloud cover. Over an actual Antarctic snow surface, there are several additional considerations involving both the measuring equipment and natural variability.
Pyranometers generally have a precision limitation related to dome heating effects. Radiation incident on the instrument induces a temperature differential between the outer dome and the thermopile detector. This differential of several kelvin can lead to a radiometric offset of several watts per square meter, and we were not able to measure this differential. Under overcast or mostly cloudy skies, when the incident radiation is almost entirely diffuse, this temperature differential and related offset are relatively small. When the direct solar beam is present, the temperature differential can be as large as 4 K, with corresponding radiometric offsets in the range 10–15 W m−2 (Bush et al. 2000). In addition to this time-varying offset, pyranometer measurements under the direct solar beam are susceptible to shading and shadowing effects from nearby structures, or the structure supporting the instruments, particularly under low solar elevation. All of these effects can yield spurious values for the surface albedo under clear skies. In our dataset, we identify data as suspect for these reasons during periods lasting 1–3 h on 30 and 31 December 2019, and 4, 7, 9, 12, 15, and 18 January. The primary indicator of suspect data is surface albedos exceeding unity as a result of these complicating effects. Most of these suspect data were recorded during the hours 1000–1300 UTC.
Over an Antarctic snow surface, the albedo is influenced by evolution of snow grain size, and, under clear skies, also strongly depends on solar zenith angle (Pirazzini 2004; Kuipers Munneke et al. 2008). Kuipers Munneke et al. (2011) show five continuous years of albedo measurements from Neumayer Station during which summertime values varied between 0.73 and 0.95. While the effects of sastrugi on Antarctic snow albedo and bidirectional reflectance have been well studied (e.g., Warren et al. 1998; Corbett and Su 2015), small-scale variability in surface roughness and snow evolution also introduces variability in the measured albedo (Yamanouchi 1983).
This field program’s albedo measurements, from the half-hourly composite dataset, are shown in Fig. 4. The observations are sorted into the five sky conditions recorded in the all-sky photographs, as described above. For the first two days, the sky condition in Fig. 4a is “unclassified” because the all-sky camera was not yet operational. The campaign-average albedo values and their standard deviations for the five sky conditions (with suspect data omitted) are 0.81 ± 0.05 (CLR), 0.81 ± 0.05 (SCT), 0.84 ± 0.04 (BKN), 0.86 ± 0.04 (TN OVC), and 0.88 ± 0.02 (OVC). The daily variation in these observations, which is consistent with expected ranges for various sky conditions, should make this dataset useful for making at least a qualitative evaluation of models that contain a sophisticated representation of cryosphere snow cover (e.g., Kuipers Munneke et al. 2011). Finally, we note that the half-hourly composite dataset does not omit any pyranometer measurements, and includes the measurements identified above as suspect data; final interpretation or rejection is left to the user.
Observed surface albedo from the composite dataset averaged at 30 min, sorted by sky condition estimated from the all-sky images. Segments of the time series specified as unclassified (UNCL) are when the all-sky camera was not operational.
Citation: Journal of Climate 36, 19; 10.1175/JCLI-D-22-0731.1
b. Cloud optical depth
We retrieve the cloud optical depth from the spectroradiometer data under all thin overcast and overcast sky conditions, using a plane-parallel radiative transfer model following Lubin and Vogelmann (2011). We use the wavelength interval 1022–1033 nm, which is considerably more transparent than at most visible wavelengths, having smaller Rayleigh scattering, no ozone absorption, and negligible water vapor absorption. Most importantly, the albedo of pristine snow at these wavelengths is considerably smaller (∼0.72) than at visible wavelengths, yielding greater sensitivity in the downwelling surface spectral irradiance to cloud optical depth. This wavelength range is in the conservative scattering regime, so the irradiance has negligible dependence on cloud particle size or thermodynamic phase (e.g., Fig. 2 in Coddington et al. 2012). For interpreting irradiance data, Grenfell and Warren (1999) and Neshyba et al. (2003) have demonstrated that equivalent spheres are a suitable representation for ice cloud particles if they are interpreted as having the same volume-to-area ratio as the nonspherical particles. Hence the single scattering albedo and asymmetry parameter are derived from Mie theory in these calculations.
Cloud optical depths throughout the campaign, averaged over 60-scan data packets having 1-min time resolution, are shown in Fig. 5. At this very high latitude, optical depths under solid overcast skies can be as large as those found over the Antarctic Peninsula (Lubin and Frederick 1991; Lubin et al. 2002) and over the North Slope of Alaska (e.g., Dong and Mace 2003). Such clouds, the result of substantial moisture intrusion over the continent, will radiate in the longwave as blackbodies, and may contribute to a thermal blanketing mechanism for inducing surface melt under the right near-surface meteorological conditions (Ghiz et al. 2021). Campaign averages and standard deviations for cloud optical depths are 15.3 ± 6.7 for solid overcast (OVC) and 4.1 ± 2.3 for thin overcast (TN OVC), these categories being manually identified from the all-sky photographs (Fig. 3).
Conservative scattering cloud optical depth, hourly averaged, for all spectroradiometer data obtained at Siple Dome under overcast skies. Data are averaged over the packets of (most often) 60 spectra as recorded by the StellarNet (Inc.) software. Error bars depict plus or minus one standard deviation.
Citation: Journal of Climate 36, 19; 10.1175/JCLI-D-22-0731.1
c. Surface energy flux components
Comparing each of the observed SEB components individually to model simulations gives insight into the model’s strengths and limitations. Here we compare the observations with corresponding ERA5 quantities averaged in the 1° × 1° grid cell cluster over the field camp. We examined the spatial variability in ERA5 surface shortwave and longwave fluxes within this 1° × 1° grid cell cluster (the original ERA5 resolution is 0.25°). Throughout the campaign, spatial standard deviations ranged from 0.9% to 14.4% for the shortwave and 0.7%–12.2% for the longwave. ERA5 spatial variability in total cloud amount is much larger, although the standard deviations tend toward smaller values of order < 14% when the area-averaged cloud amount is 90% or larger.
