SEARCH

SEARCH BY CITATION

Keywords:

  • Antarctic Peninsula;
  • cloud microphysics;
  • FTIR;
  • thermodynamic phase

Abstract

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Theoretical Background
  5. 3. Data
  6. 4. Results of Multisensor Analysis
  7. 5. Conclusions
  8. Acknowledgments
  9. References

[1] A Fourier Transform Infrared (FTIR) spectroradiometer was deployed at Palmer Station, Antarctica, from 1 September to 17 November 1991. This instrument is similar to the Atmospheric Emitted Radiance Interferometer (AERI) deployed with the U.S. Department of Energy Atmospheric Radiation Measurement (ARM) program. The instrument measured downwelling zenith radiance in the spectral interval 400–2000 cm−1, at a resolution of 1 cm−1. The spectral radiance measurements, which can be expressed as spectral brightness temperature Tb(ν), contain information about cloud optical properties in the middle infrared window (800–1200 cm−1, 8.3–12.5 μm). In this study, this information is exploited to (1) provide additional characterization of Antarctic cloud radiative properties, and (2) demonstrate how multisensor analysis of ARM data can potentially retrieve cloud thermodynamic phase. Radiative transfer simulations demonstrate how Tb(ν) is a function of cloud optical depth τ, effective particle radius re, and thermodynamic phase. For typical values of τ and re, the effect of increasing the ice fraction of the total optical depth is to flatten the slope of Tb(ν) between 800–1000 cm−1. For optically thin clouds (τ ∼ 3) and larger ice particles (re (ice) > 50 μm) the behavior of Tb(ν) in this interval switches from a decrease with increasing wavenumber ν to an increase with ν, once the ice fraction of the total optical depth exceeds ∼0.7. The FTIR spectra alone cannot be interpreted to obtain thermodynamic phase, because a relatively small slope in Tb(ν) between 800–1000 cm−1 could represent either an optically thin cloud in the ice or mixed phase, or an optically thick cloud radiating as a blackbody. Sky observations and ancillary radiometric data are needed to sort the FTIR spectra into categories of small cloud optical depth, where the mid-IR window data can be interpreted; and larger cloud optical depth, where the FTIR data contain information only about cloud base temperature. Spectral solar ultraviolet (UV) irradiance measurements from the U.S. National Science Foundation's UV Monitor at Palmer Station are used to estimate area-averaged effective cloud optical depth τsw, and these estimates are used to sort the FTIR data. FTIR measurements with colocated τsw < 16 are interpreted to estimate cloud thermodynamic phase. Precipitating cloud decks generally show flatter slopes in Tb(ν), consistent with the ice or mixed phase. Altostratus decks show a larger range in Tb(ν) slope than low cloud decks, including increasing slopes with ν, suggesting a more likely occurrence of the ice phase. This study illustrates how cloud thermodynamic phase can be defensibly retrieved from FTIR data if high quality shortwave radiometric data are also available to sort the FTIR measurements by cloud opacity, and both data types are available at the ARM sites.

1. Introduction

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Theoretical Background
  5. 3. Data
  6. 4. Results of Multisensor Analysis
  7. 5. Conclusions
  8. Acknowledgments
  9. References

[2] This study has two objectives: first, to provide some additional characterization of cloud radiative properties in the sparsely sampled Antarctic maritime environment, and second, to begin the framework for multisensor analysis of cloud optical and microphysical processes using instruments deployed at the U.S. Department of Energy Atmospheric Radiation Measurement (ARM) Program's North Slope of Alaska (NSA) research station [Stamnes et al., 1999]. The former objective has relevance to the warming trend in the western Antarctic Peninsula (WAP) [Vaughan et al., 2001], related dramatic ecological impacts [Smith et al., 1999], and a possible explanation of this warming in terms of the Southern Annular Mode [Thompson and Solomon, 2002]. Observations of all components of a high latitude climate system, including the possible response of cloud amount and opacity to dynamics related to the Annular Mode [e.g., Wang and Key, 2003], are important for a complete characterization of the climate system and prediction of future change. Measurements of cloud microphysical and radiative properties have been very sparse in the WAP, although the British Antarctic Survey is commencing some in situ instrumental programs.

[3] With respect to the second objective, one of the most versatile ARM program instruments is the Atmospheric Emitted Radiance Interferometer (AERI). Spectral radiance measurements from the AERI contain information about cloud radiative and microphysical properties, and some of this information has been successfully exploited [Turner, 2003; Turner et al., 2003]. However, to isolate the various cloud and atmospheric properties that influence a given AERI measurement, it may be necessary to colocate AERI data with other measurements. This study represents a first step in a program to derive value-added product (VAP) information from NSA ARM data using the AERI as the primary source of data, and supplemented by other measurements to help constrain the retrieval.

[4] The AERI measures downwelling radiances from the zenith direction, with a spectral resolution of 1 cm−1, in a substantial portion of the middle infrared (mid-IR, nominally 500–1800 cm−1 for AERI band 2), and a useful portion of the near-infrared (1800–3030 cm−1 for AERI band 2). The instrument is a Fourier transform infrared (FTIR) spectroradiometer based on the rugged and field-proven Bomem (Inc.) Michelson interferometer, with radiometric calibration engineering designed by the Space Science and Engineering Center at the University of Wisconsin [Revercomb et al., 1988, 1993]. In the clear sky case, AERI and similar FTIR data have been used to validate and improve the representation of trace gas emission (line strength and continuum) in radiative transfer models used for global climate simulation, in extreme cold and dry conditions [Walden et al., 1997; Tobin et al., 1998]. Under cloudy skies, radiance measurements in the mid-IR window (800–1200 cm−1, 8.3–12.5 μm) can be used to infer cloud properties, as the contributions to the radiance from trace gases in this wavelength range are both small and readily identified. For clarity, the downwelling radiance measured by an FTIR instrument (Watts per square meter per steradian per inverse centimeter) is often expressed in effective scene temperature, or “brightness temperature”, Tb(ν), by inverting the Planck function. Unless the measured emission comes from an ideal blackbody source, the brightness temperature generally varies with wavenumber ν (cm−1).

