Thèse de doctorat
Résumé : Visual sunspot observations are at the base of the single longest scientific record ofsolar activity, spanning four centuries (Owens, 2013). This primary reference index, theSunspot Number, was submitted to a full revision and re-calibration (Clette et al., 2016). Anew, improved series was released in July 2015, shedding new light on our understanding ofthe long-term variations and instabilities of the 11-year solar cycle. However, uncertaintiesremain in the revised series, and errors in past historical data must be further determined.A good overview of this ongoing effort can be found in Clette et al., 2016, 2023. Thecurrent revival of long-term solar studies aims to improve our ability to predict the futureevolution of the solar cycle, a primary quest in solar physics, constrain the latest physicalmodels of the solar dynamo, and improve our understanding of the solar influence on theEarth’s climate.The purpose of this thesis is to exploit all available historical sunspot data, including thefull database of raw sunspot counts maintained by the World Data Center SILSO (https://www.sidc.be/SILSO/home), which contains more than 500.000 observations spanningseveral centuries, to derive a better understanding of the scale differences between pastobservers, starting from the beginning of modern data produced by our worldwide SILSOobserving network, going back in time. Indeed, a key issue when building such a longterm record is to bring all observations to the same normalization scale by diagnosingand compensating for various inhomogeneity factors (instrumentation, observing practices,etc.). The level of solar activity can then be compared on a constant scale across multiplecenturies back to 1610 (the invention of the telescope).An in-depth statistical study of the most recent part of the data (35 years, 280 stationssince 1981) was realized through the VAL-U-SUN project that ended in 2021 (Brain. befederal funding; https://www.sidc.be/valusun/). This thesis extends these investigations toearlier historical data over a much longer time interval (Bhattacharya et al., 2023; Bhattacharya et al., 2021) and exploits the results from the VAL-U-SUN project (Mathieu etal., 2019).