1. Towards Assimilation of GOES Hourly winds in the NCEP Global Forecast System (GFS) Xiujuan Su, Jaime Daniels, John Derber, Yangrong Lin, Andy Bailey, Wayne Bresky, Hongming Qi JCSDA 10 th Workshop on Satellite Data Assimilation

2. Outline Review of Effort and Goals Evaluation of GOES hourly winds: data quality Experimental Setup and Approach Next Steps

3. Assimilate hourly GOES winds in the NCEP/GFS −Infrared cloud drift and water vapor winds (Task 1) −Visible and short-wave infrared winds (Task 2) Improve the quality control procedures associated with assimilation of satellite derived winds Transition updates into the next official GFS build for operations 3 Review of Effort and Goals

4. Same winds retrieval algorithm with a few updates −Improved height assignment of low level AMVs over ocean when a low level temperature inversion is detected −Image scan line time defines the time for each satellite wind observation −BUFR files contain the Expected Error (EE) and QI quality indicators More timely, reduced latency (15-30 minutes) 4 GOES Hourly Winds GOES-14 GOES-15

5. Time window about cycle time used in GSI −Half hour window for current operational GOES winds (3-hourly) −Three hour window used for GOES hourly winds −Data count comparison (IR winds, for example) −10543 for current operational GOES winds −52326 total number for GOES hourly winds Evaluation of GOES hourly winds Two week GSI run with GOES hourly winds from July 11 to 15, 2012 was conducted for this purpose 5 Evaluation of GOES hourly winds

6. 6 Data Quality: Initial Findings Hourly Winds 3-Hourly Winds

7. RMS of O-B vs relative hours to cycle time

8. Data count and quality vs. EE, QI Higher Quality

10. Vector RMS vs. EE and QI Higher Quality

12. Absolute Observation vs. EE, QI Higher Quality

14. GOES hourly winds have similar quality as current operational GOES winds The relative hour from cycle time do not impact the data quality The RMS of O-B winds decrease with higher values of quality marks of EE and QI at all levels Generally speaking, the absolute U and V values decrease with higher EE values, and increase slightly with higher QI values Most winds have QI values higher than 85 (about 75%) at all levels. Distribution of numbers of observations for EE varies with heights. 14 Data Quality: Summary

15. Based on above results, a parallel experiment was set up to assimilate GOES hourly winds from July 1 to August 15 with current GSI version (trunk) and T574L64 model resolution on zeus super computer. Filter out winds with QI scores below 85 Set observation error in GSI for GOES hourly wind double current operational error. 15 Current Assimilation Strategy

16. Complete summer assimilation experiment Another experiment for winter season will be set up if current strategy works If assimilation strategy fails for summer experiment then, −Revisit optimal usage of winds quality marks (QI and EE) to filter data −Explore different thinning strategies will be tried to get better forecast impacts 16 Next Steps