[5] The cloud properties we must first diagnose are (1) the extent to which a cloud radiates as a blackbody, and (2) the thermodynamic phase (i.e., primarily liquid water, primarily, or a “mixed phase” with important optical depth contributions from both). Once these cloud properties are identified, detailed radiative transfer calculations can be used to estimate cloud liquid or ice water path and effective particle size. If a cloud has sufficient water content (and hence opacity) that the cloud base radiates like a blackbody, then the longwave cloud forcing [Ramanathan et al., 1989] is at a maximum for the given meteorological condition. At the same time, a longwave radiance measurement under such a cloud, when expressed in terms of brightness temperature, will show no spectral dependence in the mid-IR window, and cannot be used to retrieve other properties such as cloud optical depth. If the cloud is optically thinner such that its longwave emission deviates from that of a blackbody, then the resulting spectral dependence in this longwave signature may permit the retrieval of cloud properties such as thermodynamic phase, spectral optical depth, and effective radius of the droplet or ice particle size distribution [Mahesh et al., 2001; Turner et al., 2003]. However, as discussed below, it is not possible to determine if a cloud is effectively a blackbody emitter from an AERI spectrum alone: a constant brightness temperature with wavenumber in the mid-IR window might also result from certain microphysical properties. Other in situ or radiometric data must be used in conjunction with the AERI to resolve this issue.

[6] Once a given AERI measurement has been identified with an optically thin cloud, it is possible to infer the thermodynamic phase from the mid-IR spectrum. A combination of surface radiometric data and depolarization lidar observations from the Surface Heat Budget of the Arctic (SHEBA) experiment has revealed cloud thermodynamic phase to be a major parameter governing the Arctic radiation balance. Not only did SHEBA reveal a surprising persistence of cloud liquid water at the coldest temperatures in the lower Arctic troposphere [Uttal et al., 2002], but the thermodynamic phase was found to correlate with the sign of the surface cloud forcing [Shupe and Intrieri, 2003]. This makes thermodynamic phase a first priority in cloud property retrieval from instruments such as AERI.

2. Theoretical Background

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Theoretical Background
  5. 3. Data
  6. 4. Results of Multisensor Analysis
  7. 5. Conclusions
  8. Acknowledgments
  9. References

[7] To learn what kind of mid-IR window emission signatures we might expect under maritime high latitude mixed phase clouds, we perform radiative transfer simulations using a spectral longwave model described by Lubin [1994] and Lubin et al. [2002a]. The model is based on the discrete ordinates radiative transfer formulation [Stamnes et al., 1988], and contains a representation for trace gas absorption at 20 cm−1 resolution using the method of exponential sum fitting of transmissions (ESFT [Wiscombe and Evans, 1977]). Ozonesonde data obtained at Palmer Station during 1988 (O. Torres, NASA Wallops Flight Facility, personal communication, 1989) are used to specify input data for atmospheric temperature, water vapor, and ozone profiles. Cloud liquid water droplet size distributions are taken from various applicable sources as described by Lubin [1994]. An altostratus size distribution described by Paltridge [1974] has an effective radius re = 4.3 μm. Two aircraft measurements from over the Ross Sea [Saxena and Ruggiero, 1990] have effective radii of 6.5 μm and 8.5 μm. An Arctic multimode size distribution described by Tsay et al. [1989] has re = 10.9 μm. These size distributions were chosen mainly to bracket the full range of re normally observed in nature, and are referred to from now on by re = 4, 7, 9, and 11 μm, respectively.

[8] In zenith longwave emission measurements, differences in Tb(ν) between liquid and ice water arise primarily from the difference in complex refractive index between the two media (although if we were studying near-IR radiances backscattered to space and measured at various viewing angles, the scattering phase function, as influenced by ice crystal shape, would also be important). We therefore model ice crystals as equivalent spheres [Grenfell and Warren, 1999], and use lognormal size distributions having effective variance parameter 0.1 [Stone et al., 1990]. Three values of re are considered for ice particles, 10, 50, and 200 μm. Over the high Antarctic Plateau, one can often find ice clouds with effective particle radius re ∼ 10 μm [Stone, 1993; Mahesh et al., 2001] in that extremely cold and dry environment. However, for purposes of this discussion we take the maritime WAP environment to be more typical of the Arctic, for which Hobbs and Rangno [1998] report ice particle effective radii in the range of 10–50 μm in mixed phase clouds, and for which Key and Intrieri [2000] suggest a typical value of 40 μm. The value re = 200 μm is included to bracket the upper limit of measurements reported by Hobbs and Rangno [1998] and also to simulate the case of precipitating clouds (snowfall). In these radiative transfer simulations, a cloud is allowed to contain both liquid water and ice, in the eight possible combinations of size distributions chosen.

[9] The total cloud opacity is specified in terms of the shortwave optical depth, τsw = 3/2 W/re where W is the total (liquid plus ice) water path [Stephens, 1978]. For terrestrial clouds, this expression for τsw is valid for wavelengths shorter than ∼1 μm, where absorption by liquid droplets or ice particles is very small. At longer wavelengths, including the mid-IR, the optical depth varies with wavelength as determined ultimately by the complex refractive index of the scattering and absorbing medium. The above expression for τsw is commonly adopted as a reference optical depth because of its simplicity and its relationship to cloud microphysical properties. In the calculations presented here, the total optical depth τsw varies from 1 to 10, and the fraction of the total cloud optical depth that is due to ice is increased from 0 to 1 in increments of 0.1. In these mixed phase calculations, the single scattering albedo and phase function moments (Mie phase function for water, Henyey-Greenstein phase function for ice) of the mixture were computed as weighted averages based on the contribution of each phase to the total cloud optical depth at each wavenumber, following Tsay et al. [1989].

