Designing an Effective Monitoring System Based on Satellite Sensor for the Oil Spills Detection in Lake Maracaibo, Venezuela

Towards Horizon 2020 Lasaponara R., Masini N., Biscione M., Editors EARSeL, 2013

Designing an effective monitoring system based on satellite sensor for the oil spills detection in lake Maracaibo, Venezuela

Giovanni Laneve, Giancarlo Santilli, Pablo Marzialetti and Lorenzo Fusilli

University of Rome, Department of Astronautics, Electrical and Energetics Engineering, Rome, Italy; [email protected]

Abstract. Lake Maracaibo (Venezuela) constantly suffers some serious environmental problems, in particular the oil spill resulting from intensive hydrocarbons extraction. The state-run oil company Petroleos de Venezuela S.A. (PDVSA) commissioned at the DIAEE (La Sapienza, Universi- ty of Rome) a feasibility study aiming at designing a monitoring system customized for the Lake Maracaibo, based on the use of satellite sensors, for the detection of the oil spill. However the peculiar meteorological, environmental and operational conditions present in the study area do not allow to realize a suitable and cost-effective monitoring system for the remote detection of the oil spills. We describe the results and the reasons respect of which we can state that such a monitoring system cannot be achieved. Measurement campaigns was carried out during the period February- March 2011 with the aim of analyzing the spectral behaviour of the oil when it floats on the lake surface and then define the channels of the electromagnetic spectrum in which it is more easy to distinguish oil contaminated water surface. The set of instruments used during the campaign was selected in order to cover the whole optical region (0.35 - 14 m) of the electromagnetic spectrum that is usually covered also by satellite sensors. Furthermore, this paper shows the results obtained by analyzing both the "in situ" data acquired during the measurement campaign, and the satellite data acquired from the satellite ENVISAT (ASAR and MERIS), MODIS and Landsat in the period of the campaign.

Keywords. Oil spill, satellite, Lake Maracaibo, water pollution, remote sensing.

1. Introduction

Usually we tend to think that the oil spills is caused by naval disaster like shipwrecks. However, several studies on marine oil pollution estimated that annually about 52% [1÷4] of the oil dispersed in the environment comes from the illegal discharges of hydrocarbons intentionally released by ships into the marine environment along transportation routes, accidents on oil platforms, rupture of pipelines or malfunctioning of oil extraction plants. In the last decade, satellite remote sensing has proved to be an effective tool in detecting oil spill. Several remote sensing technologies to detect oil spills have been largely documented and some in- teresting results are reported in [5÷10]. The ability to detect and monitor oil spills is determined by several factors, such as the characteristics of the sensor, type and composition of the spilled oil, environmental and meteorological conditions, chemical reaction and mixing with the surrounding water, dimension and duration of the spill. All these factors contribute to the complexity for remote detection of oil spills. Accidental events as rupture of oil pipelines, malfunctioning of oil extraction platforms, spills from floating oil storage and transfer stations are the primary cause of pollution in the Lake Mara- caibo that from long time constantly suffers of some serious environmental problems resulting from intensive hydrocarbons extraction [11]. The state-run oil company Petroleos de Venezuela S.A. (PDVSA), in the framework of a policy of environmental protection of the lake water, commis- Laneve et al.: Designing an effective monitoring system based on satellite sensor for the oil spills detection in lake Mara- caibo, Venezuela sioned at the DIAEE (La Sapienza, University of Rome) a feasibility study aiming at designing a monitoring system customized for the Lake Maracaibo, based on the use of satellite sensors, for the detection of the oil spill. A measurement campaign was carried out during the period February-March 2011 with the aim of analyzing the spectral behaviour of the oil when it floats on the lake surface and then define the channels of the electromagnetic spectrum (EMS) in which it is more easy to distinguish oil contaminated water surface. The set of instruments used during the campaign was selected in order to cover the whole optical region (0.35 - 14 m) of the EMS that is usually covered also by satellite sensors. However the peculiar meteorological, environmental and operational conditions present in this study area make difficult to realize a suitable and cost-effective monitoring system for the remote detection of the oil spills. We describe the results and the reasons respect of which we can state/demonstrate that such a monitoring system cannot be achieved. For sake of brevity this paper shows the results obtained by analyzing the satellite data acquired from the satellite ENVISAT (ASAR and MERIS), MODIS and Landsat in the period of the field campaign.

Figure 1. Map of the sites where the spectral and radiometric measurements were performed.

