S'K' Roy Bhowmik

1
Regional NWP Modelling at IMD

• S.K. Roy Bhowmik
• NWP, IMD, New Delhi

2
Overview
3
Assimilation of Doppler Weather RADAR (DWR)
Observation

• Processing for Nowcasting Applications
• Ingest into assimilation cycle of NWP models

Parameters radial wind, reflectivity and
spectrum width DWR Stations Chennai,
Machalipatnam, Vishakapatnam and Kolkata,
Sriharikota (ISRO)
4
NHEC (Telecom server,IMD)
DWR Net-work for data processing
NCMRWF
5
A RADAR mosaic creation from reflectivity
observations DWR Chennai and Machhilipatnam of
28 September 2005
6
Nov08 Ccyclone Khai Muk
A RADAR mosaic creation from reflectivity
observations
Well marked on 13 Nov 2008 over south Bay of
Bengal, concentrated into a depression in the
evening. Moved in a northwesterly direction,
intensified into a intensified into a cyclonic
storm, Khai Muk It reached its maximum
intensity near near lat. 14.5 N and long. 83.0
E around 0230 hours IST of 15th with estimated
sustained maximum wind speed of 40 knots and
estimated central pressure of 994 hpa.
7
Example of the 30 minute rainfall (mm) estimates
for the rainstorm of 2 September 2005 from a
single DWR at Chennai and the corresponding
automatic rain gauges used to validate the data.
8
(No Transcript)
9
Numerical experiments for assimilation of DWR
(radial wind and reflectivity) data of Chennai
with ARPS model for cyclone Ogni of October 2006
Cyclone Ogni of Oct06
10
Background and boundary values from GFS model
into the ARPS grid. The Diagram is showing ½
hourly assimilation cycle ( first 3 hours) then
21 hours ARPS Model forecast -
11

• 88D2ARPS -
• Doppler Weather Radar data in up8 format has
  been collected from IMD Chennai.
• DWR data in up8 format has been converted in to
  netcdf format.
• 88D2ARPS has been used for remapping IMD Chennai
  radar data ( Radial Wind
• Reflectivity) to a Cartesian grid .
• For three hourly Data assimilation , half hourly
  Radar data file containing both reflectivity
  radial velocity, starting from 00 UTC, has been
  generated. Ex -
• Name of file
  Time duration of Data
• (1) IMDS.20061029.0004 2346 UTC
  28-10- 2006 TO 0015 UTC 29-10-2006
• (2) IMDS.20061029.0034 0016 UTC
  29-10- 2006 TO 0045 UTC 29-10-2006
• (3) IMDS.20061029.0104 0046 UTC
  29-10- 2006 TO 0115 UTC 29-10-2006
• (4) IMDS.20061029.0134 0116 UTC
  29-10- 2006 TO 0145 UTC 29-10-2006
• (5) IMDS.20061029.0204 0146 UTC
  29-10- 2006 TO 0215 UTC 29-10-2006
• (6) IMDS.20061029.0234 0216 UTC
  29-10- 2006 TO 0245 UTC 29-10-2006

12
GFS ANALYSIS
13
ARPS FORECAST
14
ARPS FORECAST
15
ARPS FORECAST
16
ARPS FORECAST
17
Nov08 Ccyclone Khai Muk
18
Bay Cyclone Khai-mukh of Nov 2008Control
ADAS
19
Control Adas
20
Control Adas
21
Experiments with WRF-Var Assimilation System

• Model WRF-ARW Model
• Assimilation 3DVAR
• Data
• Observation (Synop, Temp, Pilot, Buoy, Ship and
  CMVs)
• First Guess and Boundary NCEP GFS
• Resolution (30 km / L51)

