NWP at NOAAs Earth System Research Laboratory, Global Systems - - PowerPoint PPT Presentation

NWP at NOAA’s Earth System Research Laboratory, Global Systems Division (ESRL/GSD): developments and applications for physics parameterizations

Georg Grell, Joe Olson, Shan Sun, Ben Green, Li Zhang, Ravan Ahmadov, Isidora Jankov, Stan Benjamin, Ligia Bernardet, many others

Overview

• Developments of ESRL’s operational storm scale and regional scale

modeling physics suite (currently using WRF)

– Overview of storm scale and regional scale physics developments – MYNN-EDMF-Shallow convection – Stochastic physics

• Some aspects of global modeling:

– Recent work with the Grell-Freitas convective parameterization – inline chemistry, seasonal forecasting experiments with a couple atmosphere/ocean/chemistry model

• Future modeling plans at ESRL/GSD

RAPid Refresh (RAP), and High Resolution Rapid Refresh (HRRR) domains

3 Expanded (new) RAP domain (13 km)

• Hourly update cycle for RAP

and HRRR – operational Additional experimental runs

• 750m nest experimental
• RAP also with full chemistry

(twice a day – experimental)

• HRRR with Smoke and other

anthropogenic emissions twice a day for 36 hr forecasts - experimental

Current Status - NOAA Hourly Updated Models RAP

HRRR

RAP - Rapid Refresh (Benjamin et al., MWR, 2016) – 13km

– NOAA “situational awareness” model for high-impact weather – New 18-hour forecast each hour – NOAA/NCEP operational – 1 May 2012 – RAPv2 implementation – 25 Feb 2014 – Hourly use by National Weather Service, SPC/AWC/WPC, FAA, private sector

HRRR – High-Resolution Rapid Refresh

• 3km - Storm/energy/aviation guidance
• Real-time operational – NCEP, and experimental-

ESRL supercomputer

• NCEP implementation HRRRv1 - 30 Sept 2014
• HRRRv2/RAPv3 -NCEP implementation- Aug 2016

4

RAP/HRRR Physical Processes & Parameterizations

5

Model Component Currently under development in RAP/HRRR Aspects of ongoing developments Stochastic approaches in progress Non-local Turbulent transport MYNN Mass-flux EDMF multi plume approach (Neggers et al), momentum transport inclusion, scale aware Stochastic entrainment Clouds - microphysics Thompson aerosol-aware Will be in WRFV3.9 Use of wildfires, dust, sea salt, other emissions for Thompson aerosol aware microphysics, prognostic application of Chaboureau-Bechtold, tuning of radiation coupling Stochastic SPP component for cloud fractions Non resolved deep convection Grell-Freitas parameterization Will be in WRFV3.9 Implementation and evaluation in HWRF, FIM, and GFS Stochastic SPP and SPPT in progress Land Surface and coupling to PBL RUC LSM/ MYNN Sfc Layer Real-time green fraction, alternatives to M-O for surface layer Stochastic SPP, SPPT in progress Chaboureau-Bechtold Multi plume approach

Joseph Olson1,2, Jaymes Kenyon1,2, Georg Grell1, John Brown1, Wayne Angevine1,2, Stan Benjamin1, Kay Suselj3

1NOAA’s Earth System Research Laboratory, Boulder, CO 2Cooperative Institute for Research in Environmental Science 3NASA’s Jet Propulsion Laboratory, Pasadena, CA

FY16-17

Development of a scale-aware parameterization of subgrid cloudiness feedback to radiation.

7

Scale-Aware Requirements for a Turbulent Mixing Scheme

Boundary Layer-Cloud Physics Development

1)Reduction of parameterized mixing as dx -> 0. 2)Change in the behavior of the scheme as dx -> 0.

• Mass-flux (shallow-cu) scheme – represent smaller plumes as dx -> 0.
• Eddy Diffusivity scheme – transforms to 3D mixing as dx -> 0.

Subgrid clouds in the MYNN-EDMF Scheme

• Stratus component from partial-condensation scheme within

the eddy diffusivity component.

• Shallow-cumulus component from mass-flux component.

MYNN Boundary Layer Scheme Modifications

8

• 1. Mass-flux component (MYNN-EDMF)

– Dynamic Multi-Plume: dynamic number/sizes of plumes.

• Adapts to different mode grid spacing
• Adapts to growth of PBL.

– Options to transport momentum, TKE, and chemical species. – Option to activate stochastic lateral entrainment rates (Suselj et al. 2013). – Total mixing (mass-flux transport & eddy diffusivity) is solved simultaneously and implicitly (Suselj et al. 2013).

