1. Evaluating Cloud Microphysical Schemes in Simulating Orographic Precipitation Events Using OLYMPEX Field Experiment Observations
Brian A. Colle1 and Aaron Naeger2
1School of Marine and Atmospheric Sciences, Stony Brook University, Stony Brook, NY
2University of Alabama at Huntsville
This work is supported by NASA PMM #NNX16AD81G

2. Motivation and Goals
Previous studies over the Pacific Northwest (e.g., IMPROVE-2; Garvert et al. 2005; Colle et al. 2005, Lin et al. 2009, …) showed many bulk micro schemes over-predict windward precipitation and snow aloft (too much cloud water lower windward slope and too little near crest).
There are large bulk microphysical parameter (BMP) uncertainties to riming and other ice characteristics (habit, size distribution, density, etc…).
Orographic precipitation is also highly sensitive to the upstream cross barrier flow, moisture, and stability.
There has been limited verification of orographic flooding events (high freezing levels) over the PNW.
Observed snow (g/kg)
Thompson BMP
Overprediction of cloud water
Garvert et al. (2005)
Grey shade: Cloud water
Black: Snow
Lin et al. 2009

3. Motivational Questions
How well do the various bulk microphysical schemes (BMPs) in WRF predict these precipitation structures and amounts for heavy precipitation events?
Which BMPs lead to the most realistic simulations according to OLYMPEX observations?
How can we use the observations to refine and improve BMPs?

4. WRF Model Setup
Three heavy precipitation cases simulated (12-13 Nov 2015, Nov 2015, and 8-9 Dec)
WRF V3.7.1 at 9, 3, and 1 km grid spacing (50 vertical levels).
IC/BCs: (GFS analyses and NARR – so far)
MYJ PBL, Grell-Freitas (9 km), RRTMG
36-h runs starting (11/12/12z, 11/16/12z,
and 12/08/00z). First 9-h spin-up.
Bulk Microphysical Schemes (BMPs)
Thompson – (2008) ~2D ice, ice size distribution from Field et al. (2005), variable riming efficiency.
Morrison (MORR, 2-moment -2009) predicts number concentration (Nx) to get snow/ice size distribution (l) and intercept (Nos). Spherical ice/snow.
Stony Brook (SBU, 2011) Uses Nos(T), ~2-D ice/snow, combines snow and graupel into one category, and a degree of riming is estimated and variations in snow density (T). Snow/rime properties not advected horizontally.
P3 (2015,2016) Four prognostic mixing ratio variables (total ice mass, rime ice mass, rime volume, and total number) predict the bulk particle properties of a single ice-phase. Advects ice/rime properties.
1 km
3 km
9 km

5. 12-13 November 2015 Event
Significant atmospheric river (~30 mm PWTR: obs and in WRF)
500Z and IR
12 UTC 13 Nov
PWTR: 14 UTC 13 Nov

6. NPOL vs WRF dBZ 1130 UTC 13 Nov P3 MORR NPOL THOMP SBU
Cross section along NPOL RHI scan
NPOL RHI scan
THOMP
SBU
WRF slightly too fast
Areal coverage of obs precip appears best represented by P3.
THOMP significantly overpredicts areal coverage of precip

7. NPOL vs WRF RHI dBZ NPOL P3 MORR THOMP SBU
All schemes show some orographic enhancement of precip, but most distinct peak in P3
NPOL indicates precip windward of terrain, and P3 predicts lower peak in this region
SBU predicts rather steady trend in precip from windward to over terrain P3
Precip sites
MORR
2-hr total from UTC
THOMP
SBU

8. WRF Micro Cross SectionsP3
UND Spiral
MORR
Total Ice – Red Contour
Rime/Graupel – Black Contour
Cloud Water - Shaded
THOMP
SBU
Orographic enhancement of precip in P3 is collocated with largest cloud water and rime/graupel mass.
MORR and THOMP show overall largest total ice (mostly snow) in mid-levels
SBU shows large snow amounts above 5 km in height.

9. UND Citation Micro Comparisons with WRF
MORR
P3 total ice shows good agreement with aircraft
Range is UND total ice content is adequately represented in P3
THOMP and especially SBU overpredict total ice content
UND variable total ice content is not well represented by MORR.
P3 shows overall underprediction in cloud water.
THOMP
SBU
LWC and IWC from Nevzorov probe

10. WRF precipitation verification
MORR P3
OBS
General underprediction in windward precip, except some slight overpredictions in P3
MORR and THOMP show overall largest underpredictions of 30-50%
THOMP
SBU

11. CRN Fishery Wynoochee Graves Creek
All schemes unable to capture large precip amounts at CRN
Smaller precip totals at Fishery and Graves adequately represented
Overall dry bias after 15 UTC is apparent at most sites, except Graves
P3 shows highest precip totals at sites
MORR and THOMP generally among the lowest
Wynoochee
Graves Creek

12. GMI vs GSDSU MORR NPOL dBZ GMI 89 GHz V GMI 166 GHz V MORR dBZ
MORR dBZ compares adequately to NPOL
GMI BT at 89 GHz 4-8 K colder than in MORR
GMI BT at 166 GHz > 15 K colder than in MORR
Ice/snow particles not realistically simulated by MORR
P3 and SBU not yet in GSDSU
NPOL dBZ
GMI 89 GHz V
GMI 166 GHz V
MORR dBZ
MORR 89 GHz V
MORR 166 GHz

13. UND Citation Micro Comparisons with MORR
TOP: LWC and IWC from Nevzorov probe
BOTTOM: LWC King Probe, IWC Heymsfield (PSDs from merged 2D-S and HVPS-3)
UND total ice helps verify that MORR underpredicts total ice/snow content, leading to warmer BTs in MORR than those observed
Linked to the PSD of snow in MORR
Too few ice-phase particles in MORR???

14. 13 Nov aircraft flight
LWC King Probe, IWC Heymsfield (PSDs from merged 2D-S and HVPS-3)
LWC and IWC from Nevzorov probe

15. Conclusions
All schemes generally underpredicted precipitation across the lower windward slopes, which is partly attributed to precipitation ending too soon.
However, WRF P3 scheme predicted the most realistic precipitation amounts and rates, which is at least partly due to more riming and precipitation fallout.
MORR and THOMP showed the largest underpredictions in precipitation of 30-50% over the lower windward slopes.
P3 predicted realistic total ice content profiles according to UND aircraft profiles.
MORR unable to represent variation in the observed ice water content
MORR BTs from GSDSU output were much too warm across much of the precipitation, which is related to issues in snow PSDs from the scheme

16. Future Work
Validate the WRF simulations for the other OLYMPEX cases (17-18 Nov, 3 Dec, and 8-9 Dec) with a focus on the intensive observations during the 3 Dec case
Emphasize the use of the GSDSU and GPM GMI observations to conduct a broader validation of the cloud microphysical schemes
Utilize field measurements, such as more aircraft data and dual-Pol observations, to put the WRF point verification in more of a spatial context 
Take advantage of the wealth of OLYMPEX data for more detailed model validation