You are on page 1of 19

Manuscript Click here to view linked References

1
2
3
4
5 An assessment on oil spill trajectory prediction - Case study on oil
6
7 spill off Ennore Port
8
9
10 (Ennore oil spill trajectory prediction assessment)
11
12
13 Prasad SJ1*, Balakrishnan Nair TM1, Hasibur Rahaman1, Shenoi SSC1 & Vijayalakshmi T2
14
15
16
17 Abstract
18
19
20 A Liquefied Petroleum Gas (LPG) tanker and a chemical tanker collided two nautical miles off
21
22 Ennore port on 28 January 2017. Around 196.4 Metric Tons (MT) of Heavy Furnace Oil (HFO)
23
24 was spilled and drifted towards the shore. Oil spill drift advisory and prediction was made by
25
26 Indian National Centre for Ocean Information Services (INCOIS) using General National
27 Oceanic and Atmospheric Administration (NOAA) Operational Modeling Environment
28
29 (GNOME), an oil spill trajectory model. The trajectory model was forced with analysed and
30
31 forecasted ocean currents from Global Ocean Data Assimilation System (GODAS) based on
32
33 Modular Ocean Model 4p1 (GM4p1). It was found that spread of HFO obtained from oil spill
34
35 trajectory model GNOME, has matched well with the observed spread from Sentinel -1A
36
37
satellite dataset. However the spread of the HFO was underestimated by the trajectory model,
38 when forced with forecasted GM4p1 currents. Additional ground truth observation from coast
39
40 guard also corroborates this finding.
41
42
43 Keywords : Ennore, oilspill, trajectory, radar, fronts, GNOME.
44
45
46 *email: prasadsj@incois.gov.in.
47
48 1
ESSO- Indian National Centre for Ocean Information Services, Hyderabad, Telangana, India.
49
50
2
51 Jawaharlal Nehru Technological University, Hyderabad, Telangana, India.
52
53
54
55
56
57
58
59
60
61
62 1
63
64
65
1
2
3
4
5 1. Introduction
6
7 On 28 January 2017, 0345 hrs, Indian Standard Time (IST), BW Maple, an outbound Liquefied
8
9 Petroleum Gas (LPG) tanker and an inbound chemical tanker MT Dawn Kanchipuram, collided
10
11 about two nautical miles (13.2282º N, 80.3633º E) off Kamarajar Port, Ennore. According to
12
13 local port authority, the hull of the vessel MT Dawn was ripped, damaging the ship's
14
15 accommodation as well as the pipelines on the deck. As per the information from Indian Coast
16
17
Guard, East, this accident has resulted in the spillage of 196.4 Metric Tons of Heavy Furnace Oil
18 (HFO). The HFO spill location is shown in Figure 1. The nowcast and predictions were made
19
20 for the drift of HFO, using an oil spill trajectory model, General National Oceanic and
21
22 Atmospheric Administration (NOAA) Operational Modeling Environment (GNOME). The
23
24 simulated and observed drift of HFO are compared and presented in this paper.
25
26
27 1.1 Request from Indian Coast Guard
28
29
30 Indian Coast Guard (ICG), Ministry of Defence, is responsible for the cleanup and mitigation of
31 the damages due to the oil spill in Indian coastal waters. A specific request came from Indian
32
33 Coast Guard, Chennai, to INCOIS for providing the advisory, on the drift of spilt HFO in
34
35 nowcast and forecast modes. As per ICG's request the trajectory prediction model was run with
36
37 ICG's initial estimated 20 Metric Tons of HFO and the advisories were disseminated as bulletins
38
39 for their clean up operations commenced after 28 January 2017 on near real time. The change in
40
41
the quantity of the HFO due to clean up operations was not considered in modeling the drift. ICG
42 after detailed analysis reported that the actual quantity of HFO spillage was 196.4 MT. Then, to
43
44 make an objective assessment, the model was re-run with this actual quantity (196.4MT) during
45
46 0400 hrs of 28012017 to 2300 hrs of 05022017 to estimate the drift of spilled HFO.
