1 / 60

Model Parameterizations: How do they affect NWP rainfall forecasts?

Model Parameterizations: How do they affect NWP rainfall forecasts?. Mike Baldwin NSSL SPC CIMMS. Skill of current operational models. To put it kindly… Current NWP models do a poor job forecasting heavy rainfall AVN (red), Eta (green), NGM (blue) for 1 Jan – 10 Oct 2001.

betrys
Télécharger la présentation

Model Parameterizations: How do they affect NWP rainfall forecasts?

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. Model Parameterizations: How do they affect NWP rainfall forecasts? Mike Baldwin NSSL SPC CIMMS

  2. Skill of current operational models • To put it kindly… • Current NWP models do a poor job forecasting heavy rainfall • AVN (red), Eta (green), NGM (blue) for 1 Jan – 10 Oct 2001

  3. Ingredients needed by a model in order to predict some phenomena of interest • Adequate grid spacing • To be able to resolve the feature • Physical processes • All those important in the development, maintenance, and decay of the feature • Dynamics • Accuracy, hydrostatic/non-hydrostatic • Adequate initial/boundary conditions • To be able to capture important forcing

  4. Ask yourself: • “Does the model I’m using have the necessary ingredients to predict the feature(s) that I’m considering or expecting?” YES: model guidance taken literally can be useful NO: model by itself is of little value (but not worthless) Either way: knowledge of model characteristics will increase the value of NWP guidance

  5. What are Model Parameterizations? • Techniques used in NWP to predict the collective effects of physical processes which cannot be explicitly resolved • Sub-grid scale or perhaps near-grid scale processes: • For example; cloud physics, convection, turbulent mixing, radiation, surface exchanges

  6. Interaction between different processes is critical • Especially for mesoscale models • Not only important to do a good job with a specific physical process • All pieces have to work well together in order for model to perform well • Several studies have shown great forecast sensitivity to subtle changes to a parameterization

  7. 12 Jun 01 03Z 12 JUN 01 02Z

  8. Operational models NSSL ETAKF 27h forecast OPNL ETA 27h forecast

  9. Outline • Quick overview of convective parameterizations currently available in NWP models (both operational and research) • Look at some research/case studies of NWP performance in heavy rain events • Talk about future

  10. Current EMC models use different approaches • RUC: Grell scheme • Eta: Betts-Miller-Janjic (Kain-Fritsch used experimentally at NSSL & SPC) • MRF/AVN: Grell-Pan scheme • Grell, Grell-Pan, and Kain-Fritsch schemes are Mass-Fluxschemes, meaning they use simple cloud models to simulate rearrangements of mass in a vertical column • Betts-Miller-Janjic adjusts to “mean post-convective profiles” based on observational studies

  11. “Mass-flux” parameterization

  12. What do all convective parameterizations do? • Predict convective precipitation • Feedback onto larger scales the effects of transports, mixing, circulations, etc. found within convective elements • Change vertical stability • Redistribute and generate heat • Redistribute and remove moisture • Make clouds that affect surface heating and atmospheric radiation

  13. How do convective schemes accomplish these tasks? • convective triggering (yes/no) • convective intensity (how much rain?) • vertical distribution of heating • vertical distribution of drying

  14. Triggering/activating • CAPE (ALL) • mass or moisture convergence exceeding a certain threshold (Kuo) • positive destabilization rate (Grell) • perturbed parcels can reach their level of free convection (KF) • sufficient cloud layer moisture (BMJ)

  15. Convective intensity • proportional to mass or moisture convergence (Kuo) • sufficient to offset large-scale destabilization rate (Grell) • sufficient to eliminate CAPE, constrained by available moisture (KF) • proportional to cloud layer moisture (BMJ)

  16. Vertical distribution of heating and drying • determined by adjusting to empirical reference profiles (BMJ, Kuo) • estimated using a simple 1-D cloud model to satisfy the constraints on intensity (Grell, KF)

  17. Different ways to classify convective schemes • Molinari and Dudek (1992) • Traditional • clear separation between convective and stratiform or grid-scale precipitation • Hybrid • direct interaction between convective and grid-scale physics • Fully Explicit • grid-scale cloud and precipitation physics ONLY Grid Spacing (km) 0.1 1 10 20 30 40 50 60 Fully explicit ??? Hybrid Traditional

  18. Example: Betts-Miller-Janjic (BMJ) scheme – Eta Model • BMJ scheme requires some CAPE • “Equilibrium-type” scheme • Deep and shallow components • Deep = precipitating • Shallow = non-precipitating • Critical factor in determining yes/no/amount of precipitation is the cloud layer moisture

  19. A few quick points… • Deep convection is given first priority • Deep convection will fail if • cloud layer is too dry • cloud is too shallow • Scheme defers to shallow convection if deep convection fails • No feedback of cloud water/ice

  20. Deep convection example Feedback from BMJ scheme • KSBN 18h forecast from 00Z 31 May 2000 Eta run D Td Cloud depth D T sounding 3h earlier

