Embed Size (px)
DESCRIPTION
DRAFT – Page 3 – January 14, 2016 Validation of the HR-EnKF system: Single Observation test (Analysis step) Initial guess at 2010 July/22/0000 UTC Ensemble mean: Temperature (degree) Given single observation : temperature at grid point (150,150) around 850hPa Innovation : 1.0 deg : 1 deg from HP f H T : 0.57 deg
Citation preview
DRAFT – Page 1 – May 3, 2023
Development of a Convective Scale Ensemble Kalman Filter at Environment Canada
Luc Fillion1, Kao-Shen Chung1, Monique Tanguay1
Weiguang Chang2
1. Meteorological Research Division, Environment Canada 2. Dept of Atmospheric and Oceanic Sciences, McGill University
DRAFT – Page 2 – May 3, 2023DRAFT – Page 2 – May 3, 2023
Initial guess Ensemble members
Add random perturbations
Data assimilation
ObservationPerturbed observations
GEM-LAM forecast for all the members.
Add random perturbations(model error)
Analysis step
Forecast step
High Resolution Ensemble Kalman Filter System ( HR-EnKF )
A: LAM15B: LAM2p5 C: LAM1 300x300 (MTL region)
DRAFT – Page 3 – May 3, 2023DRAFT – Page 3 – May 3, 2023
Validation of the HR-EnKF system:Single Observation test
(Analysis step)
Initial guess at2010 July/22/0000 UTC
Ensemble mean: Temperature (degree)
Given single observation : temperature at grid point (150,150) around 850hPa
Innovation : 1.0 deg : 1 deg from HPfHT : 0.57 degob
DRAFT – Page 4 – May 3, 2023DRAFT – Page 4 – May 3, 2023
Horizontal Correlations of initial perturbations (80 members)
Temperature (degree) 850hPa
With limited members:Localization is needed
Perturbations are:Homogeneous & Isotropic
DRAFT – Page 5 – May 3, 2023DRAFT – Page 5 – May 3, 2023
: 1 from HPfHT : 0.57bo
25.0
)57.0(11
1)0.1(1
1)0.1(
)(
2
f
oof
f
ffa
xzKxx
Increment: Xa-Xf
Localization radius(60 km)
0.2479
Analysis step
DRAFT – Page 6 – May 3, 2023DRAFT – Page 6 – May 3, 2023
Flow dependent single observation test
Analysis step(single obs)
Forecast step( 30-min )
Analysis step(single obs)
Temperature analysis increment
Innovation : 1.0 degree : 1 degreeo
DRAFT – Page 7 – May 3, 2023DRAFT – Page 7 – May 3, 2023
The performance of ensemble predictions
Current set up
1. Initial perturbations: U, V, T, HU, TG and P0
2. Do not consider the model errors
3. No perturbations in hydrometeor variables
4. Cycling hydrometeor variables
The forecasting error at mesoscale
DRAFT – Page 8 – May 3, 2023DRAFT – Page 8 – May 3, 2023
Microphysical scheme: Milbrandt and Yau (double moment scheme)
QB ( cloud mixing ratio )
QL ( rain mixing ratio )
QN ( snow mixing ratio )
QI ( ice mixing ratio )
QJ ( graupel mixing ratio )
QH ( hail mixing ratio )
DRAFT – Page 9 – May 3, 2023DRAFT – Page 9 – May 3, 2023
Canada/U.S. Radar Reflectivity
0030 UTC 0130 UTC
0230 UTC 0330 UTC
Case Study: 2010 July 22
DRAFT – Page 10 – May 3, 2023DRAFT – Page 10 – May 3, 2023GEM-LAM 1-km Precipitation GEM-LAM 2.5km
Radar observations (reflectivity)11μm (observes the temperature of clouds, land and sea surface)
DRAFT – Page 11 – May 3, 2023DRAFT – Page 11 – May 3, 2023
15-min Forecast Error Correlations (800mb)U V
T HU
precipitation
DRAFT – Page 12 – May 3, 2023DRAFT – Page 12 – May 3, 2023
U V
T HU
precipitation
30-min Forecast Error Correlations (800mb)
DRAFT – Page 13 – May 3, 2023DRAFT – Page 13 – May 3, 2023
21 24
16
11
6 7 10
1
Sub-7 Sub-10
Sub-24
Error correlation in vertical (30-min forecast)
Single Obs. test
DRAFT – Page 14 – May 3, 2023DRAFT – Page 14 – May 3, 2023
400mb 600mb
800mb
(dynamics)T_tendency(physics)T_tendency
Ratio
physics versus dynamics
Physical processes could be as important as dynamics.
DRAFT – Page 15 – May 3, 2023DRAFT – Page 15 – May 3, 2023
Profile of single observation test En_KF T analysis increment
Ensemble mean of physical temperature tendency
DRAFT – Page 16 – May 3, 2023DRAFT – Page 16 – May 3, 2023
Time step = 0
21 24
16
11
6 7 10
1
Time step = 2
Sub-24
Sub-10
DRAFT – Page 17 – May 3, 2023DRAFT – Page 17 – May 3, 2023
Error correlation of TT profile V.S. Vertical correlation of TT tendency ( Ensemble Forecasts) (stochastic perturbation of SCM)
DRAFT – Page 18 – May 3, 2023DRAFT – Page 18 – May 3, 2023
21
16
11
6
1Cloud mixing ratio(600mb)
PR
DRAFT – Page 19 – May 3, 2023DRAFT – Page 19 – May 3, 2023
Summary and Discussion of the next steps
1. The EnKF system has been modified from global to local area
2. The single observation validation is done
3. The results from ensemble forecasts (errors) showed strong flow dependency and revealed the importance of physical processes over precipitation areas
4. Ready to assimilate radar observations (radial winds)
The forward model (observation operator) of Doppler wind
5. Currently, McGill radar group provides us 15-20 cases to study
Vr u xr v yr (w Vt )
zr
DRAFT – Page 20 – May 3, 2023DRAFT – Page 20 – May 3, 2023
Comments and DiscussionsComments and Discussions
DRAFT – Page 21 – May 3, 2023DRAFT – Page 21 – May 3, 2023
Summer case: July / 09 / 2010Summer case: July / 21 / 2010
REF
DOP
(elv.#4)
DRAFT – Page 22 – May 3, 2023DRAFT – Page 22 – May 3, 2023
Winter case: Feb / 05 / 2011Winter case: Dec. / 12 / 2010
REF
DOP
(elv.#4)
DRAFT – Page 23 – May 3, 2023DRAFT – Page 23 – May 3, 2023
Temperature increment vertical cross-section
DRAFT – Page 24 – May 3, 2023DRAFT – Page 24 – May 3, 2023
Features of the system
Sequential processing of batches of observations
DRAFT – Page 25 – May 3, 2023DRAFT – Page 25 – May 3, 2023
Sub-ensemble 1
Sub-ensemble 2
Sub-ensemble 3
Ensemble members (80)
Sub-ensemble 4
Gain matrix K1K1 Gain matrix K2K2 Gain matrix K3K3 Gain matrix KK
Partitioning the ensemble
DRAFT – Page 26 – May 3, 2023DRAFT – Page 26 – May 3, 2023
RegGEM15 forecast
LAM15 forecast
LAM2.5 forecast
12 UTC 00 UTC 12 UTC 18 UTC
Model configuration: Optimal Nested scheme
Operational model output
T+3030-h run
IC + LBC
18 UTC
6-h
Spin-up
12-h run 6-h
Spin-up 00 UTC LAM1 forecast
6h run
T+12
T+6
Archive output :1. Control run2. Prepare for EnKF test
