The Thirteenth Workshop of OMISAR
Development of a ship-boarding operational satellite remote sensing system
Zhihua Mao， Qiankun Zhu, Delu Pan
LOPSO, Second Institute of Oceanography, State Oceanic Administration, Hangzhou, 310012, China
Abstract surface temperature, Temperature error control
Satellite remote sensing technique can provide technique
ocean environmental information in a large scale
and in near real-time. It has been a key tool to study
the ocean. A land-based satellite receiving station 1. Introduction
can receive remote sensing data covering a large Many fishery companies are interested in
area, while the area is still very small comparing to harvesting sleeve-fish in the North Pacific and send
the global scale. Most of the Ocean is far away hundreds of boats to fish there every year. They
beyond the data coverage of a land-based satellite need routinely operational ocean environmental
receiving station. When the ocean environmental charts to locate fishing areas. Satellite remote
information is needed, a nice idea is to install the Sensing can provide some sea surface
satellite remote sensing system in a boat. When the environmental products in large scale in near
boat moves to a ocean location, the station can real-time. As the North Pacific is for away beyond a
receive the satellite remote sensing data in a large HRPT land-based ground station can receive
area around the ship. This kind of the system is very satellite data which cover the areas. A good solution
useful in many fields. The fishermen need the is to establish a ship-boarding satellite telemetry
system to help them locate the best candidate of receiving station installed on a fishing boat. The
fishery areas. One satellite remote sensing system system can receive the remote sensing data about
was developed and installed in a fishing boat and 4000×4000km2 areas around the boat. It covers all
served for fishing in North Pacific fishery areas the fishery areas when the boat stays in the North
when the boat stayed there. The system can provide Pacific. The system was established to serve for
some oceanic environmental charts such as sea many fishery companies to fish in the North Pacific
surface temperature (SST) and relevant derived fishery.
products which are the most popular use in fishery
industry.
The accuracy of SST is the most important 2. The Operational Fisheries Oceanography
System
and affects the performance of the operational
The operational fisheries oceanography system
satellite remote sensing system, which is found to
was established on a boat named ZhouYu 1301 to
be dissatisfactory. Many factors affect the accuracy
receive several kinds of satellite data covering the
of SST and it is difficult to increase the accuracy by
North Pacific fishery areas. As the ship-boarding
SST retrieval algorithms and clouds detection
system is expensive and it can receive the remote
techniques. A new technique of temperature error
sensing data covering the whole fishery areas, the
control is developed to detect the abnormity of
data should be used for many fishery companies
satellite SST. The performance of the technique is
besides the boat. To do this, the data must be
evaluated to change the temperature bias from
transmitted to the main land in real-time. A
–3.04 to 0.05℃ and the root mean square (RMS)
ship-boarding satellite communication system was
from 5.71 to 1.75℃. It is very useful to be
installed on the same boat to transmit the remote
employed in an operational satellite- measured SST
sensing data to a land-based data processing system
system.
in real-time. The land-based system processes the
data to produce fishery environmental charts which
Keywords Satellite remote sensing system, Sea
serve for many fishery companies.
14-1
Development of a ship-boarding operational satellite remote sensing system
The whole operational fishery system receiving station by a hub and takes in the remote
composes of three sub-systems which are a sensing data by ftp. It composes of three modules
ship-boarding satellite telemetry receiving station, a which are oceanography environmental maps
ship-boarding satellite communication system and a producing module, computer- aided fishing decision
land-based ocean fishery charts producing system. module, and weather monitoring module.
It routinely receives three kinds of satellite remote The maps module can process about twenty
sensing data which are the Advanced very High tracks of satellite data received in a day to generate
Resolution Radiometer (AVHRR) on the NOAA a SST image. The image is occupied by many
series of weather satellites, Chinese meteorology patches of empty data because too many clouds
satellites called Fen-Yun (FY-1C/FY-1D) and always stay in the atmosphere of the North Pacific.