Comparing the observed net shortwave flux with ERA5 (Fig. 6a), we notice several sets of consecutive days showing good qualitative agreement between observations and the model (e.g., days 358–360 of 2019, days 13–16 of 2020), but also sets of consecutive days when the observed net fluxes are smaller compared to ERA5 (e.g., days 361–364 of 2019). This discrepancy would result if the model underpredicts cloud amount or optical depth. Examining the net longwave flux (Fig. 6b), we see that the clusters of days showing the shortwave discrepancy also show large (negative) simulated longwave flux, while the observed net longwave flux is close to zero (days 361–364 of 2019). This suggests that the field camp is actually fully covered with opaque clouds, radiating close to a blackbody, while the model cloud cover has much lower optical depth and allows considerable longwave radiation to escape from the surface to space. However, on other days identified in the all-sky photographs as having mainly scattered clouds or clear skies, we find agreement between ERA5 and observations within 20 W m−2.
Surface energy balance components and related meteorological quantities observed throughout the Siple Dome campaign, compared with ERA5 reanalysis.
Citation: Journal of Climate 36, 19; 10.1175/JCLI-D-22-0731.1
Comparing the observed surface sensible heat flux (Fig. 6c) and latent heat flux (Fig. 6d) with ERA5 reveals good qualitative consistency in both the diurnal cycles and overall trend throughout the field program. The observations do show numerous apparent outliers compared with ERA5, which may signify some lower-quality or less reliable data collected in the harsh environment (and which could be omitted for model comparisons). These eddy covariance measurements might therefore require, in some cases, averaging or smoothing, instead of using them for point-by-point comparisons. Generally speaking, this eddy covariance dataset suggests that ERA5 turbulent flux diagnostics are realistic overall. In comparing the observed skin temperature with ERA5 (Fig. 6e), we also see good qualitative consistency in both the diurnal cycles and overall trend throughout the field program. However, we also see two clusters of days where the ERA5 diurnal minimum is at least 2 K below the observation (days 361–363 of 2019 and days 5–7 of 2020). These days correspond to the days with largest net longwave discrepancy (Fig. 6b), suggesting that the model-simulated cloud cover is not adequately warming the surface, as compared with the observations. Finally, comparison of 2-m wind speed (Fig. 6f) and wind direction (Fig. 6g) reveals that ERA5 largely provides excellent meteorological fidelity. This excellent agreement may be due to the Automatic Weather Station (AWS; Lazzara et al. 2012) at Siple Dome, which has World Meteorological Organization Identifier 89345, and was likely feeding data into the Global Telecommunication System (GTS) that informs the major global reanalysis programs including ERA5 (M. A. Lazzara 2023, personal communication).
4. Case studies
a. Squall, 24–27 December 2019
At least two case studies in this dataset enable more specific evaluation of model deficiencies. A brief period of bad weather with moderately high near-surface winds began on 24 December and gradually subsided over the next two days. Figure 7 shows the synoptic conditions at 0000 UTC 25 December. High pressure over the East Antarctic plateau and low pressure over the eastern Ross Sea give rise to moderate cyclonic flow over Siple Dome, with midtropospheric steering wind speeds in the range 15–25 m s−1.
ERA5 (a) 700-hPa geopotential height and (b) vector wind over Antarctica at 0000 UTC 25 Dec 2019. The location of Siple Dome is depicted by a white star.
Citation: Journal of Climate 36, 19; 10.1175/JCLI-D-22-0731.1
Examining the near-surface wind speed (Fig. 8a) and wind direction (Fig. 8b) shows good agreement between ERA5 and observations, with ERA5 showing consistently northerly winds during the squall, and then showing the subsequent backing to slightly west of north (days 360–361). Examining the ERA5 spatial variability in the 1° × 1° grid cell, ERA5 total cloud cover is nearly unity between 24 and 26 December, and 0.71 with a standard deviation of 15.4% on 27 December.
Comparison of observations from the squall case study 24–27 Dec 2019 with ERA5 reanalysis: (a) wind speed, (b) wind direction, (c) net shortwave flux, (d) net longwave flux, (e) net turbulent flux, (f) estimated melt energy flux, (g) skin temperature, and (h) time series of ERA5 cloud liquid water path and ice water path and observed cloud optical depth. Cloud optical depth retrieved from the spectroradiometer are unavailable before day 360.
Citation: Journal of Climate 36, 19; 10.1175/JCLI-D-22-0731.1
Comparing the observed and ERA5 simulated net shortwave flux (Fig. 8c) and net longwave flux (Fig. 8d) reveals good agreement, typically within 20 W m−2, during the period of high winds (days 358–360), although the consistent longwave discrepancies within the time period 358.5–359.25 suggests that ERA5 underestimates cloud optical depth or cloud fraction, or overestimates cloud base height. This is followed by steadily increasing discrepancy from day 360 onward as the squall subsides. In contrast, the simulated turbulent flux (sum of sensible and latent heat; Fig. 8e) shows the greatest discrepancy just as the squall begins (first half of day 358), but for the next three days there is good agreement between ERA5 and observations in sign and diurnal cycle. There is a discrepancy in magnitude late in day 359, which may be related to some intermittent cloud clearing noted in the all-sky photographs (figure not shown). The melt energy flux derived from observations and Eq. (1) (Fig. 8f) shows good agreement with that estimated using ERA5 surface energy balance terms throughout the high wind period, with two minima in diurnal cycle near zero on days 359 and 360. Subsequently the ERA5-derived melt energy flux diverges considerably from that derived from observations, showing energy loss from the surface throughout most of day 361 while the melt energy flux derived from observations remains positive and warms the surface. This is reflected in the time series of skin temperature (Fig. 8g). Between the second half of day 358 and early day 361, ERA5 and observed skin temperatures agree to mostly within 2 K, after which the large drop in ERA5 skin temperature relative to the observation is consistent with the increasing discrepancy in melt energy flux. The large skin temperature discrepancy early in day 358 (up to 4 K) may be related to large temporary discrepancies in both sensible and latent heat fluxes (Figs. 8e and 4c,d) at the onset of the squall.