• 2. Subgrid-scale clouds

– Chaboureau and Bechtold (2002 & 2005) convective & stratus components. – Diagnostic-decay method implemented. – Coupled to the radiation schemes.

8

Boundary Layer-Cloud Physics Development

9

Dynamic Multi-Plume Model

Boundary Layer-Cloud Physics Development

Model grid column LCL

10

Dynamic Multi-Plume (DMP)

A) The maximum number of plumes available (Nmax) is determined by the model grid spacing. Max plume width = 0.75*dx

Boundary Layer-Cloud Physics Development

1 2 3 4 5 6 7 8 9 10 (#)

100 200 300 400 500 600 700 800 900 1000 (m) 1 2 3 4 5 6 7 (#)

100 200 300 400 500 600 700 (m)

B) Number of plumes (N) is further limited by the PBLH. For example, at dx = 1000 meters, a maximum of 7 plumes are available, but the number used grows as the PBLH grows:

11

Scale-Aware Tapering of Mass-Flux Scheme

Boundary Layer-Cloud Physics Development

• Taken from Honnert et al. (2011, JAS, their figure 5):

ShCu: TKE in the entrainment layer PBL: TKE in boundary layer

12

Δx = 16 km Δx = 1 km Δx = 2 km Δx = 4 km Δx = 8 km Original; shallow-cumulus scheme activated

Above figure taken from Field et al (2013) – 12 UTC 31 Jan 2010.

Comparison of Original and New Physics

Shortw ave up at TOA

New MYNN-EDMF scheme with subgrid clouds

RAP/HRRR Physics

• Aerosol aware microphysics and radiation need aerosols: Should we really

use an aerosol climatology in the presence of strong aerosol sources?

• Strong sources such as wildfires or dust can decrease SW radiation drastically

as well as change CCN by orders of magnitudes

HRRR-Smoke: VIIRS Fire Radiative Pow er, 3 prognostic aerosols

28-29 Sept 2016

HRRR-Smoke: 3km horizontal resolution, used for aerosol aware

microphysics

2016- HRRR-Smoke will include FRP data from VIIRS and MODIS, Thompson aerosol-aware microphysics (water friendly and ice friendly aerosols), including anthropogenic emissions Direct and indirect effect:

• nly small additional

computer resources needed

Injection layer

Freitas et al., GRL 2006, ACP 2007, 2010

Plumerise in HRRR: The 1-d in-line cloud model: governing equations

• W equation
• U equation
• 1st law of

thermodynamic

• water vapor

conservation

• cloud water

conservation

• rain/ice

conservation

• equation for radius

size

Example of injection height with heat flux of 30 and 80 kW/m2 aaa

aaa

Modeled vertically integrated aerosol concentrations VIIRS AOD

HRRR-Smoke simulated vertically integrated aerosol concentrations and aerosol optical depth from VIIRS for August 27, 2015 VIIRS data also very useful for independent verification!

Quantitative evaluation with retro runs: comparison of two HRRR- smoke retro periods ( 10 days) with and without feedback: RAOB verification over HRRR domain

climatology difference “real” emissions

Temperature BIAS

Surface temperature verification over HRRR domain Ceiling < 3000 ft verification over HRRR domain

SFC TEMPBIAS TSS Skill Score

climatology difference “real” emissions

Example of HRRR-Smoke forecast during 2016 fire season

AUG 19, 00Z Short wave radiation differences for one particular time in comparison to integrated smoke

Summary and future plans for aerosols and microphysics

• 1. With a double moment aerosol aware microphyics scheme only 2 additional variables

are used, including smoke in an operational version of the HRRR with cycling does not degrade the forecast – indications are it might improve forecasts

• 2. Need an extended testing period (1 year) to validate (1)
• 3. Dust and sea salt parameterization should be included
• 4. Add more fire satellite detection data (MODIS, GOES-R) and smoke boundary

conditions in future

• 5. Radiative impact versus microphysics impact

Focus on MYNN PBL

• Parameters
• Mixing length 30%
• Aerodynamic roughness length 30%
• Thermal/moisture roughness length 30%
• Mass fluxes 20%
• Prandtl number limit 2.5 +/- 1 (only for stable conditions)
• Cloud fraction 20%
• Temporal and spatial lengths
• 150km and 6hr
• 300km and 12hr
• 600km and 24hr
• Combination of MYNN PBL SPP with SPPT and SKEB
• 8-members
• 4 cases initialized at 06Z
• Green positive correlation
• Red negative correlation
• Figure presents Spread/Skill for SPP, SPP+SPPT and SPP+SPPT+SKEB

