시간차분 변수의 분산측도를 이용한 시계열 용존산소 자료의 단기 변동 분석

Abstract

This study presents a method for analyzing short-term fluctuations in time-series (TS) data, which is essential for ocean monitoring. The proposed approach utilizes statistical measures from time-differenced variables defined across various time lags, linking the standard deviation (SD) with the autocorrelation function (ACF). This connection facilitates the exploration of variability and stationarity. Focusing on variations over a 7-day period, this method is applied to dissolved oxygen concentration data collected from the marine water quality monitoring network at the Nakdong Myeongji and Eulsuk stations, provided by the Korea Marine Environment Management Corporation (KOEM). The findings suggest that the standard deviation of differenced variables exhibits fluctuation patterns consistent with those predicted by the autocorrelation theory. Rapid increases are observed within a time frame of less than 6 hours, followed by gradual changes over longer time scales. Additionally, a periodic decline is noted after approximately 12 hours on a daily scale, indicating cyclic fluctuations within the daily cycle. In conjunction with the standard deviation, robust dispersion measures are introduced to highlight the differences, with a specific emphasis on the median absolute deviation to evaluate the normality of ocean observation data. Consequently, the proposed technique serves as a valuable quantitative analytical tool for assessing diurnal variability in dissolved oxygen concentrations and for testing the assumptions of stationarity and normality in time series data.