This web page describes engineering measurements using the ROOF system, and is linked to by the main web page that describes the ROOF observing program.

The MTP observes at three local oscillator frequencies, 54.0, 55.47 and 58.80 GHz.  IF passbands extend from approximately 0.25 to 0.41 GHz on both sides of the LO frequency.  For observations at the surface, the predicted applicable ranges for the three channels are 2062, 816 and 323 meters.  The MTP scans 7 elevation angles:  0.0, 5.7, 9.0, 14.4, 23.3, 39.0 and 90.0 degrees.  The following table summarizes applicalbe altitudes versus elevation angle for each channel.

 Elev   Ch#1  Ch#2  Ch#3   Reciprocal  Air
  LO =  54.0  55.47 58.80  Air Mass    Mass
 
  90.0  2062   816   323    1.000     1.000
  39.0  1298   514   203    0.629     1.59
  23.3   816   323   128    0.396     2.53
  14.4   514   203    80    0.249     4.01
   9.0   323   128    51    0.157     6.38
   5.7   203    80    32    0.099    10.1
   0.0     0     0     0    0.000    infinity
 

Calibration of MTP Radiometer

The MTP radiometer must undergo an elaborate calibration whenever it is modified.  Since we are using a new set of frequencies, having unkown gain properties, and since the instrument is mounted in a new box with a new radome, with unknown absorption and reflection properties versus viewing angle, these aspectso f the MTP instrument must be determined.

Engineering (Instrument) Temperatures

The first consideration for any new deployment of the MTP is its safety.  Overheating is difficult to predict ahead of time, even in a laboratory setting.  The first field observations are therefore focused on temperatures within the instrument.  The following figure shows intrument temperatures for the second day of observations.

Figure 1.  MTP instrument temperatures for the 25-hour period of measurements starting at 2001.06.05, 00Z.

It should be pointed out that although the OAT points in the above figure are affected by the "heat island" effect because the sensor is close to the ROOF site, these temperatures are appropriate for use in understanding the physical environment of the MTP instrument.

Figure 2.  More MTP instrument temperatures for the 25-hour period of measurements starting at 2001.06.05, 00Z.

During this day engineering temperatures were benign.  It was clear most of the time (a few scattered clouds in the early afternoon).  We believe that the frequency synthesizer will be the most likely measurement point to exceed safe limtis, and as long as it does not exceed 65 C the instrument will be OK.  Since Fig. 8 shows that the synthesizer reached only 59 C we can be assured that during clear conditions, when the ambient air temperaure does not exceed 25 C, the MTP instrument temperatures will be within safe limits.  On hotter days, when the synthesizer temperature exceeds 60 C, the plan is to lift the MTP radome lid at the non-viewing end in order to allow a greater ventilation of outside air.  If temperatures reach 65 C the plan is to turn off the MTP.

Gains - Using Hot and Base Targets

Gains for the three MTP channels are not known accuratley, since we do not know the emissivity of the hot target. If it is assumed that for all channels the hot target emissivity is 98%, the following horizon TBs are obtained for the 24 hour period of June 2.

Figure 3.  Horizon TBs versus time for the 24-hour period 010602, using gains based on hot target emissivity = 98% for all channels. (Groups of 4 cycles have been averaged to reduce scatter.)

Figure 4.  Horizon TBs versus time for the 24-hour period 010602, using gains based on hot target emissivities = 89%, 99% and 96% for channels 1, 2 and 3.  (Groups of 4 cycles have been averaged to reduce scatter.)