[10] Figure 1 shows model simulations of downwelling zenith radiance at the surface in the mid-IR window, for a cloud with physical temperature 263 K, and total optical depth τsw = 3, expressed in brightness temperature units Tb(ν). At some wavenumbers near the edge of the mid-IR window, the surface brightness temperature exceeds the physical cloud temperature due to additional emission from water vapor. The individual curves within each plot pertain to specific fractions of the total shortwave optical depth τsw that is due to ice. For most combinations of liquid water and ice re, Tb(ν) at most wavenumbers increases with increasing ice fraction of the total shortwave optical depth τsw. (note that the total optical depth in the infrared varies with both wavelength and microphysics). This increase is most pronounced when the liquid water re is small (4 μm), where we note that the brightness temperature difference between the entirely liquid and entirely ice water cloud cases can be greater than 15 K between 900–1000 cm−1. For small ice particle re (10 μm), and either small (4 μm) or large (9 μm) liquid water re, Tb(ν) increases with increasing ice fraction, but always decreases with wavenumber between 800–1000 cm−1. However, for larger ice particle re (50 and 200 μm), the gradient in Tb(ν) from 800–1000 cm−1 transitions to an increase with wavenumber at larger ice fractions of the total optical depth. The simulations for liquid water re = 7 μm (not shown) show dependencies on wavenumber and ice fraction that are intermediate between the 4 and 9 μm cases of Figure 1. The simulations for liquid water re = 11 μm (also not shown) show similar behavior to the 9 μm case, but with less overall variability in Tb(ν).

image

Figure 1. Radiative transfer model simulations, at 20 cm−1 resolution, of the spectral brightness temperature one would expect to measure from the zenith with a narrow field-of-view FTIR instrument, under plane-parallel mixed phase clouds of various fixed liquid water and ice particle effective radii. The family of curves within each panel depicts the changing Tb(ν) as the ice fraction of the total cloud optical depth increases from zero to unity in increments of 0.1. At 950 cm−1 in each panel, the curve with the lowest Tb(ν) is the 100% liquid water cloud, and the curve with the highest Tb(ν) is the 100% ice cloud.

Download figure to PowerPoint

[11] To summarize the most important behavior of Tb(ν) in the mid-IR window, as a function of cloud microphysical properties, we can evaluate the average slope of Tb(ν) across the spectral interval 800–1000 cm−1, and plot this slope as a function of ice fraction of the total cloud optical depth, for all integer values of total cloud optical depth (Figure 2). For small liquid water re (4 μm) and small ice particle re (10 μm, Figure 2a), the Tb(ν) slope is always negative, increasing toward zero for increasing total cloud optical depth (i.e., more blackbody-like emission) and also increasing toward zero with increasing ice fraction of the total cloud optical depth. For larger liquid water re (9 μm) and small ice particle re (10 μm, Figure 2d), the Tb(ν) slope is very close to zero for larger cloud optical depths, but becomes more negative with increasing ice fraction for total cloud optical depths between 1–4. Thus, the case of small ice particle effective radius yields a variety of spectral signatures depending on liquid water effective radius and total cloud optical depth.

image

Figure 2. Summary of the major points from the radiative transfer simulations, showing the slope of the zenith brightness temperature Tb(ν) in the spectral interval 800–1000 cm−1, for plane parallel mixed phase clouds of various fixed liquid water and ice particle effective radii, as a function of total cloud optical depth and ice fraction of the total optical depth. In each panel the family of curves depicts the change in the slope of Tb(ν) as the total cloud optical depth increases from 1 to 10 in unit increments.

Download figure to PowerPoint

[12] There is more consistency for larger ice particle effective radius (50 or 200 μm) as shown in the remaining panels of Figure 2. Tb(ν) slopes are most negative for clouds that are mostly liquid water with small total optical depth. For larger total optical depths the slopes increase to effectively zero as the ice fraction exceeds 0.6–0.7. For smaller total cloud optical depths (1–4) and substantial ice fraction (greater than 0.4–0.6), the Tb(ν) slope becomes positive. This property may represent an effective method for phase discrimination, for clouds with small total optical depth. The theoretical Tb(ν) slopes for liquid water re = 7 μm are intermediate in behavior between the 4 and 9 μm cases shown in Figure 2, and the theoretical Tb(ν) slopes for liquid water re = 11 μm are similar to those for 9 μm, but with slightly less negative slopes in the applicable ranges of ice fraction and total cloud optical depth (figures omitted).

3. Data

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Theoretical Background
  5. 3. Data
  6. 4. Results of Multisensor Analysis
  7. 5. Conclusions
  8. Acknowledgments
  9. References

[13] The FTIR instrument deployed at Palmer Station measured downwelling zenith radiance with a spectral resolution of 1 cm−1, from 400–2000 cm−1, several times daily between 01 September and 17 November 1991. Radiometric calibration was provided by a Mikron (Inc.) model M-340 blackbody. Details of the calibration procedure and noise equivalent temperature change (NEΔT) evaluation are given in Lubin [1994] and Lubin et al. [1995]. Essentially, this instrument's NEΔT is ∼1 K. The current generation of the AERI instrument has NEΔT closer to ∼0.1 K, and is thus more capable of measuring the subtle spectral differences between various cloud microphysical scenarios. The somewhat prototypical instrument discussed here was built and deployed at Palmer Station before the AERI science team developed these important radiometric calibration advances [Revercomb et al., 1993]. Nevertheless, this data set from Palmer Station illustrates some of the major spectral features of ice and liquid water cloud emission in the mid-IR window.

[14] The FTIR measurements from Palmer Station were usually timed to coincide with NOAA or Defense Meteorological Satellite Program (DMSP) satellite overpasses or with the hourly spectral shortwave measurements discussed below. Sky conditions were recorded using whole sky photography, and cloud base heights were estimated visually with reference to the mountains on the Antarctic Peninsula, for which a topographic map was available at the station. These visual estimates of cloud base height are subject to the usual uncertainties with such observations, although the unobstructed view of a distant mountain range provides confidence. The photographic record reveals that when the cloud base was estimated higher than 3000 feet, most or all of the mountain peaks were visible. For cloud base estimates below 3000 feet, the mountain peaks were obscured.

