20 04 _winte r
15
16
12
5
8
0
4
-5
-10
0
0
3 20
NCDC/NOAA
36 0
40 0
0
4 40
10
WIndSpeed
5
2000
20
10
0
0
5
10
15
20
Observed Surf. windspeed(m/s)
25
0
4
8
12
Liquid Water Content(g/cm**3x10)
16
Estimated precipitation compare with surface observation in past 5
years, most satellite retrieved rain rate are less than ground truth.
STN-46695 all day
6
Day's parameters 3pct
8
4
Obse rved Pre cipitation (mm/hr)
5
Night's parameters 3pct
4
3
2
1
12
20
16
16
12
12
Ocean allday's 7TB
8
8
Day's 7TB
4
4
0
0
4
8
Estimated P re cipitatio n(mm/hr)
20
0
2
4
Estima te d Precip ita tion (mm/hr)
0
0
6
4
8
12
16
Estimated P recipita ti on(mm/hr)
20
0
nnssmi
-4
8
-8
8
-12
4
Day's Tb
4
-16
0
0
0
4
8
12
Estimated P re cipitatio n(mm/hr)
16
-20
0
4
8
Estima te d Precip ita tion (mm/hr)
12
4
8
12
16
20
16
4
12
12
0
Bias of Estim ated R ain ra te(mm/h r)
12
16
Neural network is used in this research, which include 7 channel
observations and 3 Polarized Correct Temperatures (PCT) are
applied.
1500
30
0
0
Algorithm
Bias of estim ated windspeed(m/s))
15
0
Monsoon area covered East Asia in winter season.
1000
Bia s for estim ated rain ra te (mm/hr)
Data
Take surface observation over ocean as ground truth, where is
weather station 46695 located on an island in East China Sea north
of Taiwan 60Km away. The winter season from Nov. to Mar. is
northeast monsoon season and raining in most part of this season.
Cloud type is stratus cloud; raining area is more uniform than
cumulus cloud type.
20
Obse rve d Precipitatio n(mm/hr)
Using SSM/I may retrieve precipitation by different ways.
Validation over ocean is a tuff task for most of researchers. The
comparison of many algorithms was studied (Nazzareno, 2004). For
a little long term monitoring from satellite might be useful on
climate research.
500
40
Observed Precipitation(mm/hr)
Introduction
Estimated Surf. Windspeed(m/s)
25
Observed Precipita tion(mm/hr)
2
20
10
40
20
Peter K.H. Wang1 , Lei Shi2
1
Central Weather Bureau
Time sequence for 3 parameters
WIndSpeed(m/sec)
Estimated Rain Rate(mm/hr)
Liquid Water Content(g/cm3)*10
20
60
Bias of WIndspeed(m/s)
Assessment of Precipitation Characters between
Ocean and Coast area during Winter Monsoon in
Taiwan
12
8
4
0
-4
0
4
8
12
16
Obse rved Precipitatio n(mm/h r)
20
0
4
8
12
16
Observed Precipitatio n(mm/hr)
20
Using a few data sets as input data in Neural Network, the
relationship between estimated and observation are shown as
Input data Day
set
All
Day
7TB
Day
2PCT
Day
PCT85
Day
7TB
Night
All
Correlation 0.16
0.255
0.439
0.268
0.367
0.519
Total precipitation (mm) from year 2000-2004
Seasonal
2000
2001
2002
2003
2004
Ground obs.
652.5
432.1
390.5
460.3
670
No of Sat day 20
44
103
106
94
Est. Rainfall
234
746
799
661
203
Conclusion
Most of the winter season, in northeast area of Taiwan is covered by
cloud and rainfall all day.
Results
50
20 00 _winter
40
20 01 _winter
40
30
30
20
20
10
10
0
0
320
36 0
40 0
440
320
360
400
440
360
400
440
40
20 03_winter
40
20 03 _winte r
30
30
20
20
10
10
0
0
3 20
360
400
4 40
320
Trace these 5 year, retrieved precipitation from SSM/I by neural
network have a reasonable results, but it cannot identify the
extremely precipitation. General speaking retrieval rain rate from
satellite is less than ground truth. Because of the satellite path go
through one place twice a day, it may miss many rainfall during one
day, take the mean value of satellite retrieval rain rate to get a day’s
precipitation is uncorrected. For at least one third of the satellite do
not go through test area, and most surface heavy rainfall do not have
SSM/I pass went through, it will cause huge variance on climate
study.
The bias of rainfall retrieval in all day’s ocean area is the best one.
If data set is divided as day and night two parts, the bias is increase.
The bias of estimated rain rate is increasing with rain rate, which is
less estimated. By the way to retrieve surface wind speed is higher
estimated than ground truth about 5 m/sec, and it is increase with
wind speed.
There are not so much evidences to prove that local Land Sea
wind may introduce more precipitation