Sea-viewing Wide-field-of-view Sensor (SeaWiFS) To produce a useful SST image, it needs to fuse all
to produce the fishery charts. A sketch map of the SST images during a week into a better image. It
system is shown in Fig.1. still has many empty data in a weekly-SST image
and a cloud replacement technique is employed to
remove empty data. It is very important to develop
a good cloud replacement technique in an
operational fishery system. The technique can use
the temperature from history and nearby areas.
Some ocean dynamic information should be
extracted from SST images since fishes often
aggregate near the ocean frontal regions. Some edge
detectors such as Laplacian of the Gaussian can be
used to extract oceanic fronts and eddies (Simpson
1992). The information of fronts is stored
Fig.1 A sketch map of the operational fisheries automatically in a database. The system is very
Oceanography system
useful because it can display the distributions of
SST in a large scale in near real-time in the fishery
The data flow of the system is as following.
fields. The fishermen can analyze fishery situations
The sensor on the remote sensing satellite such as
from the SST images and the changes of SST, and
NOAA and FY-1C receives the radiance emitted
find the best candidate fishery centers according to
from sea surface and translates to digital signals
the ocean environmental conditions.
which are broadcast down. The ship-boarding
The computer-aided fishing decision module is
HRPT receiving station receives the signals and
a useful tool in scheduling the fishing plan. It can
transmitted to the land-based data processing
input the locations of fishing boats and display the
system through a communication satellite such as
information of the boats on the SST images. As the
Intelsat. The land-based system produces fishery
candidate fishery centers are relevant to the fronts
charts and faxes to the fishing boats in the North
and eddies which are clearly shown in the SST
Pacific. The system can serve for many fishing
images, the relationship between the boats and
boats.
fronts are directly shown in the images. The lines
As the charts received by the boat become a
from the boat to the candidate fishery fields can be
simple paper and the remotely sensed data can be
drawn on the image, then the distance and direction
used directly in the boat in real-time, Another
are automatically calculated. The fishermen can
subsystem was established in the same boat, called
easily evaluate the situations of several candidate
the ship-boarding real-time fishery analysis system.
fishery centers and decide which one is the best.
The subsystem is connected to the satellite
The system will tell them the time and money to go
14-2
The Thirteenth Workshop of OMISAR
there. It is only take a short time to do for many normal for a system to stop working occasionally
boats. The fishermen can benefit from the system if and the important thing is to recover it as soon as
they want to change fishery places. possible. A remote trouble tracking technique is
The safety of the boat is the most important employed to quickly locate the problem. The way is
and a typhoon often destroys the boats. In other side, to record the status of some important parts of the
fishermen often catch much larger amount of fish system to a file and send the file back every day. We
just before the typhoon comes. It is always difficult can see the working situations of the key parts such
to decide when to leave and where to go to avoid as the receiver antenna, gyro, GPS, UPS, computer,
the typhoon. As the system can display clearly tape driver, software, etc from the file. It helps us to
many information of the typhoon such as the find out the problem and tell the operator how to
location, the range and the distribution, the solve it. We tried to solve the abnormal status of the
movements of the typhoon can also be tracked by a system three times and keep the system running
series of clouds images, it can help to forecast the during the fishing season. It is a good way to
development of the typhoon and make the schedule improve the reliability of the system by recording
before the typhoon comes. The system is very the working situations of key parts of the system in
useful in monitoring the catastrophic weather events a file and sent back to be checked for a
to ensure the safety of the boats. ship-boarding operational system.
