Transcript Assimilation of TRMM Precipitation into Reanalysis and its Impact on

Assimilation of TRMM Precipitation
into Reanalysis and its Impact on
Tropical Divergence Fields
Baek-Min Kim(金伯珉 )
J.-H. Oh, and G.-H. Lim
Seoul National University
Steven Cocke and D.-W. Shin
FSU/COAPS
ERA40-GPCP PRCP.
From ECMWF Web site(http:://www.ecmwf.int)
No Direct PRCP. ASSIM.
Despite the importance of the diabatic heating related to
precipitations to climate studies, the data assimilation system of longterm reanalyses do not exploit the benefit of precipitation
assimilation.
Inconsistency between the precipitations and divergences in
existing reanalyses over tropic and subtropic has been
demonstrated.
(Newman et al, BAMS 2000)
Mean precipitation in winter
Contour interval=1.5mm/day
Mean divergence in winter
Contour interval= 1× 10− 6s− 1
Motivation
●
●
●
●
Different models produce wide range of mean
seasonal precipitation, yet tropical SST/diabatic
heating forcing is a main ingredient for
understanding climate variability and predictability.
Reanalyses (NCEP, ERA40, NASA etc) have
widely varying representation of intraseasonal
oscillations such as MJO (Newman et al, 2000).
Tropical divergence for tropical cyclones
significantly under-represented.
Models have difficulty simulating ISOs.
Procedure & Object
Data Preparation
(TRMM,RII,ERA40)
Preprocessing
For
Assimilation
Assimilate the precipitation data to reanalysis
using physical initialization.
-
Test Run
And
Tuning of
Nudging Coeff.
Examination on the convectively-coupled variability over
tropics(i.e.: MJO) with high quality PAReanalysis
Production of
PAReanalysis
(Current)
Diagnostic
Studies
Model
Assimilation Method
●
●
●
●
Based on physical initialization procedure using
FSUGSM
Rain rate determined by Kuo scheme
Prognostic variables nudged using Newtonian
relaxation toward NCEP II reanalysis
Continuous assimilation for many months
Schematics of Analysis Cycle
Observation
Simplified Physical Initialization
TRMM
precipitarion rate
Background
RII
AVHRR OLR
Rain Matching
Dynamic Nudging
Reanalized atmospheric
variables
Relaxation
improved moisture vertical profile
Model Run
00Z
06Z
12Z
Analysis Cycle
18Z
Rainfall Matching Technique
●
●
●
The rainfall nudging basically modifies the
vertical profile of humidity as a function of the
observed and predicted model rain rates.
Through the continuous application, the model
rain is brought closer to the observed rain rate.
Humidity profile is modified using simple
structure function to match model rain rate
against TRMM:
Pros/Cons
●
●
●
●
●
PA is crude, but effective. Simple, easy to
implement.
Can be used to diagnose model errors (Jeuken
et al, 1996)
No cost function to distinguish quality of data
Big assumptions are made (2D field determines
3D field) – results could be largely
model/cumulus parameterization dependent
Choice of nudging coefficients rely on painful
procedure-”Try and See”
Preliminary Focus Areas
●
Impact in the Western Pacific Warm Pool
Region
●
Tropical cyclone circulation(Preliminary stage)
●
Intraseasonal oscillation(Preliminary stage)
–
Eastward propagating mode (winter)
Feasibility Experiment
●
●
●
●
Continuous assimilation for JJA 2000,
NOV2004
Sensitivity to nudging coefficients
6 hourly analyses and 3 hourly rain rates
interpolated to model time step
FSUGSM at T63L27 resolution
Data Sets
●
●
●
●
NCEP-2 Reanalysis (Kanamitsu et al 2002), 6
hourly interval
Tropical Rainfall Measuring Mission(TRMM)
Multi-satellite Precipitation Analysis(MPA;3B42)
is used.
3B42 is a merged product of passive
microwave-only product(3B40) and microwavecalibrated IR(3B41).
Temporal resolution is 3 hour and spatial
resolution is 0.25 deg.