2. Area of interest and data

Lake Maracaibo is a large inlet of the Caribbean Sea lying in the Maracaibo Basin of northwest- ern Venezuela. It is the largest natural lake in South America, covering an area of about 13280 km2, extending southward for 210 km from the Gulf of Venezuela and reaching a width of 121 km. Lake Maracaibo is one of the world’s richest and most centrally located petroleum-producing regions. Thousands of derricks protrude from the water and many more line the shore, while underwater 386 Laneve et al.: Designing an effective monitoring system based on satellite sensor for the oil spills detection in lake Mara- caibo, Venezuela pipelines transport the petroleum to storage tanks on the land. The lake’s basin supplies about two- thirds of the total Venezuelan petroleum output. Most of the industry was developed by foreign (chiefly American, British, and Dutch) investment, with very few locally owned wells, but in 1975 the petroleum industry was nationalized. The lake ecosystem is seriously compromised by several kinds of pollution, among which: large amount of solid waste from human settlements are transported by casual drainage canals in the lake; the duckweed (Lemna obscura (Austin) Daubs) infestation caused by the high concentration of ni- trogen in the water. Although efforts to remove this plant have been underway, the plant occupies over 130 million cubic metres of the lake. The only way to remove the weed, which can double its size every 48 hours, is to pull it out of the lake physically, no chemical or biological method has been found to treat the weed. Furthermore, numerous oil spills, attributable at least in part to poor maintenance chores and indiscriminate discharge of untreated sewage, significantly deteriorate water quality to the point that in some places the water has levels of pollution dangerous on health. The data used include both those obtained during the measurement campaign and the images captured from optical and radar satellite sensors, trough an ESA CAT 01 opportunity, on the base of our requests. In particular, the field campaign instruments include: . a FieldSpec ASD (a portable spectroradiometer operating between 350 and 2500 nm), . a USB 2000 (a highly portable spectrometer operating in the region between 200 and 1100 nm), . a Micro FT-IR spectrometer (a portable instrument measuring the radiation emitted by a body in the 2 to 16  m of the electromagnetic spectrum), . a FLIR SC660 (a camera operating in the thermal infrared region (7.5 - 13.0 m), the image consists of 640 x 480 pixels. The system is able to detect differences in temperatures down to 0.03 K. The satellite images considered for the study are listed in Table 1. The points where measurements of the spectral response of the pollution free lake surfaces or of the surfaces with the presence of oil were made are shown in Figure 1. The positions have been measured by using a GPS. 80 sites have been sampled by covering about 820 km in the lake. Table 1. A List of the lake Maracaibo satellite sensors images requested or used for this study.

Sensor name Number of images Type of sensor Characteristics

ASAR 3 Radar Images C-band PALSAR X Radar Images L-band AVNIR X MERIS 3 Hyperspectral Images Range 0.4-1.05 µm CHRIS 4 Hyperspectral Images Range 0.4 – 1 µm ASTER X Multispectral Images Range 0.4-13 µm TM 3 Multispectral Images Range 0.4-2.5 µm

MODIS 14 Multispectral Images Range 0.4-13 µm

387 Laneve et al.: Designing an effective monitoring system based on satellite sensor for the oil spills detection in lake Mara- caibo, Venezuela

3. Theoretical basis

First of all it is needed to define the theoretical basis upon which the procedure for detecting oil spills from the water background would be build. Obviously, the parameter being measured is different if we are operating in the reflective part of the electromagnetic spectrum (visible, near infrared and mid-range, 0.4 - 2.5 m) or when the measures are carried out in the thermal infrared region (5 - 15 m) or when the microwave range (3 - 20 cm) is exploited. In the reflective region of the spectrum the separation between the different types of surfaces is made on the basis of spectral reflectance that, in general, is different from element to element. Fig- ure 2 shows the comparison between the spectral reflectance of water and oil films in different concentration floating on water. It is clear that the major differences are present in the UV/Blue spectral region and in particular in the range from 0.3 to 0.46 m.

Figure 2. Comparison among the spectral responses of water and oil films in different concentrations In fact, the effect of the presence of oil on the water surface is to increase the reflectance in the above said region. It can be also observed how the lake’s water has behaviour substantially different from that typically found in clear water free from suspended matter, since it exhibits a peak of the spectral reflectance in the green band. The fact that the major differences are present in the region of the UV/Blue could compromise the ability to detect oil spills by satellite because of the signifi- cant effect of the atmospheric scattering on the electromagnetic radiation at these wavelengths. In the thermal region of the spectrum the parameter of interest is the brightness temperature de- fined as the temperature that the surface would have if behaves like a blackbody (i.e. unitary emissivity) and in the absence of interaction with the atmosphere. Two bodies can show a different brightness temperature because they have a different kinetic temperature and, at the same kinetic temperature, because they have different emissivity. In the case of the presence of oil on water, both conditions can exist. In fact, if the thickness of the oil is thick enough it can have a temperature different from the surrounding water due to the different reflection/absorption in the region of maximum solar radiation. On the other side, when the thickness is thin (film) its temperature is presuma- bly in equilibrium with the surrounding water and what generates the different brightness temperature is the different emissivity. This effect is clearly shown in Figure 3, where the spectral emissiv-

