Medium Range Forecasting with NCMRWF Unified ModellingSystem

Transcription

Medium Range Forecasting with NCMRWF Unified ModellingSystem
Medium Range Forecasting
with NCMRWF Unified
Modelling System
Presented by
Ashis K. Mitra
[ with inputs from many ncmrwf colleagues]
National Centre for Medium Range Weather Forecasting
A-50, Sector 62, NOIDA-201309, India
www.ncmrwf.gov.in
Outline
1. Operational Set-up
2. Data Usage
3. Monsoon 2015 : Model Bias
4. R&D and new applications
5. Summary
Assimialtion-Forecast System of NCMRWF
Data
Global Observations Reception
SURFACE
from land
stations
GTS
~600mb/dy
RTH, IMD
SHIP
BUOY
24x7
Upper Air
RSRW/
PIBAL
Aircraft
Satellite
NKN
ISRO
(MT)
NKN
NCMRWF
OBSERVATION
PROCESSING
NKN
High Resolution
Satellite Obsn
NKN
Internet (FTP)
Observation
quality
checks &
monitoring
proposed
dedicated
link
Global Model T574L64
UM-N512L70
10 day FCST
Users
IMD
INCOIS
IITM
SASE
Visualisation
Global
Analysis:
(Hy GSI)
GEFS M20
4D VAR UM
T190L28
Initial state
10 day Fcst
RIMES
UM Based
EPS M45
33km
NGFS
4 times a day for
00,06,12,18 UTC
EUMETSAT
Forecast
Models
BARC
Global Fcst
Models:NCUM,
~ 9 Gb/dy
NESDIS
Global Data
Assimilation
Other
sectors
once in a day
for 00 UTC
IC & FCST
IMD
Central/
State GOVT
Public
New
Applications
Wind
energy
Water
Cycle …
ISRO
INCOIS
IITM
value added product
Media
First
Guess
feed back on
observation
qual.
satellite obs.
(Local & Global)
NCMRWF
Numerical Modelling of
Weather & Climate
IC
observations
IC &
FCST
ESSO
Linkage of NCMRWF with Various Organizations
IAF
Sectoral
Users
(Agriculture ,
Aviation ….)
SASE
Ocean and
Fishery
services
Capacity
Building on
NWP
*Being Implemented  NCAOR Real Time Forecasts: Southern Ocean, Antarctic, Arctic
Data assimilation at NCMRWF
• NCMRWF has developed a robust data monitoring mechanism for
all observations that include conventional, satellite and radar.
• It not only helps in maintaining observational network but also helps
in producing good forecasts.
• Observational System Experiments (OSE) to study impact of data
• Assimilation becomes operational if impact of data on forecast
system is neutral or +ve.
Operational DA
UM Short
Forecast
Output
(previous
cycle)
NCUM 4-D Var Data Assimilation System
Background
Error
Observations
Observation
Processing
Obs
Process_Screen
Screen Analysis
(3DVAR)
UM2Jule
s
UM Short
Forecast
Output
(previous
cycle)
ASC
AT
SM
Obs
ASCAT SM
Obs.
Processing
Configure_LS_N1
44
VAR_N320
(4DVAR)
VAR_N144 (4DVAR)
SST &
SeaIce
Analysis
(12 UTC)
Configure_LS_N32
0
Snow
Observatio
n
(12 UTC)
JULES
SST & Sea Ice Data
(analysis) Preparation
Snow
Analysis
Atmospheric
Analysis
EKF based Land
Assimilation
System
Soil
Moisture
Analysis
NCUM Short forecast (17 km)
Types of observations Assimilated at NCMRWF
Observation category
Name of Observation.
