1 / 53

Addressing Climate Model Errors Through Observation Integration

Explore the impact of observational data on climate model accuracy. Discuss errors, spin-up, and observational proposals for better model performance in climate studies.

marycyoung
Download Presentation

Addressing Climate Model Errors Through Observation Integration

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. GPCI spin-up issues • Advertisement: CEOP cloud observations – coming up • Spin-up: Why interesting? • Errors: RMSE, bias, anomaly correlation, variabiltiy • 5day evolutions • Model errors: Hadley circulation, cirrus cloud, PBL height • Some observations: w, IWP • A very speculative fix of a low resolution bias in climate models. GCSS meeting 9-2006 Martin Köhler 1

  2. CEOP observation proposal(with Rachel Pinker, U. Maryland and Mike Bosilovich, GMAO, NASA) Linderberg lidar+radar conv. cloud high cloud med. cloud low cloud Dec 2004 • CEOP compares reference point and gridded data from observations and models • Yet, there are NO cloud observations in CEOP database • Add point data: TCC from ISCCP and cloud profiles form CLOUDNET lidar/radar • Add gridded data: 5km / 1hour resolution data from METEOSAT and GOES • Most interesting for diurnal cycle and Monsoon studies. GCSS meeting 9-2006 Martin Köhler 2

  3. Initial errors - Schematic state variable (e.g. TCWV) Forecast < 1kg/m2 1mm Analysis Observations True state 1day 5days Time Initial tendency reflects the errors of physical parameterizations assuming that analysis errors are small. Balance between analysis and forecast model: Both have identical biases. Makes it hard to detect any errors. GCSS meeting 9-2006 Martin Köhler 3

  4. History of spin-down: precip TCWV AN-FC48h(20N-20S) 29R2: Assim. Rain, limit q-incr 28R1: Conv. changes 24R3: Conv.+supersat 26R3: AIRS, New q-anal 23R4: 29R1: Moist BL, num. 1st step 25R3: Mult. Incr. 4DVAR, SSMI-rad, changes clds+cnv 28R3: num. cld+cnv, FG from 6+18, No SYNOP-q at night 25R1 28R2: Early delivery GCSS meeting 9-2006 Martin Köhler 4

  5. Spin-up Profile split into physical processesMark Rodwell, ECMWF Temperature tendencies over first 6h. Amazon, Jan 2005. GCSS meeting 9-2006 Martin Köhler 5

  6. Forecast setup • Analysis: JJA 2003 ECMWF (e-suite with AIRS) • Resolution: T42 (NCAR), T240 (UKMO), T399 (ECMWF) • UKMO: SST climatology instead of SST JJA2003 GCSS meeting 9-2006 Martin Köhler 6

  7. RMSE, anomaly correlation, BIAS, variability if ANC=1: variability2 ANC: anomaly correlation BIAS2 GCSS meeting 9-2006 Martin Köhler 7

  8. U 850hPa RMSE, BIAS, AC, variability GCSS meeting 9-2006 Martin Köhler 8

  9. W 500hPa RMSE, BIAS, AC, variability RMSE [Pa/s] GCSS meeting 9-2006 Martin Köhler 10

  10. RMSE 900hPa Q, PT, R, CC RMSE [K] GCSS meeting 9-2006 Martin Köhler 11

  11. BIAS 900hPa Q, PT, R, CC GCSS meeting 9-2006 Martin Köhler 12

  12. RMSE 800hPa Q, PT, R, CC GCSS meeting 9-2006 Martin Köhler 13

  13. BIAS 800hPa Q, PT, R, CC GCSS meeting 9-2006 Martin Köhler 14

  14. GPCI spinup: Precipitation NCAR ECMWF UKMO ITCZ precip spin-down: < factor 2 GCSS meeting 9-2006 Martin Köhler 15

  15. GPCI spinup: PBL height NCAR ECMWF UKMO UKMO, ECMWF: strcu zi spin-up: 100-200m NCAR: strcu zi spin-down: 80m GCSS meeting 9-2006 Martin Köhler 16

  16. GPCI spinup: LWP NCAR ECMWF UKMO UKMO, ECMWF: strcu LWP spin-up: 20-40 g/m2 GCSS meeting 9-2006 Martin Köhler 17

  17. GPCI spinup: IWP NCAR ECMWF ECMWF: IWP spin-down: 20 g/m2 NCAR: IWP only 10 g/m2 GCSS meeting 9-2006 Martin Köhler 18

  18. Ice Content compared to Microwave Limb SounderLi, Jiang, Waliser, Tompkins GCSS meeting 9-2006 Martin Köhler 20

  19. GPCI cross-section: Omega GCSS meeting 9-2006 Martin Köhler 21

  20. GCSS Pacific Cross-section Project Omega Bias at 500hPa sfc vector wind: QuikSCAT vs ECMWF FG Equator ITCZ V-Bias: METEOSAT-8 vs ECMWF FG IWC and W decay in 10d FC 6.2 μm Thanks to Adrian Tompkins, Hans Herbach, Claire Delsol! GCSS meeting 9-2006 Martin Köhler 22

