Search

We present a detailed analysis of the airborne passive microwave remote-sensing data that were collected at a broad range of microwave bands and at a high spatial resolution during the 2002 and 2003 National Aeronautics and Space Administration Cold Land Processes Experiment (CLPX). An accurate measurement of snowpack properties using passive microwave observations requires the detailed knowledge of the relationship between snowpack geophysical parameters and the upwelling polarimetric brightness signature. The principle microwave instrument used for the CLPX was the polarimetric scanning radiometer (PSR), which provided ~100-m resolution maps of the snow emissivity at all Advanced Microwave Scanning Radiometer for the Earth Observing System (AMSR-E) bands during several intensive observation periods over the Colorado Rocky Mountains. The observed conditions included drought, normal snowpack, and spring snowmelt. The PSR and related ground-based observations of snowpack properties made during the 2002 and 2003 CLPX campaigns provide a comprehensive high-resolution passive microwave data set. Results show that the high-resolution PSR data exhibit emissivity modes that are similar to those observed in the historical data sets, and that the empirical relationships between the emissivity and the snow water equivalent (SWE), after the effects of macrovegetation are removed, closely match those found in the past theoretical studies. The use of the 89-GHz channel in the empirical relationships provides improved accuracy under dry snow conditions and a small SWE; however, the variability of the SWE--emissivity relationships increases with an increasing SWE. A summary of the observed relationships between the emissivity spectra of snow and snowpack properties is presented. Comparison of the total water content from the AMSR-E and PSR observations shows that the satellite measurements underestimated the total volume of water storage from airborne observations on the average by a factor of five. Index Terms--Airborne polarimetric scanning radiometer (PSR), beam polarization correction, georegistration, microwave emissivity, passive microwave remote sensing, retrieval algorithm development, snowpack, snow water equivalent (SWE), topography, validation.

Ground-based observations at millimeter (mm) and submillimeter (submm) wavelengths were collected at the Atmospheric Radiation Measurement Program site at Barrow, AK, during the Arctic winter by a new 25-channel radiometer. A weighting function analysis is presented to demonstrate the enhanced sensitivity of mm- and submm-wave (50--400 GHz) radiometers to low vapor and liquid water contents with respect to conventional instruments such as the ones operating at centimeter (cm) wavelengths (20-30 GHz). In addition, based on measurements, we carried out a quantitative analysis of mm-and submm-wavelength sensitivity, yielding improvement factors from 1.5 to 69 for precipitable water vapor (PWV) and 3 to 4 for liquid water path (LWP) when compared to 20-30 GHz radiometers. Furthermore, using a simulated data set, we evaluate the effect of hydrometeor scattering: Given the conditions occurring during the experiment, the scattering contribution is within the instrumental noise for most, but not all, of the considered channels. With the same data set, we demonstrate that in the dry conditions of the Arctic, a simple linear regression yields satisfactory results when applied on selected mm- and submm-wave channels. For a dual-channel combination, the expected accuracy is ~0.23 (0.007) mm for PWV (LWP), when using mm- and submm-wavelengths, whereas it is 0.37 (0.012) mm using cm-wave channels. When the retrieval is applied to real observations, the accuracy is found in agreement with theoretical expectations. Index Terms--Atmospheric measurements, microwave radiometry, remote sensing.

Part of the Earth Observing Sysytem Aqua Advanced Microwave Scanning Radiometer (AMSR-E) Arctic sea ice validation campaign in March 2003 was dedicated to the validation of snow depth on sea ice and ice temperature products. The difficulty with validating these two variables is that neither can currently be measured other than in situ. For this reason, two aircraft flights on March 13 and 19, 2003, were dedicated to these products, and flight lines were coordinated with in situ measurements of snow and sea ice physical properties. One flight was in the vicinity of Barrow, AK, covering Elson Lagoon and the adjacent Chukchi and Beaufort Seas. The other flight was farther north in the Beaufort Sea (about 73[degrees]N, 147.5[degrees]W) and was coordinated with a Navy ice camp. The results confirm the AMSR-E snow depth algorithm and its coefficients for first-year ice when it is relatively smooth. For rough first-year ice and for multiyear ice, there is still a relationship between the spectral gradient ratio of 19 and 37 GHz, but a different set of algorithm coefficients is necessary. Comparisons using other AMSR-E channels did not provide a clear signature of sea ice characteristics and, hence, could not provide guidance for the choice of algorithm coefficients. The limited comparison of in situ snow-ice interface and surface temperatures with 6-GHz brightness temperatures, which are used for the retrieval of ice temperature, shows that the 6-GHz temperature is correlated with the snow-ice interface temperature to only a limited extent. For strong temperature gradients within the snow layer, it is clear that the 6-GHz temperature is a weighted average of the entire snow layer. Index Terms--Advanced Microwave Scanning Radiometer (AMSR), passive microwave, sea ice, snow on sea ice, validation.