If there were no horizontal gradients of air temperature along the horizontal line of sight, and if none of the antenna pattern intercepted the ground or trees, and if the effect of slightly varying beamwidth versus frequency is small, then we would be justified in requiring all horizon TBs to be the same for all 3 MTP channels.  This requirement can be met by allowing the hot target emissivities to be different for each channel.  This shouldn't be surprising, since the hot target was designed for maximum emissivity at one frequency.  I don't know what that frequency is, but it must be close to the center of the satellite SCAMS frequencies.  My highest emissivity is at Channel #2, or 55.47 GHz.  The SCAMS frequencies were 52.8, 53.9 and 55.3 GHz, so the target should be optimized for the 54 GHz region, where I find a low emissivity value.  I cannot explain the pattern of my estimated hot target emissivities for the ROOF MTP, but will assume for now that they are correct and differ at the three frequencies for reasons that I do not understand.

Figure 5.  Hot target-based gains for Channel 1, for the 25-hour period starting 2001.06.05, 0Z, assuming EmHot1 = 89%.  (Groups of 4 cycles have been averaged to reduce scatter.)

Figure 6.  Hot target-based gains for Channel 2, for the 25-hour period starting at 2001.06.05, 0Z, assuming EmHot2= 99%.

Figure 7.  Hot target-based gains for Channel 3, for the 25-hour period starting at 2001.06.05, 0Z, assuming EmHot3 = 93%.

There's an odd shape of the Ch#3 gains versus tMXR which will be discussed below.

The fitted gains for the "hand-fitted" hot target emmisivities are:

    G1 = 28.49 - 0.307 * (tMXR - 44.4 C) [cts/K], assuming EmHot1=89%
    G2 = 31.60 - 0.372 * (tMXR - 44.4 C) [cts/K], assuming EmHot1=99%
    G3 = 29.31 - 0.925 * (tMXR - 44.4 C) [cts/K], assuming EmHot1=96%

Notice that gain for Channel #3 has a greater sensitivity to mixer temperature.  Also, as mentioned above, the channel #3 gains don't fall on a straight line; they form two straight lines, with the inflection at tMXR = 44.9 C.  The departures of the plotted Ch#3 gains from the fit go from +3% to -6%.

This misbehavior of channel #3 gain promted me to check the "noise diode counts" (NDC) during the misbehaving period.  NDC can be affected by both gain changes and "noise diode output" (NDO) changes.  If the peculiar Ch#3 gain behavior is real, and if only the gain is changing in a peculiar way (i.e., the NDO doesn't change), then NDC will also exhibit the peculiar +3% to -6% behavior.  And this is what happens; NDC3 exhibits the same peculiar behavior.  This is more easily shown by calculating NDO = NDC/G.  If NDO is constant, and NDC changes in the same way as G, then NDO will be unaffected during the "peculiar period."  The follwoing graph shows NDO versus noise diode temperature.

Figure 8.  Noise Diode Output, NDO, based on gains derived from the hot and base targets and assumed hot target emissivities 88, 97 and 92%, for RF010604.

Although the figure is for a different date, a similar plot was obtained for RF010605 data (the 010604 data are shown because they represent a better determination of the NDO values).  The two Ch#3 gain "features" in this figure exhibit departures from the fit of +4% and -0% (the range is 4% instead of 9%).  This confirms that the peculiar gain behavior for Channel #3 is real, and it shows that a simple gain equation for this channel cannot be used.  Rather, gain values will have to be obtained from the noise diode, using equations for NDO versus tND.  The task remains to obtain good values for NDO for each channel.  For this, it will be necessary to compare TBs for hypothetical NDO values.  This topic will be returned to later.

The NDO values in the previous figure can be described by the following fits:

    NDO1 = 73.70 +0.254 * (tND - 42)
    NDO2 = 84.45 +0.211 * (tND - 42)
    NDO3 = 95.48 +0.061 * (tND - 42)

Another method for determining gains is to compare gain-dependent TBs for all elevation angles with credible physical temperatures averaged along the line-of-sight.  The inferred TBs are to be adjusted until they agree with the expected TBs, and one adjustment parameter is the assumed gain. However, there are other parameters that are free to be adjusted within certain limits, and these must be taken into account while determining the best gain estimate.  Recall that the horizon views produce "antenna temperature," not brightness temperature, and these two can differ slightly due to such radome effects as 1) radome material reflection, 2) radome material absorption, and 3) reflections and absorptions caused by water that has collected on the radome.  In order to proceed with this method for determining gains it will be necessary to first describe a method for removing "radome water" effects.  The topic of determining good gains will be returned to in a later section.

