Joint observations from the TRMM mission are used to examine the contribution which lightning observations can make towards estimation of vertical structure and rainfall from passive microwave data. Passive microwave (TMI) observations are binned into distinct vertical structure categories based on cluster analysis of radar (PR) vertical profiles and convective/stratiform (C/S) classifiers. TMI rain estimates are high-biased relative to radar for stratiform and mixed convective/stratiform vertical structures and low-biased relative to radar for convective structures. Significant ambiguity exists in the TMI brightness temperature space between midlevel stratiform and convective profile types which are primary contributors to tropical rainfall. The ambiguity is worst for convective profile types with radar echo tops within and just above the mixed-phase region.

The ability of TMI data (including all low and high frequency polarized brightness temperatures; 19V, 19H, 21V, 37V, 37H, 85V, 85H) to predict vertical structure is assessed. The incremental benefit of including TMI convective/stratiform (C/S) classifiers (both 85 GHz polarization and 19, 37 and 85 GHz texture-based) and LIS lightning observations is then considered. For both TMI and LIS convective/stratiform identification, it is shown that the highest skill occurs for deep convective profile types which are already well-separated in the brightness temperature space. Contribution to midlevel C/S profile separation is far more modest. However, due to the importance of these profile types to total rainfall, even a modest incremental benefit in vertical structure prediction is important. The ability of LIS observations to contribute to midlevel separation may be limited by its short viewing time; convective profiles with tops in the mixed phase region may often be "incipient" lightning producers with low flash rates (below the LIS dwell time-based flash rate detection threshold). Longer dwell times (e.g., from geostationary orbit) would improve the usefulness of lightning data in midlevel profile separation.

The use of all parameters (brightness temperatures, TMI C/S classifiers, lightning data, predicted vertical structure) to reduce TMI/PR rainfall estimate scatter is demonstrated. For both vertical structure and rainfall estimation, inclusion of C/S classifiers and lightning observations provide statistical skill improvements but do not completely alleviate critical midlevel profile C/S ambiguity and related rain estimate errors.