The brightness temperature measurements from the Airborne Imaging Microwave Radiometer (AIMR) microwave radiometer 37 GHz and 90 GHz channels are used to determine the column water vapor amount under clear-sky conditions during the Indian Ocean Experiment (INDOEX) campaign in 1999. The retrieval algorithm is based on the integrated relative humidity profiles measured by numerous dropsondes released from the National Center for Atmospheric Research (NCAR) C-130 aircraft that operated during the 2 month Intensive Field Phase (IFP) of INDOEX. The highest correlations of the linear relationship between the AIMR brightness temperature measurements and the dropsonde column water vapor values are for the nadir brightness temperatures of the 37 GHz and 90 GHz AIMR channels. The column water vapor, w, given with units of kg m−2, may be expressed in terms of the 90 GHz nadir brightness temperature, TB, via the expression, w = −209.4 + 1.025 TB, whereas in terms of the 37 GHz AIMR nadir brightness temperature, w = −277.8 + 1.835 TB. These expressions are valid when AIMR brightness temperature measurements about the NCAR C-130 are made at high altitudes (above 5 km) and when the atmospheric column is cloud-free below the aircraft. Other relationships between the polarized components of the AIMR brightness temperature measurements as well as various combinations (differences) of these polarized components with the column water vapor did not identify any parameterizations to the algorithm better than that seen with the mean of the raw brightness temperature components.
 Water vapor is a critical component of the Earth's atmosphere. Its presence and abundance in the atmosphere is a dominant source of the energy budget of the Earth in its entirety. Water vapor is one of the most important atmospheric greenhouse gases due to its ubiquity and its molecular absorption bands in the infrared [Marsden and Valero, 2004]. There is a strong radiative coupling between atmospheric water vapor and infrared emissions from the Earth's land and sea surface. Water vapor plays a decisive role in the transfer of radiation through the atmosphere and is important to the transport and release of latent heat [Stephens, 1990]. The ability to accurately quantify the amount of water vapor within the atmospheric column including its distribution over the oceans and relationship to the sea surface temperature will greatly enhance the current understanding of the role it plays in both regional and global climate and climate change.
 The water vapor greenhouse effect has been extensively studied for many years [see, e.g., Raval and Ramanathan , Stephens , Ramanathan and Collins , Lubin , and references therein]. The Earth's atmosphere absorbs energy both from external sources (the Sun) as well as internally (longwave emission from the land, oceans, and the atmosphere itself). It also re-radiates some of this energy back to space as well as down to the Earth's surface. The infrared energy that is trapped by the atmosphere is popularly known as the greenhouse effect. As defined by Ramanathan and Collins , the total greenhouse effect is given as G = E − F, where F is the radiation emitted to space, and E = σT4 is the energy emitted by the ocean surface at temperature, T, as given by the Stefan-Boltzmann Law with constant σ = 5.67 × 10−8 W m−2 K−4. For clear-sky conditions, the total greenhouse effect described above simply becomes the atmospheric greenhouse absorption, Ga, given by Ga = E − Fc. Here, Fc is just the radiation emitted to space in the clear-sky portions of the atmosphere. The difference between the total greenhouse effect and the atmospheric greenhouse portion, G − Ga, is also known as the cloud longwave forcing.
 Elements in the climate system can feed back with one another producing various effects. For instance, the water vapor feedback, introduced by Manabe and Wetherald , identifies a mechanism between the sea surface temperature and consequent emission with the atmospheric water vapor absorption and re-radiation of longwave energy. An increase in the surface and atmosphere temperatures will cause more H2O to evaporate into the atmosphere. As a result, because the water vapor in the atmosphere is a strong greenhouse absorber, more longwave radiation is captured resulting in a further increase in temperature, thus feeding back into the cycle again.
 In this study, we propose a method of quantifying the amount and horizontal distribution of water vapor via airborne brightness temperature measurements using a microwave radiometer in the Indian Ocean region. Since this radiometer does not have a channel located on a water vapor absorption line, standard retrieval methods used for SSM/I and/or TMI water vapor retrievals [e.g., Petty, 1994] cannot be applied. The method developed here is generic in its approach and applicable to other measurement campaigns with similar instrumentation, although new regression coefficients must be derived. The retrieval algorithm described below is powerful in that it extends the capabilities of airborne microwave radiometer measurements at window frequencies, providing an additional tool in greenhouse and various atmospheric studies. The resulting column water vapor values are represented at a finer spatial and temporal resolution in the local measurement region than is currently available via large-grid and temporally averaged satellite retrievals.