47
48
49 1.2 Oil spill trajectory prediction set up at INCOIS
50
51
52 Indian National Centre for Ocean Information Services (INCOIS), Hyderabad developed a
53
54 numerical modeling setup to predict the trajectory of the spilled HFO, based on prevailing,
55 forecasted winds, currents and waves in the region of spill. The core model is GNOME, an oil
56
57 spill trajectory model developed by NOAA (Beegle Krause 2011). This model was adopted from
58
59 NOAA and operationalised at INCOIS by customizing it in diagnostic mode for Indian ocean.
60
61
62 2
63
64
65
1
2
3
4 The diagnostic mode allows the user to set his own scenario (Beegle Krause 2011). The model
5
6 estimates the movement of oil spill as the vector sum of wind speed, current speed and diffusion
7
8 turbulence. The resultant of these forcings at each and every time step gives the impetus for the
9
10 oil to drift. GNOME uses three phased algorithm, in which the pollutant is treated as three
11
12 component substances with independent half lives (Zelenke et al. 2012). Prasad et al. (2014)
13
14
explained the detailed methodology adopted for predicting the trajectory of spilled oil and its
15 validation for Mumbai oil spills. In the present study the wind drift was obtained from the
16
17 forecasted wind fields of European Centre for Medium weather Range Forecast (ECMWF).
18
19 ECMWF winds were validated with in situ data obtained from the Automatic Weather Stations
20
21 (AWS) installed on board ships. The Root Mean Square error was found to be less than 2.6 m s-1
22
23 (Harikumar et al. 2012). As per the trajectory analysis handbook of NOAA,3% of wind speed is
24
25
considered for the drift of the pollutant parcel.
26 (http://response.restoration.noaa.gov/sites/default/files/Trajectory_Analysis_Handbook.pdf). The
27
28 diffusion turbulence is based on the horizontal eddy diffusivity of ocean water and value of
29
30 100,000 cm2 s-1 (Zelenke et al. 2012) was used by the present GNOME trajectory model. In the
31
32 present study we have used ocean surface currents from a global assimilation system. Global
33
34 Ocean Data Assimilation System (GODAS) based on MOM4p0d adopted from NOAA/NCEP
35
and operationalised at INCOIS in 2013. This system uses state of the art ocean general
36
37 circulation model called Modular Ocean Model (MOM) version-4p0d and 3D-VAR assimilation
38
39 technique to assimilate in situ data in the model. Temperature and salinity profiles from all in-
40
41 situ observations over the global ocean are assimilated to produce the best analysis products
42
43 (Ravichandran et al. 2013, Siva Reddy et al. 2015). Recently, Rahaman et al. (2015) upgraded
44
45 the GODAS with MOM4p1 and have further improved the ocean analysis with this upgraded
46 version. We used this upgraded GODAS analysed and forecasted surface currents. Hereafter, the
47
48 GODAS based MOM4p1 version will be referred as GM4p1.
49
50
51 2.0 The Simulations of trajectory prediction model
52
53
54 The simulations of the trajectory model, GNOME, was made with 196.4 MT of HFO. The
55
56 trajectory model was forced with analysed forcing fields of GM4p1 on daily basis until 2300 hrs
57
58 of 01022017. Figure 2 shows the nowcasted status (floating as blue dots and beached as black x
59
symbols) of the spilled HFO from 0400 hrs of 28012017 to 2300 hrs of 01022017. It can be seen
60
61
62 3
63
64
65
1
2
3
4 that ~ 6 km of coastal stretch was affected due to the HFO drift on 1400hrs of 28012017 (Figure
5
6 2a). This affected area further extended towards South upto ~18 km along the coast on 2300hrs
7
8 of 29012017 (Figure 2b). In the next two days, the HFO spread further South and reached near
9
10 Thiruvanmiyur and Palavakkam (Figure 2c,d). Afterwards, the trajectory model was run with
11
12 forecasted GM4p1 currents from 0000 hrs of 02022017 to 2300 hrs of 05022017, for which the
13
14
forecast was issued from our laboratory on 0000 hrs of 02022017. Figure 3 shows the predicted
15 status (floating as red dots and beached as black x symbols) of the spilled HFO from 0000 hrs of
16
17 02022017 to 2300 hrs of 05022017. It can be seen from Figure 3 that, HFO will drift further
18
19 South of Palavakkam during 2300hrs of 02022017 to 2300hrs of 05022017. Further, ICG has
20
21 reported the remnants of HFO at various coastal locations such as Ennore, Thiruvottiyur,
22
23 Thiruvanmiyur and Palavakkam on 05022017. These locations were very much within the
24
25
predicted area over which the oil spill will move. Further we have evaluated the nowcasted and
26 forecasted HFO spread with the independent satellite observation.
