GSI EXPERIMENTS FOR RUA REFLECTIVITY/CLOUD AND CONVENTIONAL OBSERVATIONS Ming Hu, Stan Benjamin, Steve Weygandt, David Dowell, Terra Ladwig, and Curtis Alexander 1 NOAA/ESRL/GSD/EMB RUA workshop, Boulder 06/03/2015
GSI overview The Global Statistical Interpolation (GSI) was developed mainly as NCEP operational data analysis system for improving model forecast: GFS, NAM, RAP/HRRR, HWRF, … GSI can analyze many kinds of observations: Conventional, radiance, radar, GPS, … GSI analysis cores: 3DVAR, ensemble-var hybrid GSI background can be: GFS, NMMB, NNM, ARW, … 2
Analysis versus initial condition GSI aims to generate better initial condition which make the forecast better: GSI analysis results may not fit to the observations closely. Analysis requires the analysis results fit to the observation closely to reflect the “true” atmosphere status. RTMA is the only function in GSI to do 2D analysis But GSI can be configured to conduct 3D analysis 3
GSI is analysis system GSI modifies the background to fit to the observations based on the ratio of observation error and background error For forecast Balance the weight between background and observation For analysis: Weight the observation more (small observation error) Reduce the impact radius Less balance Benefit of using GSI: Work with WRF-arw, nmm, nmmb, and GFS Advanced analysis method: 3DVAR, hybrid Observation operators available for many kinds of obs Observations are QCed inside and outside GSI Community code 4
GSI for forecast: RAP example 5 Background (01) and analysis (03) fit to observations Wind o-g 01 all count o-g 01 all bias o-g 01 all rms o-g 03 all count o-g 03 all bias o-g 03 all rms Temperature: o-g 01 all count o-g 01 all bias o-g 01 all rms o-g 03 all count o-g 03 all bias o-g 03 all rms Moisture: o-g 01 all count o-g 01 all bias o-g 01 all rms o-g 03 all count o-g 03 all bias o-g 03 all rms Wind RMS reduced 19.6% after analysis; Temperature reduced 11.8%; Moisture reduced 24.5%. Different observation type has different fit rate, such as soudning T reduced 21.1%.
U Increment from single obs test — Different horizontal impact scale 6 zoom in Change ‘hzscl_op’ not only change the horizontal influence scale, but also the weight (how much analysis results fit to the observations)! 0.6~ ~1.0 Figures from courtesy of Min Sun
U Increment from single obs test —— — Different vertical impact scales 7 0.6~0.7 X-Z Plane Figures from courtesy of Min Sun
GSI for Analysis with hzscl_op=1/8 and VS=1/2 8 Background (01) and analysis (03) fit to observations Wind o-g 01 all count o-g 01 all bias o-g 01 all rms o-g 03 all count o-g 03 all bias o-g 03 all rms Temperature: o-g 01 all count o-g 01 all bias o-g 01 all rms o-g 03 all count o-g 03 all bias o-g 03 all rms Moisture: o-g 01 all count o-g 01 all bias o-g 01 all rms o-g 03 all count o-g 03 all bias o-g 03 all rms Wind RMS reduced 61.1% after analysis; Temperature reduced 18.4%; Moisture reduced 30.8%. Much less than the single observation test because redundant data and balance in BE.
GSI for Analysis with hzscl_op=1/8 and VS=1/2 + ¼ of observation error 9 Background (01) and analysis (03) fit to observations Wind o-g 01 all count o-g 01 all bias o-g 01 all rms o-g 03 all count o-g 03 all bias o-g 03 all rms Temperature: o-g 01 all count o-g 01 all bias o-g 01 all rms o-g 03 all count o-g 03 all bias o-g 03 all rms Moisture: o-g 01 all count o-g 01 all bias o-g 01 all rms o-g 03 all count o-g 03 all bias o-g 03 all rms Wind RMS reduced 64% after analysis; Temperature reduced 27.6%; Moisture reduced 31.7%.
GSI for forecast: RAP surface analysis increment 10 Wind RMS reduced 19.6% after analysis; Temperature reduced 11.8%; Moisture reduced 24.5%.
GSI for Analysis with hzscl_op=1/8 and VS=1/2 11 Surface analysis increment Wind RMS reduced 61.1% after analysis; Temperature reduced 18.4%; Moisture reduced 30.8%.
GSI for Analysis with hzscl_op=1/8 and VS=1/2 + ¼ of observation error 12 Wind RMS reduced 64% after analysis; Temperature reduced 27.6%; Moisture reduced 31.7%. Surface analysis increment
GSI Cloud Analysis: RAP at 18Z 06/02/ background analysis Cloud top Cloud base Cloud top
Analysis PODy 1000 feet ceilingPODy 3000 feet ceilingPODy 500 feet ceiling With cloud analysis Without cloud analysis 6h fcst RAP Cloud Ceiling Verification: PODY To keep cloud in forecast: adjust moisture and temperature in cloudy area and clear area 14
GSI reflectivity analysis 15 backgroun d RAP analysis 2D reflectivity RUA analysis 3D reflectivity 00Z 08/10/2014 Composite ref
GSI reflectivity analysis cross section at j=224 and I=385: X XX Z Z Z backgroun d RAP analysis ref RUA analysis ref QSNOW at column (i=411,j=224)
Summary GSI is ready to be used for the analysis GSI analysis can include conventional observation, cloud, radar and other observations Background error covariance and observation errors should tuned for the analysis GSI has many advantages, most important one is function to use ensemble to add flow depend information into the analysis (David’s talk) 17