Explanations for most of these discrepancies can be found by examining the ERA5 cloud liquid water path (LWP) and ice water path (IWP), as shown in Fig. 8h. Between days 358 and the first half of day 361, the simulated total cloud water path would be sufficient to yield a melt energy flux consistent with that derived from the observations. However, the high IWP values are probably artificial (Silber et al. 2019a) and compensate for low LWP values that alone would not yield enough net longwave surface flux to give a mostly positive melt energy flux as observed. By the end of day 360 and throughout day 361 the ERA5 cloud microphysical scheme essentially stops producing ice or liquid water. During that period, the all-sky photographs throughout most of day 361 show solid overcast skies in which the direct solar beam is obscured (figure not shown). In addition, the cloud optical depths retrieved from the spectroradiometer data are substantial (the spectroradiometer was not operated during the preceding squall period). Assuming a typical cloud effective droplet radius of ∼12 μm as measured during AWARE by Wilson et al. (2018) at WAIS Divide in December 2015 and January 2016, the retrieved cloud optical depths are consistent with LWP in the range 60–140 g m−2. Thus, the latter half of this case study illustrates an episodic failure of the model to produce cloud cover and properties. One possible contribution to this episodic failure may be related to synoptic meteorology in ERA5. By the beginning of day 361 (27 December), the low pressure over the eastern Ross Ice Shelf shown in Fig. 7 has shifted eastward over the West Antarctic continent (figure not shown), such that Siple Dome is now at the very center of this cyclonic flow, and wind speed has greatly decreased in both ERA5 and the observations. The ERA5 microphysical scheme produces broken rather than overcast sky conditions in the center of this cyclonic flow, with dramatically reduced cloud water content.
b. Mixed-phase cloud life cycle, 10–13 January 2020
During 11–13 January 2020 we noticed a subtle diurnal effect in the SEB, driven by a diurnal transition in cloud microphysical properties. Figure 9 illustrates the synoptic conditions at 0000 UTC 12 January that give rise to relatively gentle moisture advection and cloud formation over the field camp. Throughout Antarctica, circulation on this day is dominated by two systems: a strong high pressure region over East Antarctica between Queen Maud Land and the Amery Ice Shelf interacting with a strong low pressure region over the Weddell Sea, and a strong high pressure region over the Amundsen Sea interacting with a strong low pressure region over the Southern Ocean adjacent to Wilkes Land. Between the two high pressure regions, a weak low over the eastern Ross Sea induces clockwise flow that advects moisture over Siple Dome from the Amundsen Sea.
ERA5 700-hPa geopotential height (a) and vector wind (b) over Antarctica at 0000 UTC 12 Jan 2020. The location of Siple Dome is depicted by a white star.
Citation: Journal of Climate 36, 19; 10.1175/JCLI-D-22-0731.1
During this time period, mixed-phase clouds from lower elevations were advected “up the dome” during the early morning. These clouds were then observed to precipitate throughout the morning hours (snowfall) until they were a thin overcast layer by afternoon. This visually inferred transition from obviously mixed-phase cloud to optically thin and potentially liquid water cloud is evident in the combination of our broadband and spectral radiometric data. Examining the ERA5 spatial variability in the 1° × 1° grid cell, ERA5 total cloud cover is 0.501, 0.978, and 0.996 for the start of 11, 12, and 13 January, respectively, with respective standard deviations of 86.2%, 4.0%, and 1.0%. So in this respect ERA5 is realistically simulating total cloud amount over the field camp.
Figure 10 shows the observed individual broadband radiative flux components. The second half of 10 January was mostly cloud-free, and the subsequent two days saw overcast skies. This is evident in the shortwave fluxes (Fig. 10a), which show peak values on 11–12 January ∼60% of that on 10 January. The subtle diurnal effect is seen in the longwave flux components (Fig. 10b) and the net (shortwave plus longwave) fluxes (Fig. 10c), which are mostly consistent with clear skies and scattered cloud cover during late 10 January and early 11 January. In the early morning hours (1300–1500 UTC) of 11 and 12 January, overcast cloud cover is optically thick enough that the net surface longwave flux is close to zero. Then gradually over the next ∼18 h the cloud cover becomes optically thinner such that the net longwave flux steadily decreases to ∼−20 W m−2. Adding the turbulent fluxes (Fig. 10d) we see that the melt energy flux derived from observations is mostly positive throughout this period, except during the hours of lowest sun elevation (i.e., lowest shortwave flux). In contrast, ERA5 simulates a diurnal cycle in melt energy flux that is negative half of the time. Examining the sensible and latent heat fluxes (Figs. 6c,d) reveals that approximately half of this discrepancy between the observed and ERA5 melt energy fluxes is due to ERA5 underestimates of ∼5 W m−2 in each of these turbulent flux components. Finally, we note here that the “melt energy flux” does not signify actual surface melting; it is the SEB from observations and Eq. (1) with the ground conduction term G set to zero, and its variability influences the diurnal cycle in surface skin temperature.
Surface energy balance components observed during the mixed-phase cloud life cycle case study 10–13 Jan 2020: (a) shortwave flux, (b) longwave flux, (c) net radiative fluxes, and (d) melt energy flux from observations and ERA5, along with observed surface skin temperature.
Citation: Journal of Climate 36, 19; 10.1175/JCLI-D-22-0731.1
In Fig. 10d we see that the observed surface skin temperature tracks the diurnal cycle in the melt energy flux derived from observations on 11 and 12 January, with the warmest temperatures occurring ∼2 h after the daily maximum melt energy flux. After the daily maximum in skin temperature, the overcast skies have precipitated and transitioned from mixed phase to mainly liquid water clouds with smaller optical depth, but they still provide enough downwelling longwave flux when they are present to limit the subsequent skin temperature cooling to ∼4°C below daily maximum on early (UTC) 12 January and ∼1°C below daily maximum on early (UTC) 13 January. All-sky photographs show that during the “nighttime” hours (short intervals around fractional days 11.4, 12.4, and 13.3) sky coverage is scattered to broken (figures not shown). This allowed the surface to rapidly cool to space such that the observed skin temperature dropped into the range from −16° to −20°C. Arrival of the overcast deck in the early morning hours caused the surface to warm rapidly after these diurnal minima. Hence the melt energy flux derived from observations can explain the corresponding changes in skin temperature, while the ERA5-simulated melt energy flux cannot.