This paper examines the sensitivity of Advanced Microwave Scanning Radiometer (AMSR-E) brightness temperatures (Tbs) to surface roughness by a using radiative transfer model to simulate AMSR-E Tbs as a function of incidence angle at which the surface is viewed. The simulated Tbs are then used to examine the influence that surface roughness has on two operational sea ice algorithms, namely: 1) the National Aeronautics and Space Administration Team (NT) algorithm and 2) the enhanced NT algorithm, as well as the impact of roughness on the AMSR-E snow depth algorithm. Surface snow and ice data collected during the AMSR-Ice03 field campaign held in March 2003 near Barrow, AK, were used to force the radiative transfer model, and resultant modeled Tbs are compared with airborne passive microwave observations from the Polarimetric Scanning Radiometer. Results indicate that passive microwave Tbs are very sensitive even to small variations in incidence angle, which can cause either an over-or underestimation of the true amount of sea ice in the pixel area viewed. For example, this paper showed that if the sea ice areas modeled in this paper were assumed to be completely smooth, sea ice concentrations were underestimated by nearly 14% using the NT sea ice algorithm and by 7% using the enhanced NT algorithm. A comparison of polarization ratios (PRs) at 10.7, 18.7, and 37 GHz indicates that each channel responds to different degrees of surface roughness and suggests that the PR at 10.7 GHz can be useful for identifying locations of heavily ridged or rubbled ice. Using the PR at 10.7 GHz to derive an 'effective' viewing angle, which is used as a proxy for surface roughness, resulted in more accurate retrievals of sea ice concentration for both algorithms. The AMSR-E snow depth algorithm was found to be extremely sensitive to instrument calibration and sensor viewing angle, and it is concluded that more work is needed to investigate the sensitivity of the gradient ratio at 37 and 18.7 GHz to these factors to improve snow depth retrievals from spaceborne passive microwave sensors. Index Terms--Passive microwave, remote sensing, sea ice.

Accurate knowledge of snow-depth distribution over sea ice is critical for polar climate studies. Current snow-depth-over-sea-ice retrieval algorithms do not sufficiently account for variations in snow and ice physical properties that can affect the accuracy of retrievals. For this reason, airborne microwave observations were coordinated with ground-based measurements of snow depth and snow properties in the vicinity of Barrow, AK, in March 2003. In this paper, the effects of snowpack properties and ice conditions on microwave signatures are examined using detailed surface-based measurements and airborne observations in conjunction with a thermal microwave-emission model. A comparison of the Microwave Emission Model of Layered Snowpacks (MEMLS) simulations with detailed snowpack and ice data from stakes along the Elson Lagoon and the Beaufort Sea and radiometer data taken from low-level flights using a Polarimetric Scanning Radiometer (PSR-A) shows that MEMLS can be used to simulate snow on sea ice and is a useful tool for understanding the limitations of the snow-depth algorithm. Analysis of radiance data taken over the Elson Lagoon and the Beaufort Sea using MEMLS suggests that the radiometric differences between the two locations are due to the differences in sea-ice emissivity. Furthermore, measured brightness temperatures suggest that the current snow-depth retrieval algorithm is sufficient for areas of smooth first-year sea ice, whereas new algorithm coefficients are needed for rough first-year sea ice. Snowpack grain size and density remain an unresolved issue for snow-depth retrievals using passive-microwave radiances. Index Terms--Advanced Microwave Scanning Radiometer (AMSR), microwave, modeling, polar regions, remote sensing, sea ice, snow.

The first estimates of the average bulk heat transfer coefficient for Arctic sea ice are presented as a function of mean ice thickness. Turbulent heat flux measurements made by the NASA P-3 over the St. Lawrence Island polynya (SLIP) and Kuskokwim Bay in the Bering Sea during AMSR-Ice03 were used to estimate the values of the heat transfer coefficient [C.sub.H]. Estimates of ice thickness were made from the algorithm of Perovich et al. using broadband albedos obtained from Moderate Resolution Imaging Spectroradiometer data. Plots of [C.sub.H] as a function of ice thickness showed a nearly linear relationship for ice thicknesses in the range of 0-14 cm in the polynyas. Previous estimates of [C.sub.H] for different cases over the SLIP were 1.2 x [10.sup.-3], but no estimates of ice thickness were available. These results will allow more accurate estimates of heat fluxes from the thin-ice areas of polynyas using satellite retrievals. Index Terms--Atmospheric measurements, covariance analysis, meteorology, remote sensing, sea ice.

