Transcript DET_Module1

DET Module 1
Ensemble Configuration
Linda Wharton1, Paula McCaslin1, Tara Jensen2
1
NOAA/GSD, Boulder, CO
2
NCAR/RAL, Boulder, CO
4/10/2016
Motivation

National Mesoscale Probabilistic
Prediction Workshop (September 2009)

Science issues: which ensemble is better?
o
o
o
o
Few members on high-resolution grid
Many members on coarse-resolution grid
Single model with perturbations or multiple
models
Choice of analysis around which
perturbations are made
Ensemble Configuration
Define the configuration (i.e. namelists and
scripts) to setup and run ensembles for
optimal results including best options of:
o
o
o
o
Grid resolution
Number of members
Perturbations
Etc.
Goals

Flexibility
o

Portability
o

A system that can emulate ensembles like
those run at GSD, NCAR, AFWA, EMC, CAPS,
University of WA, FSU, RENCI, etc.
Multiple computing platforms to perform
testing and evaluation
Transfer to operations
o
o
Work toward NEMS compatibility
First R2O: NCEP operations
Basic Infrastructure
Begin with community codes DTC
already supports
 Define requirements:

o
o
o
o

Domains
Horizontal and vertical spacing
Output formats
Output fields
Provide scripts to setup and run an
ensemble so that testing may begin to
address science issues
Configuration Module – Overview
Module 1
Configuration Module
Run Phase
Member 1
Member 2
Setup Phase
Member 3
Writes
namelists
& configs
WPS
namelist
Sets Paths
WRF
namelist
WPP
namelist
etc…
MET
configs
Member 4
Generates
Static Files
Member N
Configuration Module - Setup Phase
Module 1
Configuration Module
Writes Workflow Manager XML
Config File
specifies
module options
Writes
namelists
& configs
Sets Paths
Generates
Static Files
Module 2 – IC/BC Perturbations
Option 1: no perturbations
Option 2: ETKF
Option 3: Simple Interpolation
etc…
Module 3 – Physics Perturbations
Option 1: no perturbations
Option 2: Different physics with each member
Option 3: Stochastic perturbations etc…
Module 4 – Statistical Post-Processing
Option 1: no post-processing
Option 2: Bias correction
Option 3: Bayesian etc…
Module 5 – Products/Display
WPS
namelist
WRF
namelist
WPP
namelist
etc…
Option 1: no products
Option 2: mean, spread, min, max
Option 3: Probability Matching etc…
Module 6 - Verification
MET
configs
Option 1: no verification
Option 2: Traditional Statistics
Option 3: Spatial Statistics
etc…
Configuration Module - Run Phase
Module 1
Configuration Module
Module 2
IC/BC Perturbations
Runs Workflow Manager
Module 3
Physics Perturbations
Module 4
Statistical
Post-Processing
Module 5
Products/
Display
Module 6
Verification
Examples of Existing Ensembles

Operational systems
o
o

NCEP Short Range Ensemble Forecast
AFWA Joint Ensemble Forecast System
Research systems
o
o
o
o
o
ESRL Hydrometeorology Testbed Ensemble
CAPS Storm Scale Ensemble Forecast (HWT)
NCAR Joint Mesoscale Ensemble
University of Washington Ensemble
RENCI Ensemble
Configure Scripts to Consider
Workflow Manager
 JME configuration scripts
 WRF Portal / Domain Wizard GUIs
 CAPS management scripts (K. Thomas)
 NCAR-RTFDDA Ensemble Member
Selection Tool

First Year Deliverables
Basic Infrastructure
 Run ensemble
 Answer scientific questions…

Questions
Are there any ensemble systems that we
should also include?
 If so, who is a contact person?
