Satellite Imagery Analysis for Environmental Monitoring: Turkmenbashi, Turkmenistan
In June 2012, AAAS initiated a project focused on Turkmenbashi, Turkmenistan (Figure 1). The period of review was from 2000 to 2012. This project was conducted in collaboration with Crude Accountability and the Turkmen Initiative for Human Rights. The goal of the project was to use satellite imagery to investigate reports of environmental pollution in the waters adjacent to the port of Turkmenbashi, Turkmenistan (Figure 1).
Figure 1: Overview Map of Study Region
In May 2005, the Baku-Tbilisi-Ceyhan (BTC) pipeline (Figure 2) opened and began transporting oil from the Caspian Sea oil fields to the Port of Ceyhan, Turkey (BBC News, 2005). Turkey first conceived the 1,768 km pipeline in the early 1990’s for the purpose of achieving secure energy independence while reducing the number of oil tankers passing through the environmentally sensitive Bosporus Strait. The United States supported construction of the pipeline in order to bring Caspian Sea oil to global markets while bypassing potentially hostile countries such as Russia, Iraq, and Iran (Zeyno, 2005). By the late 1990’s, tentative agreements were in place between Georgia, Turkey, and Azerbaijan. In September 2001, Georgia and Azerbaijan signed an agreement finalizing the transit tariffs that Georgia could levy on petroleum shipped along the pipeline (BBC News, 2001). In 2002 an international consortium of oil companies, led by BP, was formed to finance and construct the pipeline (Figure 2).
Figure 2: Map of BTC Pipeline Route
The pipeline has led to an increase in oil production in the region, including in the waters of Turkmenistan. Oil from fields around Turkmenbashi is routed through the Turkmenbashi port across the Caspian Sea to Baku. As a result of this increased exploration and production, groups such as Crude Accountability have reported an increase in petroleum related pollution in and around Turkmenbashi Bay.
Figure 3: Map of Turkmenbashi
Crude Accountability and their local colleagues have reported that the increased volume of oil being shipped from Turkmenbashi’s two oil terminals, “Port West” and “Port East” has led to local contamination around the terminals (Figure 3). According to these reports, Port East processes the majority of the region’s oil and has been implicated as a primary source of pollution. In addition, possible leakages from both new oil infrastructure and decaying Soviet-era facilities is a concern. A number of bodies of water have been separated from the larger Turkmenbashi Bay by man-made berms and dams that form part of the oil facilities. The largest of the water bodies is the Soimonov Bay, west of Turkmenbashi. This bay was separated in 1940 by a dam and has been used as a reservoir for effluent from the Krasnovodsk oil refinery. Reports from partner organizations indicate that leakage of toxic chemicals could be occurring from Soimonov Bay, and other bodies of water, into the larger bay.
This study uses multiple sources of remotely sensed imagery to detect and monitor oil pollution in and around Turkmenbashi Bay over a period of 12 years, from 2000 to 2012. The purpose of this monitoring was to create a record of the scope of oil-related pollution in the region, informing organizations regarding the state of the environment in the region.
Historically, efforts to monitor oil spills with remotely sensed imagery have relied on synthetic aperture radar (SAR) or high-resolution imagery. Methods using SAR identify areas of the water that are unusually free of turbulence due to the viscosity of oil on the water’s surface. These methods are widely considered to provide the most accurate detections (Brekke & Solberg, 2004). SAR is also capable of penetrating cloud cover allowing for monitoring in all weather conditions. SAR methods are reliant on accurate wind speed and direction data and most SAR sensors have low temporal resolutions and high acquisition costs. As a result, SAR imagery, though well suited to analyzing known spills, is less appropriate for long-term monitoring, as required by this project. Similarly, the high acquisition costs, low temporal resolution, and small archives of most commercial high-resolution sensors limits the utility of high-resolution data for long-term monitoring of environmental phenomena (Brekke & Solberg, 2004).