An assessment of Advanced Microwave Scanning Radiometer Earth Observing System (AMSR-E) sea ice concentrations under winter conditions using ice concentrations derived from Landsat-7 Enhanced Thematic Mapper Plus (ETM+) imagery obtained during the March 2003 Arctic sea ice validation field campaign is presented. The National Oceanic and Atmospheric Administration Environmental Technology Laboratory's Airborne Polarimetric Scanning Radiometer Measurements, which were made from the National Aeronautics and Space Administration P 3B aircraft during the campaign, were used primarily as a diagnostic tool to understand the comparative results and to suggest improvements to the AMSR-E ice concentration algorithm. Based on the AMSR-E/ETM+ comparisons, a good overall agreement with little bias (~1%) for areas of first year and young sea ice was found. Areas of new ice production result in a negative bias of about 5% in the AMSR-E ice concentration retrievals, with a root mean square error of 8%. Some areas of deep snow also resulted in an underestimate of the ice concentration (~10%). For all ice types combined and for the full range of ice concentrations, the bias ranged from 0% to 3%, and the rms errors ranged from 1% to 7%, depending on the region. The new-ice and deep-snow biases are expected to be reduced through an adjustment of the new-ice and ice-type C algorithm tie points. Index Terms--Advanced Microwave Scanning Radiometer for the Earth Observing System (AMSR-E), Arctic, sea ice concentration, validation.

Aircraft-acquired passive microwave data, laser radar height observations, RADARSAT synthetic aperture radar imagery, and in situ measurements obtained during the AMSR-Ice03 experiment are used to investigate relationships between microwave emission and ice characteristics over several space scales. The data fusion allows delineation of the shore-fast ice and pack ice in the Barrow area, AK, into several ice classes. Results show good agreement between observed and Polarimetric Scanning Radiometer (PSR)-derived snow depths over relatively smooth ice, with larger differences over ridged and rubbled ice. The PSR results are consistent with the effects on snow depth of the spatial distribution and nature of ice roughness, ridging, and other factors such as ice age. Apparent relationships exist between ice roughness and the degree of depolarization of emission at 10, 19, and 37 GHz. This depolarization would yield overestimates of total ice concentration using polarization-based algorithms, with indications of this seen when the NT-2 algorithm is applied to the PSR data. Other characteristics of the microwave data, such as effects of grounding of sea ice and large contrast between sea ice and adjacent land, are also apparent in the PSR data. Overall, the results further demonstrate the importance of macroscale ice roughness conditions such as ridging and rubbling on snow depth and microwave emissivity. Index Terms--Passive microwave, roughness, sea ice, shore-fast ice, snow depth.

Corruption of C-band microwave brightness observations by radio-frequency interference (RFI) has been reported in recent data from orbiting radiometers; methods for mitigating these effects are of great importance for the design of future spaceborne microwave radiometers. One approach that has been suggested involves the use of multiple subchannels at C-band as opposed to a single channel; the use of multiple subchannels allows RFI to be detected and mitigated by analyzing relationships among subchannel brightnesses. While this approach has been utilized in previous airborne measurements, demonstrations of the RFI mitigation performance achieved have been difficult to obtain. To address this issue, an enhanced airborne system for observing radio-frequency interference effects on C-band microwave radiometers was developed, and is described in this paper. The system includes a traditional microwave radiometer with four C-band subchannels, so that RFI removal is possible using a subchannel mitigation algorithm. In addition, the system includes a digital receiver with the capability of providing high temporal and spectral resolution observations of interference. This high-resolution data allows improved understanding of RFI sources to be obtained, and also allows analysis of subchannel mitigation algorithm performance. Observations using the system in a test flight near Wallops Island, VA are described. Results show the four subchannel approach generally to be effective in mitigating the observed RFI sources, although examples are also illustrated using the digital receiver data to demonstrate failure of this approach. While studies of the digital receiver data alone could be performed to demonstrate further improvements in RFI mitigation, issues with this initial dataset limit the extent of such studies. Nevertheless, the results obtained still demonstrate qualitatively the improved RFI mitigation that can be achieved in brightness observations through the use of digital receivers. Index Terms--Microwave radiometry, radio-frequency interference (RFI).