[15] Figure 3 shows four examples of mid-IR emission spectra from Palmer Station. Comparing these measurements with the radiative transfer calculations of Figures 1 and 2, we notice that the measurement under an overcast layer with base height 2500′ (Figure 3a) appears to be consistent with a liquid water cloud of larger effective radius and moderate optical depth, although a variety of other microphysical combinations could yield this particular spectral signature. A measurement under a precipitating overcast layer (Figure 3c) has a flatter slope in Tb(ν) between 800–1000 cm−1, suggesting either a blackbody-like behavior or a signature expected from a mixed phase cloud. The measurement taken under the altostratus (Figure 3b) layer has a positive slope in Tb(ν), which from Figure 2 suggests a cloud of small optical depth primarily in the ice phase. A clear sky spectrum is shown for comparison (Figure 3d), which exhibits much lower brightness temperatures in the mid-IR window that arise from water vapor, CO2, ozone and other trace gas emission, and a much steeper positive Tb(ν) slope between 800–1000 cm−1.

image

Figure 3. Examples of zenith brightness temperature measured by the FTIR instrument at Palmer Station, Antarctica, during the austral spring of 1991.

Download figure to PowerPoint

[16] Ancillary shortwave downwelling surface flux measurements were provided by the U.S. National Science Foundation solar ultraviolet (UV) radiation monitoring system at Palmer Station. The UV Monitor consists of a Biospherical Instruments Inc. scanning spectroradiometer (model SUV-100) configured to measure the downwelling hemispheric flux in the spectral interval 280–600 nm with 1 nm resolution. Optical system and radiometric calibration details are given in Booth et al. [1994]. In 1991, the spectral UV flux measurements were made at the top of every daylight hour. The UV Monitor was originally deployed at the three U.S. Antarctic research stations and at Ushuaia, Argentina, in response to the discovery of the springtime Antarctic ozone decrease [Lubin et al., 1989; Stamnes et al., 1990]. Its usefulness the present analysis lies in its ability to provide shortwave flux measurements in the spectral interval 350–400 nm, in which trace gas absorption plays a negligible role and the radiation is governed mainly by multiple scattering from the clear atmosphere, clouds, and the surface.

4. Results of Multisensor Analysis

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Theoretical Background
  5. 3. Data
  6. 4. Results of Multisensor Analysis
  7. 5. Conclusions
  8. Acknowledgments
  9. References

[17] Our first objective is to identify which cloud decks sampled by the FTIR instrument are optically thick enough that they radiate essentially as a blackbody, versus optically thinner cloud decks that may lend themselves to optical property retrieval from the FTIR emission spectra. We therefore used the UV Monitor downwelling flux measurements at 350 nm to estimate an area-averaged “effective shortwave optical depth” τsw, following Lubin and Frederick [1991] and Stamnes et al. [1990]. This estimation proceeded in two steps. First, UV Monitor measurements identified with cloud-free conditions were compared with delta-Eddington radiative transfer model calculations [Joseph et al., 1976; Briegleb, 1992] to determine a useful estimate of the surface albedo for this location and time period. Due to multiple reflection of photons between a high albedo surface (snow or ice) and the cloud base [Gardiner, 1987], it is critical to make a defensible estimate of the surface albedo before proceeding with any cloud analysis. The delta-Eddington radiative transfer analysis here yielded an average surface albedo estimate of 0.75, with a standard deviation of 0.25, which is consistent with other experimental data from Palmer Station during this season [Ricchiazzi et al., 1995].

[18] Second, with this surface albedo estimate, the delta-Eddington model was iterated for each cloudy-sky UV Monitor measurement to find a value of τsw that matched the calculation to the measurement. Figure 4 illustrates this procedure, showing three diurnal samples of UV Monitor data and radiative transfer calculations of downwelling 350-nm flux, as a function of solar zenith angle and cloud optical depth. For each FTIR emission spectrum under overcast skies (cloud coverage >90%), a value of τsw was assigned based on the UV Monitor estimate closest in time to the FTIR measurement.

image

Figure 4. Hourly measurements of downwelling solar ultraviolet surface flux at 350 nm measured by the NSF UV Monitor scanning spectroradiometer at Palmer Station, Antarctica (diamonds). The curves in each panel depict the diurnal variability in downwelling 350 nm surface flux as calculated by a delta-Eddington radiative transfer model under clear skies and under plane-parallel clouds of varying total optical depth. In these calculations, the surface albedo is 0.75.

Download figure to PowerPoint

[19] Several cautions must be kept in mind with this approach. First, despite the superficial appearance of polar stratiform clouds as ideal “plane parallel” scattering layers, there is actually enormous spatial and temporal variability in their optical properties on several scales. This variability appears directly in UV radiation measurements from Palmer Station reported by Lubin et al. [2002b]. Alternatively, one can notice the large variability in Arctic cloud water content with distance along a research aircraft flight track reported in Hobbs and Rangno [1998]. The UV flux measurements cover the entire 2π steradian upward field of view, while the FTIR instrument views only a 40-milliradian field of view directly overhead. Second, Palmer Station is located on a coastline, with open water (low surface albedo) appearing within 1–3 km of the UV Monitor, depending on sea ice conditions. The large discontinuity in surface albedo between the sea ice or glacier and the open ocean induces a gradient in downwelling shortwave surface flux perpendicular to the discontinuity, and extending both several kilometers out to sea and in over the land. This gradient results from multiple reflection effects between the high albedo surface and the cloud base, and has been measured directly at coastal Antarctic stations by Smolskaia et al. [1999] and Lubin et al. [2002b]. In the presence of this spatial gradient in downwelling shortwave flux, an attempt to retrieve τsw from a downwelling flux measurement near the coast may not yield the cloud's true physical optical depth (i.e., the opacity implied by its water content and particle size distribution), even if the cloud for an instant in time behaves like a horizontally homogeneous plane parallel layer. Third, if the surface albedo during any given cloudy-sky measurement is different than the value of 0.75 assumed to be representative of the entire season, the retrieved cloud optical depth will be different from the cloud's true physical optical depth - too small if the surface albedo is actually larger than the assumed value, and vice versa.

[20] One obvious manifestation of uncertainties related to non-plane-parallel radiative transfer is seen in Figure 4c. On Day 294 there was broken cloud cover in the morning which became a solid overcast deck in the afternoon. The UV Monitor 350 nm flux measurements for 1200 and 1300 UTC on Day 294 are actually larger than the theoretical (plane-parallel) UV flux calculated for clear skies. This implies significant cloud-edge enhancement of the downwelling flux.