To overcome these limitations, researchers have begun to explore using the National Aeronautics and Space Administration’s (NASA) publicly available Moderate-Resolution Imaging Spectroradiometer (MODIS) data to detect slicks (Chuanmin et al., 2003, Lotliker et al., 2008). Launched aboard NASA’s Terra (EOS AM) satellite in 1999 and duplicated on NASA’s Aqua (EOS PM) satellite in 2002, MODIS is a moderate- to low-resolution sensor designed to monitor large-scale global phenomena. MODIS captures data in 36 bands, with spatial resolutions of 250 meters (2 bands), 500 meters (5 bands), and 1000 meters (29 bands). Each of the satellites image the globe daily, resulting in two images per day (Esaias et al., 1998). This high temporal resolution allows for continuous monitoring of environmental phenomena, including oil spills.
Studies using MODIS data for oil slick detection have identified the 650nm and 859nm bands as the best for detecting oil spills (Lotliker et al., 2008). These bands detect radiations of 620-670 nm and 841-876 nm, respectively, and have spatial resolutions of 250m. Spill detections are thus influenced by the sun angle at the time of image acquisition. In addition, as optical sensors, cloud cover can obscure MODIS imagery. Despite these limitations, MODIS’s high temporal resolution provides a unique potential to continuously monitor a specific area over an extended period of time.
This report relies on MODIS data to observe a portion of the Caspian Sea and Turkmenbashi Bay for the detection of oil spills. The study area lies between latitudes 390N and 40.30N and longitudes 520E and 540E. This area includes the Soimonov Bay, Turkmenbashi Harbor, the Turkmenbashi port and oil terminal, oil rigs off the Cheleken Peninsula, and the Avaza resort area (Figure 3). AAAS acquired all available MODIS data from both Terra and Aqua sensors of this area, resulting in the acquisition of 6809 images from Terra between 2000 and 2012 and 4852 images from Aqua between 2002 and 2012. Since research indicated that NASA’s standard atmospheric correction process can obscure the signal of oil slicks (Lotliker et al., 2008), the unprocessed Level 0 (L0) product was acquired.
NASA’s SeaWIFS Data Analysis System (SEADAS) software was used to process the L0 images up to L1A and each image was cropped to the study area. This process adds geolocation information to each image but does not attempt to correct for any atmospheric conditions. Instead, raw reflectance values are maintained. The 650nm and 859nm band of each image was then visually analyzed. Each image was assigned a category of “not a candidate,” “poor candidate,” “mild candidate,” or “strong candidate.”
The resulting list of candidate slicks was used to identify dates for the acquisition of high-resolution and SAR imagery. Near the date of one strong candidate, which occurred on 27 March 2011, both high-resolution and SAR imagery were available (Table 1). The high resolution, SAR, and MODIS images were analyzed using established methods of oil spill detection. This data set served as a reference and enabled analysis to go forward by establishing how the same oil slick appeared in all three types of imagery.
|Table 1: 27 March Imagery|
|27 March 2011||MODIS-Terra||T2011086072000|
|27 March 2011||WorldView-2||103001000A90C500|
|25 March 2011||Envisat-Image Mode||ASA_IMP_1PNIPA20110325_063437_000000163100_00408_47405_1414.N1|
Multiple SAR images from the ENVISAT satellite were also available and AAAS acquired 15 images (Table 2). This imagery was processed using the European Space Agency’s Next ESA SAR Toolbox (NEST). All ENVISAT data was radiometrically corrected and filtered for speckle and noise, then subjected to NEST’s automated oil spill classification algorithm.
|Table 2: SAR Imagery|
|25 May 2004||Envisat
|24 July 2004||Envisat
|5 December 2004||Envisat
|18 April 2005||Envisat
|25 May 2005||Envisat
|12 July 2005||Envisat
|12 October 2005||Envisat
|14 May 2006||Envisat
|30 May 2006||Envisat
|15 June 2006||Envisat
|21 June 2010||Envisat
|24 December 2010||Envisat
|25 March 2011||Envisat
|24 May 2011||Envisat
|4 June 2011||Envisat
From February 2000 through December 2012, spill candidates were observed 701 times in MODIS imagery, of which 147 instances were considered to be “strong” candidates. Comparing the number of spill candidates observed during the first three years of the study (2000-2002) with those seen over the subsequent period (2003-2012) is problematic, as only one of the two spacecraft that carry the MODIS instrument was in orbit during the early phase. The total number of candidates observed per year from 2003 to 2012 held reasonably steady throughout the study period (Figure 4), ranging from 43 in 2004 to 64 in 2010 with a standard deviation from the mean of 6.0. Variation in the number of strong candidates over the same period was slightly lower, ranging from 5 in 2004 to 21 in 2008, with a standard deviation of 5.0.
