基于时序相关性分析方法的浮标异常数据识别

海洋生态浮标异常数据的实时早期监测识别是保证观测数据质量的关键。本研究通过对浙江沿海浮标多年数据的分析，发现了与传统跳变异常数据不同的渐变异常数据类型。该异常类型呈现出在时序变化过程中连续平稳，但随时间逐渐偏移，最后整体偏离正常的分布特征，并且单一参数的分析方法无法对此异常进行有效识别。因此本研究利用海洋环境参数中酸碱度(pH)、溶解氧(DO)和叶绿素(Chla)三者的多参数相关性规律，提出了在一定时序上两两参数间相关性是稳定甚至是一致的假设，将8 天时间窗口的两两相关系数(R_{8 d})和前后两天R_{8 d}之差的绝对值(ΔR)作为相关性和稳定性核心指标，建立了基于相关性的渐变异常数据自动识别方法。为浮标传感器渐变异常的早期识别提供了一个新的思路，有助于提升海洋生态浮标异常数据的自动化监测能力。

Identification of abnormal buoy data based on time series correlation analysis method

The identification of abnormal marine ecological buoy data is the key to ensure the quality of buoy data. In this study, we found that the gradual abnormal data type is different from the traditional jump abnormal data through analysis of the coastal buoy data in Zhejiang for many years. With a single parameter analysis method, it is difficult to work out accurately the new gradual abnormal data type of stable and gradual deviation from the normal data. Therefore, multiple parameters correlation coefficient method is proposed based on the relationships between pH, dissolved oxygen and chlorophyll a on the condition of that the correlation between two parameters is stable or even consistent at a certain time series. There are two simple statistical parameters of the cross-correlation coefficient of 8-day time window (R_{8 d}) and the difference of R_{8 d} (ΔR) in this method. Those could be used to automatically detect the gradual abnormal buoy data and do very well. The multiple parameters correlation coefficient method provides a new idea for the gradual abnormal data identification, and also improves the automatic monitoring capability of marine ecological buoy abnormal data.