Barrow, AK Station Version Notes

Back to Barrow Home »
Back to Browser »

Retrievals are Retrievals, Not Measurements!

The most important thing to remember when using the cloud microphysical values on this CD is that they are observation-based-retrievals, not direct measurements. All of the retrievals use numerous assumptions. These may include such approximations as spherical ice particles, gamma, exponential, or lognormal droplet/ice particle size distributions, ice particle densities parameterized on particle size, and parameterization relating radar reflectivities to cloud properties based on aircraft data. It should also be considered that the radar-radiometer technique that utilizes the Liquid Water Path (LWP) from the microwave radiometer is a retrieval-based-on-a-retrieval, since LWP is itself retrieved from radiometric brightness temperatures.

Several studies have addressed the issue of errors associated with radar-based cloud retrievals. While not definitive, the ice retrieval uncertainties are expected to be 30-50% for particle size and 50-100% for ice water content and the liquid retrieval uncertainties are expected to be 20-40% for droplet size and 20-60% for liquid water content.

In the following sections, for cases where iterative or polynomial fits were not utilized, reduced equations have been provided to show how the cloud microphysics were calculated. However, it should be noted that these seemingly simple equations were derived through a complex line of reasoning, and the full derivations should be carefully examined in the references indicated.

Radar Data

All retrievals are based on continous, vertically-pointing, 35-GHz radar data that was originally collected in 4 different operating modes with variable sensitivies optimized for different cloud and precipitation situations. Typically these modes cycle every 9 s, and data are collected at 45 m or 90 m range resolutions. The modes were optimally combined to a single radar product using the DOE/ARM ARSCL (Active Remotely-Sensed Clouds Locations) program, and interpolated to a 1 min-45 meter time-height grid. The existence of an ARSCL file does not necessarily indicate that the radar was operating on a given day. Thus, the measurements panel in the browser has a "No Radar Data" label on days when there was not a single pixel of radar return, which is considered to be unlikely even in clear sky situations. The radiometers often indicate cloudiness when there is no apparent radar return, and this filter can be additionally used to distinquish missing-data-days from clear-days. Existence of an ARSCL file with radar returns does not always guarantee reliable results, as an ARSCL radar file will be generated even if the radar was not operating optimally. Due to this ambiguity, there is an extended NO DATA period between April 1 and July 31 in 2001 when it is known that the radar was measuring degraded returns due to equipment problems. There may be other intermittent periods that have not been identified. A missing hyperlink in the dates menu indicates that no ARSCL file existed in the ARM archive for that day.

Radiometer Data

The microwave radiometer measures brightness temperatures at 23.8 GHz and 31.4 GHz. Starting on April 25, 2002 this CD utilizes LWP values that implement oxygen and water vapor absorption models that are more appropriate for the supercooled liquid clouds found in the Arctic. Prior to that date, the absorption coefficients were those developed for warmer, above freezing clouds (since the ARM program was still in the midst of updating the LWP data archive when this CD was the published). Studies have been conducted to investigate the magnitude of the difference. In general, it can be expected that the LWP values will beon the order of 15-20% less with the new absorption coefficents, with the offset increasing for smaller values of LWP. Note that this change in LWP retrieval coefficients can be expected to cause a corresponding offsets in any cloud microphysical retrievals that utilize LWP data derived from the microwave radiometer. Both the new and the old LWP retrievals are based on a statistical approach which utilizes a climatological set of rawinsonde measurements. A more accurate approach would be to use radar data to derive liquid water-weighted cloud temperature as described by Liljegren et al. (2001). The error in the absolute LWP values is about 25 g/m2 (Westwater et al. 2001).

To determine infrared cloud brightness temperatures, the spectral measurements from the Atmospheric Emitted Radiance Interferometer (AERI) were integrated over 10.96-11.27 microns.