Figure 4: Spill Candidate Observations by Year
As shown in Figure 5, candidates were far more likely to be observed during the summer than the winter. While the increased prevalence of clouds in the area during the colder months partially explains this phenomenon, at least one radar image acquired during the winter shows a spill candidate clearly despite its near-invisibility in MODIS data, suggesting that sun angle may be critical to observability, as suggested by Lotliker et al. (2008).
Figure 5: Number of Days on Which Candidates Were Observed by Month
During MODIS analysis, it became apparent that an irregular serpentine feature frequently appeared in the imagery in an area located approximately 55 kilometers west-northwest of the Cheleken Peninsula (Figure 6). This feature, nicknamed “the squiggle” by the authors, represented over half of all sightings, and over 65% of strong candidate spills. It usually manifested as a darker patch spread across the surface of the water (Figure 6a & 6d). Periodically it was also visible as a light-colored meandering streak (Figure 6c); occasionally both variants were present at the same time (Figure 6b). The extent and complexity of the feature also varied considerably. At times barely visible, in some instances its effects extended outward from the source point to cover hundreds of square kilometers. Likewise, while often only moderately sinuous, the light-colored variant was occasionally observed in configurations of extreme tortuosity.
Four examples of a serpentine feature visible in numerous MODIS images throughout the study period. Clockwise from top: 7 April 2005, 18 April 2005, 21 April 2010, 8 June 2010. On 18 April 2005, a number of additional spill candidates are visible in the oilfields between the squiggle and the Cheleken Peninsula. Imagery: NASA.
As with most spill candidates observed in MODIS data, it was most frequently observed during the summer months, with detections declining considerably in the fall and winter. The frequently misshapen nature of the feature made its precise origin difficult to establish in many images. However due to MODIS’s rapid revisit cycle, plenty of imagery was available of smaller, more regular manifestations of the phenomenon. When these images were compared to determine a source point, there was remarkable consistency in the observations, which enabled the pinpointing of the squiggle’s origin at coordinates 39.5ºN, 52.6ºE (Figure 7).
By overlaying the outlines of eight observed spill candidates, a plausible source point (red arrow) was identified at 39.5ºN, 52.6ºE, which was confirmed by subsequent observations. Base image date: 8 June 2010. Imagery: NASA.
Radar imagery proved consistent with these observations. As shown in Figure 8, for example, a smooth dark area consistent with the wave-dampening action of an oil slick extends southeast from the same location shown above in Figure 7. When the automated oil spill detection algorithms described in Solberg et al. (2004) were applied to this image, they likewise identified the feature as a probable oil slick. The enhanced resolution of ENVISAT-ASAR imagery enabled further refinement of the coordinates of the feature’s origin to 39.548ºN, 52.616ºE. The bright radar reflections to the south, east, and west of the feature were presumed to be offshore oil rigs, due to their high reflectivity and lack of movement from image to image.
In this ENVISAT image from 24 May 2011, a dark patch extends from the same area as was identified in MODIS data. As discussed below, the blue and green zones outline another candidate observed in high-resolution imagery from 27 March 2011. Bright points represent oil rigs. Imagery ©2013 ESA.
High-resolution imagery from DigitalGlobe’s WorldView-2 satellite confirms that the feature in question is an oil slick, with a source point at the location suggested by radar imagery, and surrounded by drilling platforms on three sides (Figure 9). Three kilometers to the west and one kilometer to the south of the source, two rigs are present which appear to be in a state of considerable disrepair, while three kilometers to the east, an intact and possibly operational platform is present (Figure 9a-d). The slick itself exhibits a silvery sheen over much of its area as it curls north and west from its source point. This, according to the standardized appearance nomenclature listed in the National Oceanic and Atmospheric Administration’s (NOAA) guide to aerial observation of open water oil slicks, corresponds to a layer of oil between 0.04 and 0.3 microns in thickness. In one denser region of the slick, the oil’s appearance took on the rainbow colors characteristic of a layer between 0.3 and 5.0 microns thick. Using these visual interpretation standards, which are an adaptation of the Bonn Agreement Oil Appearance Code, a rough estimate of the volume of oil released can be derived from these observations, albeit one with a high degree of uncertainty due to variables such as view angle, oil type, and water conditions (NOAA, 2012).
