CIRA » Home » Research » Project » NASA Assimilation and downscaling of the GPM-like satellite precipitation
Project Title: Ensemble-based assimilation and downscaling of the GPM-like satellite precipitation information
Principal Investigator: Milija Zupanski (Colorado State University)
Co-Principal Investigator: Sara Q. Zhang (NASA Goddard Space Flight Center)
Sponsor: NASA
Duration: April 1, 2010 - March 31, 2013
Funded amount: $289,988
Summary: In a near future Global Precipitation Measurement (GPM) Mission will provide precipitation observations with unprecedented accuracy and spatial/temporal coverage of the globe. Currently operational and research experiences in using precipitation information have mostly focused on a global model resolution with prescribed static forecast error statistics, while a cloud-resolving high resolution and flow-dependent forecast error information are needed for many GPM scientific applications such as hydrology and precipitation estimates downscaled from rain-sensitive radiances. Under this project we develop an ensemble-based data assimilation system for assimilation and downscaling of precipitation information from GPM observations. The project seeks to bring a variety of observations from different instruments and platforms, a cloud-resolving model and an ensemble assimilation methodology together to produce accurate and dynamically downscaled precipitation estimates.
An ensemble data assimilation system at cloud-resolving scales has been developed jointly by NASA/GSFC and Colorado State University (CSU), named WRF-EDAS (WRF Ensemble Data Assimilation System). The system consists of components: (i) Weather Research and Forecasting (WRF) model with multiple nesting capability and NASA cloud-microphysics (ii) NASA Satellite Data Simulation Unit (SDSU), (iii) NOAA/NCEP"s Gridpoint Statistical Interpolation (GSI) forward observation operators and (iv) the CSU Maximum Likelihood Ensemble Filter (MLEF) data assimilation algorithm. WRF-EDAS has capabilities of assimilating precipitation-sensitive microwave radiances at pixel scale and estimating flow-dependent and terrain-dependent forecast error covariance.
The research conducted under this project is as follows: (i) develop the prototype data assimilation system into a validated operational-comparable system, (ii) produce downscaled precipitation analyses by assimilating precipitation information into high-resolution WRF model, using currently available TRMM TMI, SSM/I, AMSR-E, AMSU and InfraRed (IR) cloudy radiances, and (iii) verify and improve the accuracy of precipitation estimates using HydroMeteorological Testbeds (HMTs).
This project belongs to the NASA Precipitation Science Research Category 2.3 "Methodology development for improved applications of satellite products". Expected benefits are in enhancing scientific and operational applications of the GPM observations and improving weather, climate and hydrological forecasts. This research project can serve as a pilot study for the development of Level 4 GPM precipitation products.
Research Highlights:
- Ensemble-based forecast error covariance in cloud-resolving WRF model
Results using WRF-EDAS in application to re-development of the tropical storm Erin (2007) over Oklahoma are shown in Fig.1. One can see spatially localized and smooth analysis responses that suggest a well-defined error covariance. Also, the cross-covariance between cloud ice and rain (Fig.1b) shows a realistic impact that cloud ice observations could have on increasing the rain. Interesting is also to see the impact of cloud ice observation on the analysis of wind (Fig.1c).
- WRF-EDAS performance in a heavy precipitation event over the Southeast HMT
We conducted an experiment characterized by heavy precipitation over the Southeast U.S. lasting for over a week in September 2009, generally covering the Southeast HMT. In addition to the WRF-EDAS experiment we conduct a control experiment that employs current NOAA operational variational data assimilation system (WRF-GSI). The results of accumulated rain in the short-term (3-hour) forecast are shown in Fig.2.