2. Instrumentation and Measurements
 The water vapor retrieval algorithms developed in this work, in part, utilize in situ atmospheric measurements made during the intensive field phase (IFP) of the Indian Ocean Experiment (INDOEX) campaign from February thru April 1999 [Ramanathan et al., 2001]. During this phase of INDOEX, a variety of measurement platforms (land-based, sea-based, and airborne) made observations acquiring aerosol, chemical, and radiation data over the tropical Indian Ocean region during the Northern Hemisphere dry monsoon season. Although massive in its scope, one of the main goals of INDOEX was to assess the role that anthropogenic aerosols play in altering the heating and radiative balance in the atmospheric, thus affecting tropical rainfall patterns and the hydrological cycle with implications to global climate [Ramanathan et al., 2001].
 One of the primary research platforms that operated during INDOEX was the National Center for Atmospheric Research (NCAR) Hercules C-130 aircraft. Flight and logistical operations for the NCAR C-130 were based out of Male (4.18°N, 73.52°E), Republic of Maldives. Throughout the INDOEX campaign, the NCAR C-130 successfully flew 18 research missions spanning both the local region as well as the more distant surroundings in the area. Research flights northward of Male into the Arabian Sea and in proximity to the subcontinent of India explored the composition of aerosols and other gaseous pollutants near their source region. In turn, flights southward of Male, away from the Indian subcontinent were aimed at understanding the same pollutant described above, but away from the source regions after varying amounts of aging took place. Finally, long transequatorial flights (ranging to about 10°S) to locales where the environment was essentially pristine and aerosol-free, took place to probe the aerosol characteristics and the effects on the environment as their magnitude gradually decreased. During INDOEX, the observations of the highly polluted atmosphere in conjunction with the neighboring pristine atmosphere provided a valuable data set for assessing the impact of aerosols in both the local region as well as globally. The data set's overall value in studying these atmospheric effects is enhanced through the compilation of a variety of measurements (like the water vapor retrievals discussed in this work) resulting from instrumentation aboard the NCAR C-130.
 The column water vapor retrievals developed in the work are limited specifically to clear-sky atmospheric conditions. Because of limitations on the operational altitude of the NCAR C-130, column water vapor values are only represented from the surface to 5 km. Using atmospheric profile measurements during a cloud-free period of the INDOEX IFP made with radiosondes originating from the Kaashidhoo Climate Observatory (a more detailed description of these measurements is given by Bush and Valero ), about 93% of the total column water vapor (surface to top-of-atmosphere) is contained below the operational altitude of the aircraft.
2.1. Airborne Imaging Microwave Radiometer (AIMR)
 The retrieval algorithm developed in this study is based on brightness temperature, TB, measurements made by the Airborne Imaging Microwave Radiometer (AIMR) that was mounted in the nadir position on the NCAR C-130 and flown during all of the 18 INDOEX research flights during the IFP. AIMR is a superheterodyne, double sideband, total power radiometer. A rotating mirror scans over a region ±60° from nadir perpendicular to the aircraft flight track. AIMR operates at 37 and 90 GHz, and collects radiance at two orthogonal polarizations from which horizontal and vertical polarizations are derived. Beause of the geometry introduced by the rotating mirror, the contributions from vertical and horizontal polarized radiance to the raw orthogonal components measured by AIMR vary with scan angle. Equations that relate raw measurements to vertical and horizontal polarizations are solved at each scan angle, except at nadir where the physical distinction between horizontal and vertical polarization disappears. The beam width is 2.5° at 37 GHz and 1.0° at 90 GHz channel. These beam widths yield footprints in the nadir direction of approximately 218 m (37 GHz) and 87 m (90 GHz) when the aircraft operates at 5000 m. An internal calibration system includes a heated target and an ambient temperature target designed to be nearly perfect absorbers at 37 and 90 GHz. Further details on the sensor and derivation of brightness temperatures are provided by Collins et al. .
