Research Of Thermocline Processing Method Based On Argo Data

Based on the ocean data of Argo and the World Ocean Atlas,this paper analyzes and studies seawater temperature that is one of the chief marine characteristics to do relative researches on thermocline.From the very beginning,we predict the data using the method of KNN regression.With the method,we managed to choose the optimal parameter for the constructing of prediction model.After that,the temperature and salinity data in BOA_Argo is refined using the regression forecast model.The original horizontal resolution is 1 ° x 1 ° and the original datasets are vertically divided into uneven 58 layers from sea surface to the 1975 meters underwater.Refine the original data into 1°×1° horizontally and in a 1-meter interval vertically.After all these processes,we found that there are more thermocline layers obtained then before,show that too much data granularity makes some layers whose thickness are within 5 meters or so,usually called the "critical",not to be selected.Therefore,when it comes to thermocline,the data refinement described in this article is of great importance.Subsequently,entropy method,decision tree method,random forest mothed and other methods are used to research the thermocline,and based on all these methods,a new criterion for the evaluation of thermocline is proposed.This selection cannot only make us quantitative analyze the thermocline,by means of finding out the samples that constitute the thermocline,but also qualitative analyze the thermocline,by means of finding out which samples are forming the thermocline.Compared with the traditional selection principle of thermocline,‘the stronger,the better',the score evaluation criterion proposed in this paper takes all the characteristic features into consideration,which results are more comprehensive and accurate.The principle of ‘the stronger,the better' is based on the definition of the strength of the thermocline,considering the temperature and depth,as well as the temperature differences and the depth differences.However,the formation of the thermocline is also related to other factors.The traditional method has not been able to take other factors into consideration,and thus has some limitations.The score selection principle based on the entropy value method makes up for the principle of ‘the stronger,the better' only consider strength while ignoring the other characteristics.The rectification of such defects makes the results more objective.In order to ensure the consistency of the traditional method of thermocline selection and the new selection method,this paper establishes a one-to-one mapping relationship from strength to score.After obtaining the mapping relationship from strength to score,we can not only quantitatively analyze the thermocline,but also qualitatively analyze the thermocline.The practicability of this method has been validated through data of BOA_Argo.