388 Laneve et al.: Designing an effective monitoring system based on satellite sensor for the oil spills detection in lake Mara- caibo, Venezuela ity in the MIR/TIR region was obtained by using samples of oil and water. It enlightens the change in the surface emissivity due to the presence of the oil on lake water that allows the detection of the former from the latter. The emissivity difference occurs around 5 m and in the 9 – 13 m spectral region. It could be also expected that, if the oil film thickness is very thin (few m), it will be difficult to distinguish the presence of oil on water due to the fact that the source of the thermal emis- sion resides in a surface layer of several microns. From Fig. 3, some considerations can be derived. First, the emissivity of oil, with respect to that of water, does not present the same behaviour along the whole spectrum of the thermal infrared. In particular, three regions can be highlighted: - the region lying between 3 and 5 m in which the emissivity of the oil is, on average, higher than that of the water; - the region between 8 and 10.5 m in which the emissivity of oil is greater than that of water; - the region between 10.5 and 14.0 m in which the emissivity of water is higher than oil. Table 2 shows the temperatures obtained by simulating measurements made at different spectral bands, by taking into account the spectral emissivity estimated using the F-TIR instrument.

Figure 3. Comparison of the spectral response in the MIR/TIR of water and oil obtained by using the F-TIR instrument. As expected, a suitable selection of the thermal infrared spectral band could enlighten the differences, given the same kinetic temperature (set at 300 K), between the brightness temperatures measured on the oil film and on the water. Moreover, these differences are affected by the type of oil. We can try to improve the results by assuming a dual-band approach, assuming that the thermal sensor may operate in two different spectral bands. Table 2. Values of the brightness temperature simulated for different surfaces (having the spectral emissivity given in Fig. 3) as function of the size of the thermal infrared band. Spectral Range Water Oil 1 Oil 2 Oil 3 8 – 13 m 292.47 293.0 292.72 291.92 9 – 13 m 293.0 293.6 292.86 292.09 10.5 – 13 m 290.66 290.92 289.48 288.78 11 – 12.5 m 289.92 289.55 288.3 287.6 8 – 10.5 m 297.0 297.7 298.25 297.40 3.5 – 4.5 m 295.8 292.76 295.64 296.30 3.5 – 4.2 m 296.15 292.52 295.84 296.29 3.5 – 5 m 297.26 295.25 297.70 298.27 4 – 5 m 295.12 293.39 295.72 296.34

389 Laneve et al.: Designing an effective monitoring system based on satellite sensor for the oil spills detection in lake Mara- caibo, Venezuela

4. Methods and results

In principle, the MODIS sensor, guaranteeing at least a daily frequency of the observation on the area of interest, could allow the revisit frequency required to monitor oil spills in the lake. This opportunity suffers of two limits: . low spatial resolution (250x250 m2 = 6.25x104 m2 = 6.25 ha); . effect of the weather conditions. During the 20 February – 8 March 2011 period (corresponding to the field campaign) only 4 MODIS images of the Lake Maracaibo not covered by clouds were available. In particular, these acquired on days 20, 23 and 25 February and the 3rd of March. In theory, the MERIS sensor, having 15 spectral bands should allow a better separation between water, oil and lemna (duckweed, floating weeds). In this case, the main limitation is the low tempo- ral frequency. In fact, during the campaign ESA (European Space Agency) was able to acquire only three MERIS images of the area of the lake and of these only one (that of March 3rd) can be used, even if only partially, since the area of the lake results mostly hidden by clouds. As for the radar images, during the campaign 3 ASAR/ENVISAT images were acquired. Radar images are not affected by the presence of clouds; nevertheless they have a revisit frequency too low to meet the requirements of a system capable to monitor oil spills in Lake Maracaibo. It was not possible to acquire satellite hyperspectral sensor images (CHRIS/PROBA) during the campaign due to a failure of the instrument which lasted from December 2010 to May 2011. From the USGS website some Landsat 5 and Landsat 7 satellites images, acquired both during the campaign and at different dates, were obtained.

