Investigating Height Assignment Type Errors in the NCEP Global Forecast System

Transcription

1 Investigating Height Assignment Type Errors in the NCEP Global Forecast System James Jung Cooperative Institute for Meteorological Satellite Studies John Le Marshall Centre for Australian Weather and Climate Research, Australia Jaime Daniels National Oceanic and Atmospheric Administration NESDIS, Center for Satellite Applications and Research Lars Peter Riishojgaard Joint Center for Satellite Data Assimilation 1

2 Outline Height Assignment types Definitions of Statistics used Speed Direction Height Assignment Statistics Water Vapor Images Infrared Images NCEP Future Work Comments and Conclusions

3 Height Assignment Types CO Slicing GOES-1 and beyond Menzel et al 1983 Water Vapor Intercept Szejwach 198 Histogram Nieman et al 1993 IR Window Nieman et al 1993 Cloud Base Le Marshall et al

4 Wind Comparison Statistics Vector Difference VD = ( U i U m) + ( V i V m) Mean Vector Difference MVD 1 N = N i= 1 ( Ui U m) + ( V i V m) Standard Deviation SD = 1 N N i = 1 [( VD i ) ( MVD )] * Nieman et al 1997 i = observation, m = model 4

5 Wind Comparison Statistics Root Mean Square Error RMSE = ( MVD) + ( SD) Normalized Root Mean Square Error NRMSE = RMSE Speed Speed Bias N 1 BIAS = + + U i V i U m N i= 1 V m * Nieman et al 1997 i = observation, m = model 5

6 New Directional Statistics (X mod,y mod ) Y B Θ A X C (X ob,y ob ) C = + ABcos A B ( Θ) Directional RMSE 1 DRMSE = N 1 Ci i= 1 Directional Bias N 1 BIAS = Ci N i= 1 N 6

7 Height Assignment Statistics Ocean only vectors used Monitored all height assignment types available for each observation. Separated by Satellite and Image type GOES-11, GOES-1 Water Vapor winds, Cloud-Drift IR winds Combined two seasons Days 1-0 in July and December 9 Stratified by Latitude NH (0N 60N) Tropics (0N 0S) SH (0S 60S) 7

8 Speed NRMSE and Bias Normalized MVD RMSE GOES-11 Speed Bias GOES Pressure [hp RMSE [m/s] Bias [m/s] H0 HIST WIN HO HIST WIN Normalized MVD RMSE GOES-1 Speed Bias GOES RMSE [m/s] Bias [m/s] CO H0 HIST WIN CO HO HIST WIN 8

9 Directional RMSE and Bias Directional RMSE GOES-11 Directional Bias GOES RMSE [degrees] HO HIST WIN Bias [degrees] HO HIST WIN Directional RMSE GOES-1 Directional Bias GOES RMSE [degrees] Bias [degrees] CO HO HIST WIN 9 CO HO HIST WIN

10 Speed NRMSE and Bias 0 Normalized MVD RMSE GOES RMSE [m/s] Speed Bias GOES-11, IR Window Bias [m/s] H0 HIST WIN H0 HIST WIN Normalized MVD RMSE GOES-1 Speed Bias GOES RMSE [m/s] Bias [m/s] CO H0 HIST WIN 10 CO H0 HIST WIN

11 Directional RMSE and Bias Directional RMSE GOES-11 Directional Bias GOES RMSE [degrees] Bias [degrees] HO HIST WIN HO HIST WIN Directional RMSE GOES-1 Directional Bias GOES RMSE [degrees] Bias [degrees] CO HO HIST WIN CO HO HIST WIN 11

12 Latitude error dependence Normalized MVD RMSE GOES-11, HO Intercept Speed Bias GOES-11, HO Intercept RMSE [m/s] Bias [m/s] SH Trop NH SH Trop NH Normalized MVD RMSE GOES-1, HO Intercept Speed Bias GOES-1, HO Intercept RMSE [m/s] Bias [m/s] SH Trop NH 1 SH Trop NH

13 Latitude Error Dependence Normalized MVD RMSE GOES-11, HO Intercept Speed Bias GOES-11, HO Intercept RMSE [m/s] SH Trop NH Bias [m/s] SH Trop NH Normalized MVD RMSE GOES-1, HO Intercept Speed Bias GOES-1, HO Intercept RMSE [m/s] Bias [m/s] SH Trop NH 13 SH Trop NH

14 NCEP Continuation Work Modify analysis code to process all height assignment types Generate O-B/O-A statistics for each height assignment type Derive assimilation techniques specific to height assignment type Investigate QC techniques using multiple height assignments 14

15 Comments and Conclusions The cloud-drift IR winds have a slow speed bias not found in the water vapor winds. Why, how can this be resolved? Are improvements possible to the CO and HO intercept methods? What are the upper and lower limits of the techniques? O-B errors in the tropics are greater than at midlatitudes This is a model problem. How can data providers help? Can other data providers include height assignment type in the BUFR file? 15

16 Questions? 16