1. A new approach to parameterize ice-phase cloud microphysics The Predicted Particle Properties (P3) Scheme WWOSC 2014 Montreal, Canada August 17, 2014 Hugh Morrison 1 and Jason Milbrandt 2 1 National Center for Atmospheric Research, Boulder, CO, USA 2 Environment Canada, Montreal, Canada

2. Environment Canada’s High Resolution (2.5-km) Deterministic Prediction System Column-maximum REFLECTIVITY* * Computed from microphysics (Milbrandt-Yau 2-moment) Experimental implementation: summer 2014 Operational implementation: 2015

3. N (D)N (D) D [  m] 100 [m -3  m -1 ] 20 40 60 80 0 10 1 10 0 10 -1 10 -2 Microphysics Parameterization Schemes ULTIMATE GOAL: Predict evolution of hydrometeor size distributions N (D)N (D) D [  m] 100 [m -3  m -1 ] 20 40 60 80 0 10 1 10 0 10 -1 10 -2 Bin-resolving: Bulk: N (D)N (D) D [  m] 100 [m -3  m -1 ] 20 40 60 80 0 10 1 10 0 10 -1 10 -2 (spectral) 1 m 3 (unit volume)

4. Radius [cm] Bin-resolving coalescence model Berry and Reinhardt (1974) RAIN CLOUD Mass Density [g m -3 (lnr) -1 ] Time [min] BACKGROUND – Representation of Hydrometeors Historical development of bulk scheme approach: start with liquid-only (“cloud” and “rain”)

5. Example of Observed Ice Particles Photos c/o Alexei Korolev BACKGROUND – Representation of Hydrometeors Historical development of bulk scheme approach: start with liquid-only (“cloud” and “rain”) add ice-phase (“ice”)

6. GRAUPEL  g = 400 kg m -3 V = a g D b g HAIL  h = 900 kg m -3 V = a h D b h “SNOW”  s = 100 kg m -3 V = a s D b s  abrupt / unphysical conversions CLOUD ICE  s = 500 kg m -3 V = a i D b i Problems with pre-defined categories: 1. Conversions between categories are ad-hoc 2. Conversions lead to large, discrete changes in particle properties etc. … Traditional approach of bulk microphysics schemes: BACKGROUND – Representation of Hydrometeors

7. The simulation of ice-containing cloud systems is often very sensitive to how ice is partitioned among categories MOR-hail (only) MY2 - hail (only) MOR-graupel (only) MY2-baseline (g + h) Morrison and Milbrandt (2011), MWR Microphysics Schemes: MOR: Morrison et al. (2005, 2009) MY2: Milbrandt and Yau (2005) idealized 1-km WRF simulations (em_quarter_ss) base reflectivity

8. There is a paradigm shift in the parameterization of ice-phase microphysics  Increased emphasis on the prediction of hydrometeor properties Recent developments : 2-moment – Ziegler (1985), Ferrier (1994), Reisner et al. (1998), etc. 3-moment – Milbrandt and Yau (2005) predicted rime fraction – Morrison and Grabowski (2008) predicted crystal habit – Harrington et al. (2013) predicted graupel density – Connolly et al. (2005), Mansell et al. (2010), Milbrandt and Morrison (2013) BACKGROUND – Representation of Hydrometeors Partial mitigation to the problems with pre-defined categories

9. SQUEAK! QUACK! has a label that says “DUCK” big, round eyes yellow, wing-like appendages plastic exterior, hollow interior no feet makes a “squeak” noise has no label small, round eyes white, wing-like appendages feathery exterior, meaty interior webbed feet makes a “quack” noise Which of the following is more duck-like? IF IT QUACKS LIKE A DUCK … DUCK

10. Based on a conceptually different approach to parameterize ice-phase. LIQUID PHASE: 2 categories, 2-moment: q c, q r, N c, N r ICE PHASE:1 category, 4 prognostic variables: q i, q rim, N i, B rim  predicts wide range of properties (and thus types of ice) Compared to traditional (ice-phase) schemes, P3: avoids some necessary evils (category conversion, pre-defined properties) has self-consistent physics is better linked to observations is more computationally efficient New Bulk Microphysics Parameterization: Predicted Particle Properties (P3) Scheme* * Morrison and Milbrandt (2014) J. Atmos. Sci (in press)

