OIL SPILL STATISTICS FROM SAR IMAGES IN THE NORTH EASTERN BALTIC SEA SHIP ROUTE IN 2007-2009

Sven Anderson, Urmas Raudsepp, Rivo Uiboupin

Marine Systems Institute at Tallinn University of Technology

ABSTRACT

A large number of illegal oil pollutions impose considerable threat to marine environment especially in marginal seas like the Baltic Sea. Illegal spills are mainly detected on essential navigation routes. The monitoring of Oil Spills (OS) using remote sensing imagery (SAR data) was performed on the northeastern Baltic Sea ship route. The pre-analyzed satellite images for detecting marine pollution were provided to marine surveillance agency in Estonia. Out of 137 detected potential pollutions 76 were confirmed by aerial surveillance missions within two-year period. OS were mainly of low confidence, had small area, low contrast with surrounding water and smeared edges. The entrance to the Gulf of Finland was classified as the area where illegal spills of oil and bilge water take place, mainly. Between 30-50% of actual oil pollutions are not detected by SAR.

Index Terms— SAR images, multisensory monitoring, oil spill detection, identification, characteristics, statistics

1. INTRODUCTION

The number of marine pollution arising from illegal oil dischargers from ship tank or bilge pumping is much greater than those spectacular ship accidents. Active microwave sensors deployed on satellites are an important tool in oil spill detection. Among the many different sensors synthetic aperture radar (SAR) is possibly the most suited for oil spill (OS) monitoring and identification, because of its high ground resolution and independence of weather condition [1]. Many studies have proved that radar images can provide information on possible location and extent of oil spills [2; 3; 4]. OS can be detected on SAR imagery because they have a dampening effect on the Bragg waves in the sea. When oil enters the water from a vessel or another facility, it initially spreads out and forms a continuous or cohesive patch on the water surface. OS patch of oil absorbs energy and dampens out surface waves, therefore backscatter in that area is reduced. The low backscatter areas on SAR imagery are defined as slicks.

Detection and identification of OS in SAR images is complex because of look-alikes, which often occur in low

wind condition. The SAR features of an OS and its surroundings depend on parameters like wind speed as well as on the amount and a type of oil. Former experience demonstrates that the best approach for operational monitoring of OS would be simultaneous employment of different remote sensing instruments (SAR, optical remote sensing, in situ measurement) [5].

In every country, marine surveillance agencies are responsible for oil spill combating and on identification of illegal polluters. They rely on information that has been provided on potential oil spills by responsible institution. Illegal spills are mainly detected on essential navigation routes. Estonia joined marine oil spill detection service on 16.04.2007. EMSA (European Maritime Safety Agency) provides the countries participating in the project and the European Commission pre-analyzed satellite images for detecting marine pollution and performing surveillance of EU waters. Satellite surveillance enables the receiving of an early warning about marine pollution (henceforth: pollution alert message) for 98% of sea areas under the responsibility of Estonia. Satellite surveillance covers the north-western waterway of the Baltic Sea in its entirety The aim of current study is to perform statistical analyses of oil pollutions in the North Eastern Baltic Sea shipping route in 2007-2009.

2. METHODS

In Estonia, the satellite image is ordered through the Rescue and Coordination Centre of the Police and Border Guard Board (JRCC). The operator receives the maps about the planned images and the exact times of the over flights from EMSA generally 1-2 months before the scheduled flights. The operator orders all images that cover Estonian waters and the waterway of northeast Baltic Sea. The satellite images cover 80-99% of Estonian waters. 95% of the images are taken from areas where the marine traffic is intense. This system of offering and ordering images has developed through practice and due to good cooperation between JRCC Tallinn and EMSA. The role of JRCC operators has increased importantly since their suggestions are taken into account about the offering of images.

