** This message was prepared by
Larry Miloshevich. Please send any questions/comments to Larry (milo@ucar.edu).
**
Dear CRYSTAL-FACE Colleagues,
If you use or intend to use any of the radiosonde data
being collected
during CRYSTAL, please read on.
A document entitled 'crystal_radiosonde_data.txt' has
been submitted to the ~/docs/ area of the Archive, and is also copied below.
Among other things, this document describes: the five sources of CRYSTAL
radiosonde data; how to acquire the raw data as soon as they are received after
a flight; info about corrections that are applied to the radiosonde humidity
data (which can be substantial in the UT); how to locate and interpret the
corrected data that are placed in the Archive; and the humidity profile plots that
are also placed in the archive.
Best Regards, and may you acquire excellent data...
Larry
Miloshevich, NCAR
*********************************************************************
filename: crystal_radiosonde_data.txt
>>>CRYSTAL RADIOSONDE DATA AND RELATIVE HUMIDITY
CORRECTIONS <<<
Larry Miloshevich, NCAR, milo@ucar.edu
This document describes the radiosonde data that will be
acquired during CRYSTAL, how to access it, and information about corrections
that are applied to the humidity data.
Radiosonde data are acquired from 5 sources: NWS stations
(KW=Key West, MI=Miami, TA=Tampa), the PNNL PARSL facility located at
Everglades City (PA=PARSL), and a mobile facility (MO=mobile). The original
data will be made available in their native format as soon as possible, and at
a later time most of the radiosonde data will be available from the ESPO
archive in exchange format 1001. A second set of radiosonde data files that
contain both the original data and the humidity corrected as described below
will also be available from the ESPO archive. The Key West NWS site uses
Sippican (VIZ) radiosondes, and no corrections are applied to the humidity data
from this site. The Miami and Tampa NWS sites use Vaisala RS80-H radiosondes,
PARSL uses Vaisala RS90 radiosondes, and the mobile facility uses Vaisala RS90
radiosondes with GPS winds. Parallel filename conventions are used for the
original and corrected data in the ESPO archive, as follows:
MI__RAW.RAOB MI20020509_1103_RAW.RAOB
(original)
MI__COR.RAOB MI20020509_1103_COR.RAOB
(corrected)
where the first two characters are the station code
(MI,TA,KW,PA,MO), followed by the UTC launch date (yyyymmdd) and UTC launch
time (hhmm).
AVAILABILITY OF ORIGINAL DATA IN THEIR NATIVE FORMAT:
NWS DATA: available via anonymous ftp in ascii format at:
eos913c.gsfc.nasa.gov/incoming/"site", where "site" is
"EYW" (Key West), "MFL" (Miami), or "TBW"
(Tampa). The thermodynamic data are in the "S" file and the winds are
in the "R" file for any given flight. Specific permission must be set
up for access to this machine. Contact Andrew Lare (GSFC,
lare@climate.gsfc.nasa.gov) to arrange access or for questions about the NWS
data.
PARSL DATA: available via anonymous ftp in ARM-style
netCDF format at: engineering.arm.gov/pub/parsl/data/datastream/sonde. Only the
corrected PARSL data will be archived at ESPO; the original PARSL data will
ultimately be archived at ARM. Contact Jim Mather (PNNL, Jim.Mather@pnl.gov)
for questions about the PARSL data.
MOBILE DATA: (availability info not yet determined).
Contact Jeff Halverson (GSFC, halverson@agnes.gsfc.nasa.gov) for questions
about data from the mobile radiosonde facility.
CORRECTED DATA: The corrected radiosonde data (_COR.RAOB)
will be placed in the ESPO archive after the original data are retrieved and
corrected. Plots of the corrected and original humidity profiles will also be
placed in the archive (e.g. ~/images/raob/MI__RH.PDF), which can be useful for
identifying bad data or cloud layers or other features. Contact Larry Miloshevich
(NCAR, milo@ucar.edu) for questions about the corrected radiosonde data that
are not answered by the discussion below.
*********************************************************************
INFORMATION ABOUT CORRECTIONS APPLIED TO VAISALA RADIOSONDE RH DATA
This discussion concerns the content of the corrected
data files, the corrections performed, and the humidity profile plots. For the
record, this is output from "version 3" of Larry Miloshevich's
Vaisala radiosonde humidity correction algorithm.
Corrected radiosonde data files contain the following
parameters, where the different RH parameters are defined below:
time T Z P RHorig RHsmooth RHfinal Wind_speed
Wind_direction
RHorig: Original RH from the original data.
RHsmooth: Smoothed original RH, needed for the time-lag
correction. The smoothing miminizes the third derivative of the time series
while ensuring that the smoothed data are everywhere consistent with the
original data within the 1% RH precision of the measurements. This parameter
MAY be a better estimate of the original measurements than the original
measurements themselves, in part because out of necessity great effort has gone
into identifying and handling noisy and other "bad" data.
Note: T and P are also smoothed in this manner, within
their respective precisions of 0.1C and 0.1 mb; then Z is recalculated from the
smoothed measurements.
