Assimilation of GPS Radio-Occultations at DWD Harald Anlauf Research and Development, Data Assimilation Section Deutscher Wetterdienst, O ff enbach, Germany IROWG 2nd Workshop, Estes Park (Colorado, USA) 28. March 2012 Harald Anlauf (DWD) Assimilation of GPSRO at DWD 28. March 2012 1 / 14
NWP Models at DWD GME COSMO-DE Global model, hydrostatic Non-hydrostatic, “convection allowing” Triangular grid, mesh size: 20 km 60 levels (top: 5 hPa) Mesh size: 2.8 km 50 levels (1474562 × 60 grid points) Forecast times: Forecast times: 174h from 00Z, 12Z; 21h from 00Z, 03Z, . . . , 21Z 48h from 06Z, 18Z COSMO-EU COSMO-DE-EPS Non-hydrostatic (Pre-operational) Ensemble prediction system Mesh size: 7 km 20 ensemble members 40 levels (operationally: 40 members) Forecast times: Forecast times: 78h from 00Z, 12Z; 21h from 00Z, 03Z, . . . , 21Z 48h from 06Z, 18Z Harald Anlauf (DWD) Assimilation of GPSRO at DWD 28. March 2012 2 / 14
Global Data Assimilation System at DWD 3D-Var-PSAS , 3-hourly update cycle Available Forward models for GPSRO I 1d bending angle operator (Implementation by Michael Gorbunov); fixed/e ff ective tangent point for profile or individual tangent point I 3d ray tracer (Michael Gorbunov) Implementations tested and evaluated in collaboration with GFZ using data from CHAMP and GRACE (Pingel and Rhodin, 2009) I Ray tracer: best in terms of std.dev. of OBS-FG, numerically expensive! Ray tracer also needs additional data not provided in BUFR messages (satellite positions and velocities) I 1d-operator (Abel integral) with e ff ective location of occultation probably good enough for initial operational implementation (still needed major optimization e ff orts for the NEC SX-9) Refractivity 3-term expression as recommended by GRAS-SAF “Initialization” MSIS-90 climatology matched at model top ( ∼ 36 km) Harald Anlauf (DWD) Assimilation of GPSRO at DWD 28. March 2012 3 / 14
Assimilation of GPS Radio-Occultations Observation errors (S. Healy): I Linear decrease from 10% to 1% for impact height from 0 to 10 km I 1% from 10 km to 30 km Quality control of observations: I Consistency checks of profiles I Observation-minus-first guess check: 4 σ (should be made stricter) I B.a. < 0.02 rad to avoid ducting (replace by condition on refractivity) I Clip lowest section of GPS-RO profiles when non-monotonous Vertical thinning to model resolution, exponential smoothing Use impact heights 3 km–30 km Exclude occultations starting above 20 km GPS Radio-Occultations operationally used since 2010-08-03 I COSMIC/FORMOSAT-3 FM 1-2, 4-6 (FM-3 dead since 2010-08-01) I GRACE-A I GRAS on METOP-A I TerraSAR-X (since 2010-12-09) I C/NOFS, SAC-C (since 2012-02-29) Harald Anlauf (DWD) Assimilation of GPSRO at DWD 28. March 2012 4 / 14
Impact of the Assimilation of GPSRO Better fit to radiosondes in upper troposphere/lower stratosphere, esp. southern hemisphere (but mixed results in Antarctic) Significant forecast improvements with assimilation of GPSRO ANOC geopotential 500 hPa SH 2010070100 till 2010073100 (31 forecasts) 1 GME routine GMEP GPSRO 0.9 Anomaly correlation 0.8 (90% c.l.) 0.7 H. Anlauf et al., Atmos. Meas. Tech., 0.6 4, 1105–1113 0.5 0.4 0 24 48 72 96 120 144 168 forecast time (h) Anomaly Correlation of Geopotential 500 hPa, Southern Hemisphere for July 2010 ⇒ gain of several hours vs. operational system Harald Anlauf (DWD) Assimilation of GPSRO at DWD 28. March 2012 5 / 14
Comparison to Radiosondes Mean departures of temperature and rel. humidity observations from radiosondes to 3-h forecasts, Southern Hemisphere Area=-20/-60 0/360 0 0 211 11 10 10 732 114 20 20 1580 -18 30 30 1964 232 50 50 2197 93 70 70 Pressure (hPa) Pressure (hPa) 2139 266 100 100 2398 73 150 150 2384 85 200 200 2407 73 250 250 2479 40 300 300 2460 50 400 400 2480 44 500 500 2561 53 700 700 2530 13 850 850 2347 11 925 925 1641 -1 H. Anlauf et al., -2 -1 0 1 2 0.0 0.5 1.0 1.5 2.0 2.5 3.0 Bias of Temperature [K] RMS error of Temperature [K] Atmos. Meas. Tech., Area=-20/-60 0/360 4, 1105–1113 200 200 0 0 250 250 2183 17 300 300 Pressure (hPa) 2216 53 Pressure (hPa) 400 400 2262 56 500 500 2127 24 700 700 2158 -20 850 850 2273 1 925 925 1666 -16 -10 -5 0 5 10 0 5 10 15 20 Bias of Relative Humidity [%] RMS error of Relative Humidity [%] (blue: ctrl (operational), red: GPSRO experiment) Harald Anlauf (DWD) Assimilation of GPSRO at DWD 28. March 2012 6 / 14