Surface
Surface, Mobile, Ship, Buoy (SYNOPs)
Upper air
TEMP (land and marine), PILOT (land and marine),
Dropsonde, Wind profiler
Aircraft
AIREP, AMDAR, ACARS,
Atmospheric Motion Vectors from
Geo-Stationary Satellites
AMV from Meteosat, GOES-11, GOES-13, MTSAT-1R,
MODIS (TERRA and AQUA), POLAR WINDS (NOAA,
MODIS, METOP)
Scatterometer winds
ASCAT winds from METOP satellite
NESDIS / POES ATOVS Sounding
radiance data
1bamua, 1bamub, 1bmhs,1bhirs3, 1bhirs4
Satellite derived Ozone data
NESDIS/POES, METOP-2 and AURA orbital ozone
data
Precipitation Rates
NASA/TRMM (Tropical Rainfall Measuring Mission)
and SSM/I precip. rates
Bending angles from GPSRO
Atmospheric profiles from radio occultation data
using GPS satellites
MeghaTropique
Saphir radiances
NASA/AQUA AIRS & METOP/ IASI
brightness temperature data
IASI,AIRS brightness temperatures
INSAT -3D radiance
Work initiated
Geostationary Satellite Radiances
GOES
Meghatropiques
Oceansat -2
INSAT- 3D
NCMRWF Data Handling system
RMSVD of INSAT-3D AMVs computed against
collocated RS/RW observations
(for April 2014)
Root Mean Square Vector Difference (RMSVD)
Northern
Hemisphere
Tropics
Southern
Hemisphere
Low Level(1000hPa – 700hPa)
INSAT-3D IR
Meteosat-7 IR
3.7
3.8
3.6
3.0
5.3
4.1
Middle Level(700hPa – 400hPa)
INSAT-3D IR
Meteosat-7 IR
5.2
5.3
3.9
2.1
5.3
5.4
High Level(400hPa – 100hPa)
INSAT-3D IR
Meteosat-7 IR
7.46
7.48
5.1
3.9
5.3
7.5
OSE with INSAT-3D AMVs
In first week of March’14
North west India and central India
region received rainfall under the
influence of middle level westerly
trough
Control Run
with INSAT AMV
shows
positive
impact
on Day 2
rain
forecast
Improvement in the quality of INSAT
AMV
Satellite
Year
Retrieval Method
RMSVD
Comparison of INSAT-3D and METEOSAT
AMV DEC 2015
(against
obsn)
METEOSAT7
Jan2011
HA - IR window..
(NWP background)
5m/s
Kalpana-1
Jan2011
HA- GA(NWP
background not
used)
11m/s
Kalpana-1
Aug2011
HA- GA SGP
(navigation
correction)
9m/s
Kalpana-1
Feb2013
HA - IR window..
(NWP background)
7m/s
INSAT-3D
Nov2013
(offline)
HA - IR
window..(NWP
background)
8.5m/s
INSAT-3D
Feb 2014
(offline)
HA - IR window..
(NWP background)
+ Modified QC
8m/s
INSAT-3D
Apr2014
(offline)
HA - IR window..
(NWP background)
+ Modified QC +
Modified processing
5.2m/s
INSAT-3D
Sep2014
(real
,,
( assimilated since
Oct’2014)
4.8 m/s
( Dec’2015
INSAT-3D)
Large
Error
Impact of INSAT-3D AMV on TC Chapala
INSAT-3D AMV Observation
NCMRWF ANALYSIS with INSAT-3D AMV
IR LOW LVL 00Z281015 IR HIGH LVL 00Z281015
Impact on the track of TC Chapala : IC 281015
Verification of Day 03 Fcst against Radiosondes over India (2005-2014)
Root Mean Square Error (RMSE) of 850 hPa winds in m/s
9
8
7
6
RMSEV
5
4
3
2
1
Jan-05
Apr-05
Jul-05
Oct-05
Jan-06
Apr-06
Jul-06
Oct-06
Jan-07
Apr-07
Jul-07
Oct-07
Jan-08
Apr-08
Jul-08
Oct-08
Jan-09
Apr-09
Jul-09
Oct-09
Jan-10
Apr-10
Jul-10
Oct-10
Jan-11
Apr-11
Jul-11
Oct-11
Jan-12
Apr-12
Jul-12
Oct-12
Jan-13
Apr-13
Jul-13
Oct-13
Jan-14
Apr-14
Jul-14
0
The decrease in the RMSE can be attributed to the increase in the resolution
of the model, increase in the amount of data being assimilated, improvements
in data assimilation techniques and model physics.