The transition from mixed-phase to mainly liquid water clouds on 11 and 12 January is evident in the spectroradiometer data, examples of which are shown in Fig. 11. The cloud phase information is contained in the wavelength range 1450–1600 nm, while irradiance around 1150 nm is less sensitive to either phase or effective particle size for clouds of small or moderate optical depth (e.g., McBride et al. 2011; LeBlanc et al. 2015). A reference spectrum obtained under the clear-sky period early on 11 January (0447 UTC) is shown in Fig. 11a. Irradiance is significantly attenuated by clouds just before solar noon on 11 January (1807 and 254 UTC 11 January), and these two spectra in Fig. 11a also show a general spectral irradiance increase with wavelength between 1450 and 1600 nm, consistent with clouds containing significant ice water content (Scott and Lubin 2014; LeBlanc et al. 2015; Wilson et al. 2018). After solar noon (maximum downwelling shortwave flux) on 12 January we see a transition in the 1450–1600-nm spectral irradiance, toward an overall pattern that is concave downward, consistent with mainly liquid water clouds. This transition repeats itself the following day (Fig. 11b), where late morning spectra (1912–2131 UTC 12 January) show 1450–1600-nm irradiance increasing with wavelength, and spectra from later that afternoon show a pattern that is concave downward. These sequences of shortwave spectral irradiance measurements therefore appear to show a transition from mixed-phase clouds in the morning to mainly liquid water clouds in the afternoon.
Examples of near-infrared downwelling spectral irradiance measured by the StellarNet (Inc.) spectroradiometer during the mixed-phase cloud life cycle case study 10–13 Jan 2020 during the (a) first diurnal cycle and (b) second diurnal cycle. The spectrum recorded at 0447 UTC 11 Jan 2020 was obtained under clear skies. The other spectra were obtained under overcast skies at various stages of cloud evolution.
Citation: Journal of Climate 36, 19; 10.1175/JCLI-D-22-0731.1
At Siple Dome there were no independent observations of cloud phase, so we examine a different dataset to demonstrate that these three parameters apply just as well to measured hemispheric downwelling irradiance as radiance. Fortunately, a very similar dataset was obtained at the WAIS Divide during AWARE using a Panalytical (Inc.) shortwave spectroradiometer, and this AWARE dataset has concurrent cloud observations from a micropulse lidar (MPL; Flynn et al. 2007) and an ARM Facility Total Sky Imager (TSI; Long et al. 2006). Observed MPL variables such as the extent of depolarization in the MPL return signal can provide essential information of cloud phase (e.g., Intrieri et al. 2002). By using MPL observations to sort the AWARE shortwave irradiance data obtained under overcast skies into categories of mainly liquid water, mainly ice, and mixed phase, as described below, we can evaluate the three parameters from all the irradiance spectra, note how these measured parameter values vary with phase as compared with the theoretical predictions in LeBlanc et al. (2015), and determine if the parameters provide any separability between these three phase categories.
Here we use an MPL cloud mask derived using the method of Silber et al. (2018a) applied to the WAIS Divide MPL data in Silber et al. (2019b). This cloud mask has an hourly time resolution, and identifies the presence or absence of water particles (hydrometeors) with a vertical range resolution of 15 m from the surface up through a maximum potential range of 30 km. At each range and time increments (cloud mask grid cell), the cloud mask identifies the total hydrometeor volume fraction and the volume fraction of hydrometeors that are in the liquid phase; the ratio of liquid to total hydrometeor fractions specifies the liquid water droplet fraction in the grid cell. Although phase discrimination in polarized lidar returns is conceptually straightforward—for example, depolarization ratios < 0.1 can be identified as liquid water clouds as in Wilson et al. (2018)—there are various sources of uncertainty. For example, specular reflection from plate-like ice crystals can yield very low lidar depolarization even in the case of scanning lidars (e.g., Noel et al. 2002; Silber et al. 2018b, their appendix A), while the signal-to-noise ratio is generally lower for returns from higher altitudes and/or after encountering hydrometeors over a certain optical depth. The processing of Silber et al. (2019b, their appendix A) mitigates some of these uncertainties by combining backscatter cross-section estimates, depolarization and backscatter signal-to-noise ratio fields, and linear depolarization ratio measurements. Here, we assume that if 75% or more of hydrometeors at all altitudes in an overcast-sky hourly time increment are liquid water droplets, then the cloud deck is considered mainly liquid water. If fewer than 10% of hydrometeors at all altitudes are liquid water droplets, the cloud deck is considered mainly ice water cloud. Between these thresholds, the cloud deck is considered mixed phase.
For all the overcast-sky AWARE WAIS Divide irradiance spectra, conservative scattering cloud optical depth was retrieved as in Lubin and Vogelmann (2011), and the three above parameters of LeBlanc et al. (2015) were evaluated. Results are sorted into the three MPL cloud mask-derived categories, as shown in Fig. 12. When evaluating these parameters from irradiance data, their dependence on cloud phase is consistent with the radiance calculations of LeBlanc et al. (2015), in that for optical depths in the range 1–15, values for η3 and η8 increase with a transition from ice to liquid water, while values for η15 decrease with a transition from ice to liquid water. Figure 12 suggests that the parameter values themselves may offer some cloud phase discrimination capability from spectral irradiance data alone.
