Using measurements uncertainties in climate applications

Measurement uncertainties are a dispersion indicator which must be quantified when the estimation of a geophysical quantity is provided. Measurement uncertainty is defined as a "parameter, associated with the result of a measurement, that characterizes the dispersion of the values that could reasonably be attributed to the measurand’' (GUM, 2018). Therefore, measurement uncertainties can discriminate more and less certain data with confidence. Observations form the basis for any evidence of climate change. However, observations themselves possess uncertainties originating from many sources including measurement error and errors imposed by the algorithms generating derived products (Matthews et al., 2013). Nevertheless, traditional approach to climate data records, either obtained from observations or from data assimilation systems, offers datasets where uncertainty information is generic, misleading or missing. 

In particular, measurement uncertainties have been often neglected adducing their self-compensation when these are propagated from raw data to geophysical products or derived products. This is also because the metadata available and the collected observations do not allow their appropriate estimation. Moreover, other sources of uncertainty (e.g. due to interpolation, representativeness, residual of homogenization algorithms, etc.) must be quantified to provide a  proper uncertainty estimation in the derived products.

 

More recently an increasing number of datasets are provided with measurement uncertainties; few satellite retrievals are generated with a quite detailed uncertainty quantification; atmospheric renalysis is provided with an uncertainty estimation, although systematic model errors not taken into account and uncertainties are assumed uncorrelated; finally, the most recent homogenized datasets are provided with an estimation of uncertainties also for the historical data.

 

The uncertainties in climate observations pose a set of methodological and practical challenges for both the analysis of long-term trends and the comparison among datasets or with theoretical thresholds. 

 

In this work examples will be provided showing the importance of quantifying uncertainties of climate data records. The aim is also to encourage the community to develop other use cases for showing the impact of using uncertainties in climate applications.