4.1. Optical images results

The optical images, with different accuracy depending on the spatial resolution of the sensor [11÷21], can detect the presence of oil on the surface of the lake, distinguishing it from the floating vegetation (lemna, etc.), that is present on the lake with rapidly variable extension. The figure 2 shows the spectral response obtained in one of the sites where measurements were carried out from which the ability to detect oil on the lake’s water, which have a predominantly green colour, can be assessed. The effect of the oil film on water is mainly to increase the spectral response in the region of the UV/Blue. But unfortunately the blue channel is strongly affected by the interaction with the atmosphere (absorption and scattering). In the following some results obtained for different type of satellite optical sensor optical images are given. Such examples aim at illustrating the variety, in terms of spatial and spectral resolution, of the available images. Hyperspectral case The figure 4 shows an image of the hyperspectral sensor CHRIS/PROBA (Tab. 1). Such sensor is characterized by a spatial resolution that can vary between 17 and 34 m, depending on the number of channels acquired, that can be of 18 or 62, respectively. The image of the figure 4 is a CHRIS image acquired in high spatial resolution mode (18 spectral channels) on the 25 of September 2008. The image was acquired by ESA, as consequence of our request in the framework of a CAT-01 project. As said above no image has been acquired during the field campaign due to the temporary failure of the instrument. It is very notable the quantity of information that can be extracted by such image as consequence of is spatial and spectral resolution. The drawback of such sensor that pre- vents its use for the purpose of the monitoring of the lake is the small dimension of the image (14 x 14 km) and the low revisit frequency. Theoretically hundred images are required to cover the whole lake. Anyway such system can be used for periodically monitoring assigned areas of the lake. Actu- ally in the period 2008-2009 a quasi-simultaneous acquisition of an optical (CHRIS) and radar (ASAR) images was unsuccessfully attempted via ESA due to the weather conditions.

390 Laneve et al.: Designing an effective monitoring system based on satellite sensor for the oil spills detection in lake Mara- caibo, Venezuela

Figure 4. Proba/CHRIS image of September 25, 2008 on Maracaibo Lake. The excellent spectral resolution (with good spatial resolution) of this hyperspectral image, allows to highlight many different materials on the scene (heavy oil, light oil, oil rigs, pipelines, etc) thanks to the accurate spectral response that only a hyperspectral sensor can show. Multispectral medium resolution case A couple of ETM+/Landsat 7 images of the lake were acquired during the field campaign. However, in the example given here an image of the May 31, 2010 has been used. The reason is the big oil spill occurred in that period, thus the capability of the sensor to detect oil spill on the water of the Lake Maracaibo can be assessed. The ETM+/Landsat 7 sensor, apart from the malfunctioning of a mechanism for compensating the satellite motion, is characterized by a spatial resolution of 30 m (15 m the panchromatic channel) and a number of 7 spectral channels. The revisit frequency of the images is 16 days. The figure 5 shows a section of the ETM+ image acquired on the 31st of May, 2010 in combination with a MODIS image acquired on the same day 30 minutes later. On the MODIS image 2 oil spills (red squares) have been identified. Such oil spills are, of course identifi- able with better details on the ETM+ image (better spatial resolution 30 m instead of 250 m). In fact, the second oil spill (bottom square) has a width of about 1 km (about 3-4 MODIS pixels, at the visibility limit), while the same phenomenon is very well identified on the corresponding Landsat 7 image (about 33 pixels). This comparison helps in enlighten the limits of the MODIS sensor in identifying small oil spills that is the typical case occurring in the Lake Maracaibo. Even if this sensor seems adequate to monitor the oil spill in the lake as consequence of the spatial resolution and the image frame size (185 x 185 km) sufficient to cover the whole lake its revisit frequency is largely inadequate for the needs of a lake monitoring system. Multispectral low resolution case MODIS sensor, on board of the Terra and Aqua satellite is a very attractive sensor as consequence of is very high revisit frequency. In fact, considering the two satellites, in average 2 images 391 Laneve et al.: Designing an effective monitoring system based on satellite sensor for the oil spills detection in lake Mara- caibo, Venezuela for day can be acquired. Its low spatial resolution represents the main limit for an operational application to monitor the Lake Maracaibo. However, the weather conditions of the region make it difficult to obtain the claimed 6 hours revisit frequency. The effect of the spatial resolution in clearly shown in Fig. 6 where the structures present on the lake’s surface, observed during a helicopter sur- veying, were not detectable on the MODIS image.

Figure 5. Terra/MODIS and Landsat 7 images of May 31, 2010 on Lake Maracaibo. On the MODIS image have been identified 2 oil spills (2 red squares), which are better seen by ETM+/Landsat 7 sensor. In fact, the second oil spill (bottom square) has a width of about 1 km (3-4 MODIS pixels, the limit of visibility), while the same phenomenon is very well identified on the corresponding Land sat 7 image (about 33 pixels). This explains as the MODIS sensor could be of little use in identifying small oil spills that is the typical case occurring in the Lake Maracaibo.