WorldView-2 imagery from 27 March 2011 confirms that the feature originating at 39.548ºN, 52.616ºE (red arrow) is an oil slick flanked by drilling rigs (a-d), three of which (b-d) are missing major structural components and appear derelict. Slick area is approximately 23km2. Southwest of the source point, a plume of material (e) is visible emanating from the surface. Imagery ©2013 DigitalGlobe.
By mapping the area of the slick covered by oil with the properties described above, the NOAA/Modified Bonn standards suggest a total volume for the 27 March 2011 spill of between 980 and 3,200 liters – equivalent in industry terminology to a range of 6 to 19 barrels of oil. To verify this estimate, an alternative evaluation of the spill was also performed using the slightly different standards described in the ExxonMobil Oil Spill Response Field Manual (ExxonMobil, 2008). This resulted in a total spill volume of 1,420 liters (approximately 9 barrels) of oil– an estimate that falls within the lower boundary of the NOAA/Modified Bonn estimates. Due to their reliance on properties such as color and texture, application of these estimation standards to spills other than the single incident that was captured by WorldView-2 was not feasible.
Between February 2000 and December 2012, similar incidents were observed over four hundred times, and often covered substantially larger areas than the 27km2 slick that was observed on 27 March 2011. On 21 April 2011, for example, a slick originating from the location was noticed in MODIS data that covered an area of nearly 150 square kilometers. If this slick’s density was similar to the 27 March 2011 spill, the volume of oil released could range into the hundreds of barrels. At other times, density may have been even higher; while WorldView imagery reveals the 27 March 2011 spill to contain both a dark, low-density area and a brighter, high-density component, MODIS imagery acquired on the same date was unable to distinguish the two, revealing only a dark patch. This suggests that higher densities of oil than those released on 27 March 2011 may be necessary in order for the frequently observed bright component of the squiggle (e.g., Figure 7) to be visible in MODIS data.
In addition to confirming the presence of an oil spill at the coordinates derived from MODIS and ENVISAT-ASAR imagery, the high-resolution imagery provided by WorldView-2 also revealed an unusual feature located approximately 2.5 kilometers southwest of the spill’s source point. Nicknamed “the plume,” this feature is characterized by an apparent upwelling of matter from beneath the surface of the water, which spreads away from its source point in two distinct lobes, one trending to the north, the other to the east (Figure 9e). Despite the feature’s obvious discharge of material, the nature of these emissions is clearly distinct from the origin of the oil slick to its northeast in the high-resolution imagery, and the plume does not appear to produce any effects on the surface beyond its immediate vicinity. This feature is visible only in high-resolution imagery; searches for signs of it in MODIS or ENVISAT-ASAR data returned no results- its location does not coincide with the origin of any identifiable spill candidates in that imagery. A number of natural and man-made interpretations were proposed to explain this phenomenon, as addressed in the “Discussion” section.
Along with the oil that was observed near the drilling platforms located offshore in the Caspian Sea, a number of spill candidates were also identified in Turkmenbashi Gulf and in the harbor of Turkmenbashi itself. Investigation of these candidates led to more equivocal results compared to the offshore site. A number of high-quality spill candidates that were observed in MODIS data failed to appear in radar imagery taken nearly simultaneously, suggesting that they may represent phenomena other than oil spills. Alternatively, the candidates’ lack of visibility in radar imagery may result from the fact that radar relies on an oil slick’s suppression of wind-driven wave action to create smoother areas that are easily distinguishable from the rougher areas outside the spill. It is possible that the calmer, protected waters of Turkmenbashi Gulf and Turkmenbashi harbor do not produce enough ambient wave action to enable consistent and reliable oil slick detection using radar imagery under most conditions (Figure 10).
Often, strong candidates were observed in MODIS imagery (left) that failed to be detected by radar (right). In the top image pair, acquired on 12 July 2005, this likely indicates that the feature is not an oil spill. In the bottom instance, acquired 21 June 2010, the feature detected by MODIS could be masked by calm waters that appear to pervade the inner Gulf. MODIS data from NASA, ENVISAT imagery ©2013 ESA.
