| You are in: Weather Forecasts >> Modeling System | Today's Forecast | DMSP Image of the Day |
|
Modeling System
Descriptions of the system hardware, and model and graphical software are provided below for our daily numerical model runs for the Los Angeles basin. Model Configuration The Weather Research and Forecasting (WRF) Model is a next-generation mesocale numerical weather prediction system designed to serve both operational forecasting and atmospheric research needs. It features multiple dynamical cores, a 3-dimensional variational (3DVAR) data assimilation system, and a software architecture allowing for computational parallelism and system extensibility. WRF is suitable for a broad spectrum of applications across scales ranging from meters to thousands of kilometers. We use the Eulerian mass coordinate core which has been dubbed the Advanced Research WRF (ARW). We use the ARW with and without 3DVAR to produce two daily 36-hour forecasts over California and adjoining areas. The runs where we use observations and 3DVAR in combination with the ARW to produce a forecast are referred to as with data assimilation and those runs where we do not use 3DVAR are referred to as without data assimilation. Unless otherwise noted the following configuration information applies to both the with- and without- data-assimilation runs.
Model Dimensions 
Model Domain
Model Vertical Profile
Data Assimilation Model - 3DVar Theory. The 3DVar (3- Dimensional VARiational) data assimilation system is developed by NCAR. We selected it because of its ability to assimilate a wide variety of observations, especially those that are not direct measures of the model state variables (e.g., satellite data). 3DVar observation operators relate the values of the model state variables at the analysis time to observed quantities. Observations are categorized by type, each with its own error statistics. The goal is to minimize the difference between the analysis and observations and a prior estimate of the model state (background). The cost function is shown below. The analysis is a “weighted fit” of all sources of information and the “optimal analysis” is that which minimizes the cost-function: 
x = a vector of the model variables at a given time y = Hx where H is the “observation operator” O = Observation (instrumental) error F = Representivity (observation operator) error B = Background (previous forecast from NAM or WRF) error
Observations (with data assimilation)
Concept of Operations with Data Assimilation
Postprocessing
|
||||||||||||||||
