High Cloud Fraction and Upper Tropospheric Water Vapor

Determined from GOES Dwell Sondes

Francis Bretherton and Don Wylie

DRAFT Aug 15, 1998 
Do not rely on this information without consulting the Author

Introduction

This study presents a new approach to statistical validation of parameterizations of cloudiness in Global Climate Models (GCMs), focusing on the thermal infra-red radiation associated with clouds in the upper troposphere ( p < 440 hPa) as detected by operational geostationary satellites. It is based upon a sample of data products¹ prepared from the archive of GOES Dwell Soundings over the continental U.S. and neighboring oceans for the period May 15-22, 1988. All data is averaged to cells comprising one degree of latitude and longitude. It was shown by Soden and Bretherton²  that the brightness temperature in the 6.7 um absorption band (GOES_VAS Channel 10) under clear sky conditions³  is closely correlated with the pressure weighted average of the relative humidity in the line of sight, so that to a first approximation the cloud free radiance for that band (product RC10 in this sample) serves as a measure⁴ of upper tropospheric relative humidity (UTH ) relative to water. Simultaneous estimates of the effective cloud fraction⁵ with radiating pressure p < 440 hPa (product CFHIGH) are available from pixel by pixel analysis of the CO2 Channels 2-5 on the same satellite, which are sensitive to temperature in various pressure ranges but strongly modulated by any cloud which may be present.

Observations

Data from the product set has been aggregated in several different ways. A date such as 1988140 indicates an average of observations over all times of day on May 18, 1988. A time such as 0830 indicates an average at that time of day over all of the 7 days for which observations were available. A letter "l" indicates Land (continental U.S ), whereas "o" indicates Ocean (principally the western Atlantic).

The links from the tables below show 9 panels.  The captions for these panels are:

Top Row : Left, Middle, Right

The normalized histogram for UTH in the indicated region. The total number of data cells is at the right of the title;
(a) CFHIGH, (b) the fraction of cells in which ANY high cloud is detected, as functions of UTH The straight lines are the least squares fit with weights proportional to the UTH histogram;
The normalized histogram for CFHIGH.

Middle Row: Left, Middle, Right

The cloud signals ( RA10-RC10, RA3-RC3, RA8-RC8) in the water vapor, CO2, and infrared window bands as a function of UTH.

Bottom Row: Left, Middle, Right

The cloud signals ( RA10-RC10, RA3-RC3, RA8-RC8) in the water vapor, CO2, and infrared window bands as a function of high cloud fraction.

Conclusions

UTH is distributed primarily in the range 10 % to 70% consistently in different times and regions. The almost complete absence of higher values may be associated with the smaller saturation values for water vapor relative to ice. The possibility of a calibration error is under investigation.

The effective cloud fraction CFHIGH for high cloud averaged over one degree cells is less than 10% in about 65 % of cases.  For the remainder, all values between 10% and 100% are approximately equiprobable with a minimum in mid-range.

There is an consistent linear relationship between CFHIGH and UTH, and a similar relationship for  the probability of high cloud being detected in a cell.  There is, however, large scatter about these relationships in individual cases.  Both linear relationships show a consistent diurnal cycle over land, characterized by enhanced cloud fraction and probabilities of detection around 1900Z (approximately 1 pm local sun time) at low values of UTH, decaying steadily to near zero around 1300Z (approximately 7 am local sun  time).  However, there is little diurnal change for large vlaues of UTH .  Over the ocean area as a whole this diurnal cycle is almost imperceptible.

The observed correlations between the cloud signals in bands 8, 3 and 10  and CFHIGH appear consistent with a priori expectations. More careful analysis of dependencies on cloud level pressure is needed to confirm this conclusion.

The relations between these same cloud signals and UTH are more complex and less consistent.

Over the ocean, the hour by hour variations in these cloud properties are markedly less than those over a week, though the qualitative features recur consistently.  This suggests that there remain significant dependencies on the synoptic situation which have not been resolved by these statistical summaries.

Global Climate Models which are correctly simulating the climate of the observed regions should be able to reproduce these statistical relationships after appropriate calculations of UTH, at least approximately.  A larger data sample is needed to narrow the uncertainties in this constraint.

Notes

1.  Title: "Radiances and Cloud Analysis from GOES Dwell Soundings"
     Authors: Bretherton, F. and Wylie, D.
     Publisher: Space Science and Engineering Center, University of Wisconsin - Madison
     Date: May 1998
     Permanent URL:
          http://www.purl.oclc.org./NET/U_WISCONSIN_SSEC/VAS_Product/EOSDIS/"

2.   Soden, B.J., and F.P. Bretherton, 1993, "Upper Tropospheric Relative Humidity for GOES 6.7 um Channel: Method and Climatolgogy for July 1987",  J. Geophys. Res., 98, 6669-16688.

3.  Clear sky conditions are defined by a cloud mask on a pixel by pixel basis.  Cloud cleared radiances (product "RCn" where n is the band number) are averages for each reporting cell of those pixels which are deemed clear.  Observed radiances ("RAn") are averages over all pixels in the cell.  The difference RAn-RCn is known as the cloud signal.  The cloud mask is determined solely using the spatial statistics of 8 km resolution data from band 8 (the infra red window), with consistency checks to previous and subsequent days.  The analysis uses no model derived estimates of temperature.

4.  For present purposes UTH is defined by the relationship
          UTH = cos(ASaZ)*exp(31.50 - 0.115*TC10)
where ASaZ is the satellite zenith angle, and the brightness temperature TC10 is related to the cloud cleared radiance RC10 by
         RC10 = c₁*k ³ /{exp( c₂*k/TC10) - 1}
with c₁ = 1.19107e-5 mW m^-2 steradian^-1, c₂ = 1.43884 K cm, and k = 1487.0 cm^-1 for channel 10.

Given a vertical profile of temperature and specific humidity as a function of pressure and a satellite zenith angle, RC10 may be computed directly using the Woolf radiation code. The description of UTH as the weighted average relative humidity (relative to water not ice) is an approximation introduced to simplify interpretation of any discrepancies between model calculations and observations, and should not be used for definitive comparisons.

5.  Effective cloud fraction is defined as unity minus the infra red transmissivity of the cloud layer. Neglecting reflection from cloud particles, it may be equated to the product of the geometric area of cloud projected normal to the line of sight and the emissivity of the cloud ,divided by the projection of the underlying Earth surface. For middle and low cloud the area under consideration comprises only those pixels for which no high cloud is detected. Thus inference of the total amount of middle and low cloud present in a cell requires additional modeling assumptions such as random overlap.

Last Modified  Aug 15, 1988

Science Evaluation               GOES VAS Product Set