Problems with the Assimilation of GPSRO over Antarctic Large temperature bias, got even worse with assimilation of GPSRO! Partially understood: poor representation of vertical correlations in operational assimilation system, revised in December 2010 Some issues with the forecast model, but investigations ongoing Harald Anlauf (DWD) Assimilation of GPSRO at DWD 28. March 2012 7 / 14
Issues with Assimilation of GPSRO Data (I) Bias, RMS di ff erences for di ff erent satellites (processing, model, . . . ) Lower troposphere: largest bias in the tropics, smaller in extratropics Statistics for Bending Angles from METOP / GPS RO EXP=rou Statistics for Bending Angles from COSMIC / GPS RO EXP=rou OBS minus FG/AN for: Surface=all Flag=used SatId= 4 OBS minus FG/AN for: Surface=all Flag=used SatId= 740 Time period = 20110101 00UTC - 20110831 21UTC, STEP=3h Time period = 20110101 00UTC - 20110831 21UTC, STEP=3h Area=20/-20 0/360 Solid=OBS-FG, Dashed=OBS-AN 30 30 9704 18 29 29 17931 12087 28 28 9915 26 27 27 21601 9308 26 26 46 22 25 25 23340 9725 24 24 62 17 23 23 24258 10308 22 22 78 88 21 21 25797 11291 20 20 Impact height (km) 22432 10233 Impact height (km) 19 19 40141 21239 18 18 30679 11084 17 17 25177 16320 16 16 47819 18774 Tropics 15 15 46080 23550 14 14 44738 23941 13 13 43359 24019 12 12 40312 23454 11 11 36484 22187 10 10 46611 32627 9 9 26022 21818 8 8 25943 32138 7 7 10805 20876 6 6 8181 30277 5 5 2607 12438 4 4 488 3451 3 3 0 0 2 2 0 0 1 1 0 0 -0.04 -0.02 0.00 0.02 0.04 0.00 0.02 0.04 0.06 0.08 0.10 0.12 Bias of (O-B)/O RMS error of (O-B)/O Harald Anlauf (DWD) Assimilation of GPSRO at DWD 28. March 2012 8 / 14
Issues with Assimilation of GPSRO Data (I) Bias, RMS di ff erences for di ff erent satellites (processing, model, . . . ) Lower troposphere: largest bias in the tropics, smaller in extratropics Statistics for Bending Angles from METOP / GPS RO EXP=rou Statistics for Bending Angles from COSMIC / GPS RO EXP=rou OBS minus FG/AN for: Surface=all Flag=used SatId= 4 OBS minus FG/AN for: Surface=all Flag=used SatId= 740 Time period = 20110101 00UTC - 20110831 21UTC, STEP=3h Time period = 20110101 00UTC - 20110831 21UTC, STEP=3h Area=60/20 0/360 Solid=OBS-FG, Dashed=OBS-AN 30 30 12320 15 29 29 27010 19829 28 28 12300 22 27 27 38088 19294 26 26 159 26 25 25 38936 19777 24 24 155 23 23 23 38273 19516 22 22 210 75 21 21 38644 19704 20 20 Impact height (km) 36634 19191 Impact height (km) 19 19 64050 38854 18 18 47034 19592 17 17 37512 27575 16 16 71664 31290 Northern 15 15 68884 38454 14 14 65466 37768 13 13 62685 36944 Extratropics 12 12 59608 36955 11 11 56549 37123 10 10 74728 57598 9 9 43452 39207 8 8 51549 57987 7 7 27397 36761 6 6 28964 53316 5 5 14349 32491 4 4 7985 24545 3 3 0 0 2 2 0 0 1 1 0 0 -0.04 -0.02 0.00 0.02 0.04 0.00 0.02 0.04 0.06 0.08 0.10 0.12 Bias of (O-B)/O RMS error of (O-B)/O Harald Anlauf (DWD) Assimilation of GPSRO at DWD 28. March 2012 8 / 14
Issues with Assimilation of GPSRO Data (II) Apparent positive bias for non-GRAS data due to bugs in first-guess check implementation (non-symmetric w.r.t. OBS and FG!) Current GRAS data are (known to be) biased below ∼ 8 km I Rising occultations (globally) I Setting occultations (notably tropics, lower troposphere) Harald Anlauf (DWD) Assimilation of GPSRO at DWD 28. March 2012 9 / 14
Issues with Assimilation of GPSRO Data (II) Apparent positive bias for non-GRAS data due to bugs in first-guess check implementation (non-symmetric w.r.t. OBS and FG!) Current GRAS data are (known to be) biased below ∼ 8 km I Rising occultations (globally) I Setting occultations (notably tropics, lower troposphere) 3dvar monitoring Bending Angle [radian] Level Max/Min: 3944.40 / 3058.10 Date: 2011070100 - 2011073121 Satellite: 4 Region: TR30 Lat = -30.0 ... 30.0 0.020 min = 0.0116 max = 0.0216 mean = 0.0154 median = 0.0151 GRAS, First Guess [radian] 100 min = 0.0121 0.018 max = 0.0200 75 mean = 0.0161 median = 0.0159 30 � S–30 � N, 50 30 bias =-0.0008 0.016 rms = 0.0019 3-4 km 25 corr = 0.4908 20 total number = 276 15 max number in bin = 9 0.014 10 slope = 1.1471 const =-0.0019 7 5 * : mean of values per column 3 0.012 1 0.012 0.014 0.016 0.018 0.020 Observation [radian] Harald Anlauf (DWD) Assimilation of GPSRO at DWD 28. March 2012 9 / 14
Recommend
More recommend