The ship-boarding operational system is The operational fisheries oceanography system
installed on the fishing boat about 500 tons. The can provide many kinds of oceanic environmental
boat is a big one as a fishing boat, but a small one to products such as sea surface temperature (SST),
install two systems. The diameter of the antenna of chlorophyll, clouds, aerosol optical depth, etc. SST
the satellite communication system is 4 meters, and relevant derived products are the most popular
almost the same as the width of the boat. The boat use in fishery industry and provided routinely by
swags heavily on the sea and make an operator feel the system. A sample image is shown in Fig. 2. The
difficult to type a word. As a fisherman, the image is a very big picture with 3900 pixels wide
operator has little knowledge of remote sensing and and 1320 pixels high. It tell the fishermen all details
computer. He can’t finish some intricate data of ocean environmental conditions. The
processing of remote sensing by himself. Automatic distributions of temperatures, fronts, currents,
data processing is the key to keep the system eddies are clearly shown in the image. The
running as an operational system in such conditions. fishermen like to see the images daily and know the
A lot of automatic data processing functions are development of ocean dynamic information by a
finished to realize the automatic operation of the series of images. As the fishermen are familiar with
system such as geometric projecting and the relationship between the fishery situations and
registration, cloud detecting, variables inversing, the SST images. The images are very useful to help
data fusing, file managing and charts producing, etc. them locate the ocean fishery centers as soon as
It only needs the operator to input a short command possible.
to finish a daily work. Automatic operation is a key
requirement for a ship-boarding operational remote
sensing system.
Reliability of the system is another key
requirement for an operational system, especially
the system is far away in the North Pacific. The
boat stays on the sea from May to December for 7 Fig. 2 A sample image of fisheries oceanography charts
produced by the system
months. It cannot return even the system breaks
down and no maintenance experts on the boat. It is
14-3
Development of a ship-boarding operational satellite remote sensing system
3. Satellite-Measured Sea Surface Temperature produce SST imagery in the operational fishery
Sea surface temperature (SST) is the most system. The accuracy of the images is the most
important variable in fishery charts. It can be important for fishermen to analyze fishery
derived from NOAA/AVHRR and Chinese situations. Hundreds of coinstantaneous
FY-1C/1D satellite data. Most operational inversion ship-measured temperatures are used to calculate
models of SST are based on split-window technique the temperature difference. The results seem to be
(McMillin 1984). Some widely used operational dissatisfactory, shown in Fig. 3. The temperature
algorithms are multiple-channels sea surface differences range from -17.2 to 5.0℃, much larger
temperature (MCSST)[3], cross-product sea surface than the reported results. The mean bias is -3.04℃
temperature (CPSST)[4], Nonlinear sea surface with RMS of 5.71℃.
temperature (NLSST) [4] and Pathfinder SST
5
(PFSST) algorithms[5]. The algorithms are based on
that the attenuation by water vapor is different in 0
Temperature bias/0C
each window channel, so that the total attenuation
can be estimated from the measured temperature -5
difference between any two channels. A common
equation is given by Walton as following[4] -10
SST AT4 BT4 T5 CSec 1T4 T5 D -15
Where A, B, C, D are the coefficients determined 0 100 200 300 400 500
Match-up points
by linear regression from the satellite-measured
temperature and in-site data. T4 and T5 are the Fig. 3 The distribution of satellite-measured
bright temperature of channel 4 and 5. is the temperature error
zenith of the satellite view, is the estimated
ocean temperature. The accuracy of SST can be achieved to be a
The accuracy of SST is a key to evaluate the nice result in some ideal conditions. In fact, many
performance of models. Thousands of match-up factors affect the accuracy of SST imagery. A
buoy-measured temperatures are used to calculate 0.05℃ bias of radiance calibration of the AVHRR
