assessment and inter-comparison of five ghrsst products in the coastal and shelf sea around china
DESCRIPTION
Assessment and inter-comparison of five GHRSST products in the coastal and shelf sea around China. Jiping Xie 1,3 , Jiang Zhu 1 , and Yan Li 2,3. 1. Institute of Atmospheric Physics, CAS 2. National Marine Environmental Forecasting Center, Beijing - PowerPoint PPT PresentationTRANSCRIPT
Assessment and inter-comparison of five GHRSST products in the coastal and
shelf sea around China
Jiping Xie1,3
, Jiang Zhu1, and Yan Li
2,3
1. Institute of Atmospheric Physics, CAS2. National Marine Environmental Forecasting Center, Beijing
3. Graduate School of the Chinese Academy of Sciences
Melbourne, Australia, May 2007
1. Background
2. Data Processing
3. Inter-comparison GHRSST products
4. Assessment by in situ observations
5. Conclusions
Outline
GRHSST L4 Products : MWOISST (RSS ), NGSST (UT, Japan), MGDSST (JMA), OSTIA (NCOF), FSTIA (FNMOC)
Annual SST:
(2005.10~2006.9)
OSTIAMGDSST
However, we notice there are some … …
Questions
I. How about the differences and errors of the five GHRSST products in the coastal and shelf sea around China?
II. How to improve the GHRSST products as a major input for data assimilation in this region?
III. Can we use the ensemble mean method to improve the presently GHRSST products?
116-135ºE, 20-42ºN
Global
Global
Global
116-166ºE, 13-63ºN
Global
Coverage
[synthesis]10 km, dailyEMSST
NOAA/AVHRR10 km, 6 hourlyFSTIA
ENVISAT/AATSR, NOAA/AVHRR, MSG/SEVIRI, AQUA/AMSR-E,
TRMM/TMI
5 km, dailyOSTIA
NOAA/AVHRR, AQUA/AMSR-E
NOAA/AVHRR, AQUA/MODIS, AQUA/AMSR-E
TRMM/TMI, AQUA/AMSR-E
Satellite/sensor
25 km, daily
2 km, daily
25 km, daily
Resolution
NGSST
MGDSST
MWOISST
116-135ºE, 20-42ºN
Global
Global
Global
116-166ºE, 13-63ºN
Global
Coverage
[synthesis]10 km, dailyEMSST
NOAA/AVHRR10 km, 6 hourlyFSTIA
ENVISAT/AATSR, NOAA/AVHRR, MSG/SEVIRI, AQUA/AMSR-E,
TRMM/TMI
5 km, dailyOSTIA
NOAA/AVHRR, AQUA/AMSR-E
NOAA/AVHRR, AQUA/MODIS, AQUA/AMSR-E
TRMM/TMI, AQUA/AMSR-E
Satellite/sensor
25 km, daily
2 km, daily
25 km, daily
Resolution
NGSST
MGDSST
MWOISST
Table 1. Specification of different GHRSST L4 products
Cover: 116-135ºE, 20~42ºN; period: 2005.10~2006.9
2. Data Processing GHRSST L4 Products
In Situ Observations
Drifting buoy and Ship reports are collected by Near-Goos (http://near-goos1.jodc.go.jp) from GTS.
DDGR: Jun 2006
I. Pre-process
1) Excluding the reports with invalid ship signal.
2) Excluding the redundant reports.
3) Position check by tracking consistency method as Kent and Challenor(2006).
Night SST : SHIP ~ 3167; BUOY ~ 7284
II. Quality Control1) Comparing the monthly SST for a ship or buoy with
climatology, if outside of 3 standard deviation, the observations in this month will be rejected.
2) Excluding any record outside of 4.5 standard deviation referring to monthly climatology from GDEM-V3.0.
3) Consistency checking by the EMSST linear regression with 99% confidence level.
SHIP : 2715 BUOY : 6835
6.2%14.3%
Depth (m) : • <40 • 40~80 • 80~200 • >200
Distributions of in situ daily SST observations through data quality control
Num:120198573
5944
Num:115426379
1795
3. Inter-comparison of GRHSST
Difference of annual SST referring to EMSST during the period of 2005.10~2006.9a.) MWOISST (b.) NGSST (c.) MGDSST (d.) OSTIA (e.) FSTIA
Spatial distribution of the annual difference spread relative to EMSST (left) and the averaged difference spread as a function of depths (right).
