The Argo program measures temperature and salinity in the upper ocean(0–2000 m).These observations are critical for weather/climate studies,ocean circulation analysis,and sea-level monitoring.To address the limitatio...The Argo program measures temperature and salinity in the upper ocean(0–2000 m).These observations are critical for weather/climate studies,ocean circulation analysis,and sea-level monitoring.To address the limitations of traditional thresholds in Argo data quality control(QC),this study proposes a novel probability distribution-based inference method(PDIM)for temperature-salinity threshold inference.By integrating historical observations with climatological data,the method utilizes historical data corresponding to latitude and longitude grids,calculates temperature/salinity frequency distributions for each depth,and determines“zero probability”boundaries through combined frequency distribution and climatology data.Then a probability distribution model is established to detect outliers automatically based on the features in the probability density function,which eliminates the traditional dependence on the normal distribution hypothesis.When applied to global Argo datasets from China Argo Real-time Data Center(CARDC),PDIM successfully identifies suspicious profiles and sensor drifts with high reliability,achieving a low false positive rate(0.55%for temperature,0.18%for salinity)while maintaining competitive true positive rate(28.29%for temperature,55.15%for salinity).This method is expected to improve the reliability of Argo data QC and has important significance for Argo QC.展开更多
基金The National Key Research and Development Program of China under contract No.2021YFC3101503the Hunan Provincial Natural Science Foundation of China under contract No.2023JJ10053+1 种基金the National Natural Science Foundation of China under contract Nos 42276205 and 42406195the Youth Independent Innovation Science Foundation under contract No.ZK24-54.
文摘The Argo program measures temperature and salinity in the upper ocean(0–2000 m).These observations are critical for weather/climate studies,ocean circulation analysis,and sea-level monitoring.To address the limitations of traditional thresholds in Argo data quality control(QC),this study proposes a novel probability distribution-based inference method(PDIM)for temperature-salinity threshold inference.By integrating historical observations with climatological data,the method utilizes historical data corresponding to latitude and longitude grids,calculates temperature/salinity frequency distributions for each depth,and determines“zero probability”boundaries through combined frequency distribution and climatology data.Then a probability distribution model is established to detect outliers automatically based on the features in the probability density function,which eliminates the traditional dependence on the normal distribution hypothesis.When applied to global Argo datasets from China Argo Real-time Data Center(CARDC),PDIM successfully identifies suspicious profiles and sensor drifts with high reliability,achieving a low false positive rate(0.55%for temperature,0.18%for salinity)while maintaining competitive true positive rate(28.29%for temperature,55.15%for salinity).This method is expected to improve the reliability of Argo data QC and has important significance for Argo QC.
