Development of 3-D Background Error Covariance Model for PATH

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner
Thursday, 6 February 2014: 4:30 PM
Room C111 (The Georgia World Congress Center )
Kun Zhang, University of Colorado, Boulder, CO; and A. Gasiewski

The Precipitation and All-weather Temperature and Humidity (PATH) mission identified by the NRC decadal survey in 2007 will use a microwave array spectrometer to provide multispectral microwave imagery with high spatial resolution (~15-30 km) and high temporal resolution (~15-30 minutes). Due to the unique combination of high cloud penetrability of the PATH microwave channels and high temporal resolution, PATH data will provide the potential for “hydrometric tracking” of Numerical Weather Prediction (NWP) models to large individual precipitation cells under conditions of rapidly evolving mesoscale convection. Under development to explore this potential is a real-time all-weather hydrometric tracking system based on iterative extended Kalman filtering and fast forward radiative transfer modeling. In order to assimilate data over conditions of widely varying opacity, the Kalman gain matrix needs to be a function of the state of the atmosphere. However, the state variable statistics embodied in the background error covariance matrix impact the Kalman gain very significantly, and incorrect statistics can result in instabilities that amplify noise in unobservable prognostic modes. Accordingly, realization of precipitation locking requires the development of a state-dependent local background error covariance model with practical computability and local inversion capability. Such a model is being developed, in part, using LEO passive microwave data from AMSU, ATMS and airborne radar, passive microwave data from GRIP. The resulting error covariance model will specifically provide both horizontal and vertical correlation products suitable for ~50 x 50 km local 3DVar assimilation of up to a nominal 25 channel PATH data cube. In addition, the study includes determination of the computation complexity of forward RT models, namely the Unified Microwave Radiative Transfer (UMRT) model with forward model Jacobian and fast Mie library. Estimates suggest that the forward RT and Kalman gain calculation fits within the PATH temporal update specification for ~1000 x 1000 km sized regions.