Bay Cyclone Rashmi of October 2008
22
Forecast Vorticity (10-5 s-1 )
24 hour forecast valid at 00 UTC of 25-10-2008
WRF-VAR (var) experiment
48 hour forecast valid at 00 UTC of 26-10-2008
23
Meridional Cross Section of Vertical Velocity
(cms-1 )
24 hour forecast valid at 00 UTC of 25-10-2008
WRF-VAR (var) experiment
No observation (cntl) experiment
48 hour forecast valid at 00 UTC of 26-10-2008
24
Performance of operational NWP models for Cyclone
Track Prediction
A depression over the SE Bay of Bengal at 0300
UTC of 27th April 200812.00 N and long. 87.00 E.,
intensified into a cyclonic storm and lay at
0000 UTC of 28th , a severe cyclonic storm at
0900 UTC of 28th and into a very severe cyclonic
storm at 0300 UTC of 29th. It moved in easterly
direction while intensifying further and crossed
southwest coast of Myanmar between 1200 to 1400
UTC of 2nd May near lat. 16.00 N
VSCS Nargis of April 08
25
ECMWF
MM5
QLM
72 hours Forecast Initial Condition 29 April 00
UTC
UKMO
WRF
26
MM5
ECMWF
QLM
WRF
48 Hrs Forecast Initial condition 30 April 00UTC
UKMO
27
QLM
ECMWF
MM5
UKMO
WRF
24 hours forecast Initial Condition 1 May 00 UTC
QLM with initial condition 2 May 00 UTC
28
Inter-comparison of Model Performance for
Naargis April08
29
Model Error statistics (km) for Nargis
30

• Statistical Dynamical model for Prediction of
• Cyclone genesis
• Intensity

31

• Cyclone Genesis Parameter
• Two Dynamical variables
• Low level relative vorticity (?850)
• Vertical wind shear (S)
• Two Thermo dynamical variables
• (i) Middle troposphere relative humidity (M)
  
• (ii) Middle-trpospheric instability (I)

Mausam (2003), Nat. Hazards (2008)
32
GPP ? 850 MI/S if ? 850 gt 0, M gt 0
and I gt 0 0
if ? 850 0, M 0 or I 0
Where , ? 850 Low level relative
vorticity (at 850 hPa) in 10-5 s-1

S Magnitude of Vertical wind shear between 200
and 850 hPa (ms-1)
RH - 40
M -------------- Middle
tropospheric
30 relative humidity Where RH is
the mean relative humidity between 700 and 500
hPa I (T850 T500) C Middle-trpospheric
instability (Temperature difference between 850
hPa and 500 hPa)
33
VSCS SIDR of Nov 2007 Comparison of composite
Genesis potential parameter (GPPx10-5) and
Genesis potential parameter of Very Severe
Cyclonic Storm (SIDR) over the Bay of Bengal of
11-15 November 2007. (T6.0).
34
GENESIS PARAMETER OF TC NISHA of Nov 2008
35
Comparison of composite Genesis potential
parameter (GPPx10-5) and Genesis potential
parameter of Cyclonic Storm over the Bay of
Bengal of 15-19 October 2000. (T2.5). The
initial low-pressure system formed over Central
Bay of Bengal and intensified into depression
(T.No. 1.5) on 0000 UTC of 15 October. The system
persisted over the Bay of Bengal for more than
four days and traveled more than 700 km
(14.5/88.5 to 14.5/82.0), but maximum intensity
never exceeded T.No. 2.5. Finally it dissipated
over the Sea.
36
Statistical Tropical Cyclone Intensity
Prediction (SCIP) Model 62 sample cases of
Tropical Cyclones (TCs) those formed over the Bay
of Bengal during the period 1981 to 2000. Fifteen
independent cyclones were used to test the model
those formed over the Bay of Bengal during the
period 2000 to 2007.
The predictors (a) Persistence (i) Initial
storm intensity (ISI) (ii) Previous 12 hours
change in the intensity (IC12)   (b)
Thermodynamical factors (i) Storm motion
speed (SMS) (ii) Sea surface temperature
(SST)   (c) Dynamical factors (i) Initial
storm latitude position (ISL) (ii) Vertical wind
shear (850-200) hPa averaged along storm track
(VWS) (iii) Vorticity at 850 hPa (V850) (iv)
Divergence at 200 hPa (D200)
Natural Hazards (2007) J. Earth Sys. Sci.
(2008), Geofizika (2008)
37
Formulation of the model The model is
developed using multiple linear regression
technique y ao a1x1 a2x2
. anxn Where y is the dependent variable
(predictant) and x1, x2, .... xn are
independent variables (predictors). The
regression coefficients a1, a2, .... an are
determined using a large data set (62 cyclones).
The SCIP model estimates changes of intensity at
12, 24, 36, 48, 60 and 72 hours. Six separate
regression analyses are carried out for forecast
interval 12, 24, 36, 48, 60 and 72 hour. 12
hours intensity change by multiple linear
regression technique is defined as dvt ao
a1 IC12 a2 SMS a3 VWS a4 D200 a5 V850a6
ISL a7 SST a8 ISI for t
forecast hour 12, 24, 36, 48, 60 and 72
38
(No Transcript)
39
Twelve-hourly Intensity Prediction up to 36 hours
for cyclone SIDR Nov 2007
Based on 14/00 UTC
Based on 15/00 UTC
40
INTENSITY PREDICTION OF TC RASHMI of October
2008
41
Performance of the model For dependent sample
of 62 cyclones (1981-2000)
For independent sample of 15 cyclones
(2000-2007)
42
IMD Multimodel Ensemble Technique
Generation of Multi-Analysis Weights
Step-1
NCEP
JMA
ECMWF
Observed Gridded Field
Weight for each grid of each Model (W)
43
Generation of Multi-model Forecasts
Step-2
NCEP
JMA
ECMWF

