Use of reprocessed AMVs in the ECMWF Interim Re-analysis Claire Delsol EUMETSAT Fellow Dick Dee and Sakari Uppala (Re-Analysis), IoannisMallas(Data), Niels Bormann, Jean-Noël Thépaut, and Peter Slide Bauer 1 (Satellite section) Leo van de Berg (EUMETSAT)
Talk Outline Introduction 1989 experiment (Met-3) 1995 experiment: XADC period (Met-3 and Met-5) Summary Slide 2
INTRODUCTION Observing systems used in: ERA-40 Interim 1957 1989 6 Aircraft data 1973 TOMS/ SBUV 1979 METEOSAT Reprocessed Cloud motion winds 1982 1988 METEOSAT Reprocessed AMVs + Clear Sky Radiances Conventional surface and upper-air observations NCAR/ NCEP, ECMWF, JMA, US Navy, Twerle, GATE, FGGE, TOGA, TAO, COADS, 1973 VTPR 1979 HIRS/ MSU/ SSU Cloud motion winds Buoy data 1987 SSM/I 1991 ERS-1 1995 ERS-2 1998 AMSU Slide 3
EUMETSAT have reprocessed the AMV product for the Interim project: - improvements in derivation techniques - fully automated use of IR, VIS, WV images (better approach for h.a.) - use of QI (allows greater control on usage of data by NWP users) - better spatial and temporal coverage (1 ½ hrly compared to 2-4 times a day). AMV monitoring and impact Study I Interim IFS configuration: CY31R2 T255 (T159) L60 (4DVAR 12hr window) ECMWF Newsletter nº110 QC for new data: Updated blacklist + quality control same as current pre-msg operational one except for tighter ics as a result of findings from previous ERA studies (Bormann, 3) + thinning at 140km x 140km + use fg-dep QI 1 Slide 4 Data: reprocessed Meteosat-3 (centred on 0º) for 3 months: 4 th Feb to 4 th May 1989
Sample of AMV coverage: 6 th Feb 1989 old 4345 reprocessed 96615 all 1460 6928 Used (after QC) Slide 5
Reprocessed: all Reprocessed: used 71659 5308 IR 19710 1182 Cloudy WV Slide 6
all Used (after QC) 5255 438 VIS increase in numbers at high levels (IR + WV contribution) Slide + low 7 levels (IR + VIS contribution) mid level constrained more by strict quality control.
Used U, V Statistics for used reprocessed amvs and used original amvs Met 8 region exp:1180 /DA (black) v. 1179/DA 1989020-1989030(12) SATOB-Uwind Met8 used U Pressure (hpa) 100 150 250 300 500 700 850 STD.DEV 0 2 4 6 exp -ref -118-1225 +1305 +14769 +24483 +21068 +1679 +2776 +70209 +46039 nobsexp 19 1094 8960 04 27785 23933 5171 11021 86017 exp:1180 /DA (black) v. 1179/DA 1989020-1989030(12) SATOB-Vwind Met8 used V 100 STD.DEV exp -ref -118 background departure o-b(ref) background departure o-b analysis departure o-a(ref) analysis departure o-a BIAS 100 150 250 300 500 700 850 48087-4 -3-2 -1 0 1 2 3 4 nobsexp 19 background departure o-b(ref) background departure o-b analysis departure o-a(ref) analysis departure o-a BIAS 100 Semitransparency method applied irregularly to IR AMVs. Bias worse at mid to high levels but standard deviation better + large increase in number of AMVs 150-1225 1094 150 Pressure (hpa) 250 300 500 700 850 +1305 +14769 +24483 +21068 +1679 +2776 +70209 8960 04 27785 23933 5171 11021 86017 Slide 8 250 300 500 700 850 0 2 4 6 +46039 48087-4 -3-2 -1 0 1 2 3 4
Statistical significance (t-test) RMS Vector wind forecast error validated against ERA-40 analysis (90 cases) Forecast Day hpa 850hPa 500hPa hpa Extra-trop 2 3 5 7 ics 2 0. 0.1% 3 0. 0.1% Extra- 5 7 2 3 1.0% No entry means that there is not enough statistical significance 5 7 Europe 2 3 5 7 Slide 9 Improved scores in black Degraded ones in orange.