Some early results for using stochastic physics

Overview

• Developments of ESRL’s operational storm scale and regional

scale modeling physics suite (currently using WRF)

– Overview of storm scale and regional scale physics developments – MYNN-EDMF-Shallow convection – Stochastic physics

• Some aspects of global modeling:

– Recent work with the Grell-Freitas convective parameterization – inline chemistry, seasonal forecasting experiments with a couple atmosphere/ocean/chemistry model Global modeling is changing at ESRL: Switch from ESRL model to NGGPS is starting, but results shown here are still with ESRL’s model

IHYCOM: Icosahedral Hybrid Coordinate Ocean Model FIM: Flow-following- finite-volume Icosahedral Model

• Icosahedral

horizontal grid

• Isentropic-sigma

hybrid vertical coordinate

• adaptive in vertical
• concentrates

around frontal zones, tropopause

Different coupling appproach: inline, the two models share the same horizontal grid.

Inline Chemistry – from WRF-Chem

• Seasalt, dust, dms emissions modules from the Goddard Chemistry Aerosol

Radiation and Transport (GOCART) model

• Anthropogenic emissions from the Hemispheric Transport of Air Pollution (HTAP)

project

• Biomass burning are Satellite derived (MODIS), injection height calculated online

with one-dimensional plumerise model

• Simple sulfate and aerosol chemistry from GOCART (more complex available)
• Wet deposition for resolved and non resolved
• Aerosol optical properties are calculated online with MIE calculations for short

wave and longwave RRTMG radiation parameterization

FIM To be replaced with new NGGPS core, once available ! Currently three different chem suites available:

• 1. GOCART
• 2. GOCART + gas-phase

chemistry

• 3. Complex

aerosols+gas-phase + secondary organic aerosols

25

• Momentum transport (as in SAS and/or ECMWF)
• Additional closure for deep convection: Diurnal cycle effect (Bechtold)
• Changed cloud water detrainment treatment
• Mass conserving tracer transport
• Additional closures for shallow convection (Boundary Layer Equilibrium (BLQE,

Raymond 1995; W*, Grant 2001, Heat Engine, Renno and Ingersoll, JAS 1996)

• PDF approach for normalized mass flux profiles was implemented
• Originally to fit LES modeling for shallow convection
• allows easy application of mass conserving stochastic perturbation of vertical

heating and moistening profiles

• Provides smooth vertical profiles
• Latest implements: memory and third type of cloud (mid-level convection)
• Stochastic part in WRF now coupled to Stochastic Parameter Perturbation (SPP),

and Stochastic Kinetic Energy Backscatter (SKEBS) approach (J. Berner )

Recent new implementations into GF scheme

Effect of cloud scale horizontal pressure gradients (Gregory et al. 1997, Zhang and Wu, 2000) is to adjust the in-cloud winds towards those of the large scale flow. For the ECMWF approach (follows Gregory et al., 1997), the entrainment rate is simply adjusted E(u,v)up=Eup +λDup D(u,v)up=Dup +λDup Where E(u,v) and D(u,v) are simply the entrainment/detrainment rates. For SAS approach equations follow directly Zhang and Wu, 2003

• The pressure gradient force across the updraft is proportional to the product of mass

flux and vertical shear of the mean wind,

• Proportionality constant is -.55 for Zhang and Wu,
• Gregory at al at first assumed the constant to be -.7

Both are very simple to implement. Proportionality constant was tested for Stochastic Parameter Perturbation (SPP)

Momentum transport

As in ECMWF, we also include an additional heat source representing dissipation of kinetic energy (Steinheimer et al 2007)

Heat source from momentum transport: dissipation if kinetic energy

Changing the vertical mass flux PDF’s

• Large changes in vertical

redistribution of heat and moisture

• Mass conserving for

stochastic approaches

• significant impact on HAC’s,
• Increases spread for

ensemble data assimilation

PDF1 PDF2 1d version of GF

• nly

Changing momentum transport constants:

• large impact on comparison of

global wind speed biases

• Improving wind bias has

significant impact on HAC’s but does not necessarily improve HAC’s Diurnal Cycle implementation, 120 hour forecasts:

• precipitation averaged over

Amazon basin is improved

• HAC’s little impacted

Impact of momentum transport and diurnal cycle implementation 30 retro FIM runs, about 30km resolution, 120hr forecasts

FIM/IHYCOM sub-seasonal hindcast experiments: 600

• ne month runs (Green et al. 2017)
• *Uncoupled atmosphere-only setup; monthly SSTs from Hadley Centre interpolated to daily.

