Community Global Observing System Simulation Experiments Package (CGOP)

Transcription

1 Community Global Observing System Simulation Experiments Package (CGOP) S.-A. Boukabara (NOAA), I. Moradi (UMD), B. Atlas(NOAA), R. Hoffman (AOML), K. Ide (UMD), T. Auligne (JCSDA), N. Shahroudi (UMD), Y. Zhou (UMD), J. Woolen (NOAA), S. Casey (UMD), L. Cucurrul (NOAA), R. Li (NOAA), P. VanDelst (NOAA), Q. Liu (NOAA), K. Kumar (NOAA), R. Errico (GMAO), N. Prive (GMAO) 2015 CICS-MD Science Meeting, College Park, MD, USA - November 23-24, 2015 Project# NA14OAR November 23, 2015

2 Introduction

3 Observing System Simulation Experiments (OSSEs) Source: Schematic description of OSSEs

4 Conventional data assimilation vs. OSSE Data assimilation using real observations True atmosphere is sampled by imperfect instruments DA system can only be validated using sparse data such as radiosonde and dropsondes Real observations are only available from current instruments Data assimilation using OSSEs Observations are simulated from a high-resolution NWP model known as Nature run using a forward model DA system can be easily validated since the true atmosphere (the Nature run ) is known Observations can be simulated for any future instruments as long as the forward model is available

5 Application of OSSEs investigate the impact of future instruments on the weather forecast and data assimilation evaluate different data assimilation techniques since the truth is well known examine the impact of any future gap in the weather satellites investigate the impact of new data types on weather forecast and data assimilation OSSEs can be used for many other applications such as product retrieval, satellite data bias correction, etc.

6 The Community Global OSSE Package

7 The Community Global OSSE Package - CGOP Community Global Observing System Simula9on Experiment Package Forward Operators GPS-RO Bending Angle and Refrac4vity Simulator Deriving Conven4onal Data from Nature Run Satellite Radiance Simulator (CRTM) Data Assimila9on Components and Scripts GFS Model GSI-Opera4onal Version GSI-Research Version (4D-Var) DA Scripts and Codes Independent Assessment Tool (IAT) for post-processing Radiance Monitoring (RadMon) Verifica4on Database (VSDB) GSI Diag Files Data Analysis Hurricane Track/ Intensity 1

8 The OSSE Package GSI stat files Assimilation using GSI Real Observations Interpolation Location of Assimilated Data Interpolating NR to the Location of Observations Nature Run Forward Models CRTM Control Files Bending Angle and Refractivity Forward Model Perfect Simulations Simulated Radiances Conventional Data GPS-RO Profiles Error Addition Adding Random and Systematic Errors Final Simulations Simulated Observations

9 Random noise addition Real observations Perfect Simulations Assimilation using GSI Assimilation using GSI σ o2 [O B] σ s2 [S B] Random Noise = σ o 2 - σ s 2 For satellite radiances, the noise is added independently for each channel with no correlation For conventional data, the random error is added separately for p,t,q,v and stratified by layers The GPS-RO random error is added separately for bending angle and refractivity

10 Simulated versus actual observations for ATMS

11 Benchmarking of the different components of CGOP Parallel runs on S4 and Theia star2ng from the same ini2al condi2ons Anomaly Correla,on RMS Error Domain Parameter Level Forecast Day Forecast Day /200 hpa Temperature 500 hpa 850 hpa Wind 250/200 hpa NH 850 hpa 250/200 hpa Geopoten2al 500 hpa 700 hpa 1000 hpa 250/200 hpa Temperature 500 hpa 850 hpa Wind 250/200 hpa SH 850 hpa 250/200 hpa Geopoten2al 500 hpa 700 hpa 1000 hpa S4 significantly bejer than theia S4 bejer than theia, nonsignificant no discernible difference S4 worse than Theia, nonsignificant S4 significantly worse than theia

12 Benchmarking of the different components of CGOP Mean temperature difference for parallel runs on S4 and Theia star7ng from the same ini7al condi7ons 00 hour 168 hours or 7 days

13 Thank you for your attention!