Transcript MODIS DATA

Study on Cloud Classifications and Shanghai
Heavy Rain Case by Terra/MODIS and
GMS-5 Satellite Data
Shi Chunxiang, Zhang Wenjian
National Satellite Meteorological Center CMA
Contents:
1.Cloud Classification Using Terra/MODIS
and GMS-5 Data
2.Case Study About Shangsai Heavy Rain
3.Conclusion
Cloud Classification Using
Terra/MODIS and GMS-5 Data
• Cloud classification using GMS-5 data with
neural network method
• Cloud mask and classification using
Terra/MODIS data with multi-spectral
threshold technique
Cloud classification using GMS-5 data
with neural network method
The field of neural networks can be thought of as
being related to artificial intelligence, machine
learning, parallel processing, statistics, and other
fields. The attraction of neural networks is that
they are best suited to solving the problems that
are the most difficult being solved by traditional
computational methods. We use the BackPropagation (BP) neural network in this cloud
classification study, which is used wildly in
many fields.
Cloud classification using GMS-5 data
with neural network method
• Sample database of clouds, land and water
is built based on GMS-5 four channel data
that include several thousands of one pixel
samples. Sample database is also collected
from the GMS-5 satellite imageries from
June to August in 1998.
• Features:20
• Trained neural network model structure is
20 input nodes, 2 hidden layers and 4 output
nodes (20-40-15-4).
海洋 陆地 低云 中云 卷云 薄卷云 多层云 积雨云
This is a clouds classification example using
neural network method, Left-down Figure is
cloud classification false color image of GMS-5 at
06:00(UTC) on July 26 in 1999，left-up is IR
image and right-up is VIS image at same time.
The region is 0。~60。N，70。~150。E.
Cloud mask and classification using
Terra/MODIS data with multi-spectral
thresholds technique
MODIS Bands Features Associated with
cloud classification
• MODIS (Moderate Resolution Imaging
Spectrometer) has 36 channels. It offers the
opportunity for multi-spectral approaches to cloud
mask and cloud classification more accurately.
The spectrum feature of the 36 channels is shown
in table 1. Many of them could be used to detect
and classify the cloud.
In the Table, Y
denotes used band,
and N denotes no
used band in cloud
mask and cloud
classification.
In the 36 bands,
band1, band2, band3,
band31 and band32
are similar to the
AVHRR bands.
Band1 , band27,
band31 and band32
are similar to GMS-5
bands.
MODIS Bands Features Associated with
cloud classification
• MODIS data has higher resolution than GMS-5 or
NOAA/AVHRR data. It is helpful for us to
understand the details of smaller scale weather
system.
• MODIS cloud classification examples will be
mentioned latter.
MODIS band1, 250M, 200108061030(BTJ)
GMS-5 VIS, 5km, 200108061032(BTJ)
The spatial resolution of the left MODIS image is 250m, is 20 times of right GMS-5
image.Through comparing this two images, we find that it included more details of cloud cluster
in the higher resolution MODIS image. It is helpful for us to predict the smaller scale rain storm
area. The area pointed by red arrow is the station which rainfall is biggest in the rain period
before several hours. There is a small vortex center, it is very clear, in the area pointed by the
yellow arrow.
Case study About Shanghai
Heavy Rain
from August 3 to August 7, 2001
Surface Observation Rain Data
Analysis
• Rain analysis:
We analyse the heavy rain occurred on August 6 in 2001 in
Shanghai of East China, one of the 973 CHeRES region,
using MODIS images and GMS-5 cloud classification
images. On August 5-6, Shanghai suddenly suffered the
biggest heavy rain since 1950. From 20:00(BJT) on August
5 to 20:00(BJT) on August 6, the rainfall of the 50003
station (south city district) in Shanghai city center is
329mm, and in several downtown area stations the rainfall
are more than 200mm and some are more than 100mm also.
The longest heavy rain which last about 10 hours from
16:00(BJT) on August 5 to 2:00(BJT) on August 6 (BJT).