Experiment name FIM-AGF FIM-CGF FIM-SAS CFSv2 Atmospheric model Dynamic core FIM FIM FIM GFS Horizontal grid (structure, resolution) (Icosahedral, G7 ~60 km) (Icosahedral, G7 ~60 km) (Icosahedral, G7 ~60 km) (Spectral, T126 ~100 km) Vertical grid 64 hybrid σ-θ layers 64 hybrid σ-θ layers 64 hybrid σ-θ layers 64 hybrid σ-p layers Deep conv. scheme Revised GF Revised GF SAS (2015 GFS) SAS (Saha et al. 2010) All other physics 2015 GFS 2015 GFS 2015 GFS Saha et al. (2014) Ocean model Dynamic core None iHYCOM iHYCOM MOM4 Horizontal grid (structure, resolution) N/A (Icosahedral, G7 ~60 km) (Icosahedral, G7 ~60 km) Variable (Saha et al. 2010, pp. 1031-1032) Vertical grid N/A 32 hybrid σ-ρ layers 32 hybrid σ-ρ layers 40 stretched height layers 30

• Top: Bivariate correlation
• Bottom: RMSE and

spread

• Left: RMM; Right: VPM
• Interesting points:

– FIM-AGF much worse than FIM-CGF (and other coupled runs); no surprise, and no more FIM-AGF results will be shown – Higher correlations (more skill) but also higher RMSE (error magnitudes) for RMM than for VPM – FIM-CGF and CFSv2 are comparable in skill and RMSE; FIM-SAS much worse across the board

• Figure 1: Single-model skill and spread

a

31

b c d

IN GFS: Comparisons of surface precipitation rate (24 h avg mm/day) SAS (operational), SAS (imfdepcnv=2), GF (v3a), v3b (tuning experiment)

Global average Average over the Tropics (20S – 20N) schemes Total Convective Convective (land+ocean) Land (Conv) Ocean (conv) Frac (%) (Conv/tot) SAS (op) 2.81 1.53 4.41 2.89 4.87 86.7 SAS (2) 2.90 1.61 4.57 4.58 4.57 87.3 GF (v3a) 2.74 1.07 3.66 2.58 3.99 70.5 GF (v3b, imid=0) 3.05 1.56 5.35 2.49 6.22 86.6 GF (v3b, imid=1) 3.03 1.54 5.29 2.42 6.17 86.5

Experimental example V3a So far best performance

First run un of GF s sch cheme in G n GFS, no no t tuni uning or d data assimilation y

• n yet – 10 d

day f for

• recasts over 3 mon

month pe period

1-day T RMSE 10-day T RMSE

SAS typically better than GFS-GF early in forecast, but GFS-GF better later. Seen in T, RH at surface and upper air

34

• Latest implementations: memory and third type of cloud (mid-level convection)
• Splitting the module into three parts:
• Driver (may be different for various physics suites)
• Module for deep convection (independent of dynamic core or physics suite)
• Module for shallow convection (also independent)
• General clean up of unused arrays, and adding comments

Final changes (not including tuning) over last month

Evaluation happening in regional as well as global models

• n timescales from storm-scale to sub-seasonal

Experimental: aerosol aw areness

Change 2: Modified evaporation of raindrops (Jiang and Feingold) based on empirical relationship Change 1: Change constant autoconversion rate to aerosol (CCN) dependent Berry conversion

Change 2 introduces a proportionality between precipitation efficiency (PE) and total normalized condensate (I1), requiring determination of the proportionality constant Cpr

Saulo Freitas, Arlindo Silva, Angela Benedetti, Georg Grell, Oriol Jorba, Morad Mokhtari, Samuel Remy and many other WGNE Members Participants

Evaluating aerosols impacts on Numerical Weather Prediction

Many questions left to ask: 1. How simple/complex does the chemistry need to be to predict aerosols with enough accuracy 2. How does (1) impact NWP for short, medium, and long range applications 3. Impact versus improvement 4. With NOAA’s Next Generation Global Prediction System (NGGPS) program this was the ideal time to start asking the question of what should be part of a state-

• f-the-art NGGPS modeling system

5. What can we afford with respect to computational requirements?

WGNE comparisons

• Full chemistry run (with

feedbacks minus meteorology only run

• Double moment

microphysics

• Average over 20 runs, 3

days, 12Z T2m differences,

Low AOD: Most of this warming caused by constant droplet number assumption in meteorology only run