4.2. Radar images results

The analysis has been performed on 5 SAR C-Band images:

Satellite-Sensor Date Polarization Acquisition Mode Radarsat-2 21-Jun-2010 VV Wide Radarsat-2 15-Jun-2010 VV Wide Envisat-ASAR 28-Feb-2011 VV Stripmap Envisat-ASAR 01-Mar-2011 VV Stripmap Envisat-ASAR 04-Mar-2011 VV Stripmap

392 Laneve et al.: Designing an effective monitoring system based on satellite sensor for the oil spills detection in lake Mara- caibo, Venezuela

Oils on the water surface are seen on SAR images as dark slicks because the damping capillary effect produced over waves and responsible for radar basckscattering [4, 8, 22÷33]. But not only oil spills produce this phenomena, but e.g. rain cells, biogenic substances, floating ice, turbulences generated by a ship, wind shadows behind infrastructures, floating vegetation as lemna, coastal mountains or islands, or directly low winds [34, 35]. Significantly relevant in order to distinguish oil from false alarms, known as look-alikes, is the information related to topography and bathymetry, historical and current winds, and infrastructure context, and several methods were applied histori- cally in order to minimize the ambiguity slick classification, even it is difficult to determine in an absolute way the dark patches membership classes from SAR imagery. Some of these methods ap- ply Support Vector Machines, Neural Networks, Fuzzy Logic or Probabilistic Approximations techniques, and shown their potentialities with different success degrees [25, 30, 36÷41].

Figure 6. Identified look-alikes in the region of interest Envisat image (left) - Radarsat2 images (center & right). Between main restrictions, wind speeds appear as one of the most important phenomenon to take into account. Several studies talks about the significance of minimum and maximum wind speeds required to detect oil slicks, which limits the environment window of application of radar -1 technology […]. Below a threshold wind speed of 3 m*s it is not produced enough surface rough- ness to contrast with the oil [35]. A higher wind speed threshold at around 14 m*s-1, could be es- tablished as a maximum limit at which and oil spill contrast from background could be useful to extract the slick to be analysed. While biogenic oils appear to disperse when wind speed exceed 7 to -1 -1 8 m*s , mineral oils may be detectable until winds exceed 10 to 14 m*s depending on the water surface state and oil heaviness [???]. At high wind speeds the damping effect can disappear in the

393 Laneve et al.: Designing an effective monitoring system based on satellite sensor for the oil spills detection in lake Mara- caibo, Venezuela background noise of wind-generated waves [42, 43], and even at higher wind speeds the oil disap- pears from the sea surface [26, 43, 44]. Each frequency band is affected in a different way by wind and slick nature and thickness. Ex- perimental works have shown that C, X and K bands generates higher contrast responses unlike L band that is weakly affected by slicks. [45] mentioned that C band seems to be the most suitable -1 frequency allowing to measure strong contrasts until wind speed around 13 m*s . Regarding polarization properties, both vertical transmission and reception achieve better feature extraction results than other configurations. In our particular case the region of interest presents prevailing low wind areas around main production zones (see figure 6), phenomenon that limits an appropriate SAR oil slick feature extraction. Although series of images give the possibility to extract further information that in a single one is not revealed, for instance environmental conditions that typically affects floating substances ap- pearance, as waves, currents or tides. Field information collected during this campaign put into evidence the limitations to detect oil spill in main extraction areas. Figure 7 shows a portion of an Envisat-ASAR image acquired on the 1st March 2011 (see figure 7 upper left) whereas an aerial photo was taken the following day, on 2nd March 2011 (see figure 7 upper right). The red box represents the area of the satellite image corresponding to the RGB photo. Within the photo an oil spill is present and no contrast evidence of it is available in the radar image, as the backscattering map clearly shows (bottom figures).

Figure 7. Analysis of ASAR image by using ground information. The red box on the radar image corresponds to the RGB photo. The backscattering map computed within the red box reveals the difficulty to discriminate or classify apparently low wind areas (high values -in white-, evidence the response from platforms that in this case acts as typical corner reflectors). Despite the amount of images acquired during the mission, potential exploitation of classification techniques mentioned above, i.e. neural networks, were still inapplicable considering the low volume of data necessary to train it properly and mainly due to unavailable verified on field data linked with imagery. However, statistical information about slicks detected in the past on radar images could be collected and taken into account in a future development of automatic classification procedures. The difficulty in detecting oil spills in the lake by using radar images can be further evidenced by observing as some slicks (Fig. 8) that apparently don’t seem to be look-alikes, on the base of the statistical properties (Fig. 9) would be classified as look-alikes. However, it is important 394 Laneve et al.: Designing an effective monitoring system based on satellite sensor for the oil spills detection in lake Mara- caibo, Venezuela to notice that these conclusions were strongly influenced by dominants low-winds meteorological conditions.

Figure 8. Sets of Radarsat images of the lake. Yellow boxes represent some slicks that could come from oil spills.