Radome Water Effects

Before good MTP TBs can be accepted they must be free of water-on-radome effects.  The following data illustrate the effect of stratus mist which tends to collect on the top of the MTP radome, but not on the side that's close to where hot air is being blown.

Figure 9.  Same applicable altitude TBs, which should be the same unless there's dew at one location and njot at the other - which was the case until I wiped the radome at 52 ks.

MTP Ch#1 has an applicable range of 2062 meters, whereas Ch#2 has an applicable range of 816 meters.  Observations are made at zenith, 39.0, 23.3, 14.4, 9.0, 5.7 and 0.0 degrees.  Applicable height is applicable range times sine of elevation angle, so Ch#1's applicable height at 23.3 degrees is the same as Ch#2's zenith applicable height.  Provided the atmosphere temperature field is horizontally stratified (and for the moment let's ignore differences in radome reflection and absorption effects), the two TBs should be the same.  The MTP observing frequencies and elevation angle sequence was chosen to permit comparisons of this type, to aid in identifying radome water effects (and to provide reality checks on adopted gains).

The above figure illustrates how radome water can be inferred from the Ch#1 and Ch#2 measurements.  From 11 to 14.6 hours UT the "TB2 zenith" values are higher than the "TB1 at 23.3 degrees" values. Sky conditions were "overcast" during this time, and my observing log states that:

      7.3 hrs UT: "V lt mist"
      8.0 hrs UT: "V lt mist; Zcb ~500' ASL"
    13.8 hrs UT: "V lt mist; Zcb ~200' ASL; streets are wet"
    14.6 hrs UT: "Mist has stopped; wiped radomes"

It is clear that the "hump" feature for "TB2 Zen" can be attributed to water accumulation on the top of the radome, with a much smaller or non-existent accumlation on the side of the radome where the 23.3 degree observations are made.  The WVR recording of Lz(t) also shows two "humps" corresponding to the two times that dew is inferred to exist on the MTP radome.  the MTP and WVR radomes have  different blower systems, and they cannot be expected to be 100% correlated in their radome water effects.  Specifically, the WVR data cannot be used to correct the MTP data for radome water effects; rather, the MTP radome water corrections must come from the MTP data itself.  This data set allows that there's a way to determine times when dew was present at the top of the MTP radome.  It also illustrates a method for substituting the Ch#1 23.3 degree observable for the Ch#2 zenith observable when MTP radome water effects are suspected or known to be present.

Discovery That OAT Measurements Are Affected by "Heat Island"

The original goal of this section was to derive MTP gains by comparing "predicted TB" with "measured TA" for the horizon observables.  The proposed sequence is to predict TB at the horizon (TBh) and compare it with "outside air temperature" (OAT), and when they differ the assumed gains should be adjusted to provide agreement.  The starting gains will be those that have been derived from the hot and base targets, described in a preceding section.

In doing this it is assumed that antenna temperature, TA, is the same as brightness temperature, TB.  TA is expected to differ only slightly from measured TB, since the emission brightness temperature of any radome reflections and absorptions are going to be at approximately the same temperature as the sky emission that it is replacing.  For now, let's assume the measured TB equals measured TA.

The goal of establishing gains by comparing measured TBh with OAT ran into a problem with the OATs not being representative of the average temperature along the horizontal line-of-sight.  The balance of this section is included to illustrate the pitfalls of using this method for gain calibration.  The OAT readings were taken close to the ROOF site, and initially I did not realize how misleading such readings can be.  They were misleading because the ROOF site is part of a "heat island."  In other words, the assumption of horizontal stratification of the temperature field cannot be made for altitudes close to ground level when the observing site is within a heat island.