Latest aerosol work, regional and global scales

Averaging in areas with significant convection, dx= 1.7km

RNW appeared unpredictable: Convection has different strength For high resolution run: CLW and ICE appear to have a signal T2M, 18Z, Sep 10

Box averaged vertical profile of CLW+ICE Lat = -4.5 to -6.5 Lon -68 to -72

1.E6*kg/kg

PM2.5 (μg/m3)

39

T2M difference fields, September 10, 1200UTC- mid-morning. Positive (red) is warmer compared to MET – simulation with convective parameterization DIR +IND Full chemistry and physics, aerosol indirect explicitly included

DX=5km

Using convective parameterization with and without aerosol awareness 1 run only! Will have to retune GF and run all 20 cases! Direct effect only Average over 20 runs!

Using chemistry and aerosol suites with different complexity: An NGGPS project that started before the dynamic core was known

• Use ESRL’s Flow following finite volume Icosahedral Model (FIM) as dynamic

core place holder

• GFS physics package, except for Grell-Freitas convective parameterization (GF

has capability of wet scavenging, aqueous phase chemistry and aerosol interactions)

• Chemistry suites:

– Simple: bulk aerosols (GOCART) with sectional dust and sea salt – 17 additional prognostic 3d variables – Not so simple: GOCART coupled with gas-phase chemistry (RACM) – 66 additional prognostic variables – Much more complex: RACM and modal aerosols with Secondary Organic Aerosols using Volatility Basis System (VBS) – > 100 additional prognostic 3d variable – Almost non-existent: ice friendly, water friendly aerosols, total pm2.5 – 3 additional prognostic 3d variables (in the works)

Can we even predict aerosols with some confidence: Evaluation of chemical composition with ATom

The Atmospheric Tomography Mission (ATom) will study the impact of human-produced air pollution on greenhouse gases and on chemically reactive gases in the atmosphere. ATom deploys an extensive gas and aerosol payload on the NASA DC-8 aircraft for systematic, global-scale sampling

• f

the atmosphere, profiling continuously from 0.2 to 12 km altitude.

8/15/16 South Atlantic, Punta Arenas to Ascension Is. 8/17/16 Equatorial towards North Atlantic, Ascension Is. to Azores

Preliminary data: comparisons of Aerosol and Gas Tracers between FIM-Chem and ATom

8/15/2016 and 8/17/2016

• The model shows good performance in reproducing the height-latitude profiles of EC and CO at the low

altitude, especially capturing the biomass burning plumes.

• Discrepancies between model predictions and measurements are mainly over the altitude above 4~5km.

EC

Preliminary data Preliminary data

CO

AOD evaluation over longer timeperiods: using AFWA version of GOCART Scheme

Dust Evaluation with data from AERONET

Similar evaluation near biomass burning

Is there an impact of aerosols on NWP? Only direct/semi- direct impact is considered here!

00 Z 12 Z Surface temperature differences Precipitation differences (convective) 00 Z Domain averaged precip and surface temperatures are very slightly lower

mm/day

0C

Is there an impact of the gas-phase chemistry on NWP?

Convective Precipitation differences Surface T differences

mm/day

0C

Future work in global modeling, collaboration with ESRL/CSD, ESRL/PSD, EMC, ARL, and EPA

• HRRR-WRF-ARW for regional storm-scale model – working with

NCEP, NCAR, other labs, switch to FV3 will be tested

• Switch to NGGPS core, FV3.
• Test of aerosol awareness in GF scheme
• Tuning of GF within GFS physics
• Sub-seasonal/seasonal impact of wildfires and aerosols with

coupled atmos/chem/ocean model

• More detailed look at 5 to 10 day height anomaly correlations and

WGNE case for South America

• Feedback to global NWP also with microphysics:

– In addition to GFS physics, this will also run with Thompson aerosol aware microphysics

• Evaluate different dust and sea salt modules

Credits also go to:

Jian-Wen Bao, Sara A. Michelson, Evelyn Grell, Cécile Penland, Stefan Tulich, Phil Pegion Ongoing Research on NWP Model Physics Parameterizations at NOAA/ESRL/PSD

• 1. Microphysical Consistency between Grid-Resolved and

subgrid Cloud Parameterizations at Gray-Zone Resolution

• 2. Coherent 3-D TKE-based subgrid mixing: development of

a scale-adaptive TKE-based subgrid mixing scheme in the WRF model

• 3. Stochastic parameterizations based on observations and

high-resolution simulations

Thank you for your attention! Questions?