RHfinal: RHsmooth corrected for the bias and time-lag
errors described below. THIS IS THE RELEVANT RH PARAMETER FOR MOST USES OF THE
DATA.
The file header describes which corrections were
performed, e.g.:
corrections: TL TD C SA
TL = Time-lag correction. This is a numerical inversion
algorithm that calculates the "ambient" humidity profile from the
measured humidity and temperature profiles, based on laboratory measurements of
the temperature-dependence of the sensor time-constant. The correction is
minimal at temperatures above -30 or -40C, but at colder temperatures is
strongly dependent on the temperature and on the instantaneous measured
humidity gradient. Further cursory description of the algorithm can be found in
Miloshevich et al., proceedings of the 11th and 12th ARM Science Team meeting,
available at: http://www.arm.gov/docs/documents/technical/conference.html
TD = Temperature-dependence correction. Accounts for
inaccuracy in the Vaisala equation that gives the temperature-dependence of the
sensor calibration. This correction is minimal at temperatures above -30C, and
increases with decreasing temperature to about 12% of the measured RH at -60C
(%, not %RH), 21% at -70C, and 32% at -80C.
C = Contamination correction. Compensates for the
tendency of non-water molecules from plastics in the radiosonde packaging
material to occupy binding sites in the sensor polymer and render them unavailable
to water molecules, leading to a dry bias in the measurements. The correction
is a function of RH and age of the radiosonde, but is set to zero for
radiosondes calibrated after May 2000 due to introduction of a sealed sensor
cap by Vaisala that supposedly eliminates the contamination problem (so this
correction will probably be zero for CRYSTAL radiosondes). The contamination
correction is empirical and statistical, and has the largest uncertainty of all
the corrections.
SA = Sensor-aging correction. Compensates for normal
long-term drift in the sensor calibration. This is a small empirical correction
that is a function of the radiosonde age.
Further information on the latter so-called "bias
corrections" can be found in Wang et al. (2002, J. Atmos. Oc. Tech.,
accepted), and in Miloshevich et al. (2001, J. Atmos. Oc. Tech., 18, 135).
***********************************************************************
HUMIDITY PROFILE PLOTS:
Plots of the humidity profiles are located in the
~/images/raob/ area of the ESPO archive, with names like MIyyyymmdd_hhmm_RH.PDF
Generally there will be a time series of soundings from a given site in a given
file (4 soundings per page), in which case the date and time refer to the first
sounding in the file. When new soundings are added to one of these files
(multiple times daily), the old file is overwritten in the archive, so its file
size increases but its name is unchanged.
The curves are defined according to the variable names
given above: RHorig (blue), RHsmooth (black), RHbias (bias corrections only,
green), RHfinal (red), ice-saturation (dashed), tropopause or thereabout
(asterisk).
***********************************************************************
LIST OF BAD DATA:
At some point I may make a list of soundings that are bad
for one reason or another, such as sensor icing. Vaisala RS80-H (not RS90)
sensors can become coated with ice if the radiosonde passes through supercooled
liquid water between 0C and -40C, in which case the "humidity" can
remain near ice-saturation well into the stratosphere. Data files and plots are
still made available for these soundings, but they will not be used in
subsequent analyses since they are totally erroneous, and they probably
shouldn't be used in anyone else's analyses either (except maybe below the
icing event). Sometimes less severe icing can be seen (e.g., 20% RH in the
stratosphere following passage through SLW between 0C and -40C) by looking
through the profile plots in the vicinity of time periods of identified severely
iced soundings.
It is clear from looking at the humidity plots that VIZ
carbon hygristor sensors (Key West) just don't function in the UT.
*********************************************************************
REQUEST:
I am very interested in learning about the effect of
these corrections on research areas that use the radiosonde humidity data, and
I humbly request that people who use these data, especially if you choose to
run a sensitivity test comparing the corrected and uncorrected data, please let
me know what you find regarding the effect of the corrections (milo@ucar.edu).
Thank you!
FYI, I am considering a future multi-author paper that
shows the impact of these humidity corrections on various research areas (e.g.,
instrument intercomparisons, retrievals, RT calculations, numerical modeling,
etc.). Contributors might provide 1-2 figures and some text. It would add depth
to an assessment of the impact of the corrections on RH and PWV, and would also
provide an illustrative reference for future papers that you write that use RH
data corrected in this manner. For now it's just an idea to keep in mind, but
please let me know if you might be interested in contributing to a project like
this...
*********************************************************************
CREDITS:
The following are the hard-working lead people that are
acquiring the radiosonde data, correcting it, and getting it to you as quickly
as possible.
Numerous other engineers and technicians are, of course,
behind this effort as well.
NWS sondes: PI: David Starr/GSFC
PI: Humidity Corrections and QC: Larry Miloshevich/NCAR
Mobile sonde: PI: David Starr/GSFC
Co-I: Jeff Halverson/JCET-UMBC
PI: Humidity Corrections and QC: Larry Miloshevich/NCAR
PARSL sondes: PI: Jim Mather/PNNL
PI: Humidity Corrections and QC: Larry Miloshevich/NCAR