Figure 10.  Example of MTP horizon view TBs (3 lower traces) and OAT measurements (squares), showing discrepancy between the two during daylight hours (13 - 25 hrs, UT).

The figure, above, suggests that the OAT measurements taken close to the ROOF site are contaminated by local temperature "heat island" effects.  I could not adjust the hot target emissivities in a way that would reconcile both daytime and nightime TBh with OAT.  The green open squares are the most affected, which is consistent with this sensor's location being between two mobile homes, whereas the red solid square is over a bank above an oleander tree where there is probably a stronger breeze to dilute the heat island air with surrounding air.  To further check this theory, that daytime surface temperature measurements close to the ROOF site are warmer than air in the distance (in the direction of the MTP horizon view), I mounted the Radio Shack sensor to the end of a tree trimmer pole and took readings 1000 feet east of the ROOF site, in the golf course, with the sensor 14 feet above ground level.  This revealed a 3.3 C difference in temperature compared with the ROOF site sensor, with the golf course temperature being cooler.  If I could have held the sensor higher, I might have noted an even larger difference, since there seemed to be a ground-based superadiabatic lapse rate over the golf course and at the ROOF site.  This shows that the daytime OAT points in the above figure cannot be trusted for calibrating the MTP.  It also implies that a greater degree of trust can be given to the red solid square measurements during the night hours.  The best solution to this challenge is to use an IR radiometer pointed at the MTP's horizon.

In order to determine if the hot target emissivities used in the analysis to this point (89, 99 and 93%) are correct, it will be necessary to validate the use of my surface temperature sensors (i.e., OAT measurements) for characterizing the temperature of the air along the MTP horizion view path. Even the night time OAT values should be challenged, since heat island effects may still be present in them at night.  Certainly, the daytime OAT measurements should be improved.  An IR radiometer should provide improved quality "OATs" which are needed to correctly establish system gains, and therefore hot target emissivities as well as NDO values.

IR Radiometer Version for OAT

An infrared radiometer was purchased from Omega Engineering Inc.  The model Omega OS530HR appears to be have a passband of 8-14 microns, and it has a coldest reading range of -30 C.  The angular reception pattern is specified as 1:20, or a circle with a diameter of about 2.9 degeres.  The IR radiometer arrived June 11, and readings of the horizon (as well as angular scans from horizon to about +8 degrees) were begun June 11.75 UT.  The Radio Shack in situ temperature readings (near the ROOF site) were continue for about 2.5 days inorder to determine if a correction could be applied to earlier RS data.  The next graph shows the RS and IR readings for the overlap period.

Figure 11. Comparison of in situ Radio Shack temperature sensor readings (blue) and IR radiometer readings (red) for a 2.5 day period.

The two temperature readings differ in a manner that seems to have a diurnal pattern.  This is more easily seen when the data is plotted versus "fraction of day," shown in the next figure.

Figure 12.  Comparison of in situ Radio Shack temperature sensor readings (blue) and IR radiometer readings (red) versus fraction of UT day (i.e., folding of 2.5 days into a single 24-hour display).

The in situ temperatures appear to not reach cold enough temperatures late at night (at 0.3 UT day fraction), yet they appear to be too cold in mid-morning (at 0.65 UT day fraction).  The next figure shows the difference between the two temperatures plotted versus UT day fraction.

Figure 13.  Difference between in situ Radio Shack temperature sensor readings and IR radiometer readings versus fraction of UT day (red).  A freehand "fit" to this data (dashed green) is presented. Times of sunrise and sunset are shown.

The freehand fit to the in situ temperature errors does a good job of representing a diurnal pattern, and this suggests that earlier data can be corercted.  The corrections range from -5 to +5 C.