[21] Despite these uncertainties, the “effective” hemispheric τsw derived from the downwelling flux measurements can serve as an index for identifying cloud systems that are optically too thick for analysis by the FTIR data. To demonstrate this, we evaluate the slope Tb(ν) in all of the cloudy sky FTIR spectra, by performing a linear regression through all brightness temperature values between 800–1000 cm−1, excluding those at wavenumbers corresponding to water vapor or other trace gas emission features. These Tb(ν) slopes are plotted in a scatter diagram against the τsw derived from the UV Monitor data in Figure 5. In Figure 5a, all overcast FTIR measurements are shown, divided into three categories: single cloud layers without precipitation, overcast skies for which the sky observations indicate the presence of more than one cloud layer, and precipitating cloud decks. We see that for most τsw >30, the slope in Tb(ν) is very close to zero, signifying that these clouds are radiating as blackbodies. Thus, the downwelling shortwave flux measurements serve to isolate these cases as being unsuitable for further FTIR-based cloud property retrieval. Many clouds with τsw ≪ 30 also have slopes close to zero, and we may assume that these small slopes result from microphysical scenarios such as those depicted in Figure 1.

image

Figure 5. Scatterplot of the mid-IR brightness temperature Tb(ν) slope, determined from the FTIR measurements, versus the near-simultaneous estimate of shortwave cloud effective optical depth τsw, determined from the NSF UV Monitor 350 nm downwelling surface flux measurements (a) under all overcast skies sampled, (b) under nonprecipitating overcast decks with only one layer visible from the ground.

Download figure to PowerPoint

[22] From theoretical considerations (Figures 1 and 2), we would also expect many of the measured Tb(ν) slopes to diverge significantly from zero for τsw < 10, as the FTIR brightness temperature spectra begin to vary with cloud microphysical properties. In Figure 5a, we do see such variability for τsw < 10, but we also see many measured Tb(ν) slopes significantly different from zero for cloud optical depths as large as 20–40. This is probably a manifestation of the uncertainties in retrieving τsw, discussed above. Indeed, if we examine the six “outlier” points for τsw between 18–40 (the data points for which Tb(ν) slopes are less than −0.01), we notice that four of them are identified with multilayer cloud systems, for which errors in retrieved τsw would be expected to be larger and for which a hemispheric flux measurement is obviously less relevant to the FTIR's narrow overhead field of view. There are also many overcast sky data points for which the retrieved τsw = 0, which is another manifestation of the above mentioned uncertainties with this technique.

[23] In an attempt to draw some conclusions about maritime Antarctic clouds from this data set, we can use a threshold value of τsw > 15 to reject cloud cover that is most likely radiating as a blackbody. Although one should expect to use a threshold of τsw < 10, from theoretical considerations, the extra range allows some leeway with the above-mentioned uncertainties with cloud and surface inhomogeneity, as suggested by Figure 5a. Figure 5b shows a scatterplot of the Tb(ν) slopes for just the cloud sky scenes without precipitation and for which only a single cloud is visible (the important caution here is that there might be a second cloud layer above the single layer observed from the surface). These data points are sorted into two categories: low level clouds with bases below 3000′, and higher clouds. In this subset, only 14 data points have τsw > 15 and are rejected from further interpretation. Most of these rejected points have Tb(ν) slopes close to zero. For the FTIR measurements having τsw < 16, the distributions of Tb(ν) in the low versus higher clouds are skewed relative to one another.

[24] This is further clarified in Figure 6a, which shows a histogram of all the Tb(ν) slope measurements in Figure 5b for which τsw < 16. For the low clouds, all but one of the Tb(ν) slopes are less than zero, signifying that clouds whose optical depth is due mainly to larger ice particles are perhaps rare. Of the 44 observations, 37 (84%) have Tb(ν) between −0.03 and zero, and these slopes could arise from mixed phase clouds, or from liquid water clouds of relatively larger optical depth. Six observations (14%) have Tb(ν) slopes less than −0.03, and theoretical considerations suggest that these clouds are most primarily in the liquid phase, and with smaller re and optical depth. For the higher single-layer clouds, there are only 17 FTIR observations from this season. Of these, nine have Tb(ν) slopes larger than zero, suggesting the presence of larger ice particles in an optically thin cloud; four have Tb(ν) slopes between −0.03 and zero, and could pertain to mixed phase clouds or liquid water clouds of relatively larger optical depth; and four have Tb(ν) slopes smaller than −0.03, suggesting liquid water clouds with smaller re and/or smaller optical depth.

image

Figure 6. Histograms of FTIR-measured brightness temperature slopes Tb(ν), for overcast decks under which the near-simultaneous UV Monitor data yielded a shortwave cloud optical depth measurement τsw < 16 (a) for all overcast skies sorted into precipitating and nonprecipitating categories, (b) for nonprecipitating clouds sorted into two altitude categories.

Download figure to PowerPoint

[25] In Figure 6b we examine the same kind of histogram, but for all overcast sky FTIR measurement for which τsw < 16. This subset is divided into categories of clouds with and without precipitation. The precipitating clouds, for which we expect the presence of the mixed phase, all have Tb(ν) slopes between −0.03 and zero, and 90% of these slopes are between −0.01 and zero. This is consistent with the theoretical considerations shown in Figures 1 and 2 for mixed phase clouds. The nonprecipitating clouds exhibit a broader distribution in Tb(ν) slopes. In some cases the Tb(ν) slopes are larger than zero (most of these are the higher clouds). For Tb(ν) slopes between −0.04 and zero, the nonprecipitating clouds exhibit a more even distribution in this range, suggesting that the mixed phase may be a less frequent occurrence in the nonprecipitating clouds.