Spectral radiometric parameters from LeBlanc et al. (2015) evaluated from the Panalytical (Inc.) shortwave spectroradiometer measurements made at WAIS Divide during AWARE, as a function of retrieved conservative-scattering cloud optical depth and sorted into liquid, mixed-phase, and mainly ice phase categories using the cloud mask of Silber et al. (2018). Student’s t-test statistics are also given for discrimination potential between the three cloud phase categories, for each of the three parameters.
Citation: Journal of Climate 36, 19; 10.1175/JCLI-D-22-0731.1
Figure 12 also includes Student’s t-test statistics for all of the optical depth bins. These statistics were calculated by adjusting the AWARE data sample sizes assuming lag-1 autocorrelation of ∼0.90 for solar irradiance measurements made at one-minute intervals (e.g., Skartveit and Olseth 1992; Hansen et al. 2010), following Bretherton et al. (1999). These statistics suggest the potential for differentiating liquid water cloud from ice cloud, and mixed-phase cloud from ice cloud, for all optical depths, using all three parameters. For η3 and η8, there appears to be differentiation between liquid water cloud and mixed-phase cloud for optical depths < 5, while for most larger optical depths the t statistics are just at the threshold of being significant at the 95% confidence level. For η15 differentiation between liquid water cloud and mixed-phase cloud appears only possible at the smallest optical depths, and differentiation most optical depths does not rise to the 95% confidence level. We note here that these t statistics are sensitive to the degree of measurement autocorrelation and related effective sample size adjustment required. Full development of a practical cloud property retrieval algorithm based on LeBlanc et al. (2015), while beyond the scope of this introductory paper, is an ongoing effort.
Nevertheless, Fig. 12 provides enough information that we can confidently describe the phase transitions between mixed-phase and liquid water cloud in this case study. Figure 13 shows the parameter values and cloud optical depths derived from the individual Siple Dome irradiance spectra during 10–13 January. During the mostly clear-sky period, from late 10 January into early 11 January, η3 and η8 values are distinctly larger than any of the cloudy-sky periods that follow (η15 values are distinctly smaller). During the two overcast days, cloud optical depth decreases steadily from morning through afternoon, and the parameter values transition from the ice water or mixed-phase range during morning to a distinctly liquid water range by afternoon. And by afternoon the cloud optical depths are in the range where liquid water cloud should be distinguishable from mixed-phase cloud. Also, under the larger optical depths between days 11.5 and 12.0, η3 and η8 values are slightly more consistent with ice clouds than mixed-phase clouds (cf. Figs. 12a,b). Thus, we see how shortwave spectroradiometer data covering near-infrared wavelengths can record common types of polar cloud evolution; in this case, the loss of ice water content through precipitation and transition to mainly liquid water cloud before the cloud fully dissipates. Figure 13c shows how the ERA5-simlutated cloud liquid water and ice water paths evolve over this same period, and although the clear-sky period appears in this simulation, the model’s water phase composition does not resemble the observed diurnal phase transition on either 12 or 13 January.
The two diurnal cycles in cloud life cycle observed at Siple Dome as shown in the optical depth and spectral radiometric parameter values (LeBlanc et al. 2015): (a) parameters 3 and 8, (b) parameter 15, and (c) the corresponding time series of ERA5 simulated total cloud liquid and ice water paths.
Citation: Journal of Climate 36, 19; 10.1175/JCLI-D-22-0731.1
5. Summary and conclusions
This study demonstrates how a suite of transportable and solar-powered radiometric and meteorological sensors can provide data capable of large-scale model assessment, including cloud occurrence, optical thickness, and phase, over remote but climatically relevant polar locations such as West Antarctica. The following conclusions and recommendations comprise our challenge to the climate modeling community, using these data from Siple Dome.
-
A first step is to assess the realism in the surface meteorology, particularly if working with a regional model. From this study, the good agreement between observations and ERA5 in surface meteorology and turbulent fluxes suggests that single-column models initialized with ERA5 and time-varying large-scale forcing, or climate models nudged toward ERA5 reanalysis, could be used to evaluate and improve cloud microphysical parameterizations.
-
Determine if the model simulates appropriate cloud cover or clear skies, consistent with the observations on each day. The best observations for this are the all-sky photographs and the downwelling longwave fluxes. With the all-sky photographs, one could perform a manual analysis of each image to identify the basic five sky conditions shown in Fig. 3; this would provide a first-order model evaluation. Or one could adapt an automated image processing algorithm to derive fractional cloud cover, as is done with the DOE ARM Facility (Long et al. 2006). Days for which the model nearly or entirely fails to produce observed cloud cover should be flagged for particular attention in the microphysical scheme’s physics or input.
-
Compare the modeled and observed surface albedo to determine if there is an appropriate response to the presence or absence of cloud cover, and also to see if model precipitation patterns are mirrored in observed albedos for fresh versus old snow.
-
Compare the surface shortwave and longwave fluxes under simulated clouds to the observed values to determine if the model is producing realistic cloud water content. It is possible in models that opposing cloud microphysical biases might cancel one another to yield artificial agreement between modeled and observed broadband radiative fluxes; however, in this dataset the independent cloud optical depth estimation capability from the spectroradiometer can provide an additional constraint. Both optically thick and optically thin clouds can warm an ice sheet surface to the point of affecting melting or refreezing, but via different radiative transfer mechanisms that have a bearing on surface snowpack evolution (Bennartz et al. 2013; Van Tricht et al. 2016; Ghiz et al. 2021). If the model produces a realistic simulation of optically thick versus optically thin cloudy sky periods, its application to current high-latitude climate change problems is promising.
-
As a final stringent test, determine if the model provides a realistic simulation of cloud life cycle. During 10–13 January in this dataset, the spectroscopic retrievals show clear phase transitions for the common polar mixed-phase cloud scenario involving steady precipitation followed by cloud dissipation.
We encourage use of this Siple Dome dataset for climate model evaluation, and also encourage future high-latitude deployment of this type of equipment at a variety of sites to eventually gather geographically diverse data for model evaluation representative of all key polar maritime and cryosphere regimes. Engagement between the climate modeling community and field researchers can also optimize future field campaigns of this type, by iteratively refining the list of atmospheric parameters to measure and improving their related instrumentation requirements.