According to EMSA schedule there are 6-9 over flights per week on an average (12-15 images during spring-

1883978-1-4244-9566-5/10/$26.00 ©2010 IEEE IGARSS 2010

figure 1 OS and look-alikes statistics on Baltic Sea ship route shared to five sectors from South 1, 2, 3, 4, 5

summer and 4-5 images during autumn-winter) in the responsibility area of Estonia and above the waterway of the northeast of the Baltic Sea. The images are taken by different satellites of the European Space Agency (Envisat, Radarsat-1, Radarsat-2) from where the information is sent to satellite receiving stations in Norway and Italy. The operators working in the receiving stations carry out the first analysis about the pollution alert message. First, the raw image is conferred and analyzed whether the image shows actual pollution or it only „looks like pollution“. Afterwards the operator analyses the size, amount, probability (low, medium, high), etc. of the pollution. The pollution alert message is forwarded to the contact points of all countries by e-mail and short mobile message.

JRCC receives the pollution alert message and performs the necessary actions to check the pollution message. Since the image is dependent on several side factors, the first image received cannot be perceived as pure truth, but must always be checked. It is possible for the country in question to direct its resources to the area of pollution in order to identify the pollution and profile the possible source of pollution. JRCC first considers the most important information: whether the pollution has low, medium or high probability. In general the surveillance method is chosen according to the probability. Routine procedures include: 1) interviewing ships close to the area of pollution; 2) in case the patrol boats of the neighboring country are close to the area of pollution, JRCC requests operative professional assistance in the identification and detection of pollution; 3) implementing an aircraft and a vessel of the Police and Border Guard Board in the identification of pollution. The operator of JRCC Tallinn is obliged to inform EMSA

contact point of whether the pollution alert message was checked and whether it was confirmed.

In total 775 SAR images were ordered for the North Eastern Baltic Sea from EMSA from the period of April 2007 - April 2009. During this period we got altogether 156 SAR image reports. Potential pollution was marked on 137 images. OS detection was performed on imagery that was acquired by ENVISAT/ASAR and RADARSAT-1/2 sensors. The presented algorithm has three main parts: detection of dark spots, feature extraction and dark spot classification [6]. The potential OS are marked on the images. In addition, the output consists of slick co-ordinates, slick area and confidence estimate. The confidence measure is computed based on the computed features and tries to resemble the rules used by operators at Kongsberg Satellite Service (KSAT). Possible confidence level assigned to each OS had low, medium or high level. High confidence was assigned if the wind level is moderate or high, the slick has strong contrast, surroundings were homogeneous and source can be near.

Out of 137 detected potential pollutions 76 were confirmed by aerial surveillance missions, 23 hits were in Latvian or Lithuanian waters where aerial surveillance missions cannot be performed and 38 were not checked because of no confirmation from the ships in the area was received and the threat to the Estonian coast was estimated negligible.

The characteristics of potential slicks (frequency, size, confidence, shape, edge, contrast) on preprocessed images for the shipping area from Liepaya (sector 1) to Vaindloo (sector 5) were analyzed (figure 1). The area was divided into 5 sub regions, as the shipping activity is different for each sub region decreasing from the Baltic Proper towards the eastern Gulf of Finland.

3. RESULTS

Potential pollution was marked on 137 images out of 156. The largest number of potential pollutions, 49, was observed in the northeastern Baltic Proper at the entrance to the Gulf of Finland, sector 3 (figure 2). The number of potential pollutions was about half of that in the western Gulf of Finland, sector 4, and in sector 1, 27 and 26 respectively. Monthly distribution of potential slicks shows well defined seasonal distribution. Illegal pollutions have the lowest frequency during January-March and have high value from April until October. During the last period July is an exception with 8 pollutions encountered. Also somehow exceptional is rather high number of pollutions in December.

Spatial distribution of monthly pollutions shows clear areal pattern (figure 2). In the northern Baltic Proper that is the southern area of our region, the oil slicks are observed during autumn, winter and early spring, while no pollutions are detected in summer. In the Gulf of Finland the slicks are more frequent in spring, summer and early autumn, but

1884

absent during winter. This reflects intensity of the shipping, as the central and eastern part of the Gulf of Finland is covered by ice in winter.

