TRACEABILITY OF OIL SPILL FROM BAYESIAN CLASSIFICATION

Oil spills over the sea resulted in the destruction of the marine environment. Therefore, the Bayesian algorithm was developed to determine the source of spilled oil for reducing the damage to the marine environment. Based on the flow data of the offshore waters of the Bohai Sea, the influencing factors of oil spill drift were used for algorithm. The main factors affecting the traceability of marine oil spills are considered: the distance from the oil point to the source point, the flow direction of water, the type of the spilled oil source, and the scale of the source. The algorithm is completed on the basis of a feasible simulation database. We not consider the relationship between the attributes by naive Bayesian classification. In this paper, we determine the result by the minimum error rates of each source. Then the performance evaluation of the model was done by cross-validation method. The experimental result shows that the Bayesian algorithm can be used to determine the source of spilled oil. It is easier and faster to determine the source by the method raised. * Corresponding author The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Volume XLII-3/W10, 2020 International Conference on Geomatics in the Big Data Era (ICGBD), 15–17 November 2019, Guilin, Guangxi, China This contribution has been peer-reviewed. https://doi.org/10.5194/isprs-archives-XLII-3-W10-133-2020 | © Authors 2020. CC BY 4.0 License. 133


INTRODUCTION
The oil spills have caused damage to the economic development and human of the affected coastal areas. In the event of marine oil spills, if the source of the oil spill cannot be directly identified, it is difficult to confirm the person responsible for the accident. Penalties for those responsible will cause public concern about the marine environmental pollution.
The existing methods are based on the "Lagrange" particle tracking method to establish a model for predictive analysis (Cao, 2016), or the use of oil fingerprint library from oil components (Zhang, Wang, 2016), or the simple distance judgment. To meet the needs for the source determination of oil spill, so we propose a naive Bayesian classifier for this practical problem in this paper.
The structure of this paper is as follows. First, we discuss the theory of Bayesian classification. Bayesian classification was applied to determine the oil spill source after analysing the factors of oil spill drift. Second, the superior performance of Bayesian classification is proved by experiments, and the correct rate of classifier is evaluated by cross-validation.
Finally, the performance of the method application is summarized.

BAYESIAN CLASSIFICATION
The Bayesian classification is characterized by the use of probabilities to represent all forms of uncertainty, and learning or reasoning is done using probability rules. When the amount of sample data is large, the frequency of each sample obtained from a large number of repeated tests tends to a stable value. Therefore, the frequency of occurrence of an event is often treated as the probability of occurrence of an event in actual problems, and there may be more than two categories in practice, which may be more complicated. Then, the Bayesian formula in the complex case is as follows:

NAIVE BAYESIAN CLASSIFIER MODEL
3.1 Study Area Bohai Sea urban agglomeration is an important economic zone in China, and the regional GDP has reached 3.8 trillion yuan, accounting for 28.2% of the national total. Therefore, The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Volume XLII-3/W10, 2020 International Conference on Geomatics in the Big Data Era (ICGBD), 15-17 November 2019, Guilin, Guangxi, China monitoring of marine oil spills is important for offshore environmental protection and economy.

Factors for Oil Spill Drift
In offshore waters, tidal currents and wind currents are important factors in determining oil spill drift. Due to the limited statistics on marine oil spill accidents, the sources of oil spill have been determined by the type and scale of the possible oil spill sites. Therefore, the analysis of the oil spill source identification based on the simulation data is reliable in this paper. The oil spill movement affected by the oil spill source and the ocean flow field. We use the distance from the oil source to oil spill point, the current direction, the type of oil source, and the size of the oil source to make a determination.

Marine environment simulation data
The distance from the oil spill zone to the oil source Current flow Oil source type Oil spill source scale

Construction of Bayesian classifier
Determine the oil source based on the minimum error rate    The research is mainly based on domestic oil spill grading, and the oil source scale is divided into three categories for simulation data processing, which respectively represent the three levels of large, medium and small oil spills.

BAYESIAN CLASSIFIER PERFORMANCE EVALUATION
We use cross-validation to test the performance of the naive  Assuming an existing example, the oil spill zone B was found in July of that year. Because it is considered that monitoring of marine oil spills will not be carried out every day, we use the flow rate of the day to establish the sample simulation data which making the model closer to the actual situation. The flow direction derived from the properties of the flow field map is the counter clockwise angle from the east direction, and the deviation value is obtained by combining the angle obtained in the above Figure 5. Table 1 Table 3. Oil spill source determination of zone A Figure 6. Result of Bayesian classifier cross-validation The source of oil spill zone A is oilfield 3, the source of oil spill zone B is oilfield 1, and the drift of oil spill zone A and B is shown by the red arrow in Figure 7. Through the analysis of the flow field of the zone during the period, the oil traceability judgment is in line with reality, and it can be concluded that: Utilizing the Naive Bayes classifier, with the principle of minimum error rate, the source of marine oil spill can be determined.
The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Volume XLII-3/W10, 2020 International Conference on Geomatics in the Big Data Era (ICGBD), 15-17 November 2019, Guilin, Guangxi, China