Soil Moisture Experiment 2003 (SMEX03) was the second in a series of field campaigns using the National Oceanic and Atmospheric Administration Polarimetric Scanning Radiometer (PSR/CX) designed to validate brightness temperature ([T.sub.B]) data and soil moisture retrieval algorithms for the Advanced Microwave Scanning Radiometer (AMSR-E) for the Earth Observing System on the Aqua satellite. Objectives related to the PSR/CX during SMEX03 included: calibration and validation of AMSR-E [T.sub.B] observations over different climate/vegetation regions of the U.S. [Alabama (AL), Georgia (GA), Oklahoma (OK)], identification of possible areas of radio-frequency interference (RFI), comparison of X-band observations from Tropical Rainfall Measurement Mission Microwave Imager (TMI), AMSR-E, and PSR/CX, and exploring the potential of soil moisture retrieval algorithms using C- and X-band imagery in diverse landscapes. In the current investigation, more than 100 flightlines of PSR/CX data were extensively processed to produce gridded [T.sub.B] products for the four study regions. Due to the lack of significant rainfall in OK, generally dry soil moisture conditions were observed. Observations obtained over AL include a wide range of soil moisture and vegetation conditions. Results from the AL site clearly showed a lack of sensitivity to rainfall/soil moisture under forest canopy cover. Quantitative comparisons made with the TMI validated that both the PSR/CX and AMSR-E X-band channels were well calibrated. Spectral analyses indicated that the PSR/CX observations at C-band also are reasonable. As expected, there were varying degrees of RFI in the AMSR-E C-band data for the study sites that will prevent further soil moisture analysis using these data. X-band comparisons of the PSR/CX high-resolution and AMSR-E and TMI low-resolution data indicated a linear scaling for the range of conditions studied in SMEX03. These results will form the basis for further soil moisture investigations. Index Terms--Advanced Microwave Scanning Radiometer (AMSR-E), passive microwave, soil moisture, validation.

The local spatial scales of tropical precipitating systems were studied using Tropical Rainfall Measuring Mission (TRMM) Microwave Imager (TMI) rain rate imagery from the TRMM satellite. Rain rates were determined from TMI data using the Goddard Profiling (GPROF) Version 5 algorithm. Following the analysis of Ricciardulli and Sardeshmukh (RS), who studied local spatial scales of tropical deep convection using global cloud imagery (GCI) data, active precipitating months were defined alternatively as those having greater than either 0.1 mm/h or 1 mm/h of rain for more than 5% of the time. Spatial autocorrelation values of rain rate were subsequently computed on a 55 x 55 km grid for convectively active months from 1998 to 2002. The results were fitted to an exponential correlation model using a nonlinear least squares routine to estimate a spatial correlation length at each grid cell. The mean spatial scale over land was 90.5 km and over oceans was 122.3 km for a threshold of 0.1 mm/h of rain with slightly higher values for a threshold of 1 mm/h of rain. An error analysis was performed which showed that the error in these determinations was of order 2% to 10%. The results of this study should be useful in the design of convective schemes for general circulation models and for precipitation error covariance models for use in numerical weather prediction and associated data assimilation schemes. The results of the TMI study also largely concur with those of RS, although the more direct relationship between the TMI data and rain rate relative to the GCI imagery provide more accurate correlation length estimates. The results also confirm the strong impact of land in producing short spatial scale convective rain. Index Terms--Convection, correlation length, microwave, precipitation.

The full utilization of satellite-based passive microwave imagery for weather forecasting rests on the ability to assimilate radiances into numerical weather prediction (NWP) models for highly scattering and absorbing hydrometeor states. State vector updates need to be performed rapidly enough to maintain pace with the sensor data stream and require, in particular, rapid calculation of the tangent linear relationship (Jacobian) between the observed antenna temperatures and the NWP prognostic hydrometeor parameters. To facilitate the use of both spaceborne and airborne passive microwave data in numerical forecasting, we present a new rapid multiple-stream discrete-ordinate algorithm for calculating the Jacobian under arbitrary scattering and absorbing conditions. The algorithm is based on the layer-adding method for a plane-parallel atmosphere for which the number of operations required to compute the solution is proportional to the number of layers. A nontrivial aspect of the problem is the stable calculation of the reflectance and transmittance operators in highly scattering layers for which a diagonalization technique and analytical factorization of specific matrices are used to ensure stability. Scaling calculations suggest that the new algorithm will be suitable for use in real-time all-weather microwave radiance assimilation. Index Terms--Assimilation, Jacobian, microwave, radiance.