Before accepting the IR temperatures for OAT in place of the Radio Shack sensor's in situ temperatures, it would be prudent to compare the two OAT candidates with MTP horizon TBs.  There are two times when the IR temperature was falling while the RS temperature was rising.  The MTP doesn't need a correct set of gain values to distinguish between these two conditions.

Figure 14.  Comparison of MTP's horizon TB (average of Ch# 1, 2 and 3), the Radio Shack temperature sensor in a white box near the ROOF site (blue stars), and the IR radiometer looking at the eastern horizon (red squares).

This figure shows that neither the nearby Radio Shack in situ sensor nor the IR radiometer are compatible with the MTP horizon brightness temperatures for the 25-hour period shown.  There is a major discrepancy between the MTP and IR trends during the period 15 to 20 hours, when MTP TB is rising but the IR temperature is decreasing.  The in situ temperatures are also unsatifactory from the standpoint of day/night differences; during the day the MTP is lwoer by about 5 C, whereas during the night MTP is lower by only 2 C.  One hypothesis is that during the day there's a stronger "heat island" effect than at night.

How can the IR differences be explained?  Recall that the IR radiometer beam is a circle about 2.9 degrees in diameter, and that when it is pointed at the horizon the reception circle includes trees about a mile away.  When the sun rises (at 12.8 hours UT) perhaps the trees are heated faster than the air, and by 14 hours UT they are ~6 C warmer than the air.  At night, maybe the trees cool faster than the air, and become ~3 C colder than the airby 6 hours UT.

It appears that both the in situ and IR temperatures are flawed in ways that render them unsuitable for calibrating the MTP!

If only we had known this before the experiment.  The "pyramid target" should have been used at the ROOF site to provide a good quality known physical temperature emission source for the MTP.  It changes temperature slowly, so it could have been stored in a location that was either warmer or colder than ambient, and quickly brought to rest upon the radome at the MTP zenith scan location.

An alternative to consider for future experiments is to locate the MTP at a location that is unlikely to be subject to heat island effects.  Flat topography is one requirement, and few heat sources (such as nearby residences) is another.  The ideal location would be a site from which radiosondes are lauched (such as Vandenberg, or Point Mugu), so that MTP TBs at all scan angles could be compared with RAOB-predicted TBs for the purpose of establishing gains.

The ROOF data base is moderately extensive, and includes several stratus formation a dissipation cycles, so there is justification in trying other techniques for calibrating the MTP as it was deployed for ROOF.  One attempt failed, and it is decribed on another web page, Using VBG RAOBs to Calibrate MTP.  It consisted of starting with T(z) from Vandenberg RAOBs and using local supporting meteorological data (Meteograms) to adjsut the VBG T(z) to arrive at an improved T(z) for the ROOF site.  This T(z) was then to be used to calculate predictions of what MTP should observe at its varous scan angles, and these would then be used to validate (and adjust) assumed MTP gains.  It turned out that the horizontal gradients within the PBL are so great that even local meteogram data was inadeqate for adjusting the VBG T(z) for use at ROOF.

Another method for obtaining T(z) for the ROOF site was developed, and is described in the next section.

Calibrating MTP Using T(z) Produced By IR Radiometer During Drive Along Nearby Mountain Road

An alternative method for calibrating the MTP is to derive a T(z) profile by driving up a nearby mountain road and stopping frequently to take horizon IR temperature readings at an azimuth that overlooks the ROOF site (which is directed away from the mountains).  If a good T(z) profile can be derived this way, then it can be used to calculate predicted MTP TBs using software programs now in place.  Before the IR radiometer is used for this purpose, however, it is necessary to understand the "applicable range" of the IR radiometer.  An investigation of this and other related IR radioemter properties is contained in the next section.

IR Radiometer Characterization

The Omega IR radiometer arrived a the ROOF observing site June 11, and was mounted in the observing room for convenient manual logging that evening.  The frist emasurements were made on the roof, just before noon, and are plotted below.

Figure 15.  IR brightness temperature, TBir, versus sine of elevation angle.