 AIMR was originally developed to observe sea ice, however its data, in conjunction with ancillary measurements, has also been used to retrieve cloud water droplet sizes [Liu et al., 2003] and cloud liquid water paths [Liu et al., 2001], both using INDOEX measurements. A sample of the AIMR brightness temperature measurements is presented in Figure 1. This figure depicts multiple AIMR instrument scans during a 30 s interval (09:18:14 – 09:18:44 UT) of the NCAR C-130 research flight #1 on 16 February 1999. The scan angles represent the AIMR observation relative to nadir and perpendicular to the flight track. The spread in the data (thickness of each curve) indicates the variation in the brightness temperature along the flight track. The shape of the curve (minimum at nadir and maximum in the wings) is due to the angular dependence of the sea surface emissivity. As discussed above, the horizontally and vertically polarized components of the brightness temperatures are undefined near nadir and thus are not depicted in Figure 1.
 In this study, we utilize the brightness temperature measurements at both frequencies, 37 and 90 GHz, in the nadir look direction and at a 30° look direction relative to nadir. In all cases, the brightness temperatures for the stated viewing angle represent averages within a ±2.5° range centered at the given value. Table 1 summarizes the relevant brightness temperature measurements utilized in this study along with the estimated measurement uncertainties in each case [Collins et al., 1996; Haggerty et al., 2000]. Because AIMR and its calibration system were originally designed for observations of sea ice, which has a significantly higher emissivity at these frequencies than open ocean, the uncertainty estimates produced by Collins et al.  were re-examined [Haggerty et al., 2000]. Measurements of the clear sky, a radiometrically cold target, were used to assess validity of the AIMR calibration procedure and resulting brightness temperature estimates for targets with low emissivity such as the sea surface. The experiment demonstrated that the calibration curve remains linear for very cold brightness temperatures, indicating that measurement uncertainty in the range of sea surface brightness temperatures (150–200 K) is on the order of 1 K.
Table 1. AIMR Brightness Temperature Measurement Parameters Along With Associated Sensitivity Estimates
30° ± 2.5°
30° ± 2.5°
 A representative sample of the AIMR brightness temperature measurements is given in Figure 2 for INDOEX research flight #18 on 25 March 1999. As in Figure 1, the nadir brightness temperatures are presented as well as the horizontally and vertically polarized components of the brightness temperatures. Since the polarized components are undefined at nadir, these brightness temperatures correspond to measurements at 30° scan angles. The top panel corresponds to the AIMR 90 GHz channel and the bottom panel corresponds to the 37 GHz channel. The data shown here are for all measurement conditions during this research flight. Later analyses (discussed later) will apply cloud-screening algorithms and restrict the viewing to high altitude flight segments in the process of determining clear-sky water vapor retrieval algorithms from this data set.
2.2. NCAR GPS Dropwindsondes
 During the INDOEX campaign over 90 dropsondes [Hock and Franklin, 1999] were released from the NCAR C-130 over the extent of the project. The NCAR GPS dropwindsondes nominally measured atmospheric profiles from the altitude of the C-130 at the release point down to the surface. The quantities measured by the dropsondes and utilized in this study include: pressure, temperature, relative humidity, latitude, longitude, and wind speed. Using the atmospheric profile measurements from each dropsonde, the relative humidity (%) was converted to water vapor density (kg m−3) and then integrated over the entire atmospheric column to give the integrated water vapor amount (kg m−2) over the atmospheric column spanning the region from the surface to the release point altitude.
 The dropsonde measurements have a two-fold purpose in this study. First, the direct integrated atmospheric column water vapor amounts obtained from the relative humidity profiles are used as the basis for the water vapor retrieval algorithms. A direct correlation is shown between these values and the AIMR brightness temperature measurements. Second, the atmospheric profiles obtained from the dropsondes are used as inputs to radiative transfer models that are used to simulate and verify the hypothesized functional form of the retrieval algorithm.
2.3. Ancillary Measurements
 Sea surface temperatures (SST) used in this study are obtained using the TMI (TRMM Microwave Imager) instruments on board the Tropical Rainfall Measuring Mission (TRMM) satellite. The TRMM satellite is the first mission dedicated to measuring tropical and subtropical rainfall through microwave, visible, and infrared sensors. TMI is a nine-channel, five-frequency microwave radiometer and is one of the three rain instruments carried aboard the TRMM satellite [Kummerow et al., 1998]. SST measurements were obtained from the TMI instrument during the INDOEX period via the TRMM data archives. The data in these archival files are provided for 3-day means over a spatial resolution of a 0.25° × 0.25° latitude/longitude grid. TMI SST data is extracted from these data archives and matched to the temporal, latitude, and longitude values of the NCAR C-130 research flights.
