South Asian Regional Reanalysis (SARR)

South Asian Regional Reanalysis (SARR)

Ashish RoutrayNational Centre for Medium Range Weather

Forecasting (NCMRWF)Ministry of Earth Sciences

Government of India

Motivation for South Asian Regional Reanalysis

Due to the direct societal impacts, interest in Regional Hydroclimate (precipitation, surface temperature, soil moisture, stream flow, drought indices, etc.) is intense and growing.

National Action Plan on Climate ChangeGovernment of India

Prime Minister’s Council on Climate Change

3.8.2 ……. Regional data reanalysis projects should be encouraged. ……..

South Asian Regional Reanalysis (SARR)

A Collaborative Project between

Ministry of Earth Sciences, Government of India

and

National Oceanic and Atmospheric Administration, Department of Commerce, United States of America

Specific SARR Goals

Refinement in methods of precipitation and radiances assimilation.

Conduct a 5-year pilot-phase reanalysis(to test and optimize data stream organization and the geographic domain and assimilating model choices)

Develop high-resolution SST analysis for the Indian ocean from satellite and in-situ observations, including moorings, drifters and Argo floats

Design techniques for assimilation of aerosols

Generate a high spatio-temporal resolution (≤25 Km, ≤3 hours) climate data set for the 1979-2009 period over the South Asian land-ocean region.

NOAA agrees to: Provide MoES full access to the archived observations used in the global reanalysis projects.

Provide technical help, training, and guidance in organization of data streams and in the implementation of the regional reanalysis model.

Provide training to MoES scientists in regional reanalysis techniques and procedures during 6-8 week annual visits to US institutions and NOAA laboratories.

Share the NCEP data processing and quality control procedures during reanalysis project with MoES scientists.

Support travel of NOAA and US university scientists to India in connection with SARR project activities.

Responsibilities of the Parties

MoES agrees to: • Provide NOAA full access to all historical and current meteorological observations as per requirement of the project over the Indian subcontinent and Indian Ocean, including those from Indian satellites.

•Execute the South Asian Regional Reanalysis project through NCMRWF.

•Provide full-time modeling scientists to develop, implement, and test numerical codes.

•Provide 4-6 full time Ph.D. scientists to design, test, and implement various assimilation schemes in the numerical model.

•Provide high-speed mainframe computer resources for execution of this computationally intensive project.

•Provide storage devices and skilled manpower (data management specialists) to organize data streams, data archival, data dissemination, and webpage design and maintenance.

•Provide continuous high-speed internet access to project scientists, including visiting ones.

•Provide lodging and boarding for visiting US project scientists.

Exchange visits

• NOAA will provide training to 2-3 MoES scientists in regional reanalysis techniques and procedures during 6-8 week annual visits to the University of Maryland and NOAA's National Centers for Environmental Prediction (NCEP).

• NCEP will seek resources and assistance from NOAA's

International Activities office in meeting its responsibilities. • NOAA and MoES scientists will meet yearly to discuss the

project's progress, and to strategize on how to best accomplish the project goals.

• NOAA and MoES will separately cover travel costs associated with exchange visits for their respective technical and scientific personnel.

Milestones

SARR IA signed in September 2008 in New Delhi

1st Annual Review by JEM held in October 2009 in New Delhi

Functional Group created at NCMRWF for SARR in November 2009

SARR Scoping Workshop held in New Delhi in February 2010

2nd Annual Review by JEM held in October 2010 in Washington DC

The SARR Project is being carried out with an objective that the SARR Products shall be useful for

Climate Diagnostics, Climate Variability, Climate Change, Model Verification/Tuning

It is expected that

The SARR project will provide an Atmosphere-Land-Ocean surface state description where consistency between circulation and hydroclimate components is assured.

To achieve the goal, assimilation of rainfall, radiance, and aerosol observations in numerical weather prediction models shall be carried out

SARR Project Team at NCMRWF

Sarat C. Kar Project Management

Ashish Routray Assimilation- Lead

Prashant Mali Modeling- Lead

Jaganabdhu Panda Modeling (worked for about 3 months and left in September 2010)

K. Sowjanya Assimilation (worked for about 1 year and left in September 2011)

Sapna Rana Diagnostics (worked for about 1 year and left in November 2011)

Domain chosen for SARR

Lat: 150S-450N (286 pts)

Lon: 400E-1200E (332 pts)

Res.: 25 km (pilot phase)

18 km (final SARR)

Cen-lat: 17.50N

Cen-lon: 80.00E

SARR OBSERVATION

DATA BANK

NCEP IMD

NCMRWF

ISROINCOIS Field

Experiments

Countries in SARR domain

70 75 80 85 90 95longitude (E)

0

5

10

15

20

25

latitu

de

(N

)

DS3,

TS2 (SK)DS4

Chennai

Paradip

TS1 (SD)

Figure 1. Cruise track and time series (TS) observation positions.