Despite these challenges, several exceptional situations did exist in which radar data were able to identify or corroborate candidate slicks in the waters near Turkmenbashi. One such event was observed by ENVISAT on 14 May 2006, in which a dark patch appeared in the southern portion of Turkmenbashi Gulf that was visible in both MODIS and ENVISAT-ASAR data (Figure 11). Automated classification agreed with visual assessments that this candidate was a probable oil slick, and subsequent inspection of the source area using Google Earth revealed oil derricks- some located within two hundred meters of the shoreline- both in high-resolution satellite imagery and in geotagged photographs uploaded by visitors to the area (MACTAK, 2007). Given these mutually supporting pieces of evidence, it is nearly certain that this feature represents the slick resulting from an oil spill. Later, on 4 June 2011, another candidate was observed by ENVISAT approximately 23 kilometers south of Turkmenbashi. In this case, while both visual analysis and automated classification identified the feature as a probable oil slick, neither MODIS nor high-resolution optical imagery of the area was available on the date in question. Because a number of phenomena can produce dark areas in radar data that appear similar to an oil slick, in the absence of such corroboration or ground-based reports, this feature can only be considered as a “potential oil slick.”
On 14 May 2006, both MODIS and ENVISAT detected a slick in the southern portion of Turkmenbashi Gulf. Subsequent investigation revealed an oil field adjacent to its apparent source point on the Cheleken peninsula. MODIS data from NASA, ENVISAT imagery ©2013 ESA.
On several occasions, phenomena were also observed that appeared consistent with reports of leakage from Soimonov Bay into the waters adjacent to the harbor of Turkmenbashi. Such an event was observed in MODIS data from 15 June 2006, in which a large dark area was seen extending southeast from the dam which separates the Bay from the harbor (Figure 12). Likewise, radar imagery from 5 December 2004 appears to show a similar plume spreading into the Bay from the same location. As before, however, in the absence of contemporaneous high-resolution imagery or ground-based reporting, these images merely suggest the likelihood that pollution is taking place; they do not constitute definitive proof.
Two examples of possible outflow from Soimonov Bay. At left, MODIS imagery from 15 June 2006; at right, ENVISAT imagery from 5 December 2004. MODIS data from NASA, ENVISAT imagery ©2013 ESA.
Imagery analysis of the waters surrounding Turkmenbashi, Turkmenistan has demonstrated the feasibility of using MODIS as an effective and inexpensive platform for long-term oil spill monitoring. Although the method does have a number of disadvantages relative to radar, namely its reliance on clear weather and appropriate viewing geometry, the technique nevertheless has substantial advantages in terms of cost, extent of area covered, and repeat cycle. At times, MODIS is able to detect spill candidates under circumstances in which radar has difficulty, such as calm days or in protected waters (e.g., Figure 10). As with radar, MODIS is also susceptible to false positives, however, and care should be taken to confirm all candidates using other sensors before making a definitive judgment concerning a spill candidate. In this regard, the two sensors complement one another.
The feature known as “the squiggle” has been definitively identified as an oil slick. These frequent, low-volume spills often spread to cover a wide area and have been occurring semi-continuously for more than a decade. A number of possibilities exist regarding the origin of this oil, as discussed below. In terms of visual appearance the source of the squiggle is highly distinct from its nearby neighbor, “the plume,” located 2.3 kilometers to the west-southwest. Unlike the squiggle, which in the high-resolution imagery disperses outward from its source in only one direction (northeast), the plume contains two distinct directional components, one of which disperses towards the east, while the other exits the source point largely to the north. This bi-directional distribution indicates that two different forces are acting on the two different components of the plume. The east-west component’s aquamarine hue and diffuse appearance suggest that it is located beneath the surface of the water. The north-south component of the plume, by contrast, exhibits a distinct billowy nature, consistent with the action of the wind on airborne particles, and its direction of motion fits with meteorological data showing that the winds were blowing steadily from the south at the time the image was taken (Weather Underground, 2011). The absence of either a visible flame or of thick black smoke associated with the plume suggests the airborne component is not smoke resulting from the burning of methane or crude oil, while the lack of an oil slick in the immediate vicinity would appear to rule out a “gusher,” an uncontrolled well failure spewing aerosolized oil droplets into the air.