2.4. Analysis Conditions
 The column water vapor retrievals developed in the work are limited specifically to high-altitude, clear-sky atmospheric conditions. Only flight segments where the C-130 altitude was about 5 km are considered. Cloud-screening is accomplished using radiance measurements from both the AIMR instrument as well as the Narrow Field of View (NFOV) radiometer as part of the Radiation Measurement System (RAMS) [Valero et al., 1997; Bush and Valero, 2002] aboard the NCAR C-130. The NFOV consisted of two separate cryogenic radiometers sensitive to upwelling nadir-viewing radiances in the 4–50 μm and 10–11 μm infrared bands. A more detailed description of the NFOV is also given by Marsden and Valero . In the scenes studied, clouds are identified and eliminated from the analysis by examining both the magnitude and the variance (with respect to time) of the radiance measurements. Nominal or baseline values for the NFOV radiances were determined using the SBDART model [Ricchiazzi et al., 1998]. Dropsonde measurements and TMI SST values, in part, were used as input parameters to this atmospheric radiative transfer model. NFOV radiance measurements indicating values colder than the nominal values (determined using a threshold of 10% relative to the SBDART calculations) in the INDOEX region are identified as being cloudy. This condition identified and eliminated most of the predominantly cloudy scenes from the analysis. As a further cloud-screening criteria, any variances of the AIMR brightness temperatures or NFOV radiances greater than approximately 3% relative to the running mean value at a given instant are identified as being characteristic of having broken or scattered cloud conditions and eliminated from the data subset used in the analysis.
3. Model Simulations
 As a proof-of-concept of the proposed water vapor correlation algorithm, the 37 and 90 GHz AIMR brightness temperatures were simulated using a numerical model [Liu, 1998]. The plane-parallel microwave radiative transfer (MWRT) model calculates absorption coefficients for atmospheric gases (water vapor and oxygen) according to that discussed in Rosenkranz . Required model inputs for calculating upwelling radiation in clear sky conditions over ocean are atmospheric profiles of temperature and humidity, sea surface temperature, and surface wind speed. Temperature and humidity profiles were obtained by dropsondes released from the NCAR C-130. Surface wind speeds were estimated from the lowest level dropsonde measurements, and sea surface temperatures came from the TMI retrievals aboard the TRMM satellite. A parameterization included in the MWRT model calculates sea surface emissivity (and the complementary reflectivity) based on surface temperature and wind speed as described by Liu and Curry .
Figures 3 and 4 depict the relationship between the atmospheric columnar water vapor and the modeled 37 and 90 GHz brightness temperatures for nadir observations including all the surface contributions (emission and reflection) to the total upwelling radiance - a true simulation of the observations. The dropsonde columnar water vapor values are calculated by integrating the water vapor density profiles determined from the dropsonde relative humidity measurements. For both the 37 and 90 GHz AIMR channels, there appears to be a strong linear correlation between the nadir brightness temperature and the column water vapor amount. The correlation coefficient is slightly better for the 90 GHz channel (0.9727) than that for the 37 GHz channel (0.9478). The correlations depicted in Figures 3 and 4 have the same linear relationships as that seen in other studies that also related measured brightness temperatures at various microwave frequencies to the integrated water vapor amount [Wang et al., 1989; Liljegren, 2000]. For conditions below saturation, the radiometric brightness temperature is directly proportional to the amount of emitting molecules (water vapor in our case) thus, the relationship between the two quantities is linear.