27
28
29 3.0 Synthetic Aperture radar (SAR) data used for validating HFO spread
30
31
32 The available VV (Vertical-Vertical) polarized data from space borne Synthetic Aperture Radar
33
34 (SAR) sensor was obtained for the pass that occurred over Ennore region at 0600 hours (IST) on
35
36 29012017 from Sentinel -1A data hub (https://scihub.esa.int/dhus/). This data was processed
37
38
using Sentinel tool box to identify the signature of oil spill. The radar reflectance values are
39 converted to dimensionless normalized Sigma0 intensity values in Sentinel tool box. This
40
41 Sigma0 values were wrote in netcdf format for identifying the oil slicks. Figure 4a shows radar
42
43 reflectance from the SAR sensor. The intensity values are plotted in logarithmic scale. The low
44
45 back scatter values are shown as dark blue and high values are in dark red. As per our
46
47 knowledge, there is no objective criteria is available to identify HFO based on radar reflectance
48
49
values. However typically the presence of oil shows low reflectance values over ocean water. In
50 order to identify the extent of HFO spread near to Ennore we have used radar reflectance values
51
52 of -2.4 (log 0.004). This value was used based on the in-situ observations of ICG. The pixels
53
54 with intensity values ≤-2.4 (≤ log 0.004), corresponds to the ICG reported locations (Ennore,
55
56 Thiruvottiyur) of beached HFO slicks on 29012017. Hence they are considered as HFO slicks.
57
58 However all the pixels with values less than -2.4 does not corresponds to HFO. The lower back
59
scatter values can also be associated with the oceanic fronts (Lorenzzetti et al. 2008, Lyzenga et
60
61
62 4
63
64
65
1
2
3
4 al. 2004). Macedo et al.(2015) noticed the ocean fronts in Multiscale Ultra High Resolution
5
6 (MUR) Sea Surface Temperature (SST) data which reflected in Envisat datasets obtained on
7
8 October 15, 2009. Hence, we have analyzed the 1 km resolution SST data, from Group of High
9
10 Resolution Sea Surface Temperature (GHRSST https://www.ghrsst.org/ghrsst-data-
11
12 services/products/) to see whether any oceanic front exists in that area. Since we could not
13
14
acquire the concurrent SST data exactly during acquisition of SAR data, we have analysed the
15 daily averaged SST data obtained on 29 January 2017 to delineate the fronts present in the
16
17 region. Figure 4b shows the spatial distribution of daily averaged SST on 29 January 2017. The
18
19 presence of prominent fronts can be seen corresponds to low radar reflectance values in Figure
20
21 4a. To delineate the spread of HFO, the areas covered by fronts were then excluded from the
22
23 SAR data. Finally the HFO's spread is identified and marked with the black contour in Figure 4a.
24
25
The SAR data shows that ~ 15 km of the coastline was affected by the HFO at 0600 hrs on
26 29012017 and the area affected by the oil spill works out to be ~ 105 km2.
27
28
29 4.0 Comparison of model nowcast/ forecast with satellite observations
30
31
32 The comparison between the trajectory obtained from analysed/forecasted GM4p1 currents and
33
34 the observed oil spill signature from SAR data is shown in Figure 5. The comparison was made
35
36 at the time of SAR acquisition, that is at 0600 hours on 29012017. The forecast on ocean
37
38
currents issued from our laboratory on 0800hrs of 27012017, was used for the comparison. The
39 observed spread zone of HFO exported from SAR data is shown as black contour in Figure 5a.