INSAT-3D Sounder Clear Sky Brightness Temperature (CSBT)
Bias Correction
Assimilation
Bias correction shifted the innovations (O-B) towards
zero: Bias correction working fine.
Observed BT-Model Background
Corrected BT - Model
Background
Impact on radiances from other instruments
Mixed impact of INSAT3D Sounder CSBT on
other radiances.
NWP Index remains
neutral: No Impact
due to the
assimilation of
INSAT-3D CSBT.
Next Operational Upgrade
Preliminary Results
Relative impact of surface observations, Jan 2015
FSO (Forecast Sensitivity to
Observation)
system
is
implemented at NCMRWF.
This
tool enables to study the impact of
various
observations
in
the
forecast
Relative impact of satellite wind obs., Jan2015
Hybrid GSI v/s GSI
Soil Moisture Analysis at NCMRWF
Data Used in the
NCMRWF Soil
Moisture
Analysis
(NCUM)
Data
Assimilation
Method &
Analysis
Resolution
Data Used for
Verification of
the NCUM
analysis
1. Screen level
(surface) air
temperature &
humidity
observations
2. ASCAT
surface soil
wetness
observations
from MetOP-A
satellite (Cband, Level2
product)
Nudging technique 1. AMSR2
satellite
~ 25 km resolution
observations
(global analysis,
(X band)
25 x 25 km grid, 4
2. SMOS satellite
times daily at 00,
observations
06, 12 & 18 UTC at
(L-Band)
4 soil levels)
3. UK Met Office
(Extented Kalaman
soil moisture
Filter based Land
analysis
Data assimilation
4. In-situ soil
system is tested
moisture
successfully. This
observations
is being
of India
implemented in
Meteorological
the operational
Department
NCUM system)
Soil Layers
10 cm
25 cm
65 cm
200 cm
Soil Layers (thickness):
0-10 cm, 10-35 cm,
35-100 cm, 100-300 cm
Soil Moisture Analysis for last week of Dec 2015
Current Operational Global Models
• NCMRWF Global Forecast System (NGFS) T574 with hybrid 3D-Var Data Assimilation (EnKF)
 10 day forecast – at 00 UTC
 3 day forecast -12 UTC for initializing WRF
(RIMES,NPCIL)
• NCMRWF Unified Model
 NCUM (17 km) -10 day forecasts at 00 UTC
Current Operational Regional Models
 4-km NCUM-R (with explicit rain processes) running with
NCUM-G inputs at 00 UTC for 72 hours
 1.5-km NCUM-R (with explicit rain processes) trial runs with
NCUM-G inputs for 72 hours
 3-km WRF runs for 8 NPCIL sites – IC & LBC from 9-km
WRF runs from12 UTC GFS inputs
 9-km-WRF for RIMES domain – Daily running for 3 days
with 12 UTC GFS inputs
Current Operational Global EPS
• NCMRWF Global Ensemble Forecast System
 NGEFS (75 km/21 Members) -10 forecasts at 00 UTC
• NCMRWF Global EPS – based on NCUM (33 km/44
members) -10 Day Forecasts at 00 UTC
(450 nodes -3.5 hrs)
The Multiple Time Scales of Climate
Weather
MJO/monsoon bursts
Time scale
increasing
Annual cycle
El Nino/La Nina
Decadal Variability
Climate Change
Some skill exists (potentially) in Intra-seasonal to interannul prediction in some regions.
