Understanding the effects of changes in the temperature scale standards through time

Since records of surface temperature started being made there have been iterations of the fixed points standards used by national metrological institutes (that is not a typo). Assuming that all meteorological measurements through time have been made to such standards (which may be a considerable stretch) this would have imparted changes to the records that are not physical in origin. As part of meteomet efforts have been made to understand this. It is a relatively small effect compared to effects of other long recognized data issues. Nevertheless it is important to properly and systematically consider all sources of potential biases as exhaustively as possible.

The work itself was led by Peter Pavlasek of the Slovak Institute of Metrology. His introduction is reproduced below:

Temperature is one of the main quantities measured in meteorology and plays a key role in weather forecasts and climate determination. The instrumental temperature recordings now spans well over a century, with some records extending back to the 17th century, and represents an invaluable tool in evaluating historic climatic trends. 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 was recorded and written down.  In particular, the very definition of the temperature scales have evolved. While they have always been based on calibration of instruments via a series of material phase transitions (fixed points), the evolution of sensors, measuring techniques and revisions of the fixed points used has introduced differences that may lead to difficulties when studying historic temperature records. The conversion program here presented deals with this issue for 20th century data by  implementing a proposed mathematical model to allow the conversion from historical scales to the currently adopted International Temperature Scale of 1990 (ITS-90). This program can convert large files of historical records to the current international temperature scale, a feature which is intended to help in the harmonisation processes of long historic series. This work is part of the project “MeteoMet” funded by the EURAMET, the European association of National Institutes of Metrology, and is part of a major general effort in identifying the several sources of uncertainty in climate and meteorological records.

Michael de Podesta, who has served on the steering committee since ISTI's inception, reviewed the software for ISTI and had the following summary:

Assuming that calibration procedures immediately spread throughout the world – homogenisation algorithms might conceivably see adjustments in 1968, with smaller adjustments in 1990.
 
If undetected, the effect would be to create a bias in the temperature record. This is difficult to calculate since the bias is temperature dependent, but if the mean land-surface temperature is ~10°C and if temperature excursions are typically ±10 °C then one might expect that the effect to be that records prior to 1968 were systematically overestimated by about 0.005 °C, and records between 1968 and 1990 by about 0.003 °C.

Michael's full summary which includes some graphical and tabular summaries can be found here.

The code package is a windows operating system based package. It is available here.