The sensitivity of simulated orographic precipitation to model

  • Slides: 19
Download presentation
The sensitivity of simulated orographic precipitation to model details other than cloud microphysics Günther

The sensitivity of simulated orographic precipitation to model details other than cloud microphysics Günther Zängl Meteorologisches Institut der Universität München

What are the primary sources of uncertainty in high-resolution simulations of orographic precipitation? h

What are the primary sources of uncertainty in high-resolution simulations of orographic precipitation? h Cloud microphysics h (In nested-domain runs) Convection parameterizations used in the outer model domains h Soil moisture / PBL parameterization h Numerical side effects (vertical coordinate specification, numerical diffusion. . . )

Test strategy Compare the spread among five different microphysical parameterizations against the effect of

Test strategy Compare the spread among five different microphysical parameterizations against the effect of changing h the convection parameterization in the coarse domains h the soil moisture specification h the PBL parameterization h the vertical coordinate formulation h the implementation of horizontal diffusion

Set-up of the simulations h Model: MM 5, version 3 h 4 nested domains,

Set-up of the simulations h Model: MM 5, version 3 h 4 nested domains, finest horizontal resolution 1. 4 km (see figure) h 38 model levels in the vertical h Case: MAP-IOP 10 (Oct. 24/25, 1999) h Initial / boundary data: Operational ECMWF analyses h Period of simulation: Oct. 24, 00 UTC Oct. 25, 18 UTC h Validation against 81 surface stations for Oct. 24, 06 UTC - Oct. 25, 18 UTC (see figure for location)

Parameterizations used for the reference run and changes for sensitivity tests h Reisner 2

Parameterizations used for the reference run and changes for sensitivity tests h Reisner 2 microphysical scheme from MM 5 version 3. 3 (Reisner 1, Goddard v 3. 3, Goddard v 3. 5, Reisner 2 v 3. 5) h Grell cumulus parameterization in D 1 (37. 8 km) and D 2 (12. 6 km) (Kain-Fritsch in D 1 and D 2; Grell in D 1 -D 3; Kain-Fritsch in D 1 -D 2 and Grell in D 3) h Gayno-Seaman PBL parameterization (Blackadar PBL, MRF PBL) h Modified horizontal diffusion scheme (Zängl 2002, MWR) computes the horizontal diffusion of temperature and the moisture variables truly horizontally rather than along the terrain-following coordinate surfaces (Original diffusion scheme for moisture only / for moisture and temperature) h Smooth-level vertical coordinate system to Schär et al. 2002, MWR) (similar

36 h-accumulated precipitation in the reference run domain-average precip station-intp. average observation (mm) >47

36 h-accumulated precipitation in the reference run domain-average precip station-intp. average observation (mm) >47 N 11. 2 8. 1 2. 1 47 N 46. 6 N-47 N 47. 4 34. 7 18. 1 46. 6 N 46. 2 N-46. 6 N 65. 9 42. 7 33. 6 46. 2 N <46. 2 N 83. 4 69. 7 53. 7 total 45. 5 38. 1 26. 3

Difference fields (sensitivity experiment - REF run) Reisner 1 -scheme new Reisner 2 -scheme

Difference fields (sensitivity experiment - REF run) Reisner 1 -scheme new Reisner 2 -scheme +5% / +4% +2% / +3% Relative difference in domain-average (station-interpolated average)

Goddard v 3. 3 +20% / +24% Goddard v 3. 5 +7% / +6%

Goddard v 3. 3 +20% / +24% Goddard v 3. 5 +7% / +6%

Cumulus parameterizations Kain-Fritsch instead of Grell in D 1 and D 2 Grell in

Cumulus parameterizations Kain-Fritsch instead of Grell in D 1 and D 2 Grell in D 1, D 2 and D 3 -10% / -7% -8% / -5%

Combination of Kain-Fritsch in D 1 / D 2 and Grell in D 3

Combination of Kain-Fritsch in D 1 / D 2 and Grell in D 3 -16% / -13%

Boundary-layer parameterization (reference: Gayno-Seaman PBL) Blackadar PBL -6% / -6% MRF PBL -16% /

Boundary-layer parameterization (reference: Gayno-Seaman PBL) Blackadar PBL -6% / -6% MRF PBL -16% / -12% Predictive soil moisture scheme instead of fixed soil moisture: <1%

Smooth-level vertical coordinate system Parameterizations as in REF run -7% / -4% Kain-Fritsch in

Smooth-level vertical coordinate system Parameterizations as in REF run -7% / -4% Kain-Fritsch in D 1, D 2; Grell in D 3 -9%/ -10% (-24%/ -21% w. r. t. REF)

Implementation of horizontal diffusion Diffusion along sigma-levels for moisture only +9% / +26% Diffusion

Implementation of horizontal diffusion Diffusion along sigma-levels for moisture only +9% / +26% Diffusion along sigma-levels for moisture and temperature +40% / +66%

Ranking list of sensitivities (disregarding the obsolete version of the Goddard parameterization) 1. Horizontal

Ranking list of sensitivities (disregarding the obsolete version of the Goddard parameterization) 1. Horizontal diffusion (40% - 65%) 2. Convection scheme (5% - 15%) Boundary-layer parameterization (5% - 15%) 3. Vertical coordinate system (7% - 10%) 4. Cloud microphysics (2% - 7%)

Which simulation performs best and worst? i Best simulation: Kain-Fritsch parameterization in D 1

Which simulation performs best and worst? i Best simulation: Kain-Fritsch parameterization in D 1 and D 2, Grell in D 3, smooth-level vertical coordinate: bias 3. 8 mm (+14%); rms error 12. 1 mm i Worst simulation: Original diffusion scheme (i. e. diffusion along sigma-levels for temperature and moisture), parameterizations as in REF: bias 36. 9 mm (+140%); rms error 48. 0 mm i For comparison: REF run bias 11. 8 mm; rms error 19. 5 mm

Conclusions - Part I h Errors in precipitation forecasts are not necessarily due to

Conclusions - Part I h Errors in precipitation forecasts are not necessarily due to errors in the cloud microphysics The side effects arising from other parameterizations and numerical errors deserve much more attention than they currently receive h Particularly large systematic errors can arise from computing the horizontal diffusion along terrain-following coordinate surfaces

ECMWF MAP Reanalysis data instead of operational analyses Reference setup -13% / -11% New

ECMWF MAP Reanalysis data instead of operational analyses Reference setup -13% / -11% New vertical coordinate -18% / -21%

New vertical coordinate and modified configuration of convection parameterizations (Kain-Fritsch in D 1 /

New vertical coordinate and modified configuration of convection parameterizations (Kain-Fritsch in D 1 / D 2 and Grell in D 3) -3% / -6%

Conclusions - Part II i The impact of using the Reanalysis data instead of

Conclusions - Part II i The impact of using the Reanalysis data instead of the operational analyses depends sensitively on the model setup i Comparison with observations reveals: Using the Reanalysis data yields a substantial improvement for the standard setup and the standard setup with the new vertical coordinate, but not for the setup which yielded the best results with the operational analysis Comparisons between operational analysis and MAP reanalysis can produce misleading results when carried out with one model setup only