[26] Here we have emphasized the usefulness of ancillary shortwave cloud optical depth estimates to determine which FTIR spectra are suitable for retrieval of microphysical properties. A possible alternate approach suggested by Figures 1 and 3 might involve using the 9.6 μm ozone emission band. If a cloud is optically thin enough to allow FTIR retrieval of microphysical properties, it might also be optically thin enough such that the 9.6 μm ozone band is noticeable in the FTIR-measured emission spectrum. If the 9.6 μm ozone band does not appear in the spectrum, perhaps the cloud effectively radiates as a blackbody. While intuitive, this approach is not entirely successful, as shown in Figure 7. Here, the brightness temperatures are averaged over the spectral intervals 1040–1080 cm−1 (within the ozone band) and 982.5–990.0 cm−1 (a very transparent part of the mid-IR window adjacent to the ozone band). The difference in these average brightness temperatures, ΔT9.6, is plotted against the values of τsw derived from UV Monitor data, for all single-layer nonprecipitating clouds, in Figure 7a. For τsw > 40, this brightness temperature difference is close to zero, indicating that the 9.6 μm ozone band is obscured. For high clouds, most spectra show a significant brightness temperature difference (1 < ΔT9.6 < 13 K), signifying a prominent 9.6 μm ozone band feature. Several spectra under low clouds also show a prominent 9.6 μm ozone band feature. However, most of the spectra obtained under low clouds with τsw < 16 show ΔT9.6 close to zero, signifying an obscured 9.6 μm ozone band. Therefore the 9.6 μm ozone band by itself is not a reliable index for cloud optical depth. This is because the 9.6 μm ozone band feature results from thermal emission, and can be obscured by a low cloud in the warmest part of the troposphere even if the cloud has only moderate optical thickness.

image

Figure 7. Scatterplot of the difference in average brightness temperatures evaluated over the spectral intervals 1040–1080 cm−1 (in the ozone emission band) and 982.5–990.0 cm−1 (adjacent to the ozone emission band), versus the near-simultaneous estimate of shortwave cloud effective optical depth τsw, determined from the NSF UV Monitor 350 nm downwelling surface flux measurements (a) under nonprecipitating overcast decks with only one layer visible from the ground, (b) under all sky conditions without precipitation, comprising clear sky, overcast, and scattered/broken categories.

Download figure to PowerPoint

[27] While not entirely successful for sorting the optical depth of plane parallel clouds, the 9.6 μm ozone band can play another useful role in isolating FTIR measurements for retrieval of cloud microphysical properties. It is possible that a cloud layer might be broken, such that the FTIR field of view is not entirely covered by cloud. If this isn't noticed, the FTIR spectrum with its radiance contributions from both clear and cloudy sky could be analyzed to yield a spurious microphysical retrieval. The 9.6 μm ozone band can be readily used to identify broken clouds in the FTIR's field of view, as shown in Figure 7b. In this figure, the above brightness temperature ΔT9.6 difference is plotted as a function of τsw, for all overcast skies without precipitation, for cloud cover identified as scattered or broken (sky coverage less than 9/10), and for clear skies. Under clear skies, ΔT9.6 is consistently larger than 20 K. For the overcast skies, ΔT9.6 is consistently less than 10 K, with most spectra showing ΔT9.6 < 4 K. Under the scattered and broken clouds, 73% of the measurements show ΔT9.6 > 4K, and 55% of the measurements show ΔT9.6 > 10 K. This figure therefore suggests that cloudy sky spectra for which ΔT9.6 > 10 K are possible candidates for rejection due to broken cloud cover within the field of view, and should be scrutinized carefully with the help of other meteorological data to determine if they are indeed suitable plane-parallel layers for microphysical retrieval.

[28] Finally, we note that Turner et al. [2003] discuss a promising method by which microwindows in the spectral interval 17–19 μm can be used in conjunction with the mid-IR window (11–12 μm) to infer cloud thermodynamic phase. In Figure 6 of Turner et al. [2003] they show theoretical radiative transfer calculations, performed in a similar manner to Figures 1 and 2 here; their figure indicates that the ratio of cloud emissivities at 17–19 μm and 11–12 μm, when plotted against the 11 μm emissivity, shows distinctly different values for liquid water versus ice clouds when the total 11 μm cloud optical depth is smaller than 3. It is conceivable that the Turner et al. [2003] method might work better than the mid-IR window method presented herein, for such optically thin clouds. The Turner et al. [2003] method could not be tested directly with the Palmer Station FTIR data, because these data from 1991 do not have matching radiosonde observations. Direct measurements of atmospheric structure and water vapor abundance are necessary to perform the clear sky radiative transfer calculations required for evaluation of the spectral cloud emissivity [cf. Turner et al., 2003, equation (5)]. However, in a practical application of the Turner et al. [2003] method to large data sets, it is possible that ancillary radiometric data may be needed, in the manner employed herein. Figure 6 of Turner et al. [2003] shows that for ice clouds (particularly those with smaller particles), their spectral emissivity ratio is lowest (mainly in the range 0.80–0.85), is relatively insensitive to the total cloud optical depth, and only decreases slightly as the cloud optical depth approaches that of an effective blackbody emitter. Given the uncertainties with real field data, including the radiosonde data used to derive the spectral cloud emissivity, it might be worthwhile to always check FTIR data against colocated retrievals of τsw derived from suitable shortwave measurements, so that obvious cases of large cloud opacity can be eliminated before applying the Turner et al. [2003] method.

5. Conclusions

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Theoretical Background
  5. 3. Data
  6. 4. Results of Multisensor Analysis
  7. 5. Conclusions
  8. Acknowledgments
  9. References

[29] Precipitating overcast decks, in which one expects a mixed phase, generally exhibit Tb(ν) in the mid-IR window that is nearly spectrally flat, irrespective of total cloud optical depth. This is consistent with radiative transfer simulations for mixed phase clouds. Non precipitating lower overcast decks usually have negative slopes in Tb(ν), in the spectral interval 800–1000 cm−1, that range from −0.08 to zero, and the slopes in Tb(ν) that are less than −0.03 are consistent with clouds in primarily the liquid phase with small optical depth (τsw < 5) and/or small re. Slopes in Tb(ν) that are negative but closer to zero could result from liquid water clouds with slightly larger optical depths, or from mixed phase clouds. Higher clouds (bases above 3000′) exhibited the largest range in Tb(ν) slopes for the spectral interval 800–1000 cm−1, including positive slopes that are consistent with ice phase clouds having re > 30 μm and small optical depth, but also including some of the most negative slopes that are consistent with liquid water clouds having small optical depth and/or effective radius.