Acknowledgments.
We thank Jenn Danis and Kyler Hale (U.S. Antarctic Program) for outstanding field camp support, and Dean Einerson (U.S. Antarctic Program) for outstanding logistical support. The field program was supported by the National Science Foundation (NSF) under OPP-1744954. Data analysis was supported by the U.S. Department of Energy under DE-SC0021974. S. Castillo was supported by the NSF under a Research Experiences for Undergraduates (REU) award to the Scripps Institution of Oceanography, OCE-1659793. I. Silber was supported by the U.S. Department of Energy Atmospheric System Research Program under Grant DE-SC0021004.
Data availability statement.
This field program’s data are archived at the United States Antarctic Program Data Center (Lubin and Ghiz 2022). The AWARE WAIS Divide data are available in the ARM Data Archive (https://www.adc.arm.gov).
REFERENCES
Adusumilli, S., H. A. Fricker, B. Medley, L. Padman, and M. R. Siegfried, 2020: Interannual variations in meltwater input to the Southern Ocean from Antarctic ice shelves. Nat. Geosci., 13, 616–620, https://doi.org/10.1038/s41561-020-0616-z.
Ahn, J., M. Wahlen, B. L. Deck, E. J. Brook, P. A. Mayewski, K. C. Taylor, and J. W. C. White, 2004: A record of atmospheric CO2 during the last 40,000 years from the Siple Dome, Antarctica ice core. J. Geophys. Res., 109, D13305, https://doi.org/10.1029/2003JD004415.
Alley, R. B., S. Anandakrishnan, K. Christianson, H. J. Horgan, A. Muto, B. R. Parizek, D. Pollard, and R. T. Walker, 2015: Oceanic forcing of ice-sheet retreat: West Antarctica and more. Annu. Rev. Earth Planet. Sci., 43, 207–231, https://doi.org/10.1146/annurev-earth-060614-105344.
Atwater, M. A., and J. T. Ball, 1981: A surface solar radiation model for cloudy atmospheres. Mon. Wea. Rev., 109, 878–888, https://doi.org/10.1175/1520-0493(1981)109<0878:ASSRMF>2.0.CO;2.
Bell, R. E., A. F. Banwell, L. D. Trusel, and J. Kingslake, 2018: Antarctic surface hydrology and impacts on ice-sheet mass balance. Nat. Climate Change, 8, 1044–1052, https://doi.org/10.1038/s41558-018-0326-3.
Bennartz, R., and Coauthors, 2013: July 2012 Greenland melt extent enhanced by low-level liquid clouds. Nature, 496, 83–86, https://doi.org/10.1038/nature12002.
Bodas-Salcedo, A., J. P. Mulcahy, T. Andrews, K. D. Williams, M. A. Ringer, P. R. Field, and G. S. Elsaesser, 2019: Strong dependence of atmospheric feedbacks on mixed-phase microphysics and aerosol-cloud interactions in HadGEM3. J. Adv. Model. Earth Syst., 11, 1735–1758, https://doi.org/10.1029/2019MS001688.
Bretherton, C. S., M. Widmann, V. P. Dymnikov, J. M. Wallace, and I. Bladé, 1999: The effective number of spatial degrees of freedom of a time-varying field. J. Climate, 12, 1990–2009, https://doi.org/10.1175/1520-0442(1999)012<1990:TENOSD>2.0.CO;2.
Bush, B. C., F. P. J. Valero, A. S. Simpson, and L. Bignone, 2000: Characterization of thermal effects in pyranometers: A data correction algorithm for improved measurement of surface insolation. J. Atmos. Oceanic Technol., 17, 165–175, https://doi.org/10.1175/1520-0426(2000)017<0165:COTEIP>2.0.CO;2.
Cadeddu, M. P., D. Cimini, V. Ghate, D. Lubin, A. M. Vogelmann, and I. Silber, 2022: Examination of humidity and ice supersaturation profiles over West Antarctica using ground-based G-band radiometer retrievals. IEEE Trans. Geosci. Remote Sens., 60, 4102116, https://doi.org/10.1109/TGRS.2021.3077088.
Catania, G. A., H. Conway, C. F. Raymond, and T. A. Scambos, 2005: Surface morphology and internal layer stratigraphy in the downstream end of Kamb Ice Stream, West Antarctica. J. Glaciol., 51, 423–431, https://doi.org/10.3189/172756505781829142.
Coddington, O., P. Pilewskie, and T. Vukicevic, 2012: The Shannon information content of hyperspectral shortwave cloud albedo measurements: Quantification and practical applications. J. Geophys. Res., 117, D04205, https://doi.org/10.1029/2011JD016771.
Corbett, J., and W. Su, 2015: Accounting for the effects of sastrugi in the CERES clear-sky Antarctic shortwave angular distribution models. Atmos. Meas. Tech., 8, 3163–3175, https://doi.org/10.5194/amt-8-3163-2015.
Das, S. B., and R. B. Alley, 2008: Rise in frequency of surface melting at Siple Dome through the Holocene: Evidence for increasing marine influence on the climate of West Antarctica. J. Geophys. Res., 113, D02112, https://doi.org/10.1029/2007JD008790.
DeConto, R. M., and D. Pollard, 2016: Contribution of Antarctica to past and future sea-level rise. Nature, 531, 591–597, https://doi.org/10.1038/nature17145.
Dong, X., and G. G. Mace, 2003: Arctic stratus cloud properties and radiative forcing derived from ground-based data collected at Barrow, Alaska. J. Climate, 16, 445–461, https://doi.org/10.1175/1520-0442(2003)016<0445:ASCPAR>2.0.CO;2.
Fisher, A. T., K. D. Mankoff, S. M. Tulaczyk, S. W. Tyler, N. Foley, and WISSARD Science Team, 2015: High geothermal heat flux measured below the West Antarctic ice sheet. Sci. Adv., 1, e1500093, https://doi.org/10.1126/sciadv.1500093.