Seasonal Fcst systems are crucial in development of adaptation strategies climate change
A host of climate risk management tools are developed for real application
Adapt to climate variability today (seasonal)~will help in adapt to climate change tomorrow
One Model: hours, days, weeks, Monthly, Seasonal, Inter-annual, Decades, Centuary
Consortium: UK, Australia, South Korea, India, , South Africa, New Zealand, (Singapore)
Unified Model at NCMRWF (NCUM)
Same Model for Global/Regional/Mesoscale/Coupled
Seamless Modeling System
1.5-km regional model
up to 72 hr forecasts
4-km regional
model up to 72 hr
forecasts
17-km global model
for 10 Day forecasts
Global Ensemble Prediction
System 33 km, 44 members
up to10 Day Forecasts
(450 nodes -3.5 hrs)
Course Resolution Coupled GS4
2014 :Day-1 to Day-5
Drying Over east coast
and Central India and
west coast
All India Obs/Model rainfall values (cm) JJAS -2015
MODELS
Day-1
Day-3
Day-5
NCUM
89.7
91.9
85.8
NGFS
72.1
73.6
76.3
OBS merged gridded rainfall = 80.3
Correlation coefficient: Observed Vs Model predicted daily rainfall
MODELS
Day-1
Day-3
Day-5
NCUM
0.43
0.38
0.34
NGFS
0.29
0.24
0.23
NCUM: west coast
rain predicted
Highest rain at each grid
2015:Central India & Orissa, AP
reasonable & realistic
4
WIND
Rainfall
WIND
Rainfall
Systematic Error
Global NCUM 17 km
Wet bias
Dry bias
Rainfall exceeding
1mm/day is rainy day
Reduction in the number
of rainy days with
increasing forecast lead
time
Crucial over core
monsoon region
Substantial over the
Peninsula
Difference (FCST-OBS)
in the number of rainy
days
Also substantial over dry
regions of peninsula
Difference (FCST-OBS)
in the number of rainy
days
Also substantial over dry
regions of peninsula
CRA method of rainfall verification
NCUM 29 July 2015
Day 1
Day 2
Day 3
GPM
~25km
NCUM-R
~ 4km
NCUM-G
~17km
NCUM-R
Day 1,2,3 Rainfall (cm/day)
Forecast valid for 16 November 2015
Day 2
Day 3
GPM
~25km
NCUM-R
~ 4km
NCUM-G
~17km
Day 1
Day 1,2,3 Rainfall (cm/day)
Forecast valid for 2 December 2015
Diurnal variations of Rainfall (mm/hr) in July 2015
Domain 4.5N to 40.5, 64.5E to 100.5E
NCUM-Regional 4km
Hours
Spin up seen in regional model
NCUM-Global 17km
Hours
The predicted diurnal cycle of
precipitation peaks too early in global
model
Diurnal variations in July 2015
Central India domain (18N to 28N, 73E to 82E)
Rainfall (mm/hr)
NCUM-Regional 4km
Hours
Spin up seen in regional model
NCUM-Global 17km
Hours
The predicted diurnal cycle of
precipitation peaks too early in global
model
UM-REG (4km) over Indian domain
Convection Experiments
Experiment Name
Details
CNTL
Convection Off
(Cape_TS=900s)
Convection on
(Cape_TS=3600s)
~convection off + New
tropical setting
EXPT1
EXPT7
NCUM-GL
~17KM (PS34)
Convection on + CAPE option
vertical velocity dependency
3.0
Day-2
Day-1
2.5
Day-3
Bias
2.0
1.5
1.0
0.5
0.0
1
2
3
4
5
Rainfall Threshold (cm/day)
6
7
1
2
6
7
1
2
3
4
5
Rainfall Threshold (cm/day)
3
4
5
Rainfall Threshold (cm/day)
6
7
6
7
1
2
3
4
5
Rainfall Threshold (cm/day)
6
7
1.0
0.8
FAR
0.6
0.4
0.2
0.0
1
2
3
4
5
Rainfall Threshold (cm/day)
1
2
3
4
5
Rainfall Threshold (cm/day)
6
7
UM Nesting suit 1.5Km
Day-1
Madhya Pradesh Day-2
(700x450 grid)
IC: 00Z 4 Aug 2014
UM Global vn8.5 (25Km)
Day-3
Nested UM (1.5Km)
Gridded rain analysis
SRTM (90m) orography
used for 1.5Km domain
Obs
NS1.5 km
Global vn8.5
UM Nesting suit 1.5Km
Day-1
Gujarat
(600x450 grid)
Day-2
IC: 00Z 28 Jul 2014