Cloud Classification

Classification of cloud scenes into 7 categories ("ice", "simple-ice", "liquid", "simple-liquid", "mixed", "drizzle", "rain", and "snow") was based on visual inspection of radar relectivity, Doppler velocity, Doppler spectral widths, microwave and IR radiometer data, and rawinsondes. This subjective classification was independantly checked and rechecked. "Ice" and "liquid" classifications indicate that radar-radiometric retrievals were possible (radar + IR radiometer for ice, radar + microwave radiometer for liquid). "Simple-ice" and "simple-liquid" classifications indicate that retrievals are based on radar data only, either because radiometers were obscured by multiple cloud layers, radiometric data was unavailable, or in the case of ice, the clouds were optically thicker than ~6.

In many cases, the clouds are classifed as mixed-phase, indicating that both ice and liquid appeared to exist simultaneously in a cloud layer. In general, the liquid tends to exist in discrete layers; however, for the purposes of classification, the entire layer is considered to be mixed-phase and no effort is made to place the liquid. Because the radar reflectivity is dominated by the larger ice crystals, the only retrievals implemented in the mixed-phase cloud regions are the non-radiometric ice retrievals. Depending on the application, these retrievals should be used cautiously since the neglected liquid component may contribute to uncertainty in the retreived ice properties and it may be the most important phase in determining the cloud radiative properties.

In addition to in-cloud temperatures and LWP from the microwave radiometer, Doppler spectral widths have been used as a qualitative indicator that clouds may be mixed-phase. This is based on the rational that increased spectral widths (indicating a wider spread in the distribution of vertical velocities) may indicate a wider range of hydrometeor sizes (fall velocities) which might occur if water droplets and ice crystals were mixed in the same volume. This is not a fully researched topic; however, preliminary insitu aircraft comparisons as well as intercomparison between coincident depolarization lidar data with radar spectral widths at SHEBA indicate that this is not an unreasonable assumption. The other candidate for producing increased spectral widths is turbulence. The relationship between turbulence and the development of mixed-phase layers is another topic of ongoing research

Precipitation classifications were most often done on the basis of the radar mean Doppler velocities. Drizzle was characterized by fall speeds that were typically larger than ~0.2 m/s and reflectivities higher than -15 dBZ. Rain was most often indicated by a clear melting layer signature (brightband) in the radar reflectivities, and velocities greater than ~2 m/s.

The netcdf files contain a mask field that indicates which retrievals were run at any given location in a cloud time-height scene. The classification codes are described below. Note that some of the equations below utilize "dBZ" (radar reflectivity factor) and some utilize "Z" (radar reflectivity in units off mm6/m3). Z is related to dBZ by the equation: Z=10^(dBZ/10).

Precipitation Retreivals


The RAIN retrievals assume the Marshall-Palmer drop size distribution and Rayleigh scattering conditions (which may at times be violated at K-band).

RainRate = 10^((dBZ-23)/16) [mm/hr]
RainDropSize = 244-RainRate^(0.21) [microns]
RainWaterContent - 0.072 * RainRate^(0.88) [g/m^3]
RainDropConcentration = 0.00195*RainRate^(0.21) [1/cm^3]


The SNOW retrievals assume the Gunn and Marshal snow size distribution.

SnowFallRate = 10^((dBZ-14.5)/9.5) [mm/hr]
SnowFlakeSize = 392*SnowFallRate^(0.48) [microns]
SnowWaterContent = 0.25*SnowFallRate^(0.9) [g/m^3]
SnowFlakeCondentration = 0.00149*SnowFallRate^(-0.39) [1/cm^3]

Liquid Retrievals


The radar-only LIQUID retrievals assume a lognormal droplet size distribution with a width of 0.31 (Frisch et al. 2002).

LiquidWaterContent = c*Z^(0.5) [g/m^3]
DropletEffectiveRadius = d*Z^(0.166) [microns]

  • c = (pi/6)*exp(-0.432)*N^(0.5)
  • d = 50*exp(-0.048)*N^(-0.166)
  • N = 75 cm^(-3) OR N is determined by a polynomial fit to reflectivity determined using aircraft measurements.