40
41 The spread of HFO based on analysed and forecasted currents are shown as blue and red dots
42
43 respectively. The beached status of HFO on 0600 hrs of 29012017 are indicated as black x
44
45 symbols. It can be seen that the offshore and the alongshore spread of HFO estimated using the
46
47 analysed fields of currents compares better with the observations using SAR, than the trajectory
48
49
predicted with the forecasted currents. The predicted spill spread captured the alongshore spread
50 of HFO, but did not capture the offshore spread. This is not surprising, considering the higher
51
52 accuracies of analysed fields compared to the forecasted fields in any model. Figure 5b, shows
53
54 the plot of comparison between the predicted and in-situ observed drift of HFO on 05022017.
55
56 The predicted drift of HFO using forecasted GM4p1 is indicated as red dots (floating) and black
57
58 x symbols (beached). The blue ellipses are the coastal locations where ICG reported the presence
59
of HFO. It can be seen that the predicted spread matches very well with the observed HFO.
60
61
62 5
63
64
65
1
2
3
4
5 5.0 Conclusions
6
7 An oil spill occurred off Ennore port, Chennai on 28 January 2017 due to the collision of two
8
9 merchant vessels. Advisories on the drift and spread of 20 MT HFO were issued, to the Indian
10
11 Coast Guard on nowcast and forecast basis during 28012017 to 05022017. This advisory was
12
13 based on a trajectory model GNOME. In-order to make an objective assessment, the simulations
14
15 were updated with the actual spill quantity of HFO (196.4 MT). Upgraded GODAS
16
17
analysed/forecasted currents and ECMWF winds are used to force the trajectory model. The
18 simulated trajectories match well with SAR and ICG in-situ observations. The observed along
19
20 and off shore spread of the slicks were well captured by the GNOME when forced with analysed
21
22 GM4p1 currents. The offshore spread was slightly underestimated while using forecasted
23
24 GM4p1 currents.
25
26
27 Acknowledgements
28
29
30 The authors thank the officials of Indian Coast Guard, Chennai for their information on oil spill
31
32 and support in validating the trajectory predictions of spilled HFO. ArcMap tool was used to plot
33
34
and generate the trajectory output in native EPS format. Ferret software in Linux environment
35 was used to interpret the SAR data. The authors thank the Editor & Reviewers for their
36
37 comments, suggestions in improving the manuscript's quality. This research paper holds the
38
39 INCOIS contribution number 314.
40
41
42 References
43
44
45 Beegle Krause J 2001 General NOAA Oil Modeling Environment (GNOME): A new spill
46
47 trajectory model; International Oil Spill Conference Proceedings, Vol. 2001, No. 2, 865-871.
48
49 Harikumar R, Balakrishnan Nair TM, Bhat GS, Shailesh Nayak, Venkat Shesu Reddem, &
50
51 Shenoi SSC 2012 Ship-Mounted Real-Time Surface Observational System on board Indian
52
53 Vessels for Validation and Refinement of Model Forcing Fields; J. Atmos. Oceanic.
54
55 Technol. 30, 626–637.
56
57
Lorenzzetti JA, Kampel M, Franca B and Sartori A 2008 An assessment of the usefulness of
58
59 SAR images to help better locating the Brazil Current surface inshore front; International
60
61
62 6
63
64
65
1
2
3
4 Workshop on Advances of SAR oceanography from Envisat and ERS Missions, Frascati
5
6 Proceedings, European Space Agency, 1-42.
7
8
9 Lyzenga DR, Marmorino GO and Johannessen JA 2004 Ocean currents and current gradients;
10
11 JACKSON, CR APEL, JR. (Eds.). Synthetic aperture radar marine user’s manual; Washington,
12
13 NOAA, 207-220.
14
15 Macedo, Carina Regina de, Lorenzzetti and Joao Antonio 2015 Numerical simulations of SAR
16
17 microwave imaging of the Brazil current surface front; Braz. J. Oceanogr. 63(4), 481-
18
19 496; doi.org/10.1590/S1679-87592015082306304.
20
21
22 Prasad SJ, Balakrishnan Nair TM, Francis PA and Vijayalaksmi T 2014 Hindcasting and
23
24
validation of mumbai oil spills using GNOME; Int. Res. J. Environ. Sci. Vol 12, 3 -12.