 The contribution of atmospheric water vapor to the upwelling radiation at the top of the column is significant and variations in water vapor are detectable due to the relatively constant small contribution from the sea surface at these frequencies. Typical values for sea surface emissivity at 37 and 90 GHz are approximately 0.45 and 0.60, respectively. Variations up to only about 10% are observed with changes in surface roughness and temperature [Janssen, 1993]. A sensitivity analysis demonstrates that changes in the upwelling radiation resulting from the variations in the surface emission and reflectance variations are relatively small in comparison to those resulting from changes in the column water vapor content. Furthermore, because the surface emissivity and reflectivity are relatively constant along the flight tracks used in this study the variations in the upwelling radiation are predominantly due to changes in the column water vapor amount along the flight track. Sensitivity studies also indicate, as described by Wentz , that the AIMR brightness temperatures are relatively insensitive to variations in the SST. For instance, model simulations show that the 37 GHz brightness temperatures only vary by about 0.13 K per degree K change in the SST, whereas the 90 GHz brightness temperatures vary even less by about 0.09 K per degree K change in the SST. Similar studies analyzing the relationship between the AIMR brightness temperatures and the surface wind speed show a slightly larger variation: the 37 GHz brightness temperatures vary by about 0.62 K per unit m s−1 change in the surface wind speed whereas the 90 GHz brightness temperatures vary by about 0.47 K per unit m s−1 change in the surface wind speed. These variations are due to increases in the emissivity of the surface as the wind speed increases. Considering the range of values of SST, wind speed, and columnar water vapor observed during INDOEX, it is evident that most of the variation seen in TB measurements can be attributed to columnar water vapor variations. Table 2 shows ΔTB for SST ranging from 300–304 K, wind speed ranging from 0–10 m s−1, and columnar water vapor ranging from 23–62 kg m−2. Other inputs were held constant while varying each of the three parameters. These results also show that the 37 GHz channel is more sensitive to variations in the SST and surface wind speed than the 90 GHz channel since higher absorption by atmospheric water vapor at 90 GHz partially obscures surface effects at this frequency. This is most likely the explanation as to why the correlation between the columnar water vapor is better for the 90 GHz AIMR channel than the 37 GHz channel.
Table 2. Variation in TB in Both the 37 Band 90 GHz AIMR Frequencies Corresponding to Range of Model Input Values Observed in INDOEX
Sea Surface Temperature, 300–304 K
Wind Speed, 0–10 m s−1
Columnar Water Vapor, 23–62 kg m−2
4. Water Vapor Retrieval Algorithm
 Following the proof-of-concept model simulation described above, the next step in the development of the water vapor retrieval algorithm is to relate the AIMR measurements to the dropsonde columnar water vapor values. The AIMR measurements along the NCAR C-130 flight track were used in this analysis. As described in Section 2.1, only scan angles in the nadir direction (±2.5°) were extracted and only cases identified as being clear (as established by the cloud-screening algorithm) were considered. Furthermore in order to limit the brightness temperature observations to the same region in which the dropsonde made its descent, AIMR measurements were limited to the 5-minute window centered at the instant of the dropsonde release event. In most cases, each dropsonde was released when the C-130 altitude was greater than approximately 5 km. When the C-130 made a climb immediately prior to the dropsonde launch or made a descent immediately after release that coincided with the 5-minute observation window, this case was excluded in the analysis.
 The columnar water vapor correlation with respect to the measured AIMR brightness temperatures is presented in Figures 5 and 6. Each rectangular box in this plot represents a single dropsonde release with the extents of the box representing the uncertainty (variance) in the columnar water vapor measurements from the dropsonde as well as in the brightness temperature during the 5-minute averaging window. The dashed line indicates the least squares fit line through the data points with the fit parameters being given in the legend. As in the model-simulated correlation (as describe earlier and summarized in Figures 3 and 4), the correlation of the water vapor amount with the 90 GHz channel is better than that for the 37 GHz channel. Although the magnitudes of the model-simulated linear-fit parameters (A and B) are slightly less than that of the experimentally determined values, they are consistent with one another given the uncertainties determined via the regression analyses.
 One possible reason for the discrepancy in the two fits is that the model-calculated brightness temperatures were consistently less than the measured values for the 37 GHz channel, whereas for the 90 GHz channel, the model simulations tended to be slightly greater than the measured values for brightness temperatures above about 245 K and slightly less than the measured values below this temperature. Another reason for the discrepancy in the intercepts between Figures 3 and 5 or Figures 4 and 6 could be explained by a potential dry bias in the dropsonde humidity measurements. Dropsonde data quality during INDOEX was apparently improved over previous field experiments. It was reported that, due to improvements in the outer tube and in the type of desiccant used, the dry bias observed in previous experiments was reduced [Wang, 2005]. Comparisons using CAMEX-4 dropsondes, which were of comparable quality to the INDOEX dropsondes, showed a dry bias of about 5%. This underestimate in the humidity measurements would change the water vapor amounts depicted in Figures 5 and 6, and correspondingly the linear-fit parameters would differ.