All Code 3 retrievals are implemented in addition to a radar-radiometer technique that utilizes the microwave radiometer-derived LWP scaled by radar reflectivity profiles to distribute liquid water contents in cloud (Frisch et al. 1995).


At present no drizzle retrievals are implemented.

Ice Retrievals

Note that there is not currently a standard ice crystal size definition across the research community. All retrievals utilized here derive the mean diameter which characterizes the assumed exponential distribution of physical particle sizes. The mean diameter is related to the median volume diameter by: MeanDiameter=MedianDiameter/3.54 for the exponential particle size distribution. For the ice particle sizes plotted in the "Particle Size" browser panel, ice particle diameters are converted to effective radii to be consistent with the measurement units for the cloud droplet sizes. The equations used for the conversion are:

EffectiveRadius = 13.74*(MeanDiameter)^0.3 for MeanDiameters >= 23.7 microns
EffectiveRadius = 1.5*(MeanDiameter) for MeanDiameters < 23.7 microns


The radar-only, empirical ICE retrieval method uses:

IceWaterContent = a*Z^b [g/m^3]
MeanDiameter = 40.5*a^(-0.53)*Z^((0.53(1-b)) [microns]

  • a = monthly values of "a" were determined from periods during which Radar-Radiometer method (See CODE 7) was implemented on single layer, optically thin ice clouds
  • b = 0.63

METHOD 2 (Matrosov et al., 2002)
The radar-only, Doppler velocity-reflectivity ICE retrieval method is appropriate for MeanDiameter >15 microns.

MeanDiameter = determined by particle characteristic size - fall velocity relationships over 20 min averages of Doppler velocity measurements [microns]
IceWaterContent = 1100*Z/(MeanDiameter)^(1.9)) [g/m^3]


All Code 6 retrievals are implemented in addition to a radar-IR radiometer ICE retrieval method that uses AERI-derived brightness temperatures to tune the coefficient "a" (Matrosov et al., 1999). The tuned method is good for MeanDiameters greater than about 15 microns.

IceWaterContent = a*Z^b [g/m^3]
MeanDiameter = 40.5*(Z/IceWaterContent)^(0.53) [microns]

  • b is scaled linearly through cloud height with a value of 0.7 at cloud base and 0.55 at cloud top
  • a = determined iteratively from AERI brightness temperature measurements.


CODE 6 ICE retrievals are implemented.


CODE 6 ICE retrievals are implemented.

Calculations of Optical Depth

Calculations of optical depth are not straight forward due to the frequent occurance of multiple cloud layers (often ice, liquid and mixed-phase combined), and a prevalence of radiometrically thick cloud layers. To approximate a combined optical depth for the total cloud column, the following procedure is used.

For Liquid Layers: The microwave radiometer value of LWP is used if available, otherwise the LWP from the code 3 radar retrieval is calculated by summing the retrieved LWC. The CODE 3 output in used in both cases to determine a layer-mean, LWC-weighted droplet effective radius. For mixed-phase clouds and drizzle regions, the layer-mean DropletEffectiveRadius is assumed to be 10 microns. OpticalDepth is then calculated using:
OpticalDepth (Liquid) = LWP*(0.029+1.3/DropletEffectiveRadius)

For Ice layers: The IWP is calculated by summing the IWC and obtaining layer-mean, IWC-weighted ParticleMeanDiameter from CODE 6, Method 1. OpticalDepth is then calculated using:
OpticalDepth (Ice)= IWP*(0.021+1.27/ParticleMeanDiameter)

TotalOpticalDepth = OpticalDepth (Liquid) + OpticalDepth (Ice)

Future Work

Future versions of this data set will include:

  • Processing of data for 1998, 1999, 2003 and future data sets.
  • Implementation of drizzle retrievals.
  • Implementation of new LWP values prior to April 25, 2002 as they become available.
  • Implementation of more current precipitation retrievals for 35-GHz radar.


The development of this data set has been a long complex process. Comments, suggestions, and/or identification of errors are much appreciated and should be directed to