SK - ORV Sagar Kanya, SD - INS Sagardhwani, DS3 & DS4 - met ocean buoys

Period: 16 July - 30 August 1999. TS1 - 13N,87E; TS2 - 17.5N, 89E.

BOBMEX

Land Surface Processes Experiment (LASPEX)

ARMEX

STORM Programme

CTCZ

DATA from FIELD EXPERIMENTS

B

CPROWNM

SARR Scoping Workshop

held in New Delhi, India (February 10-11, 2010)

9 scientists from USA and about 20 scientists from India participated.

Analysis method and the model as well as domain of analysis finalized.

WRF model (3.1 version) and WRF-3DVar shall be used to carry out SARR Pilot phase.

The Workshop recommended an implementation strategy for success of the SARR project.

Work plan at NCEP• Training on methodology for assimilation of the

radiance data (mainly the older period radiance data) using the GSI system so that a similar technique can be developed later for the WRF-3DVAR analysis system.

• As part of the training, experiments using radiance data assimilation for Indian summer monsoon seasons (mainly for older period) using the NCEP GSI system and document impact assessment.

• Familiarization with the available diagnosis package for monitoring and for calculation of statistics of the radiance data utilized in the assimilation cycle.

SARR Pilot Phase Experiments(1999-2003)

Analysis Scheme & Model for SARR Pilot Phase

WRF 3.1 and WRF-VAR (3.1) has been chosen for SARR Pilot phase experiments

Several modeling and assimilation experiments have been carried out using past data.

Most of the experiments are for July 1999 using NCEP & NCMRWF observation datasets

0

50

100

150

200

250

300

Total TEMP WIND PILOT

00Z

06Z

12Z

18Z

Average Number of Observations per day in July 1999

0

50

100

150

200

250

300

>800hPa 800-450hPa 450-100hPa <100hPa

00Z 12Z

Av. Number of TEMP observation per day reaching particular height in July 1999

Blocks- 42 and 43

Challenging regions for obs. data

Sound

Mean RMSE of O-B and O-A for U (m/s)

2.5

3

3.5

4

sound pilot geoamv airep synop ship

Types of Obs.

U (m

/s)

O-B O-A

Mean RMSE of O-B and O-A for V (m/s)

0

1

2

3

4

5

sound pilot geoamv airep synop ship

Types of Obs.

V (

m/s

)

O-B O-A

Mean RMSE of wind components from different observations at model initial

time

Mean of RMSE of O-B for U (m/s)

1

2

3

4

5

6

sound pilot airep synop ship

Types of Obs.

U (m

/s)

NCMRWF PrepBufr

Mean of RMSE of O-B for V (m/s)

1

2

3

4

5

sound pilot airep synop ship

Types of Obs.

V(m

/s

NCMRWF PrepBufr

Mean RMSE of O-B for t (k)

0.5

1

1.5

2

2.5

sound airep synop ship

Types of Obs.

t (k)

NCMRWF PrepBufrMean RMSE of OBS-FG

Mean RMSE of O-A for U (m/s)

1

2

3

4

5

6

sound pilot airep synop ship

Types of Obs.

U (m

/s)

NCMRWF PrepBufr

Mean RMSE of O-A for V (m/s)

1

2

3

4

5

sound pilot airep synop ship

Types of Obs.

V (m

/s)

NCMRWF PrepBufr

Mean RMSE of O-A for t (k)

0.1

0.3

0.5

0.7

0.9

1.1

1.3

1.5

sound airep synop ship

Types Obs.

t (k)

NCMRWF PrepBufrMean RMSE of OBS-ANA

SARR Test runs with NCEP & NCMRWF data

SARR Pilot Phase Experiments

(i) with various Physics OptionsDynamic Downscaling using WRF

(ii) with various Physics OptionsAssimilation using WRF & WRF-VAR

Most of the experiments are for July 1999 using NCEP & NCMRWF observation datasets

All Experiments were done for July 01- 31 1999.

