Criteria for Identifying Severe Convective Cells in the TRMM Domain

The TRMM satellite has now amassed over a decade of tropical and extra-tropical observations of precipitation and lightning, and a number of data-mining techniques have been used to assimilate this body of data into subsets for a variety of purposes. We have developed a four year convective cell scale database to examine the microphysical environment of lightning production. Using these existing tools, along with the body of knowledge derived from 13 years of TRMM studies, we can insightfully revisit longstanding scientific questions.

The authors will present a study analogous to Zipser, et al.'s (2006) precipitation features-centric work investigating where in the TRMM domain the most intense storms occur. Our cell scale work uses the same intensity proxies (maximum height of the 40dBZ contour, lowest 37- and 85-GHz Polarization Corrected Temperatures (PCT) per cell, Lightning Imaging Sensor-identified lightning flash count per cell) and the addition of a second radar proxy, maximum reflectivity profile for heights as a function of temperature in the mixed phase region. This modification, and the use of the cell scale database, makes tenable the assumption that the observed proxies correspond to a physical convective process associated with a single storm. Preliminary results diverge from the Zipser study primarily in that the radar profile proxy over oceanic and island-driven convective cells is more aligned with microwave proxies, and that PCTs and reflectivities corresponding to the most extreme 1% are vastly different on the cell scale than on the PF scale. For example, the coldest 1% of minimum PCTs in convective cells in this database are 223K or lower, versus 159.7K in the PF database, and indeed at each percent, the convective cells have warmer minimum PCTs than the PFs. This is likely an artifact of the analysis methods; the minimum PCT of a PF will represent only the cell with the lowest minimum PCT out of possibly many embedded convective cells. Also, 40dBZ echo top heights in the cell analysis were roughly 3 kilometers higher at the same percentage thresholds as compared to the PF results. Again, because the convective cell database is tightly constrained to include events that are most likely severe, our results on the convective scale would include far fewer cells that don't reach 40dBZ, as compared to the reflectivity distribution of the PF database, resulting in the top 1%, 0.1%, etc. of cells exhibiting higher 40dBZ echo top heights. These inconsistencies encourage cell scale investigations in order to relate severity parameters to the discrete dynamic processes that produce them.