This first set of data from the IR radiometer after being mounted in the "observing room" shows good behavior of TBir versus elevation angle.  The entire IR beam is in clear sky for elevation angles above 2.5 degrees, where "sine(EL)" is 0.044.  It may be important to take readings at a range of elevation angles so that an extrapolation to the horizon can be made using the higher elevation data, since the horizon view includes trees about a mile away.  This needs to be determined.

Assuming that there is negligible effect due to trees in the IR beam, the IR brightness temperature profile, above, would be linear if T(z) were linear from the ground up. There is some evidence for a higher vertical temperature gradient layer based at ground level in this figure.  This feature may or may not be real.  It should be possible to estimate the "applicable range" for the IR radiometer using a plot like this, provided the true temperature profile is known.  If T(z) were adiabatic, for example, the oserved slope of TBir versus sine(EL) in the above figure could be interpreted as requiring an applicable range of ~12 km.  If T(z) is subadiabatic, then the IR applicable range would be greater than ~12 km.  As a first approximation, the IR radiometer has a sea level applicable range of about 14 km.  This is much greater than the MTP Channel #1, at 54.0 GHz, which has an applicable range of 2.0 km (Channels #2 and #3 have applicable ranges of 0.85 and 0.323 km).

Figure 16.  Sample of data from IR radiometer from before sunset to several hours afterward.  Strong "sundowner" winds created an adiabatic condition during the first two (warmer) readings.  A ground-based inversion can be seen forming in the last profile.

This figure can be treated like the MTP, in that T(z) can be approximated by plotting TBir versus applicable altitude (applicable range times the sine of the elevation angle).  Since applicable range is about 14 km, TBir at 7 degrees elevation angle corresponds to air temperature at 1.7 km (5600 feet).  TBir(Hb) profiles can provide an additional check on MTP-derived atmospheric behaviors.

Another use to which the Omega IR radiometer can be put is measurement of T(z) using horizon readings at various altitudes along Highway 154 and the East Camino Cielo Road, both of which are close to the ROOF site.  The following figure illustrates this.

Figure 17.  IR radiometer horizin TB versus observing altitude (red), compared with the VBG RAOB (dashed green).

The red trace in this figure was obtained by driving to the 3000 foot altitude along the East Camino Cielo Road, approximately 4 miles north of the ROOF site, and then driving back and parking at several sites to make IR readings of the southern horizon with the IR radiometer.  Since the applicable range of the IR radiometer is of the order 14 km (~9 miles), and since most of the sites were 2 to 5 miles northwest of the ROOF site, the applicable locations for these data are approximately 4 to 7 miles southeast of the ROOF site.  The MTP scans along the southeast azimuth, and has applicable ranges of 1060, 2700 and 6800 feet, or about 1 mile southeast of the ROOF site, the IR measurements are only 3 to 6 miles southeast of the MTP ROOF measurements.  The VBG launch site, in contrast, is 50 miles west-northwest of the ROOF site.  Therefore, the temperature profile made with the IR radiometer along Hwy 154, etc, is much more appropriate to use for calibration or evaluation of the MTP compared with the VBG sounding.  The large difference between the IR-based T(z) and the VBG sounding taken 5 hours earlier (just before sunrise) illustrates the inadequacyof trying tio use VBG RAOBs for calibrations at the ROOF site.