With Assimilation- Cyclic, Four times a day (6-hourly)

No Assimilation- only Model run Four-times a day (6-hourly). (Similar to downscaling experiments)

Precipitation in July 1999 CMAP, TRMM (3B42) and IMD Observed Rain

SARR Pilot Phase Sensitivity Experiments

Precipitation from Global Reanalysis datasets for July 1999

As can be seen, the global reanalysis has failed to bring out details of rainfall distribution over India and higher rainfall amounts are placed at incorrect locations

CU schemes PBL Schemes SFC Schemes Expt. Names

Kain-Fritsch (KF)

Yonsei University (YSU)

Noah Land surface

KF-YSU-Noah

Betts-Miller-Janjic (BMJ)

BMJ-YSU-Noah

Grell Devenyi (GD) GD-YSU-Noah

KF

Mellor-Yamada-Janjic (MYJ)

KF-YSU-Noah

BMJ BMJ-YSU-Noah

GD GD-YSU-Noah

KF

YSU

Thermal Diffusion (TD)

KF-YSU-TD

BMJ BMJ-YSU-TD

GD GD-YSU-TD

KF

MYJ

KF-MYJ-TD

BMJ BMJ-MYJ-TD

GD GD-MYJ-TD

EXPERIMENTAL DESIGN

No Assimilation With AssimilationSARR Pilot phase Sensitivity Experiments

It has been shown that

just downscaling of coarse resolution global reanalysis (No Assimilation runs) is not sufficient for accurate representation of the Indian monsoon hydroclimate.

When regional assimilation is carried out, such representation is improved.

Experiments have been carried out using ISRO derived vegetation data instead of USGS climatological vegetation available with the WRF model.

Results indicate that hydroclimate representation over India is sensitive to such specifications.

SARR Pilot Phase Sensitivity Experiments

Impact of Field phase Experiments- BOBMEX data

70 75 80 85 90 95

longitude (E)

0

5

10

15

20

25

latitu

de

(N

)

DS3,

TS2 (SK)DS4

Chennai

Paradip

TS1 (SD)

Figure 1. Cruise track and time series (TS) observation positions.

SK - ORV Sagar Kanya, SD - INS Sagardhwani, DS3 & DS4 - met ocean buoys

Period: 16 July - 30 August 1999. TS1 - 13N,87E; TS2 - 17.5N, 89E.

Bay of Bengal Monsoon Experiment(BOBMEX)

July-August 1999

Impact of Field phase Experiments- BOBMEX data (00Z 12 August 1999)

Assim- Control Assim- with BOBMEX Difference

U at 850hPa

T at 850hPa

Parallel Assimilation from May 2001 to Sept 2001. Need of Overlapping period

Pilot phase Assimilation with conventional data has been completed from 1999-2003.

Assimilation with Radiance data and conventional data is being carried out for the same period. Parallel run period is also being extended.

Comparison between CFSR, SARR and Observation(1-31 July 2000)

OBS

CFSR

SARR

SARR Production Runs Five simultaneous Streams

Jan. 1979 - Dec. 1985 7 years Apr. 1985 - Dec. 1991 7 years

Apr. 1991 - Dec. 1997 7 years Apr. 1997 - Dec. 2003 7 years Apr. 2003 - Dec. 2009 7 years

9-month overlap for each stream

Total 35 years of Reanalysis Computation

SARR Products Archival and Distribution

Archival Format (Reanalysis):IEEE (suitable for GrADS)NetCDF GRIB2

Archival Format (Observed data):ASCII (GTS)PrepBUFRlittle-ROriginal format of data

Archival online/nearline disk, Tapes

Available to Partner Organizations: Immediately

Tasks Aug 2010

Dec 2010

Apr 2011

Aug 2011

Dec 2011

Apr 2012

Aug 2012

Dec 2012

Apr 2013

Aug 2013

Pilot phase reanalysis production (1999-2003)

Evaluation of pilot phase reanalysis data

Refinement of assimilation techniques

Collection of data from countries in SARR domain

Level-I SARR Production for 1979-2009 period

Evaluation of Level-I reanalysis data

Final SARR Production for 1979-2009 period

Reanalysis data- public

SARR – What next?

SARR -II

After the successful completion of SARR’s present project, We propose to carryout SARR-60

SARR-60 From 1950 to 2009 at 9 km resolution Regional Ocean-Atmosphere coupling - shall be the comprehensive dataset for

climate studies in South Asia.

IMPACT of BACKGROUND ERRORS (BE) & ASSIMILATION

Numerical Experiments• The objective of the study is to evaluate the impact of the different

back ground errors (Global and Regional) towards simulation of four Monsoon Depressions (MDs) over Indian region during SARR pilot phase period.

• 27-29 July 1999 (Case-1)• 17-18 June 1999 (Case-2)• 11-12 June 1999 (Case-3)• 6-8 August 1999 (Case-4)

• For this purpose three numerical experiments are carried with WRF-3DVAR as follows:

1) CNTL: Without data assimilation using NCEP re- analyses as IC and BC.

2) BG-3DV: Data assimilation using NCEP global Background Error (BE).

3) BR-3DV: Data assimilation using own calculated BE over SARR region.

• The additional observations viz. SYNOP, SHIP, TEMP, BUOYS, PILOT, GEOMV, AIREP etc. are used to improve the model initial condition derived from coarse resolution large scale global analysis.