Figure 9. Examples of statistical information extracted from Envisat Slick-Sets Finally, a comment on the spatial resolution limitations is needed. In fact, figure 10 shows a photo taken on the 7th of March 2011, showing a typical Lake Maracaibo oil spill with its means dimensions of approximately 100 x 1000 meters. Thus, assuming that the object under study has the mean dimensions given in figure 10, the use of remote sensing imagery having a spatial resolution greater than 20 meters could add complexity to the potential detections of these types of spills that frequently occur inside the lake production area. In general radar systems should be preferred due to the weather characteristics of the region (frequently cloudy) which often preclude a clear remote observation of the lake surface. From SAR sensor data analysis and as shown from the reported examples, the detection of the substances floating on the water lake (oil, lemna, etc.) can result difficult to be achieved since the results are strongly depending from the wind conditions. In this paragraph an analysis of several satellite images (optical and radar) has been carried out in order to underline the advantage and the disadvantage of each satellite sensor. In general radar systems should be preferred due to the weather characteristics of the region (frequently cloudy) which often preclude a clear remote observation of the lake surface. From SAR sensor data analysis and as shown from the reported examples, the detection of the substances floating on the water lake (oil, lemna, etc.) response can result difficult to be achieved since the results are strongly depending from the wind conditions. Medium resolution multi-spectral satellite optical images are able to provide an accurate map of the lake status but the low frequency of the images, due to orbital and weather causes, makes such systems not suitable for the monitoring of the Lake Maracaibo. In any case, a periodic acquisition of this kind of images (ETM+ or TM) can allow a statistical assessment of the oil spill occurrences in the lake.

395 Laneve et al.: Designing an effective monitoring system based on satellite sensor for the oil spills detection in lake Mara- caibo, Venezuela

Figure 10. Helicopter RGB photo of an oil slick. The image has been acquired the 7th of March 2011.

5. Conclusions

The paper discuss the problem of assessing the possibility of developing an oil spill monitoring system for the lake Maracaibo based on satellite images. To reach this objective a field campaign has been carried out in the period 21 February - 8 March 2011. Many examples of satellite images (optical and radar) of the lake, some of which have been acquired during the campaign, are describe with the objective to illustrate the difficulties to monitor oil spill in the lake due to is peculiar meteorological, environmental and operational conditions. Regarding the 'in situ' data acquired during the field campaign, for sake of brevity just of a couple of examples is given, deferring to another paper the presentation of all the data acquired. Satellite images suffer of the limitation that in the high revisit frequency optical sensor (MODIS) images the lake surface results often covered by clouds. Radar images can overcome this problem but in this case the low revisit frequency (days/weeks) makes such instrument adapt to monitor very important and dangerous events. Also in this case, the success in the detection of the oil spill is strongly related to the weather conditions (mostly wind regime on the lake). Therefore, rejected, for the moment, the possibility of using satellite sensors, we conclude that the only system that could allow the oil spill monitoring into the particular environment of the Lake Maracaibo should be based on an observation sensor mounted on an airplane or helicopter. Such a monitoring system should be equipped with a multi/hyperspectral camera operating in the MIR/TIR range of the electromagnetic spectrum or a VIS/NIR/SWIR imaging system and a processing system capable to analyze images in real-time providing, at least, information on the presence of oil spill for alert reasons. Of course, this kind of system will inevitably be more expensive than a system based on satellite images. However, satellite images can still be taken into account for some analysis: to carry on a statistical analysis of the periodic (six-months, annual) trend of the oil spill based on a, for instance, be- weekly acquisition of radar images of the area of interest; to analyze the impact, if any, of the slicks on the lake environment based on optical satellite images periodically acquired capable to provide updated land cover maps of the area surrounding the lake.

Acknowledgements

The authors thank PDVSA for having supported this study.

396 Laneve et al.: Designing an effective monitoring system based on satellite sensor for the oil spills detection in lake Mara- caibo, Venezuela

