Resolution dependence in the Zhang-McFarlane deep ...Resolution dependence in the Zhang-McFarlane...
1
Resolution dependence in the Zhang-McFarlane deep convection parameterization and impact of CAPE calculation William I. Gustafson Jr. 1 , Heng Xiao 1 , Samson M. Hagos 1 , Chien-Min Wu 2 , and Hui Wan 1 1 Atmospheric Sciences and Global Change Division, PNNL; 2 Department of Atmospheric Sciences, National Taiwan University Poster ASR PI Mtg 201503.indd | March 10, 2015 12:26 PM Funding provided by a U.S. Department of Energy Early Career grant to Dr. Gustafson. For more information contact [email protected] or [email protected]. http://www.pnnl.gov/atmospheric/staff/staff_info.asp?staff_num=5716; http://www.researcherid.com/rid/A-7732-2008 Objective To better understand what drives the resolution dependence of quasi-equilibrium (QE) based convection parameterizations by analyzing the subgrid-scale vertical transport of moist static energy, . Methodology Use two CRM simulations to generate meteorological conditions to drive the Zhang-McFarlane convection parameterization (ZM) in an offline, diagnostic mode. 1) Ideal case: a radiative-convective equilibrium configuration using constant Q1 and Q2 forcing; the shear case from Arakawa and Wu (2013, JAS, doi:10.1175/jas-d-12-0330.1). 2) Real case: a 28-day realistic maritime tropical case with time-varying boundary conditions for the AMIE/DYNAMO period. Average the 2-km CRM grid boxes to generate coarser grids ranging from 8×8 to 256×256 km 2 , which are then used to drive ZM. The resulting diagnosed convective transport of moist static energy, ZM , is directly compared to the simulated equivalent from the CRMs, CRM . November 2011 Precipitation rate (mm h -1 ) REAL TRMM obs. IDEAL mean Compare and w ' h ' CRM w ' h ' ZM CRM simula1ons (2x2 km 2 resolu1on , 10 or 20 minute snapshots) Evaluate gridmean state variables and tendencies due to gridscale advec1on, radia1on and PBL processes at different spa1al resolu1ons (Δx = 8, 16, 32, 64, 128, 256 km) Calculate unresolved ver1cal transport of moist sta1c energy ( ) on different spa1al resolu1ons (Δx = 8, 16, 32, 64, 128, 256 km) w ' h ' CRM Calculate (dA/dt) g and F for each sub domain Calculate M b and parameterized convec1ve transport of moist sta1c energy for each subdomain ( ) w ' h ' ZM Predictive capability of QE declines with increasing resolution 256 km 32 km 64 km 16 km 8 km r=0.890 r=0.888 r=0.866 r=0.831 r=0.739 r=0.535 w’h’ ZM at 6 km (K m/s) w’h’ CRM at 6 km (K m/s) 128 km PDF (%) Correlation decreases with increasing resolution CRM-Predicted vs. ZM-Diagnosed Convective MSE Flux for Various ∆x Implications: parameterizing convection QE-based closures need to consider scales large enough to encompase a range of cloud scales or else their predictive capabilities decline. Local convective calculations that include non-local, larger scale information improve the ability to predict the overall convective state across a range of resolutions. This implies closures should be aware of regions larger than one column. Details for scales smaller than the QE closure region could be treated stochastically. Xiao, H., W. I. Gustafson, S. M. Hagos, C.-M. Wu, and H. Wan, 2015: Resolution-dependent behavior of subgrid-scale vertical transport in the Zhang-McFarlane convection parameteriza- tion, J. Adv. Model. Earth Sys., accepted. The CRM is a proxy for realistic convective motions to compare against the diagnostic ZM calculated grid-scale values of CAPE tendency, (dA/dt) g , and convective MSE flux. The ZM closure only uses positive CAPE tendencies to determine convective strength. A diagnostic equivalent connecting the CAPE tendency to the cloud base mass flux, M b , is Resolution dependence of grid-scale CAPE tendency depends on convective intensity (quartiles in plots). This directly leads to the resolution-dependent and convective-intensity-dependent convective MSE flux. normalized value Log ΔX 2 <w’h’> ZM at 6 km <(dA/dt) > g+ IDEAL REAL total REAL quartile 4 REAL quartile 3 REAL quartile 2 REAL quartile 1 Log ΔX 2 Increasing convective intensity Normalized by value at ∆x = 128 km CAPE Tendency for Default Closure Convective MSE Flux w/ Default Closure Reality sometimes has negative CAPE tendencies in the presence of convection, which alters the overall resolution depenence. Including negative CAPE tendencies in the ensemble results in similar resolution depen- dence across convective intensities. The resulting behavior mimics the CRM resolu- tion dependence across convective intensities. Resolution dependence can be improved by modifying the closure to average over an area where QE holds, i.e., where negative CAPE tendencies cannot dominate the ensemble. normalized value Log ΔX 2 Log ΔX 2 <w’h’ > CRM at 6 km <(dA/dt) > g IDEAL REAL total REAL quartile 4 REAL quartile 3 REAL quartile 2 REAL quartile 1 CAPE Tendency Including Negative Values “True” Convective MSE Flux from CRM Closure assumption impacts the resolution dependent behavior normalized value Log ΔX 2 < w’h’ > ZM,m at 6 km IDEAL REAL total REAL quartile 4 REAL quartile 3 REAL quartile 2 REAL quartile 1 Convective MSE Flux w/ Modified Closure
Transcript of Resolution dependence in the Zhang-McFarlane deep ...Resolution dependence in the Zhang-McFarlane...