PA-Reanalysis
TRMM
Total Precipitation for November 2004
TRMM
PAR
November 15 0Z
NCEP R2
Correlation with TRMM
Area-average rain in Eastern Pacific ITCZ
Black – TRMM
Green – PA Rean
Red – NCEP R2
Impact in Western Pacific Warm
Pool (B.-M. Kim et al, 2007)
●
●
Assimilation of TRMM 3B42 into NCEP R2
Seasonal assimilation from 1 May to 31 Aug
2000
●
Sensitivity of nudging coefficients
●
Compare with ERA40, GPCP
●
●
Statistics(Mean, Var., Corr., RMSE) are
compared.
This region is interesting because of large
differences in the reanalyses
Dynamic relaxation(nudging)
●
●
●
●
Dynamic nudging provides
background(reanalysis) of precipitation
assimilation.
A is model variable (vor., Psfc, div., T, q ).
A* (model prediction prior to relaxation) gets
closer to A0 (reanalysis)
Bigger N forces more toward reanalysis.
Sensitivity of nudging coeff. (N)
Exp.
Vor.
Psfc
Div.
T
q
NUDG1 8
8
8
8
8
NUDG2 8
8
4
4
4
NUDG3 8
8
2
4
4
NUDG4 8
8
4
2
4
NUDG5 8
8
4
4
2
Unit=day^-1
Statistics over the Western Warm Pool Region (15S-25N, 60-180E)
Variable
TRMM
GPCP
ERA40
R2
NUDG1
NUDG2
NUDG3
NUDG4
RMSE
0
1.5
4.9
4.0
2.0
1.8
1.7
1.7
Std.
8.5
7.9
8.4
9.5
8.7
8.4
8.4
8.4
TCOR
1.00
0.62
0.49
0.27
0.91
0.92
0.93
0.92
GCOR
0.62
1.00
0.34
0.23
0.56
0.56
0.57
0.56
Mean PRCP(TRMM vs RII)
Mean PRCP(TRMM vs NUDG3)
Mean Div.(NUDG3 vs RII)
Contour:1e-6 s^-1
Tropical Cyclones(Steven Cocke)
●
●
Examine impact on circulation/intensity of
tropical cyclones
Current experiments used T63 resolution, but
future experiments need to be done a higher
resolution:
–
Hurricane-like vortices are exaggerated in size at
low resolutions
–
TRMM is available at 0.25 deg. -don’t waste it.
TS Bonnie & Charlie (2004/08/11)
GOES-12 RGB=CH1,CH1,CH4
08/11/2004 14:45 UTC
Intraseasonal Oscillation
(J.-H. Oh)
●
Assimilation period: 1 Jan 1998 to 31 Dec 2005
●
NUDG3 setting (vor:8,Psfc:8,div:2,T:4,q:4)
●
●
Composite based on Wheeler and
Hendon(2004)
Eastward propagating mode (winter MJO)
MJJA rainfall variability (98~05)
GPCP
New reanalysis
NCEP R2
Upper: MJJA ISO rainfall standard deviation, lower: MJJA rainfall standard deviation
Eastward propagating ISO
Composite strategy: based on
Wheeler and Hendon(2004) -MJO index
Naritime continent
West.Hem. And Africa
Western Pacfic
Indian Ocean
NOAA OLR
NCEP R2
New reanalysis
New reanalysis rainfall
NCEP rainfall
-0.821
-0.501
Pattern correlation
OLR
New reanalysis-wind & divergence (200hPa)
New reanalysis-wind(850hPa) & vertical velocity(500hPa)
NCEP R2-wind & divergence (200hPa)
New reanalysis-wind(850hPa) & vertical velocity(500hPa)
Summary
●
●
●
●
The TRMM data is very well assimilated, and
the associated model variables are consistent
with the assimilated precipitation.
Tropical Pacific rainfall and variability and its
associated divergent circulation appear to be
improved.
Tropical Cyclone circulation appears improved
and more consistent with independent
observations.
ISO signature is much more pronounced in the
PA-reanalysis when compared to NCEP R2.
Summary
●
The meteorological variables such as
divergence and vertical velocity of the
reanalyses are consistent with their respective
reanalysis precipitation – so, if the reanalysis
rain is not correct, then.....
One Possiblity….
PAReanalysis
Reanalysis
Div
.
Div.
Model
precipitation
Vorticity
budget
equation:
Observed precipitation
Vorticity
budget
equation:
Imbalanced
More balanced