Statistical Quality Control Algorithm for In Situ Sea Surface Temperatures

<p>In situ sea surface temperature (SST) observations—collected from drifting buoys, moored systems, and ships—are essential for the calibration and validation of satellite SST products, climate monitoring, and operational oceanography. However, the quality of these observations varies significantly across platforms, time, and location, leading to frequent anomalies and outliers. NOAA’s operational quality control (QC) system, iQuam v2.10, provides a baseline method for flagging questionable data, but it suffers from excessive false positives, especially during daytime periods, due to rigid thresholds and outdated external reference comparisons.</p> <p>This thesis presents a refined statistical quality control algorithm that enhances the accuracy and reliability of outlier detection in in situ SST datasets. The proposed method introduces three complementary strategies: (1) a physically motivated consecutive differences test to detect abrupt SST jumps, (2) a signal decomposition and correlation filter based on STL ("seasonal"-Trend decomposition using LOESS) to identify observations inconsistent with expected diurnal behavior, and (3) a revised external reference check that compares smoothed satellite SST fields to platform-specific trend baselines. The approach is modular and platform-aware, enabling tailored thresholds and logic for drifting buoys, tropical moorings, and coastal moored buoys.</p> <p>Signal characterization, including the decomposition of SST into trend and “seasonal” components, was critical to the algorithm design. Evaluation over more than three million observations (January–April 2024) demonstrates a reduction of up to 40% in false positive flags compared to the iQuam external consistency check, particularly in drifters and tropical moorings. The model preserves physically consistent diurnal SST cycles—avoiding iQuam’s daytime over-flagging—and is better suited to detecting persistent anomalies such as slow sensor drift, thanks to trend-level comparisons with external references.</p> <p>This work delivers a more adaptive, statistically robust, and physically interpretable QC system, improving the utility of in situ SST data for both research and operational oceanography. It lays the groundwork for next-generation QC frameworks capable of integrating domain knowledge with data-driven methodologies.</p>

Name thesis:degree_name

Master of Science (M.S.)

Level thesis:degree_level

Thesis

Discipline thesis:degree_discipline

Computer Science

Year dc:date.available

2025

Author dc:creator

• Estaire, Laura

Contributors dc:contributor

• Irina Gladkova

• sea surface temperature
• quality control
• in situ observations
• iQuam
• STL decomposition
• satellite reference
• outlier detection
• data cleaning
• oceanography
• signal processing
• outliers
• Computer Engineering

Repository record dc:identifier

https://academicworks.cuny.edu/cc_etds_theses/1267

OAI identifier oai:identifier

oai:academicworks.cuny.edu:cc_etds_theses-2333