Web page designed by Jon Moskaitis (Penn State) and Sharan Majumdar (RSMAS/MPO, Univ. of Miami)
 
 
 

WSR 2000

Statistical Tests

ET KF Signal Variance vs NCEP Signal Variance at Targeting Time (3 bins)
ET KF Signal Variance vs NCEP Signal Variance at Targeting Time (5 bins)
ET KF Signal Variance vs NCEP Signal Variance at Verification Time
NCEP Signal Variance vs Reduction in NCEP Forecast Error Variance
ET KF Rescaled Signal Variance vs Reduction in NCEP Forecast Error Variance

Targeting Time:  Time of observation via dropsondes released from aircraft

MRF - ETKF Comparison  at Targeting Time : The top panel is the MRF squared signal in 3 -level u,v, and T derived from analyses of two parallel MRF forecasts initialized at the targeting time.  The parellel runs are a control run without targeted observations included, and the operational MRF, which includes the targeted obs.  Thus the signal is (control - operational).  The bottom panel shows the ETKF signal variance in  3-level u,v, and T calculated using MRF and ECMWF ensemble data from runs initialized 24 and 36 hours before the targeting time.  The positions where the dropsodes were released is shown on both panels.

MRF - ETKF Comparison at the Verification Time:  The top panel is the MRF squared signal , as described above, but this time using the parallel forecast data valid at the verification time.   The bottom panel shows the ETKF signal variance.  The verification region is a 1000 km radius circle, chosen during the experiment by HPC forecasters as a time and place where a storm is predicted.  For some flights, there are more than one verification regions.
Note:  In parentheses next to the verification time is a code pertaining to the location of the verification region.  A stands for Alaska, W for the West Coast and E for anywhere east of the Rockies.

MRF RSFE with MRF Squared Signal : The top panel is the MRF squared signal, exactly as in the MRF - ETKF Comparison.  The bottom panel is the Reduction in Squared Forecast Error (RSFE) in 3-level u,v,T achieved by the operational MRF (using targeted obs) relative to the control run of the MRF (does not use targeted obs).  The RSFE is derived by subtracting the Squared Forecast Error of the operational run from the Squared Forecast Error of the control run.  The forecast errors are calculated by subtracting the operational (or control) forecast initialized at the targeting time from the operational MRF analysis valid at the verification time.

MRF Squared Signal Evolution:  These show how the MRF squared signal evolves in 12 hour time intervals.

ETKF Signal Variance Evolution:  These show how the ETKF signal variance evolves in 12 hour time intervals.  Dropsonde release positions are drawn at the targeting time and verification regions are drawn at the verification times.
 

WSR 2001

Statistical Tests

Verification Cases EAST of Rockies
ET KF Signal Variance vs NCEP Signal Variance at Verification Time

Verification Cases WEST of Rockies
ET KF Signal Variance vs NCEP Signal Variance at Verification Time

Links

NCEP Ensemble products
NCEP Targeted Observations page
University of Miami Meteorology
Penn State Meteorology