Repository logo

Improving the quality of extreme precipitation estimates using satellite passive microwave rainfall retrievals


Satellite rainfall estimates are invaluable in assessing global precipitation. As a part of the Global Precipitation Measurement (GPM) mission, a constellation of orbiting sensors, dominated by passive microwave imagers, provides a full coverage of the planet approximately every 2-3 hours. Several decades of development have resulted in passive microwave rainfall retrievals that are indispensable in addressing global precipitation climatology. However, this prominent achievement is often overshadowed by the retrieval's performance at finer spatial and temporal scales, where large variability in cloud morphology poses an obstacle for accurate rainfall measurements. This is especially true over land, where rainfall estimates are based on an observed mean relationship between high frequency (e.g., 89 GHz) brightness temperature (Tb) depression (i.e., the ice-scattering signature) and rainfall rate. In the first part of this study, an extreme precipitation event that caused historical flooding over south-east Europe is analyzed using the GPM constellation. Performance of the rainfall retrieval is evaluated against ground radar and gage reference. It is concluded that satellite observations fully address the temporal evolution of the event but greatly underestimate total rainfall accumulation (by factor of 2.5). A primary limitation of the rainfall algorithm is found to be its inability to recognize variability in precipitating system structure. This variability is closely related to the structure of the precipitation regime and the large-scale environment. To address this influence of rainfall physics on the overall retrieval bias, the second part of this study utilizes TRMM radar (PR) and radiometer (TMI) observations to first confirm that the Tb-to-rain-rate relationship is governed by the amount of ice in the atmospheric column. Then, using the Amazon and Central African regions as testbeds, it demonstrates that the amount of ice aloft is strongly linked to a precipitation regime. A correlation found between the large-scale environment and precipitation regimes is then further examined. Variables such as Convective Available Potential Energy (CAPE), Cloud Condensation Nuclei (CCN), wind shear, and vertical humidity profiles are found to be capable of predicting a precipitation regime and explaining up to 40% of climatological biases. Dry over moist air conditions are favorable for developing intense, well organized systems such as MCSs in West Africa and the Sahel. These systems are characterized by strong Tb depressions and above average amounts of ice aloft. As a consequence, microwave retrieval algorithms misinterpret these non-typical systems assigning them unrealistically high rainfall rates. The opposite is true in the Amazon region, where observed raining systems exhibit relatively little ice while producing high rainfall rates. Based on these findings, in the last part of the study, the GPM operational retrieval (GPROF) for the GMI sensor is modified to offer additional information on atmospheric conditions to its Bayesian-based algorithm. When forming an estimate, the modified algorithm is allowed to use this ancillary information to filter out a priori states that do not match the general environmental condition relevant to the observation and thus reduce the difference between the assumed and observed variability in ice-to-rain ratio. The results are compared to the ground Multi-Radar Multi-Sensor (MRMS) network over the US at various spatial and temporal scales demonstrating outstanding potentials in improving the accuracy of rainfall estimates from satellite-borne passive microwave sensors over land.


Rights Access


regional bias
microwave radiometer
Bayesian retrieval
rainfall regimes


Associated Publications