 The linear-fit regressions relating the column water vapor amounts to the 37 and 90 GHz brightness temperatures were validated using the following method. The A and B regression coefficients were recalculated using all but one of the data points depicted in Figures 5 and 6 as the data sample. For the point that was eliminated, the column water vapor amount was retrieved using the linear-fit expression and the corresponding AIMR brightness temperature. This retrieved column water vapor value was then compared to that of the dropsonde measurement. This process was completed for each point in the overall data sample. In each case, the best fit line fell well within the limits of the regressions shown in Figures 5 and 6 as determined by the uncertainties in the A and B parameters. By removing a single data point in the regression analysis and then comparing this point to the retrieval results, a circular comparison during the validation process is avoided.
 The differences in the retrieved and measured column water vapor amounts for each of the data points calculated using the validation method above reflect the overall accuracy of the algorithm. For both the 37 and 90 GHz channels, the mean difference was less than 0.03 kg m−2, which is quite small and is not surprising because one would expect a roughly equal number of the retrieved points to be both greater than and less than the measured values. The standard deviation of the difference was roughly 2.5 kg m−2 (about 6.5% of the column water vapor amount) and 2.3 kg m−2 (about 6.1%) for the 37 and 90 GHz channels, respectively. These standard deviation values are indicative of the accuracy of the overall regression technique used here for determining the column water vapor amounts.
 The discrepancy between the model simulations in Figures 3 and 4 and the empirical calculations in Figure 5 and 6 are irrelevant in the final water vapor retrieval. These differences may be a result of possible calibration errors in the channels of the AIMR instrument or improper modeling due to bandpass simulation discrepancies, etc. The model-tested correlation is most important for verifying the linear relationship between the columnar water vapor and the brightness temperature values. The absolute magnitude is not critical in this procedure. The empirically determined correlations presented in Figures 5 and 6, determined entirely from in situ observations and bounded by the dropsonde and AIMR measurements, are used in the final water vapor retrievals using this instrument.
 This study was not limited to relating the columnar atmospheric water vapor amounts from the dropsonde measurements solely to the mean 37 and 90 GHz AIMR nadir brightness temperature measurements. An analysis, similar to that presented above for these quantities was completed for a variety of combinations of the horizontal and vertical polarization components of the two AIMR channels. A brief summary of these analyses is given in Table 3. The parameter being tested is given in the left column and the linear correlation coefficient of the fit to the dropsonde water vapor amount is given in the right column. The top two entries in this table correspond to the case presented in Figures 5 and 6, respectively.
Table 3. Parameters Used to Investigate Columnar Water Vapor and Brightness Temperature Correlations
 Clearly, the mean nadir brightness temperatures from both the 37 and 90 GHz AIMR channels give the highest correlation fit to the water vapor amounts. The individual horizontally or vertically polarized components (or the polarization differences) of either of these channels do not provide more highly correlated fits to the water vapor data. Of the parameters tested and shown in Table 3, the combination of the brightness temperatures that comes closest to the correlations shown in Figures 5 and 6 is the brightness temperature difference between the mean nadir 90 GHz and 37 GHz channels.
5. Water Vapor Retrievals
 As determined above, during the entire INDOEX campaign, the water vapor, w, may be expressed in terms of the AIMR brightness temperature measurements at 37 GHz, TB (37 GHz), or 90 GHz, TB (90 GHz), via the following expressions:
Either Equations (1) or (2) are equivalent for extracting the columnar water vapor value, however, the higher correlation and smaller standard deviation of the 90 GHz channel indicate that a better retrieval will occur using this channel. It should also be emphasized that these expressions are only valid under conditions in which the atmospheric column is free of clouds from the surface to the observation altitude. A further requirement that the aircraft be at high altitudes (greater than about 5 km) must also be met.