Flynn, C. J., 2016: Shortwave Array Spectroradiometer–Hemispheric (SASHe) instrument handbook. US Department of Energy, Office of Science Tech. Rep. DOE/SC-ARM-TR-172, 26 pp., https://doi.org/10.2172/1251414.
Flynn, C. J., A. Mendoza, Y. Zheng, and S. Mathur, 2007: Novel polarization-sensitive micropulse lidar measurement technique. Opt. Express, 15, 2785–2790, https://doi.org/10.1364/OE.15.002785.
Fürst, J. J., G. Durand, F. Gillet-Chaulet, L. Tavard, M. Rankl, M. Braun, and O. Gagliardini, 2016: The safety band of Antarctic ice shelves. Nat. Climate Change, 6, 479–482, https://doi.org/10.1038/nclimate2912.
Gades, A. M., C. F. Raymond, H. Conway, and R. W. Jacobel, 2000: Bed properties of Siple Dome and adjacent ice streams, West Antarctica, inferred from radio-echo sounding measurements. J. Glaciol., 46, 88–94, https://doi.org/10.3189/172756500781833467.
Gettelman, A., and Coauthors, 2020: Simulating observations of Southern Ocean clouds and implications for climate. J. Geophys. Res. Atmos., 125, e2020JD032619, https://doi.org/10.1029/2020JD032619.
Ghiz, M. L., R. C. Scott, A. M. Vogelmann, J. T. M. Lenaerts, M. Lazzara, and D. Lubin, 2021: Energetics of surface melt in West Antarctica. Cryosphere, 15, 3459–3494, https://doi.org/10.5194/tc-15-3459-2021.
Gilbert, E., and C. Kittel, 2021: Surface melt and runoff on Antarctic ice shelves at 1.5°C, 2°C, and 4°C of future warming. Geophys. Res. Lett., 48, e2020GL091733, https://doi.org/10.1029/2020GL091733.
Gilbert, E., A. Orr, I. A. Renfrew, J. C. King, and T. Lachlan-Cope, 2022: A 20-year study of melt processes over Larsen C ice shelf using a high-resolution regional atmospheric model: 2. Drivers of surface melting. J. Geophys. Res. Atmos., 127, e2021JD036012, https://doi.org/10.1029/2021JD036012.
Gorodetskaya, I. V., and Coauthors, 2015: Cloud and precipitation properties from ground-based remote-sensing instruments in East Antarctica. Cryosphere, 9, 285–304, https://doi.org/10.5194/tc-9-285-2015.
Grant, R. H., and G. M. Heisler, 1997: Obscured overcast sky radiance distributions for ultraviolet and photosynthetically active radiation. J. Appl. Meteor., 36, 1336–1345, https://doi.org/10.1175/1520-0450(1997)036<1336:OOSRDF>2.0.CO;2.
Grazioli, J., J.-B. Madeleine, H. Gallée, R. M. Forbes, C. Genthon, G. Krinner, and A. Berne, 2017: Katabatic winds diminish precipitation contribution to the Antarctic ice mass balance. Proc. Natl. Acad. Sci. USA, 114, 10 858–10 863, https://doi.org/10.1073/pnas.1707633114.
Grenfell, T. C., and S. G. Warren, 1999: Representation of a nonspherical ice particle by a collection of independent spheres for scattering and absorption of radiation. J. Geophys. Res., 104, 31 697–31 709, https://doi.org/10.1029/1999JD900496.
Hansen, C. W., J. S. Stein, and A. Ellis, 2010: Statistical criteria for characterizing irradiance time series. Sandia Tech. Rep. SAND2010-7314, Sandia National Laboratories, 51 pp., https://doi.org/10.2172/1002108.
Haynes, J. M., C. Jakob, W. B. Rossow, G. Tselioudis, and J. Brown, 2011: Major characteristics of Southern Ocean cloud regimes and their effects on the energy budget. J. Climate, 24, 5061–5080, https://doi.org/10.1175/2011JCLI4052.1.
Hersbach, H., and Coauthors, 2020: The ERA5 global reanalysis. Quart. J. Roy. Meteor. Soc., 146, 1–51, https://doi.org/10.1002/qj.3803.
Hines, K. M., D. H. Bromwich, S.-H. Wang, I. Silber, J. Verlinde, and D. Lubin, 2019: Microphysics of summer clouds in central West Antarctica simulated by the Polar Weather Research and Forecasting Model (WRF) and the Antarctic Mesoscale Prediction System (AMPS). Atmos. Chem. Phys., 19, 12 431–12 454, https://doi.org/10.5194/acp-19-12431-2019.
Hines, K. M., D. H. Bromwich, I. Silber, L. M. Russell, and L. Bai, 2021: Predicting frigid mixed-phase clouds for pristine coastal Antarctica. J. Geophys. Res. Atmos., 126, e2021JD035112, https://doi.org/10.1029/2021JD035112.
Hirasawa, N., H. Konishi, Y. Fujiyoshi, K. Shibata, and K. Iwamoto, 2022: Snowfall detection by a newly deployed X-band Doppler radar at Syowa Station, Antarctica. Okhotsk Sea Polar Oceans Res., 6, 36–41, https://doi.org/10.57287/ospor.6.36.
Hyder, P., and Coauthors, 2018: Critical Southern Ocean climate model biases traced to atmospheric model cloud errors. Nat. Commun., 9, 3625, https://doi.org/10.1038/s41467-018-05634-2.
Intrieri, J. M., M. D. Shupe, T. Uttal, and B. J. McCarty, 2002: An annual cycle of Arctic cloud characteristics observed by radar and lidar at SHEBA. J. Geophys. Res., 107, 8030, https://doi.org/10.1029/2000JC000423.
Jacobel, W. R., B. C. Welch, D. Osterhouse, R. Pettersson, and J. A. MacGregor, 2009: Spatial variation of radar-derived basal conditions on Kamb Ice Stream, West Antarctica. Ann. Glaciol., 50, 10–16, https://doi.org/10.3189/172756409789097504.
