Search Results
Enhancing PMW Satellite Precipitation Estimation: Detecting Convective ClassAbstract
A decades-long effort in observing precipitation from space has led to continuous improvements of satellite-derived passive microwave (PMW) large-scale precipitation products. However, due to a limited ability to relate observed radiometric signatures to precipitation type (convective and stratiform) and associated precipitation rate variability, PMW retrievals are prone to large systematic errors at instantaneous scales. The present study explores the use of deep learning approach in extracting the information content from PMW observation vectors to help identify precipitation types. A deep learning neural network model (DNN) is developed to retrieve the convective type in precipitating systems from PMW observations. A 12-month period of Global Precipitation Measurement mission Microwave Imager (GMI) observations is used as a dataset for model development and verification. The proposed DNN model is shown to accurately predict precipitation types for 85% of total precipitation volume. The model reduces precipitation rate bias associated with convective and stratiform precipitation in the GPM operational algorithm by a factor of 2 while preserving the correlation with reference precipitation rates, and is insensitive to surface type variability. Based on comparisons against currently used convective schemes, it is concluded that the neural network approach has the potential to address regime-specific PMW satellite precipitation biases affecting GPM operations.
Abstract
A decades-long effort in observing precipitation from space has led to continuous improvements of satellite-derived passive microwave (PMW) large-scale precipitation products. However, due to a limited ability to relate observed radiometric signatures to precipitation type (convective and stratiform) and associated precipitation rate variability, PMW retrievals are prone to large systematic errors at instantaneous scales. The present study explores the use of deep learning approach in extracting the information content from PMW observation vectors to help identify precipitation types. A deep learning neural network model (DNN) is developed to retrieve the convective type in precipitating systems from PMW observations. A 12-month period of Global Precipitation Measurement mission Microwave Imager (GMI) observations is used as a dataset for model development and verification. The proposed DNN model is shown to accurately predict precipitation types for 85% of total precipitation volume. The model reduces precipitation rate bias associated with convective and stratiform precipitation in the GPM operational algorithm by a factor of 2 while preserving the correlation with reference precipitation rates, and is insensitive to surface type variability. Based on comparisons against currently used convective schemes, it is concluded that the neural network approach has the potential to address regime-specific PMW satellite precipitation biases affecting GPM operations.
The Evolution of the Goddard Profiling Algorithm to a Fully Parametric SchemeAbstract
The Goddard profiling algorithm has evolved from a pseudoparametric algorithm used in the current TRMM operational product (GPROF 2010) to a fully parametric approach used operationally in the GPM era (GPROF 2014). The fully parametric approach uses a Bayesian inversion for all surface types. The algorithm thus abandons rainfall screening procedures and instead uses the full brightness temperature vector to obtain the most likely precipitation state. This paper offers a complete description of the GPROF 2010 and GPROF 2014 algorithms and assesses the sensitivity of the algorithm to assumptions related to channel uncertainty as well as ancillary data. Uncertainties in precipitation are generally less than 1%–2% for realistic assumptions in channel uncertainties. Consistency among different radiometers is extremely good over oceans. Consistency over land is also good if the diurnal cycle is accounted for by sampling GMI product only at the time of day that different sensors operate. While accounting for only a modest amount of the total precipitation, snow-covered surfaces exhibit differences of up to 25% between sensors traceable to the availability of high-frequency (166 and 183 GHz) channels. In general, comparisons against early versions of GPM’s Ku-band radar precipitation estimates are fairly consistent but absolute differences will be more carefully evaluated once GPROF 2014 is upgraded to use the full GPM-combined radar–radiometer product for its a priori database. The combined algorithm represents a physically constructed database that is consistent with both the GPM radars and the GMI observations, and thus it is the ideal basis for a Bayesian approach that can be extended to an arbitrary passive microwave sensor.
Abstract
The Goddard profiling algorithm has evolved from a pseudoparametric algorithm used in the current TRMM operational product (GPROF 2010) to a fully parametric approach used operationally in the GPM era (GPROF 2014). The fully parametric approach uses a Bayesian inversion for all surface types. The algorithm thus abandons rainfall screening procedures and instead uses the full brightness temperature vector to obtain the most likely precipitation state. This paper offers a complete description of the GPROF 2010 and GPROF 2014 algorithms and assesses the sensitivity of the algorithm to assumptions related to channel uncertainty as well as ancillary data. Uncertainties in precipitation are generally less than 1%–2% for realistic assumptions in channel uncertainties. Consistency among different radiometers is extremely good over oceans. Consistency over land is also good if the diurnal cycle is accounted for by sampling GMI product only at the time of day that different sensors operate. While accounting for only a modest amount of the total precipitation, snow-covered surfaces exhibit differences of up to 25% between sensors traceable to the availability of high-frequency (166 and 183 GHz) channels. In general, comparisons against early versions of GPM’s Ku-band radar precipitation estimates are fairly consistent but absolute differences will be more carefully evaluated once GPROF 2014 is upgraded to use the full GPM-combined radar–radiometer product for its a priori database. The combined algorithm represents a physically constructed database that is consistent with both the GPM radars and the GMI observations, and thus it is the ideal basis for a Bayesian approach that can be extended to an arbitrary passive microwave sensor.
