Description: | The 'tuning functions' used to adjust the equivalent brightness
temperatures (TBs) measured by the ARM microwave radiometers (MWRs) are
now believed to be both incorrect and unnecessary. They should no longer
be used and the data (going back to 1992) that incorporated them should
be reprocessed. By eliminating these tuning functions the radiometer
retrievals would be independent of the soundings.
BACKGROUND
A recent comparison by Barry Lesht (ANL) of the precipitable water
vapor (PWV) retrieved from the MWR-measured brightness temperatures
against PWV derived by integrating along the trajectory of radiosonde
ascents has revealed that the MWR values are about 90% of those derived
from the soundings. This is directly attributable to the slope of the
tuning function for the vapor-sensing channel (23.8 GHz) of 0.915 which
is applied to the measured brightness temperatures prior to retrieval
of PWV.
The rationale behind the use of the tuning functions is that the
radiation model (Liebe 87), on which the retrieval is based, is
imperfect whereas the radiosondes represent 'ground truth.' Thus the
observed brightness temperatures must be adjusted to match those
calculated with the model using co-located soundings so that the
retrieval yields precipitable vapor amounts that agree with the
soundings.
Tuning functions were developed for the present ARM MWRs using
co-located soundings launched between October 1992 and December 1993.
These were adjusted slightly in January 1995 to account for the effects
of the 1-point calibration check performed prior to launch (see DQR
P950110.1):
23.8 GHz: TB_model = 0.789 + 0.915 TB_measured (R2 = 0.998)
31.4 GHz: TB_model = 1.142 + 0.910 TB_measured (R2 = 0.984)
However, repeating this exercise for soundings launched during 1994 and
1995 (excepting those that were mis-calibrated by the manufacturer; see
D960229.1) it now appears that the model-calculated
brightness temperatures are in much closer agreement with the measured
values and that the tuning functions account more for variations in the
radiosonde calibration than for any deficiencies in the radiation
model.
Consequently, it appears that the present tuning functions are
incorrect and bias the retrieved PWV low by 10%. In addition, given
the present agreement between measured and modeled brightness
temperatures, the tuning functions are also unnecessary.
METHODOLOGY
Brightness temperatures measured with microwave radiometer (MWR) serial
number 10, which was deployed at the central facility in December 1993,
have been compared against calculations using measurements from the
co-located Balloon-Borne Sounding System (BBSS). The results are
summarized in two tables. In each table, the calibration dates of the
sondes and MWR are listed as well as the time period and number of
samples included in each regression. Each MWR sample is a 40-minute
average, centered on the time of the sonde launch, of the microwave
brightness temperature. In order to include only clear sky conditions,
samples for which the standard deviation of the liquid-sensing (31.4
GHz) channel exceeded 0.3 K were eliminated. To assure that the water
vapor was reasonably homogeneous horizontally, samples for which the
standard deviation in the vapor-sensing (23.8 GHz) channel exceeded 0.4
K (in 1995) or 0.5 K (in 1994) were eliminated. The 1994 threshold is
larger in order to increase the number of samples and reduce the
standard error in the results.
The microwave radiometer measurements used in this comparison have been
reprocessed to account for calibration changes and other problems (see
P940813.1)
TB vs PWV
The first table is a comparison of microwave brightness temperature
(TB_mwr) regressed against the precipitable water vapor (PWV) computed
by integrating along the trajectory of the radiosonde ascent. The
sondes launched during May - December 1994 are compared against two
sets of MWR data; the first uses the May 1994 calibration, and the
second uses the calibration of July 1994. A comparison is also made of
TB_model vs PWV ('Liebe87') for reference.
The intercepts indicate the contribution due to molecular oxygen (i.e.
the tail of the 60 GHz line) which is affected by temperature and
pressure. Note that the 'Liebe87' intercepts vary seasonally as the
temperature changes. Note also that the effect of MWR calibration
changes is most evident in the intercept: offsets of 1-2 K are
observed. Because the MWR calibration values represent the slope of
the radiometer equation (see Appendix), the magnitude of the offset is
largest at 0 K (i.e. the intercept) and declines to zero at ambient
temperature (~290 K).
The slope of the regression is essentially unaffected by the MWR
calibration. Variations in the slope of the regression correlate with
sonde calibration date. The sondes calibrated in May 1994 or later
appear to yield much closer agreement between the measured brightness
temperatures and those calculated with the Liebe 87 model than those
calibrated in January 1994 or earlier, with which the present tuning
functions were developed.
TABLE 1. Microwave brightness temperature vs. precipitable water vapor
Relationship: TB_mwr (K) = intercept (K) + slope (K/cm) * PW_sonde (cm)
Standard Error of the intercepts and slopes are given in parentheses.
Date of Date of Period ------ 23.8 GHz ----- ----- 31.4 GHz -----
Sonde Cal MWR tip Covered N intercept slope intercept slope
1991-93 92-93 Oct92-Dec93 91 6.7 14.7 8.1 5.3
1992,93 Dec 93 Jan-Feb 94 85 6.9(0.19) 15.8(0.26) 8.8(0.13) 5.6(0.17)
1992,93 Liebe87 Jan-Feb 94 85 6.5(0.02) 13.8(0.03) 8.9(0.07) 4.5(0.09)
Jun 93 Dec 93 Apr 94 16 10.6(1.11) 14.8(0.55) 10.1(0.51) 5.6(0.25)
Jun 93 Liebe87 Apr 94 16 6.9(0.05) 13.6(0.02) 8.1(0.09) 5.0(0.05)
1992,93 May 94 May-Jun 94 48 7.0(1.03) 14.9(0.45) 7.8(0.41) 5.7(0.17)
1992,93 Jul 94 May-Jun 94 48 5.1(1.03) 14.9(0.44) 6.6(0.39) 5.7(0.17)
1992,93 Liebe87 May-Jun 94 48 7.1(0.11) 13.5(0.05) 8.4(0.16) 4.9(0.07)
Jan 94 Dec 93 Feb-May 94 95 7.6(0.27) 14.3(0.14) 8.5(0.15) 5.5(0.08)
Jan 94 Liebe87 Feb-May 94 95 6.9(0.05) 13.6(0.02) 8.1(0.09) 5.0(0.05)
May 94 May 94 Jun-Aug 94 78 12.3(1.04) 13.0(0.34) 11.0(0.39) 4.8(0.13)
May 94 Jul 94 Jun-Aug 94 78 10.3(1.04) 13.1(0.34) 9.8(0.39) 4.8(0.13)
May 94 Liebe87 Jun-Aug 94 78 7.8(0.22) 13.3(0.07) 8.6(0.29) 4.9(0.10)
Jun 94 May 94 Jul-Dec 94 57 8.3(0.37) 13.6(0.21) 8.8(0.19) 5.2(0.11)
Jun 94 Jul 94 Jul-Dec 94 57 6.4(0.37) 13.6(0.21) 7.7(0.18) 5.2(0.10)
Jun 94 Liebe87 Jul-Dec 94 57 6.9(0.08) 13.5(0.04) 8.3(0.10) 4.9(0.06)
Aug 94 May 94 Sep-Dec 94 90 7.4(0.15) 13.5(0.09) 8.8(0.11) 5.1(0.07)
Aug 94 Jul 94 Sep-Dec 94 90 5.5(0.14) 13.6(0.09) 7.8(0.12) 5.0(0.07)
Aug 94 Liebe87 Sep-Dec 94 90 6.8(0.05) 13.6(0.03) 8.6(0.09) 4.9(0.06) |