the temperature difference and root mean square channel 4 and 5 can lead to a 0.33℃ SST error by a
(RMS) between satellite-measured SST and buoy CPSST algorithm[6]. The sensor noises are also a
temperature. McClain evaluated the accuracy of source of temperature error, which are found to be
MCSST with 0.3-0.4℃ bias and 0.5-0.6℃ RMS scattered in the SST images. Different models
value[3]. Walton reported that the CPSST had a produce SST with different accuracy using the same
mean offset of 0.1℃ with RMS of 0.6℃ and data. The RMS values of 8 models range from 1.04
calculated the accuracy of NLSST with the bias of to 1.71℃[2]. Even the some model has different
0.05℃ and the RMS of around 0.6℃[4]. The accurate when it is applied in different regions,
accuracy of PFSST also has a nice result with the Emery applied PFSST in tropic areas and polar
bias of 0.02℃ and the standard deviation of regions with the RMS values larger than 2℃,
0.53℃[5]. The statistic results show that all above though the RMS was about 0.5℃ on the global
operational algorithms can retrieve SST images in a data[7]. The satellite measures the ocean skin
nice accuracy. Unfortunately, all algorithms are temperature and the in-situ data are bulk
used to calculate SST in cloud-free conditions, temperature. The differences range from 0.1 to
which is very difficult to be met in an operational 0.3℃ with variations of -0.5℃ to 1.8℃ depending
satellite remote sensing system. on wind speed and other environmental
Some published NLSST algorithms are used to conditions[6]. Even the in-situ temperature itself has
14-4
The Thirteenth Workshop of OMISAR
a RMS of 0.5℃[8]. The changes of atmospheric and satellite retrieval SST by comparing the temperature
oceanic conditions also affect seriously the difference. A suitable threshold is important to be
accuracy of SST. Undetected clouds are really the set to decide whether the SST is abnormal or not.
biggest source of SST errors[9] and it is very difficult The threshold values of 10,6,3,1 and 0.5℃ are
to be avoided. The operational algorithms are based tested. A 10℃ threshold can detect some abnormal
on a cloud-free condition and developed to only SST and the mean temperature bias decreases from
correct the effects of the attenuation caused by -2.43 to -1.62℃. The values of 6,3 and 1℃
water vapor, other factors can also easily cause continue to decrease the bias of -1.0, -0.66 and
temperature errors larger than 1℃. It is really -0.49℃ respectively, while 0.5℃ threshold
difficult to limit the accuracy of SST within 1℃ in increases the bias to -0.65℃. A small threshold will
an operational satellite fishery system. increase the accuracy of SST.
As the accuracy of ship-measured temperature Threshold also affects the oceanic fronts. As
is ensured within 0.5℃, the temperature errors in the standard temperature maps are derived from
Fig.3 come from satellite derived SST. Since the 15-year SST data and become spatial homogeneous,
errors seriously affect the performance of the which weak the magnitude of oceanic fronts. The
operational fishery system, they must be removed real-time SST images are very sensitive to the
as much as possible. Clouds make about 80% oceanic fronts which are very useful in locating the
satellite infrared data invalid in calculating SST and fishery center. Some threshold values are tested on
undetected clouds cause big satellite derived the effects of the oceanic fronts. If the threshold is
temperature errors. As a 100-metre thickness clouds larger than 6℃, no changes are found in the frontal
can almost absorb the whole infrared radiance region of the SST images. A small threshold will
emitted from sea surface, the satellite will measure take the oceanic front temperature as an abnormity.
the temperature of clouds instead of ocean If the threshold is smaller than 6℃, some
temperature. Many techniques have developed to temperature data of oceanic fronts are removed and
detect clouds. Still it is difficult to mask a 10-metre a 1℃ threshold will remove most of data in the
thickness cloud or a sub-pixel cloud by a nice cloud front regions. Since oceanic fronts are the most
detecting technique, while any residual cloud important information in the fishery charts, we set
contamination will lead a big satellite-measured the threshold of 6℃ in detecting the SST
SST error. It is obvious that the cloud detection abnormity.