The figure in next page shows the distribution of 24
rainfall from 20:00(BJT) on August 5 to 20:00(BJT) on
August 6, 2001.
This figure shows the distribution of 24 rainfall from 20 on
August 5 to 20 on August 6, 2001. We know from this figure
the largest rainfall in these period is 329mm pointed by black
arrow.
Some single station hour rainfall in the period
from 20 on August 5 to 20 on August 6, 2001.
It shows the periods of rain on this day.
These figures show
that the largest hour
rainfall is about
60mm.
Every hour GMS-5 cloud
classification image analysis
The following animation is evolvement
of cloud system by GMS-5 hourly
measurement from 8 on August 3 to 20
on August 7(BJT).
It needs about more then one minutes.
Please wait patiently.
This is the 24 hours temperature difference and
Shanghai is located negative area. It shows that there are
cold air came from the north.
The evolvement of cloud system from
GMS-5 hourly measurement reveals that:
1. The tropical cyclone cloud system born on
August 3 in East China Sea moved to west and
landed on the boundary between zhejiang and
fujiang on August 4 and than moved to northeast.
2. Then It went up north and suddenly turn to noutheast direction to move. And when it arrive the
Shanghai downtown on August 5 late, this system
suddenly and severely develop.
3. The tropical cyclone cloud system move very
slowly because subtropical high is vary stable and
so it stayed in the Shanghai region long period
and resulted in heavy rain-storm.
4. After the tropical cyclone cloud system land, it
move very slowly and weaken gradually. In this
process the cold air from the north systems merge
with tropical low-pressure system and with warm
and moist air from south-west monsoon. So the
different system interwork result in tropical lowpressure system enhance again and again.
5. The heavy rain lasts about 8 hours from 18
August 5 to 02 August 6 because the interwork
among tropical low-pressure system, cold air came
from north system and front system and warm and
moist air from south-west monsoon.
Analysis ShangHai heavy rain
using MODIS data
MODIS view the
Shanghai area
The figure is
combined image of
ch1/ch4/ch3.
The spatial
resolution is 250m.
The blue arrow
points Shanghai
city district.
This figure is cloud classification
image using MODIS data at 10:30 on
August 6, 2001(BJT)
In image, the green regions are land,
blue are water, red are cumulonimbus,
white are cirrus, and dark yellow are
low-level cloud.
MODIS combined image
of ch01/31/06
a t 10:30 on August 6,
2001(BJT)
The arrow represents the rain
area of ShangHai.
The different of cloud classification
methods among the GMS-5, AVHRR,
FY-1C and MODIS is that we can use
more useful information.
Ch1--0.695
Ch31--11.03
Ch6--1.64
Such as band6(1.64μm), because the differences in reflected solar
radiation between the 0.645 and 1.64μm bands contain information
regarding cloud particle phase due to distinct differences in bulk
absorption characteristics between water and ice at the longer
wavelength. if the cloud is composed of ice, or if the surface is snow
covered (similar in effect to large ice particles), then the reflectance
of the cloud at 1.64μm will be smaller than for an otherwise identical
liquid water cloud. In the left image the darker areas pointed to by
the red arrows are high level clouds.
Ch1--0.695
Ch31--11.03
Ch35--13.935
CO2 slicing is a useful method for sensing cloud amount and the height
of clouds. Simple tests using the CO2 channels are useful for cloud
classification, particularly high clouds. MODIS band35 (13.9μm)
provides good sensitivity to the relatively cold regions of the
atmosphere. Only clouds above 500 hPa will have strong contributions
to the radiance to space observed at 13.9μm; negligible contributions
come from the earth’s surface. In the left image the light areas arrows
are high level clouds. The red arrows represent Shanghai rain area.
In conclusion, we used much more
information from MODIS data to classify
cloud, for example: BT11, BT11- BT12, BT11BT8.6, BT13.9, R0.66, R0.87, R0.87/ R0.66, R1.6, etc.