References

[1] European Space Agency, (1998). Oil pollution monitoring. ESA brochure: ERS and its applications Marine, BR-128, 1. [2] Fingas, M. (2001). The basics of oil spill cleanup. Lewis Publishers. [3] REMPEC, 2002. Protecting the Mediterranean against maritime accidents and illegal discharges from ships. Malta, 2002. [4] Huijer, K., 2005. Trends in Oil Spills from Tanker Ships 1995-2004. Vol. 1 Proceedings of the Twenty-eighth Arctic and Marine Oil spill Program (AMOP) Technical Seminar. 7-9 June 2005. Calgary, Canada. Pp 319-332. [5] Brown, C. E., & Fingas, M. F, 2001. New Space-Borne Sensors For Oil Spill Response. International Oil Spill Con- ference, 911-916. [6] Brown, C. E., Fingas, M. F, & Goodman, R.H., 1998. Oil-spill Remote Sensors: New tools that provide solutions to old problems. Proc. XXI Arctic and Marine Oil spill Program Technical Seminar. [7] Fingas, M. F., & Brown, C. E., 1998. Review of the visibility of oil slicks and oil discharges on water. Proc. XXI Arctic and Marine Oil spill Program Technical Seminar. [8] Fingas, M. F., & Brown, C. E., 1997. Review of oil spill remote sensing. Spill Science and Technology Bulletin, 4, 199– 208. [9] Sherman, J. W.III., 1992. Historical Perspectives: Marine Oil Spill Detection by Remote Sensing. The First Themat- ic Conference on Remote Sensing for Marine and Coastal Environment, New Orleans, Louisiana, 1992, pp. 15-17. [10] Lodge, A.E., 1989). The Remote Sensing of Oil Slicks. Institute of Petroleum, London, Wiley and Sons, New York, 1989. [11] Hu, C., Mqller-Krager, F. E., Taylor, C. J., Myhre, D.,Murch, B., Odriozola, A. L., 2003. MODIS detects oil spills in Lake Maracaibo, Venezuela. EOS, Transactions, American Geophysical Union, 84 (33), 313-319. [12] AI-Ghunaim, I., Abuzar, M. & AI-Qurnas, F. S., 1992. Delineation and monitoring of oil spill in the Arabian Gulf using Landsat thematic mapper (TM) Data. Proceedings of the First Thematic Conference on Remote Sensing for Marinc and Coastal Environments, ERIM Conferences, Ann Arbor, Michigan, pp. 1151-1160. [13] A1-Hinai, K. G., Khan, M. A., Dabbagh, A. E. and Badcr, T. A., 1993. Analysis of Landsat thematic mapper data for mapping oil slick concentrations-Arabian Gulf oil spill 1991. The Arabian Journal for Science and Engineering, 18 (2), 85-93. [14] Cross, A., 1992. Monitoring marine oil pollution using AVHRR data: observations off the coast of Kuwait and Saudi Arabia during January 1991. International Journal of Remote Sensing, 13, 781-788. 141 [15] Grimaldi, C.L.S, Casciello, D., Coviello, I., Lacava, T., Pergola, N., and Tramutoli, V., 2009). Near Real Time Oil Spill Detection And Monitoring Using Satellite Optical Data. IGARSS 2009, 4, 709-712. [16] Tseng, W. Y., and Chiu, L. S., 1994. AVHRR observations of Persian Gulf oil spills. Proc. IGARSS’94, 2, 779– 782. [17] Voloshina, I. P. and Sochnev, O. Y., 1992. Observations of surface contaminations of the region of the Kol’shii Gulf from IR measurements. Soviet Journal of Remote Sensing, 9 (6), 996-1000. [18] Borzelli, G., Ulivieri, C., Laneve, G., Antonelli, G. and Ciappa, A., 1996. A new perspective on oil slick detection from space by NOAA satellites. Int. Journal of Remote Sensing, Vol. 17, N. 7, pag. 1279. [19] Taravat, A. and Del Frate. F., 2012. Development of band rationing algorithms and neural network to detection of oil spills using Landsat ETM+ data. EURASIP Journal on Advances in Signal Processing. [20] Alam, M.S. and Sidike, P., 2012. Trends in oil spill detection via hyperspectral imaging. 7th Int. Conf. on Electrical and Computer Engineering, Dhaka, Bangladesh, December 2012. [21] Adamo, M., De Carolis, G., De Pasquale V. and Pasquariello, G., 2009. Detection and tracking of oil slicks on sun- glittered visible and near infrared satellite imagery. Int. Journal of Remote Sensing, Vol. 30, N. 24, pp. 6403-6427. [22] Byfiel, V. and Boxal, B., 1999). Thickness estimated and classification of surface oil using passive sensing at visible and near-infrared wavelength. Proc. IEEE International Geoscience and Remote Sensing Symposium, 1475- 1477. [23] Bern, T-I., Wahl, T., Anderssen, T. and Olsen, R., 1993. Oil spill detection using satellite based SAR: Experience from a field experiment. Photogrammetric Engineering and Remote Sensing. 59 (3): 423:428. [24] Brekke, C. and Solberg, A. H. S., 2005. Oil spill detection by remote sensing. Remote Sensing of Environment 95, 1-13. [25] Del Frate, F., Petrocchi, A., Lichtenegger, J., Calabresi, G., 2000. Neural networks for oil spill detection using ERS-SAR data. IEEE Transactions on Geoscience and Remote Sensing 38, 2282–2287. [26] Espedal, H., 1999. Detection of oil spill and natural film in the marine environment by spaceborne SAR. Proc. IGARSS, vol. 3 (pp. 1478– 1480). [27] Espedal, H. A. and Johannessen, O. M., 2000. Detection of oil spills near offshore installations using synthetic aperture radar (SAR). International Journal of Remote Sensing, 21(11), 2141–2144.