The observations are, however, consistent with steam, such as would be expected from an underwater volcanic eruption. Mud volcanoes are known to be active in the area, and often release a mixture of mud, methane, and hydrocarbons (Buryakovsky et al., 2001). Although neither the squiggle nor the plume correspond to any of the seamounts visible in bathymetric data (the nearest of which is located 8 kilometers to the west), when compared to imagery of the surface manifestations of submarine volcanoes, their similarity to the plume is striking (Associated Press, 2005). The steam plume from an erupting submarine mud volcano and its associated trail of waterborne sediment is therefore the most likely explanation for this feature.
The identification of the plume as a submarine mud volcano has a number of implications regarding the origin of the squiggle. The behavior of the two features is certainly very different; the squiggle releases oil that spreads out over a large area, yet exhibits no visible manifestation on the surface other than a slightly denser oil slick, while the plume violently erupts from beneath the surface yet shows no sign of contributing to an oil slick in over ten years of continuously collected data. In light of these facts, three possible explanations for the origin of the squiggle’s oil appear to exist:
1. The oil originates from another submarine mud volcano that was not erupting at the time that WorldView-2 passed overhead and acquired its imagery.
2. The oil originates from a leaking pipe, coupling, or other piece of submarine oil infrastructure.
3. The oil originates from a natural seep, possibly associated with the same geologic structure that caused the plume.
The first of these scenarios appears unlikely, as the location of the source of the squiggle in high-resolution optical imagery perfectly corresponds to the source observed in numerous MODIS and ENVISAT-ASAR images over the past decade. Given the tendency of a slick to drift with the current, one would expect that if the source of the oil had shut off at the time of the WorldView-2 pass, the entire slick would have displaced itself to the northeast, leaving clear water over its point of origin. Instead, while the slick did indeed drift in that direction, its origin remained fixed, suggestive of ongoing release. Additionally, the lack of any visible slicks associated with the plume in MODIS data indicates that the squiggle’s long-term behavior is distinct from that of a volcano.
The second possibility is circumstantially supported by the proximity of the squiggle’s source to confirmed oil rigs, and by its precise alignment with them. The source is nearly equidistant from the two rigs to its east and west, and is perfectly aligned with the axis of the two to its south. This alignment appears more than coincidental, and suggests that the oil slick and the oil rigs may be related, for example if submarine piping connecting the rigs to one-another has sprung a leak. There is evidence to support this hypothesis: according to an environmental impact assessment published one year before the spill first appears in MODIS imagery, submarine piping is known to exist in the area and to be in poor condition:
“There is a series of gathering pipelines from the individual wells/platforms which eventually connect to the main trunk pipelines to shore. These vary in diameter and flow under natural formation pressure. There are three main field export lines to shore, each approximately 37 km in length, but only two lines are in use. The Zhdanov trunk line is currently shut in pending re-routing of the landfall and integrity checking. The flowlines are not cathodically protected and there are no leak detection systems or regular inspections. The pipelines are exposed and their landfalls, are heavily corroded on the surface and coastal erosion has left them unsupported in some sections. Anti-corrosion wrapping was installed in early 1999, to prevent further corrosion, but was partially removed in an act of vandalism or theft” (Dragon Oil, 1999).
The area from which the squiggle originates appears to be a reasonable location for a three-way junction between the drilling rigs- a mechanism that could also be a plausible point of failure (Figure 13).
The symmetry of the source of the “squiggle” oil slick (red arrow) with respect to nearby drilling platforms is striking, as is its location near a logical junction point for underwater pipes. The Westernmost Platform is located beyond the edges of the figure, but is visible in ENVISAT imagery in Figure 8. Imagery ©2013 DigitalGlobe.
Despite this, however, the third possibility -a natural hydrocarbon seep- cannot be ruled out with the available data. While the placement of the source appears conveniently aligned with existing infrastructure, a natural seep would nevertheless be entirely consistent with observations and could also explain the squiggle’s longevity, as managing such a natural phenomenon might fall outside the mandate of the petroleum industry or environmental regulators. Until a detailed accounting of the area’s submarine geology and drilling infrastructure is available, a definitive explanation will remain unavailable.