[30] From the standpoint of multisensor data analysis, we see that downwelling flux measurements, when used to estimate τsw, can be used to identify many cloud decks with too large an opacity for FTIR analysis. Most of the clouds with large τsw estimated from the UV Monitor data radiated like blackbodies, having Tb(ν) near zero. At the ARM NSA site, the well-calibrated broadband (pyranometer) and spectrally resolved multifilter rotating shadowband radiometer (MFRSR) shortwave flux measurements can be used to estimate τsw [Leontyeva and Stamnes, 1994], and the technique for using these estimates to isolate blackbody-like clouds should be applicable to the NSA AERI data. However, there are uncertainties in retrieving τsw over high albedo surfaces from downwelling hemispheric flux data, and related issues about the relevance of a flux-based retrieval to radiance measured in a narrow field of view directly overhead. As a consequence of these uncertainties, the use of flux-based estimates of τsw as a filter for FTIR analysis may accidentally reject some FTIR spectra for which the sampled cloud area actually had a suitably small opacity, and may retain other FTIR spectra for which the sampled cloud area actually had too large an opacity. This may introduce errors in long-term climatological analysis of FTIR data. Ideally, one would prefer a shortwave radiometer for the ancillary τsw retrieval that measures radiance in exactly the same narrow field of view as the FTIR instrument, so that there is greater certainty that the retrieved τsw is relevant to the measured FTIR spectrum.

[31] However, the NSA site offers a better location for applying the multisensor retrieval technique presented here. Although the NSA site is located near a coastline, which potentially imposes the same uncertainty in determining τsw as found in the Palmer Station data, additional instruments at the NSA site provide ways to mitigate this uncertainty. The Whole Sky Imager (WSI) data reveals not only cloud amount but also sky radiance patterns that can be interpreted to isolate two dimensional surface albedo effects, with the help of Monte Carlo radiative transfer simulations [Ricchiazzi and Gautier, 1998]. In addition, the surface albedo at the NSA site is monitored using a network of upward and downward looking pyranometers, and these instantaneous measurements of surface albedo should reduce one source of uncertainty in the present work - that associated with a seasonally averaged surface albedo estimated from just the downwelling shortwave measurements. Provided the operational parameters and measurement uncertainties of each instrument are understood, the use of two or more colocated radiometers can be an effective strategy for the more difficult atmospheric retrievals, such as cloud microphysics.

Acknowledgments

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Theoretical Background
  5. 3. Data
  6. 4. Results of Multisensor Analysis
  7. 5. Conclusions
  8. Acknowledgments
  9. References

[32] This research was supported by the U.S. Department of Energy under DOE FG0303-ER63538. The FTIR data collection at Palmer Station was supported by the U.S. Antarctic Program under NSF DPP-9018207.

