Radiance assimilation in JMA’s Meso-scale Analysis

Radiance assimilation in JMA’sMeso-scale Analysis

Masahiro Kazumori

Izumi Okabe

Japan Meteorological Agency

AMSR-E Science Team Meeting, Asheville, North Carolina, U.S.A.June 28-29, 2011

Outline• Introduction

– JMA Meso-scale Analysis

• Assimilation experiments– Case study 1: Heavy precipitation in Baiu season– Case study 2: Typhoon

• Summary and Plan

2

Introduction• The main objective of JMA’s Meso-Scale Model

(MSM) is to provide guidance for issuing warnings or very short-range forecasts of precipitation covering Japan and its surrounding areas.

• JMA’s Meso-scale Analysis (4D-Var) requires a lot of observations to produce accurate initial condition for the forecast model.

• Total column water vapor and Rain rate from AMSR-E, TMI, and Temperature profiles from ATOVS had been assimilated together with other observation data.

• On Dec. 13, 2010, direct radiance assimilation was introduced in JMA operational Meso-scale Analysis as the replacement of the retrieval assimilation.

3

JMA Meso-scale AnalysisJapan is an island country surrounded by ocean. Moisture information over the ocean is a key for accurate precipitation forecasting.

Retrieval assimilation of TCWV is out-of-date. Most operational NWP centers use observed radiances directly in data assimilation system.

Direct radiance assimilation enable us to use the observations without any retrieval process and retrieval error contamination.

Early use of satellite data into operational NWP is possible after the L1 data release.

Fast radiative transfer model (e.g. RTTOV) is necessary for the forward and adjoint calculation in the variational data assimilation.

Meso-Scale Model domainHorizontal res. 5km (3600x2880km)

50 vertical layers up to 22km15-hours forecast

from 00,06,12,18UTC initial 33-hours forecast

from 03,09,15,21UTC Initial

4

Data coverage in Meso-scale AnalysisIn Situ Observations

03UTC(Daytime)

Available observation data depend on the analysis time.

21UTC(Night time)

5

Remote Sensing Observations

03UTC(Daytime)

01 July, 2010

21UTC(Nighttime)

Also available polar orbiting satellite data depend on the analysis time.

6

2009/07/22/00 UTC

Addition of F-16,F17 SSMIS TbChange to F13 SSMI Tb

F13 SSMI TCWV Ground based GPS TCWV

Ground based GPS TCWV

Addition of F-16,F17 SSMIS Rain Rate

F13 SSMI Rain Rate Rain Rate from ground-based Radar.

F13 SSMI Rain RateRain Rate from Radar

Retrieval Assimilation Radiance assimilation

7

Bias correction for Tb• Scan bias correction

– Biases dependent on scan position

– Scan biases were corrected by fixed coefficient tables for each channels and sensors

• Air-mass bias correction (VarBC)– In the JMA global DA system, the biases in O-B are corrected by variational bias

correction scheme (VarBC). The biases are estimated by using a linear function with some predictors and those coefficients are optimized inside the 4D-Var analysis and updated every analysis cycle.

– Predictors : Integrated weighted lapse rate, surface temperature, cloud liquid water, zenith angle.

Red: Mean Bias, Green: Std, Blue: Data counts (after thinning)

F-17 SSMIS 19V 22V 37V 92V

[K]

8

Configuration of Assimilation Experiments

Microwave Imager (AMSR-E, TMI)TCWV, Rain Rate

Microwave SounderTemperature

Microwave ImagerAMSR-E, TMI Radiance,

Rain RateMicrowave Sounder

Radiance

Microwave ImagerSSMIS F16 F17

Radiance, Rain Rate

Microwave Humidity Sounder(MHS, AMSU-

B)RadianceMTSAT-1R

IR Clear Sky Radiance

ControlTest

Retrieval Assimilation(Same as operational as of Oct. 2010)

Radiance Assimilation(addition of other available radiance data) 9

Case study 1MTSAT IR

7/2-7/4 Total rainfall amount 10

Comparison of data coverage

Retrieval Assimilation (Control)

Radiance Assimilation (Test )11

Radar obs. vs. MSM precipitation forecasts

6-hr forecast

12-hr forecast

18-hr forecast

Retrieval assimilation(Control)

RA observation

Valid time: 12JST 03 July, 2010

Weak rain in forecasts

3-hr accumulated rainfall

12

Radar obs. vs. MSM precipitation forecasts

6-hr forecast

12-hr forecast

18-hr forecast

Radiance assimilation (Addition of F-16,17 SSMIS Imagers)

RA observation

Improvement in short-range precipitation forecast

3-hr accumulated rainfall

Valid time: 12JST 03 July, 201013

Retrieval assimilation

Radiance assimilation

Difference (Test-Control)

2010/07/02 21UTC

The reason of the precipitation forecast improvement is the difference of analyzed TCWV field

Moisture flow from southwest around Kyushu area was strengthened in the radiance assimilation’s analysis

[mm][mm]

[mm]

14

Observation Retrieval assimilation

Radiance assimilation

Valid time: 03UTC 3 July, 2010, 6-hour forecast from 21UTC 2 July, 2010 initial time.