The automotive industry is currently considering the introduction of short-range radars (SRR) operating near 24 GHz for improving road traffic safety. SSRs are intended to observe the full azimuthal space cover around a vehicle using up to eight sensors. The sensors would operate in an ultrawideband (UWB) mode, occupying 3-5 GHz of bandwidth. Interference from SRR transmitters with passive microwave remote sensing satellites used for weather and climate monitoring could occur as the result of several coupling mechanisms, including direct coupling via the transmit antenna beam and scattering and diffraction of the transmitted signals from leading vehicles, buildings, and other nearby objects. In this study, we estimate the amount of coupling anticipated to occur from SRRs, including the direct and scattered contributions. The calculations are based on bistatic scattering measurements of a typical automobile and ray optical simulations of reflection and propagation in an urban environment. Using these calculations, the maximum allowable SRR transmitted power for interference levels acceptable for meteorological and climatological remote sensing applications are quantified. The study provides criteria for SRR operation with the Earth Exploration Satellite Service on a noninterference basis. Index Terms--Automotive radar, compatibility, Earth Exploration Satellite Service (EESS), interference, radiometry.

A two-scale ocean surface emissivity model tuned to WindSat and Special Sensor Microwave/Imager (SSM/I) polarimetric brightness temperatures for general passive microwave applications is detailed. The model provides a full Stokes vector emissivity calculation at arbitrary microwave frequencies and observation angles for wind speeds of up to 15 m/s. During model development, it was found that the untuned two-scale model generally produced plausible azimuthal behavior in the ocean surface emissivity vector; however, large discrepancies between the untuned model and WindSat and SSM/I observations were observed, in particular for the zeroth-azimuthal-harmonic coefficients. These discrepancies can be ascribed to inaccuracies in contemporary ocean foam coverage and emissivity models. Accordingly, foam influences were treated using machine-tunable correction parameters incorporated as a means of improving and extending the physically based two-scale model. In addition, a hydrodynamic modulation function and the lower cutoff wavenumber for small-scale perturbation integration were treated as empirically tunable. Model tuning was performed by minimizing the $\chi ^{2}$ metric over all available wind bins, channel frequencies, polarizations, and azimuthal harmonics. The result is an approximately eightfold reduction in $\chi ^{2}$ from its initial untuned model value, indicating that machine tuning can considerably reduce model errors inherent in the two-scale model to levels acceptable for oceanic passive microwave remote sensing applications. The tuned model is independently validated against NASA Global Precipitation Measurement Microwave Instrument (GMI) measurements. The tuned model and GMI observed emissivities, when scaled to the surface temperature, are in agreement to within ±0.3 K root-mean-square (rms) error, thus suggesting good applicability of the model over a wide range of microwave frequencies and wind speeds. [ABSTRACT FROM AUTHOR]

The Advanced Microwave Scanning Radiometer (AMSR) holds promise for retrieving soil moisture in regions with low levels of vegetation. Algorithms for this purpose have been proposed, but none have been rigorously evaluated due to a lack of datasets. Accordingly, the Southern Great Plains 1999 Experiment (SGP99) was designed to provide C-band datasets for AMSR algorithm development and validation. Ground observations of soil moisture and related variables were collected in conjunction with aircraft measurements using a C-band radiometer similar to the AMSR sensor (6.92 GHz), the Polarimetric Scanning Radiometer with its C-band scanhead (PSR/C). The study region has been the focus of several previous remote sensing field experiments and contains vegetation conditions compatible with the expected capabilities of C-band for soil moisture retrieval. Flights were conducted under a wide range of soil moisture conditions, thus providing a robust dataset for validation. A significant issue found in data processing was the removal of anthropogenic radio-frequency interference. Several approaches to estimating the parameters of a single-channel soil moisture retrieval algorithm were used. PSR/C soil moisture images show spatial and temporal patterns consistent with meteorological and soil conditions, and the dynamic range of the PSR/C observations indicates that the AMSR instrument can provide useful soil moisture information. Index Terms--Advanced microwave scanning radiometer (AMSR), passive microwave, soil moisture.