Mean wind analysis (control) Vector difference of mean wind analysis between ctl and expt (new Met3) 20m/s 2.5m/s hpa 50m/s 2.5m/s 300hPa Slide 10
AMV monitoring and impact study II: XADC period This corresponds to period when Meteosat-5 was operational at 0º and Meteosat-3 was leant to the US due to a faulty GOES satellite. Reprocessing of both datasets gives us an opportunity to look at the impact of having more reprocessed datasets simultaneously. Interim IFS configuration: CY31R2 T255 (T159) L60 (4DVAR 12hr window) Data: reprocessed Meteosat-5 (0º) and Meteosat-3 (75º W) for 3 months: 1 st Jan to 31 st Mar 1995 QC: as for the 1989 experiment. Example of coverage: 19950102 Slide 11 Reprocessed Met3 and Met5 (satid: 50 and 52) Original Met5 (satid: 5)
Met-5 and Met-3 Used U Reprocessed AMVs:black Old AMVs: red Northern Extra-tropics +ve bias still present in extra-tropics: semitransparency correction method irregularity? Highlights large increase in data used in assimilation (note no Met3 in original operations) ics Expect bias/std to be different Slide 12 Large bias not as pronounced in 1989 experiment
P (hpa) zonal mean windspeed bias (obs-fg) 60S Eq 60N Met5 (used) Met3 (used) 1-15 Jan 1995 600 800 WVcloudy (45641) 600 800 IR (37218) 600 800 WVcloudy (25316) 600 800 IR (84751) 60S 0 60N 60S 0 60N 60S 0 60N 60S 0 60N 600 VIS (25352) VIS (29080) 600 800 60S 0 60N Slide 13 800 60S 0 60N
3 EXPERIMENTs CTL: original operational set-up EXPT1 Met3 Met5 Met3 Met5 EXPT2 Met3 Met5 EXPT3 Met3 Met5 Slide 14 reprocessed AMVs original AMVs no original AMVs
EXPT1 Statistical significance RMS Vector wind forecast error validated against ERA-40 analysis Forecast Day 2 3 4 5 6 7 hpa 0. 0.1% 850hPa 0. 1% 0. 500hPa 0. 1% hpa 0. 0. Slide 15 VERY good scores except for ics! Improved scores in black Degraded ones in orange. Smaller % = more significant impact
EXPT2 Statistical significance RMS Vector wind forecast error validated against ERA-40 analysis Reprocessed Met-5 (no Met-3) Forecast Day 2 3 4 5 6 7 hpa 850hPa 500hPa 1% 0. 0.1% hpa 1% 0. Slide 16 extra-tr not as good extra-tr particularly good at high levels ics degraded Improved scores in black Degraded ones in orange.
EXPT 3 Statistical significance RMS Vector wind forecast error validated against ERA-40 analysis old Met-5 reprocessed Met-3 Forecast Day 2 3 4 5 6 7 hpa 1% 850hPa 0. 500hPa 1% 0. hpa 0.1% 1% 1% 1% Slide 17 extra-tropics better ics not as bad! Improved scores in black Degraded ones in orange.
Negative impact in ics: WHY? ics difficult area to validate against other observations Subtropical jets area: sensitive (location + intensity) Tested blacklisting more strictly (remove the mid-to-high-level biases between 30ºS and 30ºN up to 300hPa) - removes negative impact locally but the negative impact still present at the very high levels (ie. hpa) Further Investigation: Difference between 1989 and 1995 experiments is in the observing system: ERS-1 scatterometer surface winds. Slide 18 Run experiment during the wet season for the ITCZ (more active season)
SUMMARY Reprocessed winds were monitored as part of the Interim Re-analysis project. EUMETSAT s support for this has been of great value. A first quality and impact study: Reprocessed Met-3 AMVs for Feb-April 1989 Large increase in the amount of AMVs + improved std dev of departures BUT biases at mid-levels semi-transparency correction method? Forecast impact: relatively neutral in extra-tropics and very positive in low-level ics. A second study: for XADC period (1995) Met-3 (75ºW) and Met-5 (0º). Mid level bias still present. Very positive impact in Extra-ics but very negative in ics at high levels Blacklisting ics more strictly reduces the ve impact locally but more ve at hpa. Slide 19 Areas to pursue: ITCZ wet season (July-Sept) + impact of the scatterometer data
Slide 20
Scatterometer surface wind - Interim expt: 1992 (May) 500hPa hpa Slide 21
exp:1323 /DA (black) v. 1332/DA 1995010100-1995040100(12) SATOB-Uwind N.Midlat used U 100 STD.DEV exp -ref -17 nobsexp 0 background departure o-b(ref) background departure o-b analysis departure o-a(ref) analysis departure o-a BIAS 100 150 +248 351 150 Pressure (hpa) 250 300 500 700 850 +2401 +16695 +41058 +30244 +3336 +11408 +59766 3492 21583 53937 40326 6732 14278 128079 250 300 500 700 850 0 2 4 6 +32373 43304-4 -3-2 -1 0 1 2 3 4 exp:1323 /DA (black) v. 1332/DA 1995010100-1995040100(12) SATOB-Uwind ics used U 100 STD.DEV exp -ref -443 nobsexp 107 background departure o-b(ref) background departure o-b analysis departure o-a(ref) analysis departure o-a BIAS 100 150-2212 7108 150 Pressure (hpa) 250 300 500 700 +16170 +75569 +38509-5103 -2508 +19070 81128 108992 56886 2613 2831 22118 250 300 500 700 850 +211495 337835 850 +107562 0 2 4 6 136465-4 -3-2 -1 0 1 2 3 4 exp:1323 /DA (black) v. 1332/DA 1995010100-1995040100(12) SATOB-Uwind S.Midlat used U 100 STD.DEV exp -ref -37 nobsexp 0 background departure o-b(ref) background departure o-b analysis departure o-a(ref) analysis departure o-a BIAS 100 150 +414 1227 150 Pressure (hpa) 250 300 500 700 850 +11291 +45690 +81977 +51856 +6996 +11783 +153402 17119 59853 103393 61995 11310 17634 265513 250 300 500 700 850 Slide 22 0 2 4 6 +79752 99420-4 -3-2 -1 0 1 2 3 4
hpa Vector difference of mean wind analysis between ctl and expt (reprocessed Met-3 and Met-5 with stricter ics) Jan/Feb/Mar 1995 300hPa hpa Slide 23
Stricter blacklist Statistical significance RMS Vector wind forecast error validated against ERA-40 analysis Forecast Day 2 3 4 5 6 7 hpa 0. 1% 850hPa 0. 1% 500hPa 0. 0. hpa 0. 1% 0. 0. 0.1% 0.1% 0. Slide 24 Better in ics : -ve impact confined to higher levels BUT ve impact in >500hPa in Improved scores in black Degraded ones in orange.