The majority of potential oil spills had area less than 1 km2 (figure 3). There were only 6 OS with the area that exceeded 10 km2. Larger spills were identified at the entrance to the Gulf of Finland, in sector 3. There was also the largest number of potential oil spill. The analysis of the OS’s operational mapping obtained from April 2007 to April 2009 shows that the meteorological and oceanographic conditions were favorable. The detected OS’s were of low or medium confidence and contrast. The spills had mainly stripe/tail shape that did not have sharp edges and strong contrast with surrounding sea. It shows that potential spills had low concentration of oil and/or had been in water over longer period.

About 94% of potential oil pollutions on SAR images were classified as low and medium confidence. In 76 cases the OS were confirmed by aerial surveillance using Side Looking Air Radar (SLAR) onboard of airplane, which is about 55% of all SAR detections. A part of the OS detection and identification problem is to distinguish oil slicks from other natural phenomena like seaweed, foam, and other organic material. Low wind area, algae, upwelling, internal waves that appear as dark structures on the sea surface.

There are still a number of actual oil pollutions that are not detected by SAR. In 2008 69 actual oil pollutions was detected in the study area (figure 4). Only 31 of them were detected by SAR and confirmed by SLAR. Meanwhile 49 potential oil pollutions were presented on SAR images. It gives that 45% of actual pollutions were detected by SAR on one side and about 40% of SAR detections were look-alikes or a very minor pollutions. The same proportions were true for the year 2009, also. Since April 2007when offering of SAR images started, the number of confirmed oil pollutions was 25 out of 70. The number of detected actual pollutions was 99, but we like to note that this number represents whole year while SAR images were available since April.

figure 2. Number of OS in different sectors from Apr 2007 to Apr 2009.

figure 3. Distribution of potential OS by area.

figure 4. Annually detected oil pollutions using different methods, potential oil pollutions identified from SAR images and confirmed by SLAR in the northeastern Baltic Sea. In 2007 the number of potential oil pollutions and confirmed pollutions refers to the period from 16 April.

4. CONCLUSIONS

During the remote sensing monitoring from 2007 to 2009 about 76 OS’s of different classification and characteristics were detected and confirmed in the northeastern Baltic Sea ship route. Potential spills were mainly of low confidence, had small area, low contrast with surrounding water and smeared edges. The entrance to the Gulf of Finland was classified as the area where illegal spills of oil and bilge water take place, mainly. Therefore, it is important to map and analyze the areas of more frequent illegal pollution and to increase aerial surveillance there. An important part of OS’s detection and identification is separation of oil slicks from look-alikes. On the other hand, the SAR imagery could

0 1 2 3 4 5 6 7 8 9 100

5

10

15

20

25

30

35

Area (km2)

Num

ber o

f OS

1885

not be the only tool for the detection of illegal oil pollutions, as less than 50% is detected by SAR.

5. ACKNOWLEDGEMENT

This work was partially supported by the Estonian Science Foundation (Grant 7633 and Grant 7283).

6. REFERENCES

[1] O. Lavrova, M. Mityagina, T. Bocharova, A. Kostianoy and V. Krovotyntsev, Multisensor approach to operational oil pollution monitoring in coastal zones, Space Research Institute, RAS, “Planeta”, Moscow, 2008. [2] M. Gade and W. Alpers, “Using ERS-2 SAR images for routine observation of marine pollution in European coastal waters”, Sci. Total Environ., 237, pp. 441 - 448, 1999.

[3] H.R. Gordon, O.B. Brown and M.M. Jacobs. “Computed relationships between the inherent and apparent optical properties of a flat, homogenous ocean”, Applied Optics, 14, pp. 417 - 427, 1975. [4] M.L. Heron and A.Prytz, “Coherence scales for microwave Bragg scatter” OCEANS '96. MTS/IEEE. Prospects for the 21st Century. Conference Proceedings, III, pp. 1400 -1405, 23-26 Sept. 1996. [5] R. Uiboupin, U. Raudsepp, L. Sipelgas, “Detection of oil spills on SAR images, identification of polluters and forecast of the slicks trajectory”, 2008 IEEE/OES US/EU-BALTIC INTERNATIONAL SYMPOSIUM, pp. 270-274, 2008. [6] A.H.S. Solberg, S.T. Dokken and R. Solberg, “Automatic detection of oil spills in Envisat, Radarsat and ERS SAR images”, EC project Oceanides, 2002.

1886