This paper presents a method for combining internal noise injection and external reference standard looks to accurately calibrate an airborne dual-channel radiometer. The method allows real-time estimation of the correct values of the radiometer gains and offsets, even for nontemperature-stabilized radiometers and with minimum loss of measurement time spent in external load measurement. Crosstalk and leakage introduced by the noise injection circuitry is also taken into account, thus providing high gain and offset estimation accuracy. The method was implemented on a National Oceanic and Atmospheric Administration airborne instrument, the Polarimetric Scanning Radiometer, which was used to obtain an extensive set of radiometric measurements over oceanic convection during CAMEX3 in August-September 1998. Index Terms--Calibration, radiometry.

The microphysical parameterization of clouds and rain cells plays a central role in atmospheric forward radiative transfer models used in calculating microwave brightness temperatures. The absorption and scattering properties of a hydrometeor-laden atmosphere are governed by particle phase, size distribution, aggregate density, shape, and dielectric constant. This study investigates the sensitivity of brightness temperatures with respect to the microphysical cloud parameterization. Calculated wideband (6-410 GHz) brightness temperatures were studied for four evolutionary stages of an oceanic convective storm using a five-phase hydrometeor model in a planar-stratified scattering-based radiative transfer model. Five other microphysical cloud parameterizations were compared to the baseline calculations to evaluate brightness temperature sensitivity to gross changes in the hydrometeor size distributions and the ice-air-water ratios in the frozen or partly frozen phase. The comparison shows that enlarging the raindrop size or adding water to the partly frozen hydrometeor mix warms brightness temperatures by as much as 55 K at 6 GHz. The cooling signature caused by ice scattering intensifies with increasing ice concentrations and at higher frequencies. An additional comparison to measured Convection and Moisture Experiment (CAMEX-3) brightness temperatures shows that in general all but two parameterizations produce calculated [T.sub.B]s that fall within the CAMEX-3 observed minima and maxima. The exceptions are for parameterizations that enhance the scattering characteristics of frozen hydrometeors. Index Terms--Clouds, electromagnetic scattering, millimeter wave radiometry, rain, remote sensing, snow.

The design, analysis, and demonstration of a digital-correlation microwave polarimeter for use in Earth remote sensing is presented. We begin with an analysis of a three-level digital correlator and develop the correlator transfer function and radiometric sensitivity. A fifth-order polynomial regression is derived for inverting the digital correlation coefficient into the analog statistic. In addition, the effects of quantizer threshold asymmetry and hysteresis are discussed. A two-look unpolarized calibration scheme is developed for identifying correlation offsets. The developed theory and calibration method are verified using a 10.7 GHz and a 37.0 GHz polarimeter. The polarimeters are based upon 1-GS/s three-level digital correlators and measure the first three Stokes parameters. Through experiment, the radiometric sensitivity is shown to approach the theoretical as derived earlier in the paper and the two-look unpolarized calibration method is successfully compared with results using a polarimetric scheme. Finally, sample data from an aircraft experiment demonstrates that the polarimeter is highly useful for ocean wind-vector measurement. Index Terms--Correlator, digital, polarimetry, radiometer, Stokes vector.

The use of passive 118-GHz [O.sub.2] observations of rain cells for precipitation cell-top altitude estimation is demonstrated by using a multilayer feedforward neural network retrieval system. Rain cell observations at 118 GHz were compared with estimates of the cell-top altitude obtained by optical stereoscopy. The observations were made with 2-4.km horizontal spatial resolution by using the millimeter-wave temperature sounder (MTS) scanning spectrometer aboard the NASA ER-2 research aircraft during the Genesis of Atlantic Lows Experiment (GALE) and the Cooperative Huntsville Meteorological Experiment (COHMEX) in 1986. The neural network estimator applied to MTS spectral differences between clouds, and nearby clear air yielded an rms discrepancy of 1.76 km for a combined cumulus, mature, and dissipating cell set and 1.44 km for the cumulus-only set. An improvement in rms discrepancy to 1.36 km was achieved by including additional MTS information on the absolute atmospheric temperature profile. An incremental method for training neural networks was developed that yielded robust results, despite the use of as few as 56 training spectra. Comparison of these results with a nonlinear statistical estimator shows that superior results can be obtained with a neural network retrieval system. Imagery of estimated cell-top altitudes was created from 118-GHz spectral imagery gathered from CAMEX, September through October 1993, and from cyclone Oliver, February 7, 1993. Index Terms - Microwave remote sensing, microwave spectra 118 GHz, neural network, precipitation estimation, rain cell-top altitude.