The Joint Ensemble Forecast System

1
The Joint Ensemble Forecast System

• JEFS

GOAL Prove the value, utility, and operational
feasibility of EF to DoD operations. FOCUS How
to best exploit EF output within forecasting and
decision processes.
Maj Tony Eckel, PhD Ensemble Projects Manager HQ
Air Force Weather Agency
Dr. Michael Sestak Meteorologist, Ensemble
Models Fleet Numerical Meteorology and
Oceanography Center
2
Joint Global Ensemble (JGE)

• Description Combination of current GFS and
  NOGAPS global, medium-range
• ensemble data. Possible
  expansion to include ensembles from CMC,
• UKMET, JMA, etc.
• Initial Conditions Ensemble Transform (GFS)
  and Breeding Modes1 (NOGAPS)
• Model Variations/Perturbations Two unique
  models, but no model perturbations
• Model Window Global
• Grid Spacing 1.0?? 1.0? (80 km)
• Number of Members 46 at 00Z
• 30 at 12Z
• Forecast Length/Interval 10 days/6 hours
• Timing
• Cycle Times 00Z and 12Z

1 Toth, Zoltan, and Eugenia Kalnay, 1997
Ensemble Forecasting at NCEP and the Breeding
Method. Monthly Weather Review Vol. 125, No.
12, pp. 32973319.
3
Joint Mesoscale Ensemble (JME)

• Description Multiple high resolution,
  mesoscale model runs at FNMOC and AFWA
• Initial Conditions Ensemble Transform Filter 2
  run on short-range (6-h),
• mesoscale data
  assimilation cycle driven by GFS and NOGAPS
• ensemble
  members
• Model variations/perturbations
• Multimodel WRF-ARW, COAMPS
• Varied-model various configurations of physics
  packages
• Perturbed-model perturbed surface boundary
  conditions (e.g., SST)
• Model Window East Asia (COPC directive, Apr
  04)
• Grid Spacing 15 km for baseline JME (summer
  06)
• 5 km nest (in summer 07)
• Number of Members 20-30 (½ AFWA, ½ FNMOC)
• Forecast Length/Interval 60 hours/3 hours

5 h production /cycle
2 Wang, Xuguang, and Craig H. Bishop, 2003 A
Comparison of Breeding and Ensemble Transform
Kalman Filter Ensemble Forecast Schemes. Journal
of the Atmospheric Sciences Vol. 60, No. 9, pp.
11401158.
4
JEFS Design
FNMOC
AFWA
lateral boundary conditions
analysis and short range forecasts
raw
post
Data Assimilation 3DVAR / NAVDAS
Joint Global Ensemble (JGE) Products ? Apply
postprocessing calibration ? Long-range products
tailored to support warfighter planning
Calibrate
Ensemble Transform ? Generate/Apply initial
condition perturbations
Observations and Analyses
6-h update cycle (32 members)
Calibrate
initial conditions

• Joint Mesoscale Ensemble (JME)
• ? 20 members, 15/5km, 60 h, 2/day
• ? One demonstration theater
• Multi model (WRF, COAMPS)
• Perturbed model varied physics
• and surface boundary conditions

post
raw

• JME Products
• ? Apply postprocessing calibration
• Short-range products tailored to
• support warfighter operations

5
JEFSTesting and Evaluation
26 OWS
6
This ensemble stuff is useless! Last time they
said 80 chance of snow we didnt get a
flurry This time they said only 40!
7
Application Tools Must Bridge the Gap
Stochastic Forecast
Binary Decisions/Actions
AR Route Clear 7
Go / No Go
T-Storm Within 5
?
IFR / VFR
GPS Scintillation
Bombs on Target
Crosswinds In / Out of Limits
Flight Hazards
8
IWEDA Today vs. WRAP--- Weather Risk Analysis
and Portrayal ---
MISSION Static line training airdrop Surface
Wind Thresholds Red gt 13 kt Yellow
10-13 kt Green 0-9 kt
IWEDA (deterministic)
WRAP (stochastic)
Wx Risk Tolerance N/A Forecast 12
kt Decision Input Yellow Decision
GO Observed 19 kt Result 13 injured personnel
Wx Risk Tolerance Low (10) Forecast 12 kt,
range 5-26 kt G (25),
Y(33), R(42) Decision Input Red Decision
NO GO Observed 19 kt Result training delayed
9
Notional Transition Plan for EF Production

• Prototype Joint Ensemble Forecast System
  (JEFS)
• RD
• Field Test
• Operational Ensemble Prediction System (EPS)
• Global Mesoscale Ensemble
• CONUS Fine-scale Ensemble
• OCONUS Fine-scale Ensemble(s)

JGE JME
17 OWS, 26 OWS
20 members from NCEP, goal of 25km
20 members from NCEP, goal of 10km
Incremental by theaters, goal of 15/5km
2007 2008 2009 2010 2011 2012 2013
Risk
Low
Moderate
High
10
JGE Ground Truthfor Verification and Calibration

• UKMet Unified Model, Global Analysis
• 0000Z and 1200Z
• 3D variational data assimilation, using 3h obs
  window
• Horizontal grid (432x325 grid), resolution
  60km
• Longitude 5/6 (0.8333)
• Latitude 5/9 (0.5555)