A few qualifiers are needed before accepting the IR-based "Hwy154" T(z) profile.  First, the coldness above 2600 feet is probably produced by adiabatic cooling of upslope winds (synoptically driven southward).  I recall seeing clouds at that altitude that did not exist at lower altitudes.  This means that perhaps the VBG profile is a better choice above 2600 feet, since the "over the mountain ridge" will be local perturbations and are not likely to be present southeast of the ROOF site.  Second, the VBG sounding occurred at 5:00 AM, whereas the Hwy154 data were taken at about 10 AM.  During this 5 hour interval the gound level air must have heated approximately 7 C (based on the SBA meteogram).  At SBA (4 miles from the ROOF site) the surface temperature at 12Z was 12.2 C, only 2.0 C warmer than VBG.  If the Hwy154 data had been taken at 12Z the two profiles would have been in better agreement.  Third, the drive down Hwy 154 took 42 minutes.  During that time the surface was heating at the rate of 1.1 C/hr (based on my ROOF measurements).  Therefore, the upper altitude temperatures should be corrected upward about 0.8 C to correspond to the end of data time (low altitude readings).  Assume, then, that the 2000 foot temperature was 11.2 C, at the time the 500 foot temperature was 17.2 C, then the vertical temperature gradient between these two altitudes would have been -13.1 K/km.  This is superadiabatic!  This implies that air at the surface was convectively circulating all the way to the 2000 foot region.  This could explain why the temperatures at 2000 feet are colder than those at VBG at the earlier time.

Several features of the Hwy154 T(z) profile are credible, provided small adjustments for likely effects are made to it.  For the purpose of calibrating the MTP in a daytime setting (one of the goals of this effort), I adopt the following T(z) profile for the epoch June 12, 1650Z.

Figure 18.  Adopted T(z) for Jun 12, 1650Z.

The black trace profile in this figure incorporates the adjustments described above, and is suitable for use at the indicated time.  This represents the first time during the ROOF field observations that a credible T(z) profile has been available for use at the ROOF site!  It should now be possible to evaluate gains, as well as the hot target emissivities.

The "adopted" T(z) profile has been converted to a format resembling the R-file format I derive from RAOB data.  Specifically, the following information is contained in an R-file:

    RSB 2001 June 12 1650Z

       P       Zp     Zg     T     VD  LWC

    1006.0     61     64   19.8   8.60   0
     996.5    140    145   17.5   8.59   0
     990.8    189    195   16.8   8.56   0
     985.0    238    244   16.5   8.52   0
     935.3    671    681   10.6   6.88   0
     921.6    792    804   13.6   7.05   0
     908.1    914    929   16.2   6.66   0
     889.2   1088   1108   16.6   5.87   0
     875.1   1219   1244   16.2   5.37   0
     835.6   1597   1636   15.2   4.13   0
     724.3   2743   2830    8.5   2.42   0
     595.3   4267   4436    3.8   1.12   0
     465.7   6096   6382   -9.1   0.52   0
     301.0   9144   9615  -31.3   0.04   0
     123.5  14631  15404  -71.0   0.02   0
      32.2  21336  23582  -59.8   0.01   0
       2.2  30480  40957  -40.9   0.00   0

The 2nd and 3rd columns are pressure altitude [meters] and geometric altitude [meters].  The VD-column is vapor density [g/m3], derived from the 12Z VBG RAOB water vapor mixing ratio profile.  I used the hydrostatic equation and vapor mixing ratio profile information to calculate geometric altitude.  (Including water vapor had negligible effect on almost all brightness temperatures, being less than 0.04 C for all but two, Ch #1 at zenith and the next lower elevation angle, where the differences were 0.24 and 0.12 K.)  (programs GAP_ROOF.BAS and TB_ROOF.BAS were used to process R-file R1612B.RSB to TBs.)

The next three figures show the best subjective fit between observed and predicted TB(elevation) that can be achieved using only gains for each channel as "free to vary" parameters.

Figure 19.  Attempted fit of observed TB1(elevation) with predicted TB1(elevation) using only channel #1 gain as a single degree of freedom for fitting.

Figure 20.  Attempted fit of observed TB2(elevation) with predicted TB2(elevation) using only channel #2 gain as a single degree of freedom for fitting.

Figure 21.  Attempted fit of observed TB3(elevation) with predicted TB3(elevation) using only channel #3 gain as a single degree of freedom for fitting.

