PSU WRF EnKF/4DVar hybrid regional data assimilation system: Technical notes

Ying, Y., X. Chen, Y. Zhang, M. Minamide, R. Nystrom, H. Chen, J. Poterjoy, C. Melhauser, Y. Weng, Z. Meng, A. Aksoy, F. Zhang

2018

Department of Meteorology and Atmospheric Science, and Center for Advanced Data Assimilation and Predictability Techniques, The Pennsylvania State University, University Park, Pennsylvania

The Pennsylvania State University (PSU) regional hybrid data assimilation system (hereafter referred to as the “PSU system”) started as a simple proof-of-concept ensemble Kalman filter (EnKF) code for radar data assimilation (Snyder and Zhang 2003). It was later adopted to work with the Weather Research and Forecasting (WRF) model to perform regional data assimilation (Zhang et al. 2006; Meng and Zhang 2007; 2008 a, b). And it was further developed into an operational hurricane data assimilation and forecast system that assimilates both conventional and aircraft reconnaissance data (Zhang et al. 2009; 2011; Weng and Zhang 2012). Over the course of development in the past decade, the PSU system becomes more versatile. Currently, the supported data assimilation methods include EnKF, 3DVar and 4DVar (using WRFDA package), as well as hybrid methods such as E4DVar and 4DEnVar. More types of observations are supported recently, including the airborne Doppler radar radial velocity, and the satellite brightness temperature.

The WRF model is a fully compressible, nonhydrostatic mesoscale model (Skamarock et al. 2005). The vertical coordinate follows the terrain using hydrostatic pressure, and the model uses an Arakawa C grid. Prognostic variables are the column mass of dry air, velocities, potential temperature, and geopotential with optional variables including turbulent kinetic energy and any number of scalars such as water vapor mixing ratio, rain/snow mixing ratio, and cloud water/ice mixing ratio. The model domain is typically configured to perform a regional simulation of the weather system of interest at a convection-permitting resolution. For example, the hurricane operational forecast setting uses three nested domain with horizontal grid spacing of 27, 9, and 3 km. It is possible to configure the model to simulate systems across a wide range of scales.

The PSU system supports a variety of ensemble and variational data assimilation methods. The EnKF was first proposed in the geophysical literature by Evensen (1994) as an approximation to the Kalman filter (1960), which provides the optimal state estimation for a linear system with Gaussian errors. The Kalman filter was derived under the Bayesian estimation framework, it combines information from model forecast and available observations and their respective uncertainties. The EnKF uses an ensemble of model forecasts to characterize a flow-dependent background error covariance, which helps better propagate observed information in space and time to those variables that are unobserved. In geoscience applications, the EnKF typically faces the challenge of limited ensemble size comparing to dimension of the state, which will cause sampling noises in the estimated error covariance and requires ad hoc localization techniques to reduce the dimensionality of the analysis. Inflation of the ensemble spread is often needed to account for unrepresented error sources in the system. A comprehensive review of the development of EnKF is provided by Houtekamer and Zhang (2015).

More recently, hybrid data assimilation methods are proposed to combine the merit from both ensemble and variational methods. Using WRFDA package as variational component, the PSU system now fully support the state-of-the-art hybrid methods, such as the E4DVar and the 4DEnVar. The PSU system uses bash control scripts to coordinate the workflow and I/O of data among modules. This document provides a detailed technical description of the PSU system.

Ying, Y., X. Chen, Y. Zhang, M. Minamide, R. Nystrom, H. Chen, J. Poterjoy, C. Melhauser, Y. Weng, Z. Meng, A. Aksoy, F. Zhang, 2018: PSU WRF EnKF/4DVar hybrid regional data assimilation system: Technical notes. Department of Meteorology and Atmospheric Science, and Center for Advanced Data Assimilation and Predictability Techniques, The Pennsylvania State University, University Park, Pennsylvania.