methods cannot screen all kinds of clouds and the The performance of temperature error control
SST algorithms perform well only under the is evaluated using ship-measured temperature,
cloud-free conditions. Both cannot ensure to derive shown in Fig.4. Fig.4a shows the comparison
SST with high accuracy in any conditions. Other between the satellite-measured SST and ship SST
methods should be developed to ensure the before using temperature error control technique
accuracy of SST in an operational fishery system. and Fig.4b is the result after the technique is
employed. From Fig.4a, most of match-up points
locate around the diagonal line, while some points
4. Temperature Error Control Technique fill in region A where the satellite-measured
A new technique is developed to control the temperatures are obviously lower than the
temperature error, which employs a series of ship-measured temperatures. The ship temperatures
standard SST maps. The maps come from NOAA range from 13 to 18℃ and the average is 16.47℃,
15-year mean SST data. As the annual changes of while the satellite SST range from 0 to 19℃ and the
ocean temperature at the same place normally limit average is 13.38℃. As the clouds contamination
within 5℃. A daily standard temperature maps are can reduce the radiance emitted from the sea
carefully prepared to judge the abnormity of surface and low down the sea surface temperature
14-5
Development of a ship-boarding operational satellite remote sensing system
received by satellite, the lower satellite-measured removes all of the satellite SST abnormity points
SSTs in region A are mainly caused by undetected distributing in region A of the Fig.4a. The technique
clouds. The characteristics of these clouds in is efficient in detecting the satellite SST abnormity
satellite data are very similar to that of the sea and improves the performance of the operational
surface and they are very difficult to be masked by fishery system. Some statistic results before and
clouds detection techniques. after using the technique are shown in table 1.
The effects of standard temperature control
technique are demonstrated in Fig.4b. It almost
30
30
28
28
26
26
24
In-situ temperature/0C
In-situ temperature/0C
24
22
20
A 22
20
18
18
16
16
14
14
12
12
10
10
8
8
6
6
4
4
2
2
0
0
0 5 10 15 20 25 30
0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30
0
Satellite measured SST/ C Satellite measured SST/0C
Fig.4 The comparison of satellite-measured SST and in-situ temperature. Fig.4a is the result before applying the
temperature error control technique, where region A is the distribution of satellite derived SST abnormity. Fig.4b is
the result after using the technique.
14-6
The Thirteenth Workshop of OMISAR
Table 1 Temperature accuracy comparison before and after employing the temperature abnormity control
technique
Before SST abnormity After SST abnormity detection No of
Match-up
Date detection
bias/℃ RMS/℃ R bias/℃ RMS/℃ R
Jun. 1 - Jun. 29 -4.14 7.26 -0.55 0.03 1.86 0.243 40
Jun. 30-Jul.15 -12.8 12.5 0.328 -1.34 2.25 0.914 18
Jul.16- Jul.29 -4.67 6.17 0.389 1.07 1.54 0.931 46
Jul.30-Aug.14 -1.07 3.23 0.508 -1.01 2.06 0.803 54
Aug.15-Sep.2 -5.44 7.01 0.427 0.89 1.57 0.873 52
Sep.3- Sep.9 -2.29 5.28 0.140 -0.19 1.47 0.809 62
Sep.10- Sep.16 -4.92 7.26 0.220 -1.23 2.00 0.764 50
Sep.17- Sep.23 -2.95 3.11 0.954 -0.30 0.95 0.936 36
Sep.24- Oct.7 -3.89 7.40 0.239 1.90 2.34 0.806 55
Oct.8- Nov.11 -0.49 1.95 0.802 -0.28 1.18 0.823 61
Average -3.04 5.71 0.399 0.05 1.75 0.821 474
From Table 1, we know that the After the technique is used, the average of
satellite-measured SST are often lower than the temperature bias changes from -3.04 to 0.05℃
in-situ temperatures and the temperature bias are and the average of RMS changes from 5.71 to
all negative before using the temperature 1.75℃.