Mean RMSE of O-A for U (m/s)

0

1

2

3

4

5

6

Sound Synop Geoamv Airep Pilot Metar Ships

Types of OBS

U (m

/s))

BG-3DV BR-3DV

Mean RMSE of O-A for V (m/s)

0

1

2

3

4

5

6

Sound Synop Geoamv Airep Pilot Metar Ships

Types of OBS

V (m

/s))

BG-3DV BR-3DV

Mean RMSE of O-A for Temperature (k)

0

0.5

1

1.5

2

Sound Synop Airep Metar Ships

Types of OBS

Tem

pera

ture

(k))

) BG-3DV BR-3DV

a) b)

c)

Mean RMSE from BR-3DV and BG-3DV of O-A for a) U (m/s), b) V (m/s) and c) T (K).

Case-1 NCEP ANA BG-3DV ANA BR-3DV ANA

OBS: 21.0/89.0

CNTL:21.8/89.8

BG-3DV:21.6/88.8

BR-3DV:20.8/89.5

NCEP ANA BG-3DV ANA BR-3DV ANACase-2

OBS:18.5/86.0

CNTL:18.5/87.0

BG-3DV:18.9/87.1

BR-3DV:19.2/86.5

Model Initial time wind fields at 850 hPa and MSLP

Case-1

Track Error

0100200300400500600700800900

0 12 24 36 48

Forecast hours

Err

ors

(km

)

CNTL

BG-3DV

BR-3DV

Track Errors

0

100

200

300

400

500

600

700

0 12 24

Forecast hours

Err

ors

(km

s)

CNTL

BG-3DV

BR-3DV

Case-2

Track Eorrors (km)

0

50

100

150

200

250

0 12 24

Forecast hrs

Err

ors

(km

)

CNTL

BG-3DV

BR-3DV

Track Errors (km)

0

100

200

300

400

0 12 24 36 48

Forecast hrsE

rrors

(km

)

CNTL

BG-3DV

BR-3DV

Case-3 Case-4

Cases RMSE CC

CNTL BG-3DV BR-3DV CNTL BG-3DV BR-3DV

Case-1(27-29 July 99)

29. 62 26. 24 22.15 0.23 0.36 0.52

Case-2(17-18 June 99)

24. 32 22. 31 18. 38 0.21 0.33 0.46

Case-3(11-12 Jun 99)

26. 93 22. 54 18. 68 0.26 0.35 0.46

Case-4(6-8 Aug 99)

40. 59 34. 25 29. 41 0.33 0.46 0.52

Mean 30. 37 26. 34 22. 16 0.26 0.38 0.49

Spatial RMSE (mm) and Correlation Co-efficient (CC) of rainfall over the area (Lat=150-250N; Lon=750-900E) for all cases.

Mean VDEs (km) and gain skill of experiments

0

100

200

300

400

00 12 24

Forecast hours (UTC)

Mea

n V

DE

s(km

s)

0

20

40

60

Ski

ll o

f E

xpts

.(%

)

CNTL BG-3DV BR-3DV

CNTL vs. BG CNTL vs. BR BG vs BR

Impact of Radiance data

GTS GTS+Rad

Diff. (Rad-GTS)

Temperature (oC) at 850 hPa

GTS GTS+Rad

Diff. (Rad-GTS)

Wind (m/s) at 850 hPa

Wind (m/s) at 500 hPaGTS GTS+Rad

Diff. (Rad-GTS)

These are accumulated 6-hrly Rainfall from the models used for Reanalysis. Every 6-hour, observed data are inserted into the data Assimilation systems, and analyses are carried out. Assumption is that models are good enough for at least 6 hour.

Rainfall Climatology

These studies show there are large uncertainties in the Global Reanalysis data over our Region.

Model Resolution? Data Quality/Quantity?

We need to carry out a Systematic Regional Reanalysis for our Region to have a consistent Hydro-climate dataset.

The Global reanalysis data are utilized for studying climate change and to develop several Application models.

Therefore, we should provide the users with a good quality data set for our Region.

Global BE

Reg. BE

a single Temperature observation 10K

Global BE

Reg. BE

a single u-wind observation 1 m/s

Response of the Analysis Increments to • A large part of tropical forecast errors can be represented by equatorial waves.

• These modes effectively reduce the mass/wind coupling at the equator.

• Daley (1996) has noted that equatorial error covariance is weaker than higher latitude and similar to that obtained by equatorial beta plane theory.

• By suppressing the erroneous tropical wind-height coupling, Daley did not find the covariance pattern to the south of central latitude in the tropical domain.

•In our study, we find that for the BE statistics, the effect of a single wind observation is consistent with theoretically derived wind correlations for non-divergent flow.