UM Global vn8.5 (25Km)
Day-3
Nested UM (1.5Km)
Gridded rain analysis
SRTM (90m) orography
used for 1.5Km domain
Obs
NS1.5 km
Global vn8.5
Land Cover and land Use
Major changes seen on urban, forest and snow tiles
NRSC/ISRO data in 1.5 km UM
Source CWET(NIWE)
NCMRWF Model wind comparison with MERRA – Aug 2015
Model validation
Wind Speed Time-series comparison with Met Mast at Jaisalmer
(July and Aug 2015)
Model validation:
Wind direction comparison with Met Mast at Jaisalmer
(July and Aug 2015)
NCUM EPS System
• Model resolution –
N400L70
(~33 km at Mid-lat)
• No. of members – 44
• IC Perturbation Method –
ETKF
• Model Perturbation Stochastic physics
EPS Products are now
available in NCMRWF web
site
Day-7
NEPS
Based
On
NCUM
Day-6
Day-5
Ensemble Value added
Products using Model
Climatology
•MOGREPS Operational Rainfall forecasts (Day-1 to Day-7)
•Rainfall Climatology (2007-2015)
•Actual (MOGREPS forecast)
•Normal (MOGREPSDaily Mean)
•Forecast Rainfall Departure
Rainfall Forecast Departure: (EnsAv-Clim)
Forecasts valid for 17th Nov 2015
Rainfall Forecast Departure: (EnsAv-Clim)
Forecasts Valid for 23rd Nov 2015
Rainfall Forecast Departure: (EnsAv-Clim)
Forecasts Valid for 2nd Dec2015
Forecast
rainfall
exceeding
the model
climatology
Rainfall
exceeding
the 90th
percentile of
model
climatology
Rainfall Analysis: Merged Satellite Gauge Data Product (Joint IMD, NCMRWF)
GPM Core Observatory
launched on 27Feb2014
DPR & GMI
Heavy rain
Moderate rain
Light rain
Improved:
Light Rain
Falling snow
Rain microphysics
Daily Merged Rainfall Data
Moving from TRMM
to GPM
Earlier 0.5 x 0.5 grid
Now 0.25 x 0.25 grid
Parallel run Aug 2015
onwards
Will be operational from
01OCT2015
GPM
Gridded Radar Rainfall of Chennai
Radar rain vs. Model obtained Rain
DELHI
Satellite Observations
Microwave
IR
ARG
AWS
Radar
Rain
Gauge
Merged Gridded Rainfall Product
Objective Analysis
Summary:
A high resolution Unified modelling real-time system (Global , Regional) and 4-D
Var data assimilation have been implemented on new HPC during 2015.
A 45 member global (33 km) UM based real-time EPS has been implemented in
Dec 2015
R&D on model particularly for monsoon rainfall has been taken up
Model systematic errors at shorter time scale are important – linked to extended
and seasonal (and beyond) scales.
Model development and DA are crucial for all scales.
Various new applications are developed for end users with high resolution
model output
An ocean-atmosphere coupled model is being implemented for real time use.
Major Modelling centers including ECMWF are moving towards Coupled Modelling
(Seamless from Days-to-Season)  Faster Model Development
NCMRWF Coupled Model
1. Stand alone versions of NEMO, CICE were implemented
Work on sensitivity tests and diagnosis of NEMO for Indian Ocean region
studied
2. A course resolution version of coupled model implemented on IBMP6 (old
HPC), 14 Seasons monsoon hind-cast have been completed with GloSea4
setup. Mean Monsoon circulation is reasonable, has intense monsoon
compared to Obs.
3. As part of the MoES Extended Range Prediction working group NCMRWF
during monsoon 2015 was issuing realtime forecasts to the group based on
UKMO outputs.
4. On new Bhaskar HPC higher resolution GloSea5 is being implemented
Coupled Modelling for MRF/Cyclone, Extended(up to month) , Seasonal Seamless
5. NEMOVar and Coupled NWP by 2016 December
Thank You