MODIS band1, 250mm, 200108061030(BTJ)
GMS-5 VIS, 5km, 200108061032(BTJ)
The spatial resolution of the left MODIS image is 250m, is 20 times of right GMS-5
data.Through comparing this two images, we find that it included more details of cloud cluster
in the higher resolution image. It is helpful for us to predict more smaller scale rain storm area.
The area pointed by red arrow is the station which rainfall is biggest in the rain period before
several hours. There is a small vortex center in the area pointed by the yellow arrow.
The right one is the combined image of MODIS ch01/31/06
at 10:30 on August 6, 2001(BJT). The left is figure that the
TBB equal line overlap with TBB image. The area pointed
by yellow arrow is the station which rainfall is biggest in
the rain period before several hours
100
80
60
40
20
20
40
60
80
The right one is the combined image of ch01/31/06 at 10:30 on August 6, 2001(BJT).
The left is figure that the TBB equal line from GMS-5 data with 5km resolution.
The area pointed by yellow arrow is the station which rainfall is biggest in the rain
period before several hours. Through comparing this two figures , we find that it included
more details of cloud cluster in the the left figure .
Ch1--0.695
Ch18--0.936
Ch2--0.865
Ch31--11.03
Ch6--1.64
Ch35--13.935
ch1
ch2
ch3
ch4
ch5
NOAA-16 AMSU-B Bright Temperature 20010805-0512 (UTC)
ch1
ch2
ch3
ch4
ch5
NOAA-16 AMSU-B Bright Temperature 20010806-0502 (UTC)
ch1
ch2
ch3
ch4
ch5
NOAA-16 AMSU-B Bright Temperature 20010807-0452(UTC)
• 雨情：从8月5日傍晚开始，上海突然遭受自解放以来的最大暴雨
袭击，到8月6日14时，上海市中心黄埔区的降雨量已经达到294
毫米，市区其他地方的降水量也大多在100毫米以上。
• 天气分析：8月3日生成于东海南部的热带低压云团在向偏西移动
中于8月4日早在江闽交界处登陆，然后北上经皖南后突然转向偏
东方向移动，5日晚到达上海市区时云团突然强烈发展。因此时
海上副高较为稳定，导致云团移动特别缓慢，长时间滞留在上海
地区，造成强烈降水。
• Weather Analysis:
When the cloud cluster of tropical low-
pressure system born on August 3 in East
China Sea moved to west, it landed on the
boundary between zhejiang and fujiang and
than moved to northeast. then It went up
north and suddenly turn face east direction
to move after wannan. And when it arrive the
Shanghai downtown on August 5 late, this
system suddenly and severely develop.
The cloud cluster move very slowly
because subtropical high is vary stable and
so it stayed in the Shanghai region long
period and resulted in heavy rain-storm.
ch1
ch2
ch3
ch4
ch5
NOAA-16 AMSU-B Bright Temperature 20010805-1312 (BJT)
ch1
ch2
ch3
ch4
ch5
NOAA-16 AMSU-B Bright Temperature 20010806-1302 (BJT)
ch1
ch2
ch3
ch4
ch5
NOAA-16 AMSU-B Bright Temperature 20010807-1252(BJT)
lassification image,
61032(BTJ)
t the Cb areas have
the Shanghai.
The right one is the combined image of MODIS ch01/31/06
at 10:30 on August 6, 2001(BJT). The left is figure that the
TBB equal line overlap with band31 TBB image. The area
pointed by yellow arrow is the station which rainfall is
biggest in the rain period before several hours
• Conclusion:
1. The 36 channel MODIS data offers the opportunity for
multi-spectral approaches to cloud mask and cloud classification
more accurately.
2. The MODIS high spatial resolution is useful for studying the
details of smaller scale system.
3. The GMS-5 high time resolution is useful for understanding
the evolvement of system.
4. The heavy rain which lasted about 8 hours from 18 August 5
to 02 August 6 in Shanghai because 1) the interwork among
tropical low-pressure system, cold front system and south-west
monsoon, 2) The tropical cyclone cloud system move very
slowly because subtropical high is vary stable and so it stayed in
the Shanghai region long period and resulted in heavy rain-storm.