  21. Speculation on resolution dependent bias Thomas Jung (2005) half # cyclones in T95 vs T255 dxy=T399 - T159 T at 400hPa Peter Janssen (2005) surface KE missing in T95 vs Quickscat v`T` transient meridional sensible heat transport GCSS meeting 9-2006 Martin Köhler 23

  22. singular vector perturbation T511 T255 T159 GCSS meeting 9-2006 Martin Köhler 24

  23. Conclusions • Errors found (compared to ECMWF analysis): • NCAR: equatorial 850hPa westward wind bias • NCAR: stratus warm dry bias– but clouds OK • UKMO: trade cumulus cold moist bias – high CC • All models: precip spin-down by factor 2 • NCAR: PBL height down • UKMO, ECMWF: PBL height (and LWP) up • ECMWF: IWP reduces from 40g/m2 to 20g/m2 (fixed by CY31R1) • UKMO, ECMWF, NCAR: Hadley circulation slows down • Interpretation: • Surface scatterometer wind and cloud vector winds support Hadley circulation problem • Hypothesis: low resolution model misses pole-ward heat transport • Possible fix: use singular vector perturbations to amplify mid-latitude eddies. GCSS meeting 9-2006 Martin Köhler 25

  24. Extra Slides GCSS meeting 9-2006 Martin Köhler 26

  25. Temperature bias change due to resolution GCSS meeting 9-2006 Martin Köhler 27

  26. IWP in tropics: CY31R1 versus CY30R1 GCSS meeting 9-2006 Martin Köhler 28

  27. GCSS meeting 9-2006 Martin Köhler 29

  28. Validation of model against Meteosat 8 (IR channel) 1 July - 31 August 2006 / 24h FC from 00Z GCSS meeting 9-2006 Martin Köhler 30

  29. 150 hPa 480 hPa 150 hPa 480 hPa Qice OPER CY30R1 Qice esuite CY31R1 Validation of model against Meteosat 8 (IR channel) 1 July / 24h FC from 00Z GCSS meeting 9-2006 Martin Köhler 31

  30. Validation of model against Meteosat 8 1 July / 24h FC from 00Z: CCOV and Qliq. esuite CY31R1 OPER CY30R1 GCSS meeting 9-2006 Martin Köhler 32

  31. GCSS meeting 9-2006 Martin Köhler 33

  32. GCSS meeting 9-2006 Martin Köhler 34

  33. GCSS meeting 9-2006 Martin Köhler 35

  34. GCSS meeting 9-2006 Martin Köhler 36

  35. GCSS meeting 9-2006 Martin Köhler 37

  36. GCSS meeting 9-2006 Martin Köhler 38

  37. GCSS meeting 9-2006 Martin Köhler 39

  38. GCSS meeting 9-2006 Martin Köhler 40

  39. GCSS meeting 9-2006 Martin Köhler 41

  40. GCSS meeting 9-2006 Martin Köhler 42

  41. GCSS meeting 9-2006 Martin Köhler 43

  42. GCSS meeting 9-2006 Martin Köhler 44

  43. GCSS meeting 9-2006 Martin Köhler 45

  44. GCSS meeting 9-2006 Martin Köhler 46

  45. History of spin-up (20N-20S) GCSS meeting 9-2006 Martin Köhler 47

  46. History of spin-up (20N-20S) 29R2: Assim. Rain, limit q-incr List with selection of model changes. In addition there were many changes to the use of satellite data. 24R3: Conv.+supersat 28R1: Conv. changes 23R4: 26R3: AIRS, New q-anal 29R1: Moist BL, num. 1st step 25R3: Mult. Incr. 4DVAR, SSMI-rad, clds+cnv changes 28R3: num. cld+cnv, FG from 6+18, No SYNOP-q at night 25R1 28R2: Early delivery GCSS meeting 9-2006 Martin Köhler 48

  47. History of spin-up (20N-20S) GCSS meeting 9-2006 Martin Köhler 49

  48. Conclusions on spin-up • Spin-up has improved with: • A modest reduction in precipitation spindown • A substantial reduction in TCWV spindown • A change from increase of evaporation during the forecast to a decrease of evaporation. (BL has become more dry in analysis). • It is difficult to make a precise link between model changes and impact on spin-up • Model changes and data assimilation changes (including use of satellite data) have contributed • It is impossible to verify TCWV within 1 kg/m2 using radio sonde data. GCSS meeting 9-2006 Martin Köhler 50

  49. End – Extra Slides GCSS meeting 9-2006 Martin Köhler 52

  50. GCSS meeting 9-2006 Martin Köhler 53

More Related