Application of multilayer feedforward neural networks to precipitation cell-top altitude estimation

The use of passive 118-GHz O₂ observations of rain cells for precipitation cell-top altitude estimation is demonstrated using a multilayer feedforward neural network retrieval system. Data was derived from a collection of 118-GHz rain cell observations along with estimates of the cell-top altitude obtained by optical stereoscopy. The observations were made 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), 1986. The neural network estimator applied to MTS spectral differences between clouds and nearby clear air yielded an RMS discrepancy of 1.77 km for a combined cumulus, mature and dissipating cell set and 1.50 km for the cumulus-only set. A slight improvement in RMS discrepancy to 1.48 km was achieved by including additional MTS information on the absolute atmospheric temperature profile. Comparison of these results with a nonlinear statistical estimator shows that superior results can be obtained with the neural network retrieval system. The neural network estimator was then used to create imagery of cell-top altitudes estimated from 118-GHz CAMEX spectral imagery gathered from September through October, 1993