Retrieval of Water Vapor Profiles via Principal Components: Options and Their Implications

View More View Less
  • 1 Department of Atmospheric Science, Colorado State University, Fort Collins, CO 80523
© Get Permissions Rent on DeepDyve
Restricted access

Abstract

Principal components have been widely used in regression retrieval of atmospheric parameters, but when applied to water vapor concentrations their use entails special problems. We discuss two of these problem and present results of retrieval experiments designed to alleviate them. The experiments employed High-resolution Infrared Radiation Sounder satellite data in conjunction with radiosonde observations. We found that mixing ratio is a less appropriate parameter for principal component-based retrieval than is a mean-saturation adjusted mixing ratio. Also, retrieval accuracy was vapor by identifying the optimum numbers of eigenvectors to use when transforming the water vapor profiles and the satellite brightness temperature, respectively, into their principal components. In our studies three eigenvectors were optimal for representation of water vapor, implying that HIRS-2 data are capable of retrieving at least third-order vertical resolution in water vapor profiles. In addition, we compared principal component-based retrieval with standard multiple regression and found that a hybrid of the two methods gave the greatest retrieval accuracy.

Abstract

Principal components have been widely used in regression retrieval of atmospheric parameters, but when applied to water vapor concentrations their use entails special problems. We discuss two of these problem and present results of retrieval experiments designed to alleviate them. The experiments employed High-resolution Infrared Radiation Sounder satellite data in conjunction with radiosonde observations. We found that mixing ratio is a less appropriate parameter for principal component-based retrieval than is a mean-saturation adjusted mixing ratio. Also, retrieval accuracy was vapor by identifying the optimum numbers of eigenvectors to use when transforming the water vapor profiles and the satellite brightness temperature, respectively, into their principal components. In our studies three eigenvectors were optimal for representation of water vapor, implying that HIRS-2 data are capable of retrieving at least third-order vertical resolution in water vapor profiles. In addition, we compared principal component-based retrieval with standard multiple regression and found that a hybrid of the two methods gave the greatest retrieval accuracy.

Save