11. Mesoscale Model Results

12. 3D Squall Line case: (June 20, 2007 central Oklahoma) WRF_v3.4.1,  x = 1 km,  z ~ 250-300 m, 112 x 612 x 24 km domain initial sounding from observations convection initiated by u-convergence no radiation, surface fluxes

13. 2007 OK Squall Line: Base Reflectivity (1 km AGL, t = 6 h) MOR-G MOR-H THO Observations WSM6 WDM6 P3 dBZ MY2 Morrison et al. (2014) J. Atmos. Sci (in press)

14. WRF Results: Line-averaged Reflectivity (t = 6 h) Observations dBZ MOR-G MOR-H THO WSM6 WDM6 P3 MY2

15. DERIVED Ice Physical Properties         F r ~ 0-0.1  ~ 900 kg m-3 V ~ 0.3 m s -1 D m ~ 100 μm  small crystals  F r ~ 0  ~ 50 kg m-3 V ~ 1 m s -1 D m ~ 3 mm  aggregates  F r ~ 1  ~ 900 kg m -3 V > 10 m s -1 D m ~ 5 mm  hail etc. FrFr VmVm DmDm ρpρp Vertical cross section of model fields (t = 6 h) P3 SIMULATION

16. Precipitation rate at 1 km height Time- averaged from 6-7 h

17. Low-density, unrimed snow Rimed snow/ low-density graupel Small, dense ice Z qiqi qrqr qcqc FrFr VmVm DmDm ρpρp Z P3 SIMULATION Vertical cross section of model fields (t = 24 h)  

18. 2014 OU CAPS Ensemble (HWT) 1 km Reflectivity, 22 UTC 8 May, 2014 22 h forecast OBSMOR-G P3 THO MY2-v2 MY2-v1 http://hwt.nssl.noaa.gov/Spring_2014

19. Scheme Squall line case (  x = 1 km) Orographic case (  x = 3 km) # prognostic variables P3*0.436 (1.043)0.686 (1.013) 7 MY20.621 (1.485)1.012 (1.495)12 MOR-H0.503 (1.203)0.813 (1.200) 9 THO0.477 (1.141)0.795 (1.174) 7 WSM60.418 (1.000)0.677 (1.000) 5 WDM60.489 (1.170)0.777 (1.148) 8 Average wall clock time per model time step (units of seconds.) Times relative to those of WSM6 are indicated parenthetically. Timing Tests for 3D WRF Simulations  P3* is one of the fastest schemes in WRF *1 ice-phase category version

20. Based on a conceptually different approach to parameterize ice-phase. LIQUID PHASE: 2 categories, 2-moment: q c, q r, N c, N r ICE PHASE:1 category, 4 prognostic variables: q i, q rim, N i, B rim  predicts wide range of properties (and thus types of ice) New Bulk Microphysics Parameterization: Predicted Particle Properties (P3) Scheme* * Morrison and Milbrandt (2014) J. Atmos. Sci (in press) The single “free (ice-phase) category” version shows very promising early results

21. Multi-category version* LIQUID PHASE:2 categories, 2-moment: q c, q r, N c, N r ICE PHASE:category 1: q i_1, q rim_1, N i_1, B rim_1 category 2: q i_2, q rim_2, N i_2, B rim_2 category 3: q i_3, q rim_3, N i_3, B rim_3…  predicts wide range of properties (and thus types of ice) – as before – BUT now allows for different types of ice in the same grid point New Bulk Microphysics Parameterization: Predicted Particle Properties (P3) Scheme * Milbrandt and Morrison (2015) (in preparation)

22. Conclusions The P3 approach introduces a conceptual departure from current bulk microphysics schemes. Preliminary results – idealized, real-case simulations, and real-time forecasts – are very promising… Further developments to the P3 scheme will include: more predicted properties + spectral dispersion (3-moment) + liquid fraction + etc…