 A sample application of the water vapor retrieval algorithm using both Equations (1) and (2) for a 20-min level flight segment of the NCAR C-130 during INDOEX research flight #18 on 25 March 1999 is presented in Figure 7. For this segment, the NCAR C-130 altitude was approximately 5.6 km and the cloud-screening algorithms indicate that the atmospheric column below the aircraft was cloud-free. The measured AIMR brightness temperatures at 37 GHz and 90 GHz are shown in the top panel. The water vapor retrievals using these brightness temperature measurements are shown in the bottom panel for the 37 GHz channel (determined using Equation (1)) and for the 90 GHz channel (determined using Equation (2)). The thickness of each curve represents the uncertainty in the water vapor retrieval as determined by the fit parameters shown in Figures 5 and 6. In this case, the water vapor retrievals from both AIMR channels are nearly identical. As indicated earlier, the uncertainty in the water vapor amount calculated using the 90 GHz algorithm is slightly less than that from the 37 GHz algorithm. This is indicative of the fact the water vapor absorption coefficient is higher at 90 GHz than at 37 GHz, thus the 90 GHz AIMR channel effectively “sees” less of the signal variation from the surface and more from within the atmosphere. The column water vapor values determined using either AIMR channel are comparable to one another within error estimates and indicate that retrievals using either expression are interchangeable. This redundancy is a benefit in that water vapor retrievals are still possible should the case arise that either of the AIMR channels is inoperable at a given time.
 The dotted vertical lines in the bottom panel of Figure 7 indicate the release times of two NCAR GPS dropsondes that were launched during this segment. Each of the black circles represents the column-integrated water vapor associated with the atmospheric profile measurements from these dropsondes. These dropsondes represent two of the points depicted in Figures 5 and 6 that were used in the correlation analysis for determining the parameters in the retrieval algorithm. The differences between the measured dropsonde column water vapor amounts and those determined using the correlation analysis given in Equations (1) and (2) is indicative of the uncertainty associated with this regression. The actual uncertainty of the column water vapor parameterization is determined by the uncertainty in the “A” and “B” linear fitting parameters given in Figures 5 and 6.
 The AIMR brightness temperature measurements and column water vapor derived values for a flight segment on 16 March 1999 (research flight #13) during INDOEX is shown in Figure 8. This flight segment was located in the Arabian Sea with the latitude of the observations ranging from about 14.7 N at 0806 UT and going to about 12.7 N at 0834 UT. The two dropsondes that were released during this segment (also shown in this figure) represent some of the smallest clear-sky, column integrated water vapor amounts (24.0 and 24.6 kg m−2) of all of the dropsondes released during the INDOEX period. In contrast, the AIMR brightness temperatures and associated column water vapor amounts (dropsonde values of 50.4 and 51.5 kg m−2) depicted in Figure 9 correspond to some of the largest clear-sky sounding measurements during INDOEX. These measurements occurred on the 24 February 1999 INDOEX research flight #4 that was located just south of the equator (6.0 S at 0813 UT and 5.0 S at 0828 UT). Based upon the available dropsonde data that is used in the dropsonde retrieval algorithm parameterizations (Equations (1) and (2)), the conditions depicted in Figures 8 and 9 correspond to the limits of the water vapor retrieval algorithm discussed in this work. However, the model simulations presented in Section 3 show a very linear relationship with a correlation coefficient close to unity between the brightness temperature measurements and the column water vapor amounts. This indicates that extrapolation of the retrieval algorithm outside of the conditions presented here is reasonable. Since the relationships presented in Figures 3–6 all rely upon the actual INDOEX dropsonde measurements, to validate the retrieval algorithm outside the parameters described therein would require further model simulations with simulated atmospheric profiles to extend the limits of the column water vapor amounts.
 As a final example of the application of the clear-sky column water vapor retrieval during INDOEX, in Figure 10, we show both the AIMR 37 and 90 GHz brightness temperatures and the derived water vapor amounts for a 3-h measurement period on INDOEX research flight #8 on 4 March 1999. During this period, the NCAR C-130 flew a high-altitude (greater than about 5 km) transect starting south of the equator at about 9 S and continued north of the equator to about 3 N. Roughly every 1 degree progression in latitude, a dropsonde was released. As in Figures 7–9, the release points are marked on the figure. Nearly all of these dropsonde releases (11 out of 13) coincided with clear-sky, high altitude, and wings-level conditions – fitting of application of the retrieval algorithm. Note that the gaps in the measurements shown in Figure 10 result in either cloud screening or missing AIMR brightness temperature measurements due to non-level flight or other conditions.
 The dropsonde column water vapor measurements as well as the values retrieved using both the 37 GHz and 90 GHz AIMR brightness temperature measurements are summarized in Table 4. The times and the corresponding latitudes at the release points are also given here. Looking at both Figure 10 and Table 4, it is clear that the column water vapor values derived using the 90 GHz AIMR channel are closer to the dropsonde measured values than that determined using the 37 GHz AIMR channel. This is consistent with the correlation coefficients determined during the best fit linear analysis shown in Figures 3–6. For the data presented on the 4 March 1999 flight in Table 4, the average deviation between the column water vapor measurements and the retrieved values is 1.3 kg m−2 and 1.9 kg m−2 for the 90 GHz and 37 GHz AIMR channels, respectively.