I consider these "fits" to be unacceptable.  Two discrepancies are apparent: 1) the adopted T(z) surface temperature appears to be too hot, and 2) the actual T(z) appears to have a stronger "elevated inversion layer" than the adopted T(z).  An alternative way of expressing the first discrepancy is to say that the MTP reference target is actually warmer than its reading (as if the sensor for the target is not close enough to the surface, where the measured microwave emission originates).  This alternative way of accounting for the first discrepancy suggests that we should allow another degree of freedom in the fitting process; namely, an offset correcction to the target temperature.  The following three figures show that an improvement is possible when this two-degrees of freedom fitting is performed.

Figure 22.  Attempted fit of observed TB1(elevation) with predicted TB1(elevation) using channel #1 gain AND a reference target offset as degrees of freedom for fitting.

Figure 23.  Attempted fit of observed TB2(elevation) with predicted TB2(elevation) using channel #2 gain AND a reference target offset as degrees of freedom for fitting.

Figure 24.  Attempted fit of observed TB3(elevation) with predicted TB3(elevation) using channel #3 gain AND a reference target offset as degrees of freedom for fitting.

Although these fits are better, they now suffer from an apparently too conservative representation of the actual T(z) elevated inversion structure.  It is acceptable for TB1 to be too cold at the horizon because of its long applicable range (2.1 km); a substantial portion of its measured emission originates from trees and other non-atmospheric material when it views the horizon.  Even at an elevation angle of 5.7 degrees a significant portion of the antenna pattern is at low elevations angles, some even below the horizon, which will influence measured TB. At 9.0 ndegrees, however, TB1 should be free of ground influence, but it will be somewhat influenced by horizontal gradients close to the surface, as well as temperature structure in the lowest km.  TB3, on the other hand, should agree much better at the horizon because of its short applicable range (0.3 km), and TB3 should be a reliable indicator of near-surface temperature structure (regardless of gain assumptions, or tTGT offsets).  Thus, the discrepancy between observed and predicted TB3(elevation) structure should be viewed as significant, and reflective of an incorrect adopted T(z) structure.

At this point in the analysis, I am reluctant to proceed along a path that is too subjective!  Yes, an improved T(z) could be adopted which would provide better agreement with observed TB(elevation), but this process would give too much freedom to the person making arbitrary adjustments, and would therefore undermine the credibility of the resultant gains.  Moreover, it would be impossible to estimate the existence or not of "window correction table" (WCT) strucuture.  This is a map of corrections needed to correct for reflections and absorption/emission from the radome cover, and for arborne work has been found to vary with elevation angle.  Since the airborne radome cover is much thicker than the ground-based radome, it would be reasonable to assume that the WCT is zero for all elevations and channels.

Lessons Learned

Calibrating an MTP is critical to success.  The WVR is self-calibrating, thanks to the use of tip curves.  But an MTP must either be built with superb engineering or undergo extensive field calibration. Here are some suggestions for next time.

Next time, we should use a high quality "ambient horn" for establishing a target temperature offset.  Gains, however, require more effort.

MTP gains can be derived by deploying an MTP at a RAOB site for a period that includes several clear-weather soundings.  The logistics for such a calibration are daunting, though we have done it a few tiems (at El Monte Airport, Point Mugu Naval Air Station, Ontario Airport, Buffalo Airport, the ship Quadra at Ocean Station P, Madrid Tracking Station and Loyala Marimount University).  If an alternative site it used, it should at least be a flat area, with an unobstructed view of a distant horizon.  There problems of a site remtoe from a RAOB site merely compound the calibration problems associated with not knowing how much low elevation observables are contaminated by nearby non-atmospheric objects.

Impasse Reached

Because of the inability to arrive at a credible calibration of the MTP I reluctantly conclude that it would be a waste of my time to continue to work with the MTP data to produce T(z) histories for the study of the atmospheric physics of stratus formation and dissipation.

Return to main page.

This site opened:  May 31, 2001.  Last Update:  September 8, 2001