CMG Research - Storm-Scale Data Assimilation

April 19, 1996 | Boundary Anchoring | Flanking Lines | Microphysics | EnKF Data Assimilation | Severe Hail
R. B. Wilhelmson | B. Jewett | M. Gilmore | G. Romine | L. Cronce | L. Orf | A. Houston | L. Wicker
UIUC Weather | Weather Links
Journal | Informal | Dissertation and Theses
Supertwister | Images and Animations | HVR Demo

Convective Modeling Group

Storm-Scale Data Assimilation of Polarimetric Radar Observations

Click for Source

Much progress has been made recently in the relatively new field of storm-scale data assimilation, particularly as a tool for improved analyses relative to more traditional approaches (e.g. thermodynamic retrieval, dual Doppler). The ensemble Kalman filter (EnKF) approach is particularly well suited to time series of remotely sensed data when a forward operator can be built to convert model fields to observation space. Further, the weak constraint system does not require building an adjoint to the model - such as 4D Variational Assimilation. Given an observation, such as radial wind, the EnKF system will adjust model fields within a given radius of influence. An ensemble of perturbed simulations is used to replicate the uncertainty in the model system (typically 100 members), while Gaussian noise is added to the observation. Covariance matrices are generated relating model fields to the observation fields. Subsequently, model fields within the radius of influence for each member of the ensemble are adjusted toward the observation via the dynamic relations provided by the covariance matrices. This process is replicated for each observation, then model members are integrated forward in time to the next observation period and the process repeats. The mean of the ensemble following each update cycle represents the best model estimate of the observed system.

Lately, this EnKF methodology has been applied to real data cases. Dowell et al. (2003) generated analyses using single Doppler radar data equivalent to that via dual Doppler radar techniques. Yet, the current system is limited to by the use of simple warm rain microphysics in the model operator - which is known to generate bias errors (e.g. Gilmore et al. 2004). Real precipitation systems contain myriad types of water particles in various and mixed phases. Recent attempts to apply an EnKF system with a more advanced microphysical model have been particularly problematic owing to the system becoming ill-posed without the addition of new observations. Polarimetric radars, which generate orthogonally oriented radar pulses, and by processing both signals over the same volume, information about the type, mass and concentrations of particles can be inferred. Theoretically, the radar data moments containing this additional information can be ingested by the EnKF system and may allow for the use of more advanced microphysical models which include ice processes. A prototype system is currently in development.

Project Members

Glen Romine - Research Lead
Lou Wicker (NSSL)
David Dowell (NSSL/CIMMS)
Robert Wilhelmson - PI


NSF ATM-99-86672, NSF ATM-0449753

Additional Project Information

May 8, 2003 Polarimetric Radar Case Study

Site Map | Contact Us | © 2005, all rights reserved | Last updated 29 January, 2005