Fisheye cloud tracking dataset and processing scripts — Réunion Island, 30-day campaign (8 October – 10 November 2024)

This dataset contains the processed measurements and analysis scripts underlyingAndrinirintsoa et al., ": A Case Study in Réunion Island" (submittedto Atmospheric Measurement Techniques). Data were acquired using a VIVOTEK FE9382-EHV-v2 fisheye camera (2048x2048 px,equidistant projection, 180° field of view, ~1 image/min) at a sewage/sky-observationsite on Réunion Island (21°S, 55.5°E), over 30 days of continuous daytime acquisition(8 October – 10 November 2024). Contents:- resultats_nuages.csv: per-image detail file (cloud fraction, optical-flow speed and direction, zenith/azimuth), one row per processed image across the 30-day campaign.- classification_nuages.csv: per-track classification file (194,546 raw tracks), giving each tracked cloud's lifetime, surface growth rate, assigned dynamic scenario (one of nine physically defined categories), and reliability score.- cloud_popup_detector_stage1_original_EN.py: the original raw-image processing pipeline (fisheye calibration, RBR segmentation, temporal tracking, Farnebäck optical flow, physical classification) that produced the two CSV files above.- replot_and_fix_english.py: a post-processing script that reads the two CSV files and applies three corrections identified during peer review -- a true-North azimuth correction, an in-cone/out-of-cone confidence-mask flag, and an upstream noise pre-filter -- then regenerates all validation figures reported in the article. License: CC-BY 4.0. Please cite the associated article if you use this dataset.