Resolution dependence in the Zhang-McFarlane deep convection parameterization and impact of CAPE calculationWilliam I. Gustafson Jr.1, Heng Xiao1, Samson M. Hagos1, Chien-Min Wu2, and Hui Wan1
1Atmospheric Sciences and Global Change Division, PNNL; 2Department of Atmospheric Sciences, National Taiwan University
Post
er A
SR P
I Mtg
201
503.
indd
| M
arch
10,
201
5 12
:26
PM
Funding provided by a U.S. Department of Energy Early Career grant to Dr. Gustafson.
For more information contact [email protected] or [email protected]. http://www.pnnl.gov/atmospheric/staff/staff_info.asp?staff_num=5716; http://www.researcherid.com/rid/A-7732-2008
ObjectiveTo better understand what drives the resolution dependence of quasi-equilibrium (QE) based convection parameterizations by analyzing the subgrid-scale vertical transport of moist static energy, .
MethodologyUse two CRM simulations to generate meteorological conditions to drive the Zhang-McFarlane convection parameterization (ZM) in an offline, diagnostic mode.
1) Ideal case: a radiative-convective equilibrium configuration using constant Q1 and Q2 forcing; the shear case from Arakawa and Wu (2013, JAS, doi:10.1175/jas-d-12-0330.1).
2) Real case: a 28-day realistic maritime tropical case with time-varying boundary conditions for the AMIE/DYNAMO period.
Average the 2-km CRM grid boxes to generate coarser grids ranging from 8×8 to 256×256 km2, which are then used to drive ZM. The resulting diagnosed convective transport of moist static energy, ZM, is directly compared to the simulated equivalent from the CRMs, CRM.
November 2011
Prec
ipita
tion
rate
(mm
h-1)
REALTRMM obs.IDEAL mean
Compare and w 'h 'CRM w 'h 'ZM
CRM simula1ons (2x2 km2 resolu1on , 10 or 20 minute snapshots)
Evaluate grid-‐mean state variables and tendencies due to grid-‐scale advec1on, radia1on and PBL processes at different
spa1al resolu1ons (Δx = 8, 16, 32, 64, 128, 256 km)
Calculate unresolved ver1cal transport of moist sta1c energy ( ) on different spa1al
resolu1ons (Δx = 8, 16, 32, 64, 128, 256 km)
w 'h 'CRM
Calculate (dA/dt)g and F for each sub-‐domain
Calculate Mb and parameterized convec1ve transport of moist sta1c energy for each sub-‐domain ( ) w 'h 'ZM
Predictive capability of QE declines with increasing resolution
256 km
32 km
64 km
16 km 8 km
r=0.890 r=0.888 r=0.866
r=0.831 r=0.739 r=0.535
w’h’ZM at 6 km (K m/s)
w’h
’ CRM
at 6
km
(K m
/s)
128 km
PDF (%)
Correlation decreaseswith increasing resolution
CRM-Predicted vs. ZM-Diagnosed Convective MSE Flux for Various ∆x
Implications: parameterizing convectionQE-based closures need to consider scales large enough to encompase a range of cloud scales or else their predictive capabilities decline.
Local convective calculations that include non-local, larger scale information improve the ability to predict the overall convective state across a range of resolutions. This implies closures should be aware of regions larger than one column.
Details for scales smaller than the QE closure region could be treated stochastically.
Xiao, H., W. I. Gustafson, S. M. Hagos, C.-M. Wu, and H. Wan, 2015: Resolution-dependent behavior of subgrid-scale vertical transport in the Zhang-McFarlane convection parameteriza-tion, J. Adv. Model. Earth Sys., accepted.
The CRM is a proxy for realistic convective motions to compare against the diagnostic ZM calculated grid-scale values of CAPE tendency, (dA/dt)g, and convective MSE flux. The ZM closure only uses positive CAPE tendencies to determine convective strength. A diagnostic equivalent connecting the CAPE tendency to the cloud base mass flux, Mb, is
Resolution dependence of grid-scale CAPE tendency depends on convective intensity (quartiles in plots).
This directly leads to the resolution-dependent and convective-intensity-dependent convective MSE flux.
norm
aliz
ed v
alue
Log ∆X2
<w’h’>ZM at 6 km<(dA/dt) >g+
IDEALREAL total
REAL quartile 4REAL quartile 3REAL quartile 2REAL quartile 1
Log ∆X2
Increasingconvective
intensity Normalized byvalue at ∆x = 128 km
CAPE Tendency for Default Closure Convective MSE Flux w/ Default Closure
Reality sometimes has negative CAPE tendencies in the presence of convection, which alters the overall resolution depenence.
Including negative CAPE tendencies in the ensemble results in similar resolution depen-dence across convective intensities.
The resulting behavior mimics the CRM resolu-tion dependence across convective intensities.
Resolution dependence can be improved by modifying the closure to average over an area where QE holds, i.e., where negative CAPE tendencies cannot dominate the ensemble.
norm
aliz
ed v
alue
Log ∆X2Log ∆X2
<w’h’ >CRM at 6 km<(dA/dt) >g
IDEALREAL total
REAL quartile 4REAL quartile 3REAL quartile 2REAL quartile 1
CAPE Tendency Including Negative Values “True” Convective MSE Flux from CRM
Closure assumption impacts the resolution dependent behavior
norm
aliz
ed v
alue
Log ∆X2
< w’h’ >ZM,m at 6 km
IDEALREAL total
REAL quartile 4REAL quartile 3REAL quartile 2REAL quartile 1
Convective MSE Flux w/ Modified Closure
