Climate Digitizer: A Citizen Science Platform for Crowdsourced Digitization of 115 Years of Analog Meteorological Records from Kandilli Observatory

Historical meteorological records are essential for understanding long-term climate dynamics, yet a vast portion of pre-digital observational data remains locked in analog formats. Kandilli Observatory in Istanbul holds an archive of over 4,000 thermograms, hygrographs, barographs, and actinographs spanning 1911 to 2005. Standard OCR and computer vision techniques fail on these records due to faded ink, torn paper, overlapping curves, and grid lines intertwined with data traces. This paper presents Climate Digitizer, a web-based citizen science platform that enables volunteers to digitize these records by tracing curves directly on scanned images. The platform assigns each record to multiple independent contributors and applies a cross-validation pipeline including median-based consensus estimation, IQR outlier detection based on Tukey’s fence method, and leaveone-out scoring. Two calibration methods convert pixel coordinates to physical values: grid linear interpolation and reference point regression with robustness filters. A gamification system sustains volunteer engagement. Across nine completed records spanning all four data types, 13 independent contributors produced 78 validated traces with a mean cross-validation score of 9.1/10 and mean deviation of 10.8 px, with the IQR pipeline successfully detecting two outlier traces. The system is built entirely on client-side technologies with Google Firestore, requiring no server-side code. This tutorial provides a complete description ofthe algorithms, data formats, and deployment procedures for replication.