Table 4. Clear-Sky Column Water Vapor Amounts and Retrieved Values During INDOEX Research Flight #8 on 4 March 1999
w (dropsonde) (kg m−2)
w (37 GHz) (kg m−2)
w (90 GHz) (kg m−2)
51.20 ± 2.93
48.26 ± 1.14
43.22 ± 2.62
41.56 ± 1.19
37.24 ± 1.77
37.73 ± 1.17
38.12 ± 1.15
38.32 ± 1.10
44.84 ± 1.82
45.47 ± 1.18
44.55 ± 1.90
44.80 ± 1.18
42.64 ± 1.11
43.90 ± 1.11
42.77 ± 1.29
43.67 ± 1.14
44.82 ± 1.46
46.57 ± 1.18
41.46 ± 2.04
42.96 ± 1.23
45.26 ± 1.62
45.99 ± 1.30
 The brightness temperature measurements from the AIMR microwave radiometer 37 GHz and 90 GHz channels are used to determine the column water vapor amount under clear-sky conditions during the INDOEX campaign in 1999. The retrieval algorithm is based on the integrated relative humidity profiles measured by numerous dropsondes released from the NCAR C-130 aircraft that operated during the 2 month IFP of INDOEX. Because the retrieval algorithm is based upon the dropsonde relative humidity measurements, its accuracy is highly dependent on the absolute accuracy of these in situ measurements. The relationship between the mean AIMR nadir brightness temperatures and the column water vapor values are shown to have a high correlation with respect to a linear parameterization. The highest correlation seen to exist is between the mean 90 GHz AIMR nadir brightness temperatures and the integrated column water vapor quantity. The next best correlation is between the 37 GHz AIMR brightness temperatures and the water vapor. The drop-off in the correlation from the 90 GHz TB and the 37 GHz TB is due to the greater sensitivity of the 37 GHz channel to variations in the SST and surface wind speeds. Relationships between the polarized components as well as various combinations (differences) of these polarized components with the column water vapor did not identify any parameterizations to the algorithm better than that expressed with the mean nadir TB given by Equations (1) and (2).
Figures 7 through 10 demonstrate the application of the retrieval algorithm developed in this work to the AIMR measurements made during INDOEX. For the clear-sky conditions shown in these figures, the column water vapor amounts are determined over the extents of the AIMR measurements. Effectively, this algorithm gives the integrated water vapor amounts “continuously” over the track of the aircraft rather than just at discrete times corresponding to the dropsonde measurements. Figures 7–10 also indicate the consistency between the retrieved water vapor amounts with the dropsonde measurements. Because the available satellite-derived column water vapor amounts did not coincide well in time and location during these INDOEX measurements, it was not possible to make a direct comparison with the retrieved values.
 This water vapor retrieval algorithm is powerful in that it greatly enhances the capabilities of the AIMR instrument package. It is general in its nature and can be applied to different seasons or regions. With the availability of calibration points via dropsonde measurements or even accurate model simulations to determine the fit parameters in the retrieval algorithm expression, column water vapor values may be determined and related to other in situ measurements made aboard the platform – the NCAR C-130 in the case here for INDOEX. The enhanced spatial and temporal coverage of the column water vapor values for the local measurement region, beyond that available via large-grid and temporally averaged satellite retrievals will provide a useful tool in greenhouse and various atmospheric studies.
 The analysis completed in this study was supported by National Science Foundation grant NSF-ATM-0422567. The acquisition of AIMR data was supported by special funds provided to the National Center for Atmospheric Research by the National Science Foundation. We thank the Joint Office for Scientific Support at the University Corporation for Atmospheric Research for their coordination efforts during the INDOEX project as well as the National Center for Atmospheric Research and the Research Aviation Facility for the opportunity to utilize instrumentation aboard and data acquired from the C-130 aircraft. We also indicate great appreciation to the government of the Republic of Maldives for hosting efforts during the field phases of the INDOEX project. We also thank Anthony Bucholtz, Shelly Pope, and Sabrina Leitner for assistance in pre-mission and post-mission calibrations of the NFOV instrument as well as data acquisition support in the field.