A Survey on Data Correction of Observation and Prediction Using Machine Learning: Preliminary Study for Optimizing Oil Spill Model

Accurate prediction of the movement of spilled oil is important for a fast and effective response to an oil spill accident. To predict the dispersion of released oil, numerical models using physiochemical properties and/or environmental conditions as input data have been applied. The environmental conditions that are provided as observation data have values that vary depending on time and space; moreover, accuracy can be compromised due to the occurrence of erroneous data or missing data during the observation. This study investigates observation and prediction data correction methods using machine learning techniques and evaluates their applicability to dispersion models for spilled oil. It is expected that the performance of these dispersion models can be improved by improving the accuracy of the data through the data correction process.