References

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Theoretical Background
  5. 3. Data
  6. 4. Results of Multisensor Analysis
  7. 5. Conclusions
  8. Acknowledgments
  9. References
  • Booth, C. R., T. B. Lucas, J. H. Morrow, C. S. Weiler, and P. A. Penhale (1994), The United States National Science Foundation's polar network for monitoring ultraviolet radiation, in Ultraviolet Radiation in Antarctica: Measurements and Biological Effects, Antarc. Res. Ser., vol. 62, edited by C. S. Weiler, and P. A. Penhale, pp. 1737, AGU, Washington, D. C.
  • Briegleb, B. P. (1992), Delta-Eddington approximation for approximation for solar radiation in the NCAR community climate model, J. Geophys. Res., 97, 76037612.
  • Gardiner, B. G. (1987), Solar radiation transmitted to the ground through cloud in relation to surface albedo, J. Geophys. Res., 92, 40104018.
  • Grenfell, T. C., and S. G. Warren (1999), Representation of a nonspherical ice particle by a collection of independent spheres for scattering and absorbing of radiation, J. Geophys. Res., 104, 31,69731,709.
  • Hobbs, P. V., and A. L. Rangno (1998), Microstructure of low and middle-level clouds over the Beaufort Sea, Q. J. R. Meteorol. Soc., 124, 20352071.
  • Joseph, J. H., W. J. Wiscombe, and J. A. Weinman (1976), The delta-Eddington approximation for radiative flux transfer, J. Atmos. Sci., 33, 24532459.
  • Key, J. R., and J. M. Intrieri (2000), Cloud particle phase determination with the AVHRR, J. Appl. Meteorol., 39, 17971804.
  • Leontyeva, E., and K. Stamnes (1994), Estimations of cloud optical thickness from ground-based measurements of incoming solar radiation in the Arctic, J. Clim., 7, 566578.
  • Lubin, D. (1994), Infrared radiative properties of the maritime Antarctic atmosphere, J. Clim., 7, 121140.
  • Lubin, D., and J. E. Frederick (1991), The ultraviolet radiation environment of the Antarctic Peninsula: The roles of ozone and cloud cover, J. Appl. Meteorol., 30, 478493.
  • Lubin, D., C. R. Booth, T. Lucas, D. Neuschuler, and J. E. Frederick (1989), Measurements of enhanced springtime ultraviolet radiation at Palmer Station, Antarctica, Geophys. Res. Lett., 16, 783785.
  • Lubin, D., D. Cutchin, W. Conant, H. Grassl, U. Schmid, and W. Biselli (1995), Spectral longwave emission in the tropics: FTIR measurement at the sea surface and comparison with fast radiation codes, J. Clim., 8, 286295.
  • Lubin, D., P. Ricchiazzi, A. Payton, and C. Gautier (2002a), The significance of multidimensional radiative transfer effects measured in surface fluxes at an Antarctic coastline, J. Geophys. Res., 107(D19), 4387, doi:10.1029/2001JD002030.
  • Lubin, D., S. K. Satheesh, G. McFarquar, and A. J. Heymsfield (2002b), The longwave radiative forcing of Indian Ocean tropospheric aerosol, J. Geophys. Res., 107(D19), doi:10.1029/2001JD001183.
  • Mahesh, A., V. P. Walden, and S. G. Warren (2001), Ground-based infrared remote sensing of cloud properties over the Antarctic Plateau, part II: Cloud optical depths and particle sizes, J. Appl. Meteorol., 40, 12791294.
  • Paltridge, G. W. (1974), Infrared emissivity, short-wave albedo, and the microphysics of stratiform water clouds, J. Geophys. Res., 79, 40534058.
  • Ramanathan, V., R. D. Cess, E. F. Harrison, P. Minnis, B. R. Barkstrom, E. Ahmad, and D. Hartman (1989), Cloud-radiative forcing and climate: Results from the Earth Radiation Budget Experiment, Science, 243, 5763.
  • Revercomb, H. E., H. Buijs, H. B. Howell, D. D. LaPorte, W. L. Smith, and L. A. Sromovsky (1988), Radiometric calibration of IR Fourier transform spectrometers: Solution to a problem with the High-Resolution Interferometer Sounder, Appl. Opt., 27, 32103218.
  • Revercomb, H. E., F. A. Best, R. G. Dedecker, T. P. Dirkx, R. A. Herbsleb, R. O. Knuteson, J. F. Short, and W. L. Smith (1993), Atmospheric emitted radiance interferometer (AERI) for ARM, paper presented at Fourth Symposium on Global Climate Change Studies, Am. Meteorol. Soc., Anaheim, Calif.
  • Ricchiazzi, P., and C. Gautier (1998), Investigation of the effect of surface heterogeneity and topography on the radiation environment of Palmer Station, Antarctica, with a hybrid 3-D radiative transfer model, J. Geophys. Res., 103, 61616176.
  • Ricchiazzi, P., C. Gautier, and D. Lubin (1995), Cloud scattering optical depth and local surface albedo in the Antarctic: Simultaneous retrieval using ground-based radiometry, J. Geophys. Res., 100, 21,09121,104.
  • Saxena, V. K., and F. H. Ruggiero (1990), Antarctic coastal stratus clouds: Microstructure and acidity, in Contributions to Antarctic Research I, Antarc. Res. Ser., vol. 50, edited by D. H. Elliot, pp. 718, AGU, Washington, D. C.
  • Shupe, M. D., and J. M. Intrieri (2003), Arctic surface cloud forcing at SHEBA, part 1: The impact of cloud properties, surface albedo, and solar zenith angle, J. Clim., in press.
  • Smith, R. C., et al. (1999), Marine ecosystem sensitivity to climate change, Bioscience, 49, 393404.
  • Smolskaia, I., M. Nunez, and K. Michael (1999), Measurements of erythemal irradiance near Davis Station, Antarctica: Effect of inhomogeneous surface albedo, Geophys. Res. Lett., 26, 13811384.
  • Stamnes, K., S.-C. Tsay, W. J. Wiscombe, and K. Jayaweera (1988), Numerically stable algorithm for discrete-ordinate-method radiative transfer in multiple scattering and emitting layered media, Appl. Opt., 27, 25022509.
  • Stamnes, K., J. Slusser, M. Bowen, C. R. Booth, and T. Lucas (1990), Biologically effective ultraviolet radiation, total ozone abundance, and cloud optical depth at McMurdo Station, Antarctica, September 15, 1988 through April 15, 1989, Geophys. Res. Lett., 17, 21812184.
  • Stamnes, K., R. G. Ellingson, J. A. Curry, J. E. Walsh, and B. D. Zak (1999), Review of science issues, deployment strategy, and status for the ARM North Slope of Alaska - Adjacent Arctic Ocean Climate Research Site, J. Clim., 12, 4663.
  • Stephens, G. L. (1978), Radiation profiles in extended water clouds. II: Parameterization schemes, J. Atmos. Sci., 35, 21232132.
  • Stone, R. S. (1993), Properties of austral winter clouds derived from radiometric profiles at the South Pole, J. Geophys. Res., 98, 12,96112,971.
  • Stone, R. S., G. L. Stephens, C. M. R. Platt, and S. Banks (1990), The remote sensing of thin cirrus clouds using satellites, lidar, and radiative transfer theory, J. Appl. Meteorol., 29, 353366.
  • Thompson, D. W. J., and S. Solomon (2002), Interpretation of recent Southern Hemisphere climate change, Science, 296, 895899.
  • Tobin, D. C., R. O. Knuteson, H. E. Revercomb, V. P. Walden, S. A. Clough, E. J. Mlawer, and R. G. Ellingson (1998), AERI-ER at the SHEBA ice station: Far infrared water vapor continuum measurements, in Proceedings of the Eighth Atmospheric Radiation Measurement (ARM) Science Team Meeting, March 23–27, 1998, Tucson, Arizona, Rep. DOE/ER-0738, pp. 753756, U.S. Dep. of Energy, Germantown, Md.
  • Tsay, S.-C., K. Stamnes, and K. Jayaweera (1989), Radiative energy budget in the cloudy and hazy Arctic, J. Atmos. Sci., 46, 10021018.
  • Turner, D. (2003), Microphysical properties of single and mixed-phase Arctic clouds derived from ground-based AERI observations, Ph.D. dissertation, 167 pp., Dep. of Atmos. and Oceanic Sci., Univ. of Wis., Madison.
  • Turner, D. D., S. A. Ackerman, B. A. Baum, H. E. Revercomb, and P. Yang (2003), Cloud phase determination using ground-based AERI observations at SHEBA, J. Appl. Meteorol., 42, 701715.
  • Uttal, T., et al. (2002), Surface heat budget of the Arctic Ocean, Bull. Am. Meteorol. Soc., 83, 255275.
  • Vaughan, D. G., G. J. Marshall, W. M. Connolley, J. C. King, and R. Mulvaney (2001), Devil in the detail, Science, 293, 17771779.
  • Walden, V. P., S. G. Warren, F. J. Murcray, and R. G. Ellingson (1997), Infrared radiance spectra for testing radiative transfer models in cold and dry atmospheres: Test case from the Antarctic Plateau, Bull. Am. Meteorol. Soc., 78, 22462247.
  • Wang, X., and J. R. Key (2003), Recent trends in Arctic surface, cloud, and radiation properties from space, Science, 299, 17251728.
  • Wiscombe, W. J., and J. W. Evans (1977), Exponential sum fitting of radiative transmission functions, J. Comput. Phys., 24, 416444.