Effect of changes in temperature scales on historical temperature data

Temperature has always been a central quantity in meteorology and plays a key role in weather forecast and climate determination. Long temperature series are a useful tool both as a direct indicator of global climatic trends and as input to climate models. However, ensuring the quality of the data records is challenging, with issues arising from the wide range of sensors used, how the sensors were calibrated, and how the data were recorded and written down. In particular, the very definition of the temperature scales has evolved. While the temperature scales have always been tied to phase transitions of substances such as the freezing and boiling points of water, the temperatures assigned to those phase transitions have been revised occasionally when new sensors and measuring techniques suggested discrepancies between the temperature scales and thermodynamic temperature. Invariably this has led to alterations in the reference scales of temperature that give rise to biases when compiling historic temperature records, an issue that appears to be neglected in the meteorological literature. The present work deals with this issue for 20th century data by proposing a mathematical model to allow the conversion from historical scales to the International Temperature Scale of 1990 (ITS-90). This work also presents the implementation of this mathematical model into a software tool, which can convert large files of historical records to the ITS-90. The correction equation is applied to example observations, and it is found that the correction becomes an issue in special cases involving aggregated temperature data. However, the corrections are significantly smaller than the global warming trend seen in the 20th century. The corrections applied have shown a level of change, which has not significantly affected the global warming trend.