397 Laneve et al.: Designing an effective monitoring system based on satellite sensor for the oil spills detection in lake Mara- caibo, Venezuela

[28] Espedal, H. A. and Wahl, T., 1999. Satellite SAR oil spill detection using wind history information. International Journal of Remote Sensing, 20(1), 49–65. [29] Ferraro, G., Bernardini, A., David, M., Meyer-Roux, S., Muellenhof, O., Perkovic, M., Tarchi, D. and Tpouzelis, K., 2007. Towards an operational use of space imagery for oil pollution monitoring in the Mediterranean basin: A demonstration in the Adriatic Sea. Marine Pollution Bulletin, 54, 403-422. [30] Fiscella, B., Giancaspro, A., Nirchio, F., Pavese, P. and Trivero, P., 2000. Oil spill detection using marine SAR images. International Journal of Remote Sensing, 21(18), 3561–3566. [31] Indregard, M., Solberg, A. and Clayton, P., 2004. D2-report on benchmarking oil spill recognition approaches and best practice. Tech. rep., Oceanides project, European Commission, Archive No. 04-10225-A-Doc, Contract No: EVK2-CT-2003-00177. [32] Kostianoy, A., Litovchenko, K., Lavrova, O., Mityagina, M., Bocharova, T., Lebedev, S., Stanichny, S., Soloviev, D., Sirota, A. and Pichuzhkina, O., 2006. Operational Satellite Monitoring of Oil Spill Pollution in the Southeast- ern Baltic Sea: 18 Months Experience. Environmental research, engineering and management, 38 (4), 70-77. l2gen, http://oceancolor.gsfc.nasa.gov/DOCS/MSL12/MSL12_prod.html [33] Nirchio, F., Pandiscia, G.V., Ruggieri, G., Santoleri, R., Tataranni, F., Trivero, P., Pinardi, N., Masini, A., Manzel- la, G. and Castellani, C., 2009. PRIMI (Pilot Project Marine Oil Pollution). 33rd ISRSE, International Symposium on Remote Sensing of environment Sustaining the Millennium Development Goals, May 4-8, 2009, Palazzo dei Congressi Stresa, Lago Maggiore, Italy, http://isrse-33.jrc.ec.europa.eu/. [34] Migliaccio, M., Tranfaglia, M., Ermakov, S. A., 2005. A Physical Approach for the Observation of Oil Spills in SAR Images. IEEE Journal of Oceanic Engineering, 30 (3), 496-507. [35] Gonzalez Vilas, L., Torres, J., Martinez-Iglesias, G., Mosquera, A., Corchado, J., 2007. Utilización de datos de ENVISAT para la deteccion de vertidos de hidrocarburos. Revista de Teledetección. Número Especial Junio, 2006, 55-59. [36] Gonzalez Vial, L., Torres, J., Yarovenko, N., Martin, J., 2009. Discriminación automática de vertidos de hidrocarburos a partir de imágenes ASAR utilizando una máquina de vectores soporte SVM. XIII Congreso de la Asociación Española de Teledetección, Calatayud, 537-540. [37] Topouzelis, K., Karathanassi, V., Pavlaki,s P., Roko,s D., 2007. A new object-oriented methodology to detect oil spills using Envisat images. Proc. Envisat Symposium 2007, Montreux, Switzerland. [38] Hang, L. and Duong, N., 2009. Oil Spill Detection and Classification by ALOS PALSAR at Vietnam Sea. 7º FIG Regional Conference. Hanoi, Vietnam. [39] Karathanassi, V., Topouzelis, K., Pavlakis, P., Rokos, D., 2006. An object-oriented methodology to detect oil spills. International Journal of Remote Sensing, 27, 5235–5251. [40] Nirchio, F., Di Tomaso, S., Biamino, W., Parisato, E., Trivero, P., Giancaspro, A., 2005. Oil spills automatic detection from sar images. Proc. of the 2004 Envisat & ERS Symposium, Salzburg, Austria (ESA SP-572). [41] Solberg, A., Brekke, C., Husoy, P., 2007. Oil spill detection in Radarsat and ENVISAT SAR Images. IEEE Transac- tions on Geoscience and Remote Sensing, 45 (3). [42] Denim, B. T., Ermakov, S. A., Pelinovsky, N. Y., Talipova, T. G., Sheremet’yeva, A. I., 1985. Study of elastic properties of sea surface-active films. Atmos. Oceanic Phys., 21, 312-316. [43] Jackson, C. R., Apel J. R., 2004. Synthetic Aperture Radar Marine User's Manual. http://www.sarusersmanual.com. [44] Scott, J. C., 1986. Surface films in oceanography. ONRL Workshop Report c-11-86, 19. [45] Girard-Ardhuin, F., Mercier, G., Garello, R., 2003. Oil slick detection by SAR imagery: potential and limitation. Proc. OCEANS vol. 1, 164–169.

398