Forecast (F) WiFi D
D Value addition
44

• Multi-model Ensemble at 0.25o resolution
• Member Models
• ECMWF at 0.25o resolution
• JMA at 1.2o resolution
• NCEP at 1.0o resolution

45
MME ECMWF, JMA, NCEP
Roy Bhowmik and Durai, 2008, Atmosfera, 21(3),
225-239
46
Performance Evaluation of MME Forecasts during
Monsoon 2008
The results of spatial correlation coefficient
for day1 to day 5 forecasts illustrating the
superiority of the MME technique over the member
models ( ECMWF, JMA,NCEP)
ECMWF
NCEP
The results of anomaly correlation coefficient
for day 3 forecast showing superiority of MME
JMA
MME
47
Sate-wise performance Day 3 Rainfall Forecast
Over-all performance of MME district level
forecasts over some major states. Performance
index is defined as the of total districts with
threat score more than 0.5 for different rainfall
thresholds. Threat score is defined as number of
correct forecasts divided by total forecast. The
threat score ranges between 0 and 1
48

• Near Future Plan Now-casting
• and mesoscale forecasting
• Real-time radar (DWR) mosaic creation
• Operation of ARPS model at 3 km resolution
  with
• assimilation of DWR data for local severe
  weather
• City forecast for Delhi as required for the
  Commonwealth
• Games 2010
• Implementation of dynamical Fog prediction model
  for visibility forecasting at the major airports
  of India.

49

• Near Future PlanRegional Models
• WRF model with 3 nested domains (at the
  resolution of 27 km, 9 km and 3 km). The nested
  model at the 3 km resolution would be operated at
  the Regional/State Met Centres at 6 hours
  interval with 3 DVAR data assimilation.
• MM5 model with 2 nested domains (at the
  resolution of 27 km and 9 km) at 12 hours
  interval with 3 DVAR data assimilation.
• For Cyclone Track Prediction, 72 hours forecast
  from Quasi Lagrangian Model (QLM) at 40 km
  resolution at six hours interval WRF (NMM) at 27
  km resolution with assimilation package of Grid
  Statistical Interpolation (GSI).
• For Cyclone track and intensity prediction
  multimodel ensemble technique and application of
  dynamical statistical approach for 72 hours
  forecasts, forecast would be updated at 12 hours
  interval.

50
Proposed triple nesting WRF model (27, 9, 3 km)
with flexible fixing of inner most domain
51
Immediate Short-range Forecasting Strategy at
RMC and MC in IMD

• RMC-MC
• Very high-resolution (3km) double nest
  operational model forecast generation for 2 days
• Strom scale model with 1km resolution for 24
  hours forecast
• 3 hourly cycle for specific event
• Assimilation of region specific special
  observations e.g. DWR

52
THANKS