11
Rank Histograms
700HPa GPH00Z forecasts72h forecastJGE
members (30 GFS 16 NOGAPS)19 Aug 11 Oct 2006

Polar Region (60-90N)
Mid-latitude Region (30-60N)
Tropical Region (0-30N)
12
Rank Histograms
6h cumulative precipitation00Z forecasts78h
forecast lead timeJGE members (30 GFS 16
NOGAPS)1 Sep 11 Oct 2006
Polar Region (60-90N)
Mid-latitude Region (30-60N)
Tropical Region (0-30N)
13
Reliability Diagrams
Probability 10m wind gt 25kt, 35kt00Z
forecasts0-120h forecast lead timeJGE members
(30 GFS 16 NOGAPS)19 Aug Sep 11 Oct 2006
14
Backup Slides
15
draft input to C-Plan on ORM in AFW Operations
Initial Assessment
Potential for mission Impacting wx?
No
Standard Weather Information (may include
uncertainty)
Yes
Risk Analysis
Does wx risk outweigh mission acceptable risk?
No
Yes
Forecast Refinement
Able to reduce wx risk below mission acceptable
risk?
Refined Weather Information (may include
uncertainty)
Yes
Updated, down-scaled weather information
No
Mission Adjustment
ORM Capable Customer?
Stochastic Weather Mitigation
Deterministic Weather Mitigation
Yes
No
16
Example 2010, Forecaster/Automated
Initial Assessment Risk Analysis Fcst.
Refinement Mission Adjustment
Mission Oriented Protective Posture (MOPP)
Operating Threshold 35?C (95?F)
17
Sample JGE Product Probability Plot
Probability of Surface Winds ? 35 kt
0.1 10 30 50
70 90
18
Sample JME Product Multimeteogram
Misawa AB, Japan
AFWA Forecast Multimeteogram JME Cycle 11Nov06,
18Z
RWY 100/280 15km Resolution
Wind Speed (kt)
11/18 12/00 06 12 18
13/00 06 12 18
14/00 06
Valid Time (UTC)
Wind Direction
19
Sample JME Product Probagram (Probabilistic
Meteogram)
Probability of Weather Warning Criteria at Osan
AB, Korea
Valid Time (Z)
20
Weather Risk Analysis and Portrayal (WRAP)
Given uncertain weather, lower acceptable risk
means less certainty in the desired outcome.
Stochastic Weather Data

• WRAP
• ORM calculations
• Stop-light style
• decision inputs
• Portrayal Techniques

Warfighter Risk Tolerance
IWEDA Rules
21
WRAP Decision Input Processing
Acceptable Risk Decision Input
Marginal
Unfavorable
Drop Zone 1
Low Med High Low Med High Low Med High
99
1
0
(Accept 10)
(Accept 40)
(Accept 70)
Surface Wind Speed
5 10 15 kt
Drop Zone 2
1
31
68
5 10 15 kt
Drop Zone 3
37
52
11
5 10 15 kt
22
Long Term EF Vision (2020)
AFWA
Microscale Ensembles Runs/Cycle O(10)
Resolution O(100m) Length 24 hours
Global Mesoscale Ensemble Runs/Cycle O(10)
Resolution O(10km) Length 10 days
United Global Mesoscale Ensemble
Runs/Cycle O(100) Resolution O(10km)
Length 10 days
Microscale Ensemble Runs/Cycle O(10)
Resolution O(100m) Length 2 days
Global Mesoscale Ensemble Runs/Cycle O(10)
Resolution O(10km) Length 15 days
Coalition Weather Centers Global Mesoscale
Ensembles
etc.
Global Mesoscale Ensemble Runs/Cycle O(10)
Resolution O(10km) Length 10 days
Microscale Ensembles Runs/Cycle O(10)
Resolution O(100m) Length 24 hours
MSC
JMA
ABM
FNMOC
23
Notional Requirements for NWS Ensemble Data (by
2010)
CONUS (WRF) Limited Area Ensemble Global Mesoscale Ensemble
Model Domain 1) CONUS (20N-55N, 135W-60W 2) Alaska (55N-73N, 170W-130W) Global
Grid Spacing 10 km 25 km or spectral equivalent (T565) with step down resolution beyond 48-h forecast lead time
Number of Levels 60 40
Ensemble Size 20 Members 20 Members
Cycle Frequency 4 per day 4 per day out to 72 h, with 2 of those cycles out to 240 h
Forecast Length/Interval 48 h / 3 h 72 h 240 h / 6 h
Delivery Schedule Incremental delivery starting NLT 3h after initialization time, complete NLT 5h after initialization time same
Analysis Perturbations Robust initial conditions (Ex Ensemble Transform Kalman Filter) same
Model Perturbations Robust accounting for model uncertainty using single model framework with multiple physics combinations, physics perturbations, and/or stochastic physics. same
Calibration Robust correction for systematic errors, both 1st moment (bias correction) and 2nd moment (spread correction) of the ensemble distribution  same
24
EPS Spatial Coverage (draft)

SWA 10/3km
Alaska 10km (NCEP)
CONUS 10km (NCEP)


East Asia 10/3km






GLOBE 25km (NCEP)
FOUO
Possible Contingency Window
25
EPS Temporal Coverage (draft)
00 06 12 18 24 36 48 60 72 84 96 h
10 days
Global 25km (NCEP)
50km
Global 25km
10 days
Global 25km
50km
Global 25km
CONUS 10km (NCEP)
CONUS 10km
CONUS 10km
CONUS 10km
OCONUS Windows 15/5km (AFWA)
OCONUS Windows 15/5km
OCONUS Windows 15/5km
OCONUS Windows 15/5km