abnormity control technique. Some bias are larger In the operational fishery system, the
than –5.0℃ and the largest one is –12.8℃. The threshold of 6.0℃ is used to detect the
RMS values are also very big. The smallest one is temperature abnormity, which cannot detect the
1.95℃ and the largest one is 12.5℃. The temperature errors with the difference lower than
correlation values are very low, most of them are 6.0℃. The oceanic fronts are very strong because
smaller than 0.5. A strange thing is getting a of Kuroshio in the North Pacific and the threshold
negative correlation value during 1-29 of June, must be set large. In other ocean regions, the
2001. The averages of bias and RMS are –3.04 oceanic fronts will be weak and it is suitable to set
and 5.71℃ respectively. The situation is improved a small threshold such as 3.0℃. It can detect
after employing the standard temperature control much more satellite retrieval SST errors and
technique. The temperature bias are around increase the accuracy of SST images. The
±1.0℃, and half of them become positive. The technique is efficient in improving the
technique is efficient in detecting some low performance of the operational satellite fishery
satellite retrieval temperatures. The RMS values system.
are much smaller now, most of them lower than
2.0℃. The correlation values become much larger
and the average changes from 0.399 to 0.821. 5. Conclusions
14-7
Development of a ship-boarding operational satellite remote sensing system
A ship-boarding operational satellite development and operational application of
remote sensing system is a good way to monitor nonlinear algorithms for the measurement of
the ocean environmental conditions in real-time sea surface temperatures with NOAA
where it is far away beyond a land-based satellite polar-orbiting environmental satellites. J
ground station can receive the data. The Geophys Res, 103: 27999–28012, 1998.
operational fishery oceanography system has been [5] Kilpatrick K A, Podesta G P, Evans R.
running for two years and is very useful for the Overview of the NOAA advaced very high
fishery industry in fast locating the fishery regions, resolution radiometer Pathfinder algorithm for
scheduling the best fishing plans and avoiding the sea surface temperature and associated
typhoon in the North Pacific. matchup database. J Geophys Res,
The accuracy of SST is the most important 106:9179-9197, 2001.
in applications and affected by many factors. [6] Brown J W, Brown O B, Evans R H.
Some advanced SST inversion algorithms and Calibration of advanced very high resolution
clouds detection techniques can improve it and it radiometer infrared channels:A new approach
is still difficult to detect some temperature to nonlinear correction. J Geophys Res, 10:
abnormity in an operational satellite SST system. 18257-18268,1993.
The temperature abnormity control technique, [7] Emery W J, Yu Y, Wick G A. Correcting
developed in the paper, is efficient in detecting the infrared satellite estimates of sea surface
satellite retrieval SST errors and increases the temperature for atmospheric water vapor
accuracy of SST images. It can improve the attenuation. J Geophys Res,99:5219-5236,
performance of a real-time operational fisheries 1994.
oceanography system. [8] Emery W J, Baldwin D J, Schlussel P,
Reynolds R W. Accuracy of in situ sea surface
temperature used to calibrate infrared satellite
Acknowledgements measurements. J Geophys Res, 106:
This research received supports from 2387-2405 2001.
National Science Foundation of China (40006011), [9] Simpson J J, Mcintire T J, Stitt J R, Hufford G
Chinese National 863 Program (818-11-02 and L. Improved cloud detecting in AVHRR
2002AA639220) and Chinese National 973 daytime and night-time scenes over the ocean.
Program (G1999043701). Int J Remote Sensing,22: 2585-2615, 2001.
References
[1] Simpson J J. Remote sensing and geographical
information system: their past, present and
future use in global marine fisheries. Fish
Oceanogr, 1: 238-280, 1992.
[2] McMillin L M, Crosby D S. Theory and
validation of the multiple window sea surface
temperature technique. J Geophys Res, 89:
3655-3661, 1984.
[3] McClain E P, Pichel W G, Walton C C.
Comparative performance of AVHRR based
multichannel sea surface temperatures. J
Geophys Res,90:11587-11601, 1985.
[4] Walton C C, Pichel W G. Sapper F J, et al. The
14-8