An Assessment of the Differential Inversion Method for Remote Sounding of Temperatures

View More View Less
  • a Department of Meteorology/CARSS, University of Utah, Salt Lake City, Utah
  • | b Division of Atmospheric Studies, Geophysics Directorate, Phillips Laboratory, Hanscom Air Force Base, Massachusetts
© Get Permissions
Full access

Abstract

We have explored the applicability of the differential inversion (DI) method to temperature retrievals in both clear and cloudy atmospheres using red satellite data. The main theme of the DI is that the local Planck intensity can be exactly expressed by a linear combination of the derivatives of radiances in the logarithmic pressure coordinate. The inversion coefficients are obtained by fitting the weighting function to a generalized form. The higher-order derivatives of radiances are determined from polynomial fittings. The satellite dataset used in this work contains collocated brightness temperatures and radiosonde data that have been collected during the period of Baseline Upper Atmospheric Network (BUAN) experiments. These data include both cloudy and clear cases. A multispectral cloud-removal method using the principle of the N* method has been developed. This method uses radiances of High-Resolution Infrared Radiation Sounder channels 6, 7, and 8 to estimate clear radiances of these channels and the surface temperature simultaneously based on radiative transfer simulations. Subsequently, the quantity N* (the ratio of effective cloud cover over adjacent pixels) and the clear radiances of the rest of the channels are evaluated.

Retrieval results are presented in terms of rms temperature differences between retrieved and sounding profiles. Considering all clear and partly cloudy cases, the rms differences in temperature of approximately 2 K for retrievals using the DI are comparable to those using the minimum-variance scheme. The rms differences in temperature for retrievals using the multispectral cloud-removal scheme are slightly larger than those using the BUAN cloud-removal scheme by approximately 0.5 K. Finally, the rms temperature differences are much smaller than those for the first guess of the minimum-variance scheme. These results indicate fire that the DJ can achieve acceptable performance without first-guess or error covariance matrices; second, that the proposed multispectral cloud-removal method is also capable of generating reasonable cloud-removed clear radiances; and finally that the DI can be used as a tool to obtain first guesses in the current operational method and to perform large-volume temperature retrievals for climate studies.

Abstract

We have explored the applicability of the differential inversion (DI) method to temperature retrievals in both clear and cloudy atmospheres using red satellite data. The main theme of the DI is that the local Planck intensity can be exactly expressed by a linear combination of the derivatives of radiances in the logarithmic pressure coordinate. The inversion coefficients are obtained by fitting the weighting function to a generalized form. The higher-order derivatives of radiances are determined from polynomial fittings. The satellite dataset used in this work contains collocated brightness temperatures and radiosonde data that have been collected during the period of Baseline Upper Atmospheric Network (BUAN) experiments. These data include both cloudy and clear cases. A multispectral cloud-removal method using the principle of the N* method has been developed. This method uses radiances of High-Resolution Infrared Radiation Sounder channels 6, 7, and 8 to estimate clear radiances of these channels and the surface temperature simultaneously based on radiative transfer simulations. Subsequently, the quantity N* (the ratio of effective cloud cover over adjacent pixels) and the clear radiances of the rest of the channels are evaluated.

Retrieval results are presented in terms of rms temperature differences between retrieved and sounding profiles. Considering all clear and partly cloudy cases, the rms differences in temperature of approximately 2 K for retrievals using the DI are comparable to those using the minimum-variance scheme. The rms differences in temperature for retrievals using the multispectral cloud-removal scheme are slightly larger than those using the BUAN cloud-removal scheme by approximately 0.5 K. Finally, the rms temperature differences are much smaller than those for the first guess of the minimum-variance scheme. These results indicate fire that the DJ can achieve acceptable performance without first-guess or error covariance matrices; second, that the proposed multispectral cloud-removal method is also capable of generating reasonable cloud-removed clear radiances; and finally that the DI can be used as a tool to obtain first guesses in the current operational method and to perform large-volume temperature retrievals for climate studies.

Save