Jenkins, A., D. Shoosmith, P. Dutrieux, S. Jacobs, T. W. Kim, S. H. Lee, H. K. Ha, and S. Stammerjohn, 2018: West Antarctic ice sheet retreat in the Amundsen Sea driven by decadal oceanic variability. Nat. Geosci., 11, 733–738, https://doi.org/10.1038/s41561-018-0207-4.
Joughin, I., B. E. Smith, and B. Medley, 2014: Marine ice sheet collapse potentially under way for the Thwaites Glacier Basin, West Antarctica. Science, 344, 735–738, https://doi.org/10.1126/science.1249055.
Kay, J. E., C. Wall, V. Yettella, B. Medeiros, C. Hannay, P. Caldwell, and C. Bitz, 2016: Global climate impacts of fixing the Southern Ocean shortwave radiation bias in the Community Earth System Model (CESM). J. Climate, 29, 4617–4636, https://doi.org/10.1175/JCLI-D-15-0358.1.
Key, J. R., X. Wang, J. C. Stroeve, and C. Fowler, 2001: Estimating the cloudy-sky albedo of sea ice and snow from space. J. Geophys. Res., 106, 12 489–12 497, https://doi.org/10.1029/2001JD900069.
Kingslake, J., J. C. Ely, I. Das, and R. E. Bell, 2017: Widespread movement of meltwater onto and across Antarctic ice shelves. Nature, 544, 349–352, https://doi.org/10.1038/nature22049.
Kittel, C., and Coauthors, 2022: Clouds drive differences in future surface melt over the Antarctic ice shelves. Cryosphere, 16, 2655–2669, https://doi.org/10.5194/tc-16-2655-2022.
Kittler, R., and P. Valko, 1993: Radiance distribution on densely overcast skies: Comparison with CIE luminance standard. Sol. Energy, 51, 349–355, https://doi.org/10.1016/0038-092X(93)90147-G.
Kuipers Munneke, P., C. H. Reijmer, M. R. van den Broeke, C. König-Langlo, P. Stammes, and W. H. Knap, 2008: Analysis of clear-sky Antarctic snow albedo using observations and radiative transfer modeling. J. Geophys. Res., 113, D17118, https://doi.org/10.1029/2007JD009653.
Kuipers Munneke, P., M. R. van den Broeke, J. T. M. Lenaerts, M. G. Flanner, A. S. Gardner, and W. J. van de Berg, 2011: A new albedo parameterization for use in climate models over the Antarctic ice sheet. J. Geophys. Res., 116, D05114, https://doi.org/10.1029/2010JD015113.
Lachlan-Cope, T., C. Listowski, and S. O’Shea, 2016: The microphysics of clouds over the Antarctic Peninsula—Part 1: Observations. Atmos. Chem. Phys., 16, 15 605–15 617, https://doi.org/10.5194/acp-16-15605-2016.
Lawson, R. P., and A. Gettelman, 2014: Impact of Antarctic mixed-phase clouds on climate. Proc. Natl. Acad. Sci. USA, 111, 18 156–18 161, https://doi.org/10.1073/pnas.1418197111.
Lazzara, M. A., G. A. Weidner, L. M. Keller, J. E. Thorn, and J. J. Cassano, 2012: Antarctic Automatic Weather Station Program: 30 years of polar observation. Bull. Amer. Meteor. Soc., 93, 1519–1537, https://doi.org/10.1175/BAMS-D-11-00015.1.
LeBlanc, S. E., P. Pilewskie, K. S. Schmidt, and O. Coddington, 2015: A spectral method for discriminating thermodynamic phase and retrieving cloud optical thickness and effective radius using transmitted solar radiance spectra. Atmos. Meas. Tech., 8, 1361–1383, https://doi.org/10.5194/amt-8-1361-2015.
Lenaerts, J. T. M., and Coauthors, 2018: Climate and surface mass balance of coastal West Antarctica resolved by regional climate modeling. Ann. Glaciol., 59, 29–41, https://doi.org/10.1017/aog.2017.42.
Long, C. N., J. M. Sabburg, J. Calbó, and D. Pagès, 2006: Retrieving cloud characteristics from ground-based daytime color all-sky images. J. Atmos. Oceanic Technol., 23, 633–652, https://doi.org/10.1175/JTECH1875.1.
Lubin, D., and J. E. Frederick, 1991: The ultraviolet radiation environment of the Antarctic Peninsula: The roles of ozone and cloud cover. J. Appl. Meteor., 30, 478–493, https://doi.org/10.1175/1520-0450(1991)030<0478:TUREOT>2.0.CO;2.
Lubin, D., and A. M. Vogelmann, 2011: The influence of mixed-phase clouds on surface shortwave irradiance during the Arctic spring. J. Geophys. Res., 116, D00T05, https://doi.org/10.1029/2011JD015761.
Lubin, D., and M. Ghiz, 2022: Siple Dome surface energy flux. U.S. Antarctic Program (USAP) Data Center, last accessed 4 August 2023, https://doi.org/10.15784/601540.
Lubin, D., P. Ricchiazzi, A. Payton, and C. Gautier, 2002: Significance of multidimensional radiative transfer effects measured in surface fluxes at an Antarctic coastline. J. Geophys. Res., 107, 4387, https://doi.org/10.1029/2001JD002030.
Lubin, D., and Coauthors, 2020: AWARE: The Atmospheric Radiation Measurement (ARM) West Antarctic Radiation Experiment. Bull. Amer. Meteor. Soc., 101, E1069–E1091, https://doi.org/10.1175/BAMS-D-18-0278.1.
Mason, S., J. K. Fletcher, J. M. Haynes, C. Franklin, A. Protat, and C. Jakob, 2015: A hybrid cloud regime methodology used to evaluate Southern Ocean cloud and shortwave radiation errors in ACCESS. J. Climate, 28, 6001–6018, https://doi.org/10.1175/JCLI-D-14-00846.1.