Note: if the sample is over 40 in a depth interval, a red circle can be shown.
5
1
25
1
)(1
i
M
jj
ij
ii
spread
i
EMSSTSSTM
DEvaluate the difference about the five products:
MWOISST NGSST MGDSST OSTIA FSTIA EMSST
Bias 0.23 0.35 0.12 -0.04 0.06 0.15
RMSE 0.80 0.94 0.88 0.31 0.26 0.52
Table 2. Comparison of Bias and RMSE of the different SST products referring to buoy observations (ºC)
4. Assessment by in situ observations
Comparison of bias and RMSE of SST products relative to buoy and ship observations in different depth interval. The histogram denotes the bias, and the error bar denotes the corresponding RMSE.
)exp( 000)( dCBAT d Correct Increment:
Depth < 40 m
Comparison of Bias and RMSE through the Correction
Comparison of bias and RMSE of SST products through correction according to the depth scale. The histogram denotes the bias, and the error bar denotes the corresponding RMSE.
Bias (ºC)OSTIA/New
0.2/-0.01
-0.03/-0.04-0.02/0.02-0.05/0.0
Bias (ºC)FSTIA/New
0.4/0.070.07/-0.040.08/0.020.06/0.0
Bias and RMSE
OSTIA FSTIA New_OSTIA New_FSTIA
Te
mp
era
ture
(o
C)
-.2
-.1
0.0
.1
.2
.3
.4
.5
.6
.7
.8
.9
1.0
1.1
1.2
1.3
1.4
Buoy Ship Bias RMSE
OSTIA -.039 0.311
FSTIA .064 0.256
New_OSTIA -.005 0.308
New_FSTIA -.003 0.247
Table 3. Changes of Bias and RMSE through correction to Buoy SST (ºC)
Bias RMSE
OSTIA -.040 1.212
FSTIA -.014 1.289
New_OSTIA -.019 1.212
New_FSTIA -.093 1.294
Table 4. Changes of Bias and RMSE through correction to Ship SST (ºC)
Referring to buoy SST, the bias can be decreased an order, but the RMSE can be decreased by 1% (OSTIA) and 4% (FSTIA) of RMSE.
Comparison of Bias and RMSE through the Correction
Coatal Station ID
11201146
21234130
41344144
41526452
74157416
74249710
RM
SE
(o
C)
1.01.52.02.53.03.54.04.55.05.56.0
OSTIANew_OSTIAFSTIANew_FSTIA
BIA
S (
oC
)
-1.8
-1.2
-.6
0.0
.6
1.2
1.8
2.4
3.0
OSTIANew_OSTIAFSTIANew_FSTIA
Validation of the corrected SSTs by night SST Observations of 12 Coastal Stations
Comparison of bias and RMSE of OSTIA and FSTIA by exponent correction relative to costal station observations. On average, the RMSE can be decreased about 5% and 3%, respectively to OSTIA and FSTIA.
Mean: (ºC)
1.61.00.8-0.2
Mean: (ºC)
3.53.32.32.2
74157416
7424
5. Conclusions
Looking for a major data source to feed our data assimilation system for preoperational ocean forecast, we investigate the five GHRSST Level 4 products in the coastal and shelf sea around China. The five GHRSST products are compared to each other and to the in situ SST observations in the period of 2005.10~2006.9. Some efforts to correct the bias near coastal regions are made. The main results are:
1) These products are agreed with very well when water depth is larger than 80m. However, near the coast around China, large differences exist. For example, as the depth is less than 40m, the annual mean spread of them is nearly to 0.9ºC.
2) Comparing to in situ observations, most of them are excellent for reasonable bias and RMSE. Especially the RMSE of OSTIA and FSTIA (reference to buoy SST observations) are about 0.3ºC. However near the coast, i.e, the depth is less than 40m, the warm bias (>0.2ºC) and the considerable RMSE (>0.6ºC) are prevalent in the five products.
3) A bias correction scheme that uses an exponent function of depth is applied to two products (OSTIA and FSTIA). The correction can reduce bias effectively, however, the RMSE doesn’t reduced so effectively.
4) The current GHRSST L4 products, especially some of them, can be safely used as a major input of data assimilation system in the shelf sea around China. However, near the coastline where depth is less than 40m, they are not ready to be assimilated. It is necessary to make further correction or improvement.
Thank you !