  21. What does BMJ deep convection do? • Stabilize the cloud layer • Typically heats mid/upper cloud • Dries lower part of cloud • Does not modify the sub-cloud layer • Feedback reduces CAPE and precipitable water

  22. Shallow convection example Feedback from BMJ scheme • KOKC 4h forecast from 12Z 1 Jun 2000 Eta run D T D Td Cloud depth sounding 4h earlier

  23. What does BMJ shallow convection do? • Mixes moisture up from cloud base to cloud top • Mixes heat down from cloud top to cloud base (destabilizes the cloud layer) • Location of cloud base & top critical for determining impact on forecast fields • Could affect lapse rates, cap strength • Does not affect precipitable water

  24. How to recognize BMJ shallow convection • No convective precipitation • Forecast sounding has a smoothly varying moisture profile up to ~200mb deep (usually concave shape) • “Straight-line” temperature profile over the same layer, just above LCL • Base of unusual mid-tropospheric inversion indicates cloud top

  25. Gallus (1999) Weather and Forecasting p. 405-426 • “Eta simulations of three extreme precipitation events: Sensitivity to resolution and convective parameterization” • Ran Eta Model at four different horizontal resolutions (78, 39, 22, and 12km) and with two convective schemes (BMJ vs. KF) • Variations in precipitation forecasts were found to be highly case dependent

  26. Gallus (1999) Figure 1a • 16-17 Jun 1996 • MCS formed over central IA in warm sector ahead of sfc low • Heavy rains also occurred north of warm front in WI

  27. Gallus (1999) Figure 3

  28. Gallus (1999) Figure 10 • BMJ runs • 78, 39, 22, and 12km res • Contours at 5mm, then every 25mm

  29. Gallus (1999) Figure 13 • KF runs • 78, 39, 22, and 12 km res • Contours at 5mm, then every 25mm

  30. Items to note for this case • BMJ runs DO NOT produce higher precip amounts as resolution increases • KF runs DO produce higher precip amounts as resolution increases • BMJ produced a broad area of precipitation that covered observed region for many hours • Peak amounts in high-res KF runs produced mainly by grid-scale precipitation scheme

  31. Gallus (1999) Figure 1b • 16-17 Jul 1996 • MCS/MCC developed north of warm front • Training cells found in region of peak rain

  32. Gallus (1999) Figure 6

  33. Gallus (1999) Figure 7 • Omaha sounding 00 UTC 17 Jul 96 • 2000 J/kg CAPE, less than 25 J/kg CIN (above inversion) • Example BMJ reference profiles (dashed)

  34. Gallus (1999) Figure 15 • BMJ runs • Peak ppt decreases as resolution increases • Areal coverage increases with resolution

  35. Gallus (1999) Figure 16 • KF runs • Location errors are large, too far to the north • Peak amounts are more reasonable and increase with resolution

  36. Gallus (1999) Figure 1c • 27 May 1997 • Jarrell TX tornado outbreak • Supercell-type heavy rain event • Boundary interaction important

  37. Gallus (1999) Figure 8

  38. Gallus (1999) Figure 9 • Estimated sounding near Jarrell at 18 UTC 27 May 97 • CAPE near 5000 J/kg, no CIN

  39. Gallus (1999) Figure 19 • BMJ runs • Peak ppt increases with resolution • Heaviest ppt produced mainly by grid-scale scheme in high-res runs

  40. Gallus (1999) Figure 21 • KF runs • Peak ppt increases slightly with resolution

  41. Gallus (1999) Figure 20 • BMJ moisture divergence at 15 UTC • Outflow-type circulation initiated in NW Texas • Moved to the southwest

  42. Lessons learned… • This study shows great sensitivity of the QPF to different convective schemes and horizontal resolutions • Heaviest QPF was produced by grid-scale parameterization • No consistent behavior by either scheme from case to case • Should expect great difficulty in developing a model to predict heavy rainfall with accuracy and consistency

  43. Sensitivity, continued… • Not only are models sensitive to different parameterizations • Changes to a single parameterization can also produce significant differences in model QPF • Spencer and Stensrud (1998) MWR for example

  44. Sensitivity of a 1km cloud model to cloud microphysics Standard LFO for “large Ice” no Precipitation Species Size Distributions [Matt Gilmore, NSSL]

  45. National Hail Research Experiment Knight et al. (1982) .12 Fraction of Total .06 104 106 108 [Matt Gilmore, NSSL] Slope Intercept (dm-3 mm-1 )

  46. Hail/graupel slope-intercept (no) (dm-3 mm-1 ) 103 to 105 (hail) 104 to 108 (graupel/hail) References Federer and Waldvogel (1975), Dennis et al. (1971), and Spahn (1976) Knight et al. (1982) Slope-Intercept Observations [Matt Gilmore, NSSL]

  47. Density (kg m-3) Hail: 700 to 900 (some suggest as low as 400) Graupel: 50 to 890 Reference Pruppacher and Klett (1978) Hail/Graupel Particle Density Observations [Matt Gilmore, NSSL]

  48. Species Size Distributions [Matt Gilmore, NSSL]

More Related