Simulated MTSAT image (WV)Observed MTSAT image (WV)

MTSAT WV image contains moisture information in the upper troposphere.

Simulated image from Test’s forecast field is close to real observation.

Verification with MTSAT cloud image

15

Diff. of TCWV(Test-Cntl)

Data coverage of newly added DMSP F16,17 SSMIS radiance

Case study 2

F-16, F17 SSMIS radiance and rain rate data were newly added in the Test run.

16

Test’s analyzed TCWV

Control’s TCWV IncrementControl’s analyzed TCWV

Test’s TCWV Increment

The first analysis

[mm] [mm]

New microwave imagers data enhanced the TCVW contrast.

17

Valid time: 09UTC 9 Aug., 2010Simulation from 09UTC 9 Aug., 2010 (initial time)

Simulated MTSAT image (IR)Observed MTSAT image (IR)Retrieval assimilation

Radiance assimilation

Separated feature is well represented in the analysis.

18

Radiance assimilation

Retrieval assimilation

Clearly separated

Radar observation

[mm/3hr]

3-hr precipitation forecast

19

Summary and Plan• Atmospheric water vapor content is one of the fundamental

amount in NWP model. The information provided by Microwave imagers is essential for the accurate forecasting of heavy precipitation and the typhoon.

• Direct radiance assimilation showed large positive impactson the analyses and forecasts. Direct radiance assimilation enable us to use a lot of satellite data without retrieval process. And new data, DMSP F-16, F-17 SSMIS were incorporated in the analysis.

• A number of MW-Imager data provide realistic moisture field in the analysis.

• It is desirable to use well calibrated Microwave radiance data as much as possible. New Microwave imagers are– TMI Ver. 7 as a replacement of current Ver. 6 data– WindSat– F-18 SSMIS 20

TMI Tb data in JMA’s NWP system• JMA assimilates TRMM Microwave Imager (TMI) observations for

their information on humidity over the ocean in Global DA system.• Variational DA assumes no bias between observed Tb and model

equivalent. Variational bias correction (VarBC) is applied for Tb.– A linear function is assumed to describe the bias by using some predictors.

Coefficients are optimized in the analysis and used in the next analysis. However, the coefficients are determined as global constants in every analysis. It is difficult to correct local biases in the current VarBC scheme.

xxβxz bphh~

y

Bias correction term is in the observation operator Coefficients: Predictors: p TCPW, TSRF, TSRF

2, WSSRF, CLW Const.

TMI 19.35GHz V pol.

Coefficients are determined in JMA global analysis.

Time evolution of coefficients

21

Comparison of Ver. 7 and 6 TMI data• TMI data (Ver. 6) is erroneous because it assumes a fixed reflector temperature in

calibration. Time varying solar biases are reported in the comparison with ECMWF first guess (A. Geer 2010).

• NASA plans to distribute Ver.7 TMI data. JMA obtains the sample data via JAXA.

• An evaluation was performed to confirm the improved calibration.

Ver. 7 TMI 19GHz V pol. Tb and the difference from Ver. 6

June 1, 2010 [K] [K]22

19V pol. May 　　　　　　　　 　 Jun. 　　　　 　 　　　 Jul. 　　　　　　　　　Aug. 　　　　　　　　 Sep. 　　　　　　　

TMI Tb data in JMA’s NWP system• Solar biases observed in TMI Ver.6 Tb.

TMI 19GHz V.pol Bias corrected O-B (observed Tb – background Tb), clear scene only

TMI V6

Data counts

O-B

Local time

Lat

[K]

23

TMI Tb data in JMA’s NWP system• TMI Ver.7 Tb showed improved data quality. Solar biases are

much reduced. 19V pol. May 　　　　　　　　 　 Jun. 　　　　 　 　　　 Jul. 　　　　　　　　　Aug. 　　　　　　　　 Sep. 　　　　　　　 TMI V6

TMI V7

Local time

Lat

[K]

24

Time sequences of VarBC coefficients

Cold start

TMI 19GHz V pol.

Dotted : Coefficients for Ver.6 TMISolid: Coefficients for Ver.7 TMI

47days

Coefficients’ change is reduced, 47 days gaps disappeared.

AMSR-E 19GHz V pol.

F-16 SSMIS 19GHz V pol.

F-17 SSMIS 19GHz V pol. TCPW

TSRF

TSRF2

WSSRF

CLW Const.

25

21V TMI-V6 TMI-V7 　　　　　　　 SSMIS16 　　 　　　　SSMIS17 　　 　　 AMSR-E 　　　　

June 2010

[K]Local time

Lat

From RSS home page

Comparison of water vapor channel’s O-B biases in JMA NWP

26

Final comments• Radiance assimilation of Microwave imagers was started in

JMA’s Meso-scale Analysis in Dec. 13, 2010.

• Direct radiance assimilation of Microwave imagers has large positive impacts in JMA NWP system. Microwave imager’s radiance data is necessary for accurate humidity analyses and precipitation forecasts for Japan.

• As direct assimilation of radiance data in NWP is major trend, the Tb’s calibration accuracy is more important than before.Our NWP system has capability to detect the calibration problem.

• NWP is expected as a powerful tool for Cal/Val process of GCOM-W1/AMSR2 and GPM/GMI.

27