25
26 Rahaman H, Behringer DW, Penny SG and Ravichandran M 2015 Impact of an upgraded model
27
28 in the NCEP Global Ocean Data Assimilation System; the tropical Indian Ocean; J. Geophys.
29
30 Res. Oceans ; doi:10.1002/2016JC012056.
31
32
33 Ravichandran M, Behringer D, Sivareddy S, Girishkumar MS, Chacko N, Harikumar R, 2013
34 Evaluation of the global ocean data assimilation system at INCOIS; the tropical Indian Ocean,
35
36 Ocean Modell. 69,123–135; doi:10.1016/j.oceamod.2013.05.003.
37
38
39 Sanikommu Sivareddy, Muthalagu Ravichandran, Madathil Sivasankaran Girishkumar, Koneru
40
41 Venkata Siva Rama Prasad 2015 Assessing the impact of various wind forcing on INCOIS-
42
43 GODAS simulated ocean currents in the equatorial Indian Ocean ; Ocean Dyn. September,
44 Volume 65, Issue 9, 1235-1247; doi 10.1007/s10236-015-0870-6.
45
46
47 Zelenke B, Connor CO, Barker C, Beegle-Krause J and Eclipse L 2012 General NOAA
48
49 Operational Modeling Environment (GNOME) Technical Documentation; U.S. Dept. of
50
51 Commerce, NOAA Technical Memorandum NOS OR&R 40 Seattle, WA, Emergency Response
52
53 Division, NOAA,pp 105.
54
55
56
57
58
59
60
61
62 7
63
64
65
1
2
3
4
5
Figure No & Captions
6
7 Figure 1
8
9
10 Map of India along with HFO spill location near Ennore port at the East coast of India. zoomed
11
12 version of the HFO spill location is also shown in the rectangular box.
13
14
15
Figure 2
16
17 Nowcasted HFO drift using analyzed GM4p1 currents. The black x (blue dot) denotes the
18
19 beached (floating) status of drifted HFO in all the panels a) Spread of HFO on 1400 hrs of
20
21 28012017 b)Spread of HFO on 2300 hrs of 29012017 c) Spread of HFO on 2300 hrs of
22
23 31012017 d) Spread of HFO on 2300 hrs of 01022017.
24
25 Figure 3
26
27
28 Forecasted HFO drift using forecasted GM4p1 currents. The black x (red dot) denotes the
29
30 beached (floating) status of drifted HFO in all the panels a) Spread of HFO on 2300 hrs of
31
32 02022017 b) Spread of HFO on 2300 hrs of 03022017 c) Spread of HFO on 2300 hrs of
33
34 04022017 d) Spread of HFO on 2300 hrs of 05022017.
35
36 Figure 4
37
38
39 a) Logarithmic plot of radar reflectance from Sentinel-1A satellite observation on 0600hrs of
40
41 29012017. The black contour line shows the HFO spread zone b) Daily averaged Sea Surface
42
43 Temperature (SST in ºC) on 29012017 from GHRSST observation.
44
45
46
Figure 5
47
48 a) Comparison of HFO drift obtained from analyzed and forecasted currents on 0600 hrs of
49
50 29012017 with satellite observation. The black contour line indicates the observed HFO spread
51
52 zone obtained from SAR dataset. The blue (red) dots indicate the spread of HFO obtained from
53
54 analyzed (forecasted) GM4p1 currents. The black x denotes the beached status of the HFO. b)
55
56
Comparison of predicted HFO drift on 2300hrs of 05022017 with the in-situ observation. The
57 black x (red dot) denotes the beached (floating) status of HFO obtained from forecasted GM4p1
58
59
60
61
62 8
63
64
65
1
2
3
4 currents. The blue ellipses are the coastal locations, where, ICG reported the presence of HFO on
5
6 05022017.
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62 9
63
64
65
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47 Figure 1
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62 10
63
64
65
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53 Figure 2
54
55
56
57
58
59
60
61
62 11
63
64
65
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53 Figure 3
54
55
56
57
58
59
60
61
62 12
63
64
65
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49 Figure 4
50
51
52
53
54
55
56
57
58
59
60
61
62 13
63
64
65
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35 Figure 5
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62 14
63
64
65
Figure1
Figure2
Figure3
Figure4
Figure5