Sustained and ongoing release of oil into the waters of the Caspian Sea near the city and port of Turkmenbashi represents a legitimate environmental concern. This study has documented hundreds of instances in which petroleum discharge has taken place near drilling platforms in the Caspian Sea and another that occurred adjacent to oilfields on the shores of Turkmenbashi Gulf. In addition, dozens of additional cases were observed in which oil pollution was highly likely, but could not be confirmed due to a lack of adequate archival data or contemporaneous ground-based reporting. These included a number of incidents in which material appeared to have escaped confinement in Soimonov Bay and spread into the harbor of Turkmenbashi itself. While the possibility remains that the perennial discharge event located at 39.548ºN, 52.616ºE represents natural seepage, the circumstances of its location are suspicious and warrant additional clarification.
Associated Press (2005). Japan: Eruption made 3,300-ft. vapor column. Retrieved from: http://www.nbcnews.com/id/8452611/#.UUsOxFt4aGg 2013 March 28
Baran, Zeyno (2005). “The Baku-Tbilisi-Ceyhan Pipeline: Implications for Turkey” . The Baku-Tbilisi-Ceyhan Pipeline: Oil Window to the West (The Central Asia-Caucasus Institute, Silk Road Studies Program): 103–118.
BBC News (2001). Caspian Pipeline Deal Signed. Retrieved from: http://news.bbc.co.uk/2/hi/europe/1571294.stm 2013 March 28
BBC News (2005). Giant Caspian Oil Pipeline Opens. Retrieved from: http://news.bbc.co.uk/2/hi/business/4577497.stm 2013 March 28
Brekke, C. and Solberg, A. (2004) Oil spill detection by satellite remote sensing, Remote Sens. Environ. 95, 1-13
Buryakovsky, L., Chilingar, G., and Aminzadeh, F., Petroleum Geology of the South Caspian Basin Woburn, MA: Gulf Professional Publishing, 2001.
Chuanmin, H., Müller-Karger, C., Myhre, D., Brock, M., Odriozola, A., and Godoy, G. (2003) MODIS Detects Oil Spills in Lake Maracaibo, Venezuela, Eos 84(33), 313-319.
Dragon Oil (1999). Block II Field Development Project: Environmental Impact Assessment, Environmental Resources management, 8 Cavendish Square, London W1M 0ER
Esaias, W., Abbot, M., Barton, I., Brown, O., Campbell, J., Carder, K., Clark, D., Evans, R., Hoge, F., Gordon, H., Balch, W., Letelier, R., and Minnett, P. (1998) An overview of MODIS capabilities for ocean science observations IEEE Transactions on Geoscience and Remote Sensing 36(4), 1250-1265.
ExxonMobil Research and Engineering Company (2008), ExxonMobil Oil Spill Response Field Manual, Retrieved 2013 March 28.
Lotliker, A., Mupparthy, R., Kumar, S., and Nayak, S. (2008) Evaluation of Hi-Resolution MODIS-Aqua data for oil spill monitoring Proc. of SPIE 7150, 71500S, doi: 10.1117/12.804907
MACTAK (2007). Geotagged photograph uploaded to Panoramio, retrieved via GoogleEarth. Available at: http://www.panoramio.com/photo/1523617 2013 March 28
National Oceanic and Atmospheric Administration, Office of Response and Restoration, Emergency Response Division Open Water Oil Identification Job Aid for Aerial Observation, Version 2. Seattle: U.S. Department of Commerce, 2012
Solberg, A., Brekke, C., and Solberg, R., Algorithms for oil spill detection in Radarsat and ENVISAT SAR images (2004) Geoscience and Remote Sensing Symposium, 2004. IGARSS ’04. Proceedings. 2004 IEEE International, 20-24(7) 4909-4912.
Weather Undergrond, History for Turkmenbashi, Turkmenistan, Sunday, March 27, 2011 Retrieved from: http://www.wunderground.com/history/airport/UTAK/2011/3/27/DailyHistory.html?req_city=NA&req_state=NA&req_statename=NA 2013 March 28.
A PDF of Satellite Imagery Analysis for Environmental Monitoring: Turkmenbashi, Turkmenistan is available here.