An improved method for quality control of in situ data from Argo floats using α convex hulls

Abstract

An improved method for detecting abnormal oceanic in situ temperature and salinity (T/S) profiles is developed. This procedure extends previous method developed by Udaya Bhaskar et al. [2017].This method utilizes World Ocean Atlas 2013 gridded climatology which is on 0.25° x 0.25° resolution to build α convex hulls. These α shapes are then used to categorize good and bad in situ T/S data profiles. This extended method classify the entire profiles instead of data for standard depths to avoid any errors introduced by interpolation to standard depths. Like in previous method, an 'n' sided polygon (convex hull) encompassing the T/S profile data is constructed using Jarvis March algorithm and Points In Polygon (PIP) principle is employed to judge the profile as good or bad. Extensive sensitivity experiments were done for arriving at the optimal α value such that false positives and true negatives are minimized. All types of issues associated with the in situ oceanographic data are identified and quality flag assigned. Examples of this improved method as applied to few Argo floats are presented. • The T/S profiles corresponding to region of interest are used to build α convex hulls. • This extended method can be effectively used for quality control of entire profile and clearly demarcate the profile as good/bad. • This method has the advantage of treating bulk of oceanographic in situ profiles data in a single go which filters out erroneous profile data from the good.

Keywords

• α Convex Hulls
• Point in polygon
• Outliers
• In situ data
• Argo floats
• Classification