Inter‐comparison of methods to homogenize daily relative humidity

Abstract

Three homogenization methods (ACMANT, MASH and HOMOP) have been evaluated for their efficiency in homogenizing daily relative humidity data. A homogeneous surrogate data set based on Austrian stations was created and perturbed to simulate inhomogeneous, realistic time series ("validation data sets"). Two validation data sets ("simple" and "complex") were created. In both data sets the magnitude of the breaks depends on the time of year and the measured values. They differ in the number of missing values and especially on whether the break signal was perturbed by white noise. In the latter case, the noise also changed to take into account changes in random measurements errors and other physical factors. The evaluation showed high agreement in statistical characteristics between the real data and the surrogate data set. The homogenization methods were compared in their ability both to detect breaks and to reproduce the homogeneous surrogate data set. For the evaluation of the final data set the distribution, trends and root-mean-square error (RMSE) were analysed. The percentage of improved time series depends on the evaluation parameter considered. Less stations were improved when using the "complex" validation data set. Because of the large number of breaks and the small signal-to-noise ratio, an improvement of the data by homogenization was non-ideal for all methods used, with each having its advantages and disadvantages. The quality of the ACMANT and HOMOP methods is comparable, with ACMANT solving less stations but declaring less stations falsely as homogeneous. To get an impression of the influence on real data, ACMANT was applied to homogenize daily Austrian time series of relative humidity. While the quality of data from some stations can be improved through the homogenization, this is not the case for all time series. A final evaluation of homogenized time series should be performed to ensure their quality before further use.

A new homogenized daily data set for temperature variability assessment in Italy

“…Both of them ranked best in the Multi‐test project, a comparative study of the performance of different homogenization methods (Multi‐test project; Domonkos and Coll, ) and their effective use is documented in several studies concerning different climate variables: humidity (Chimani et al ., ), temperature (Mamara et al ., ; Hunziker et al ., ), precipitation (Luna et al ., ; Mamara et al ., ; Hunziker et al ., ), wind (Azorin‐Molina et al ., ; ).…”

A new method to study inhomogeneities in climate records: Brownian Motion or Random Deviations?

“…Later studies used both processes. A study on daily temperature have generated data that is mostly Brownian Motion (Killick, 2016), while newer studies on monthly temperature and daily humidity inserted breaks as Random Deviations (Guijarro et al, 2017;Chimani et al 2018).…”

Towards a global land surface climate fiducial reference measurements network

“…The effects of inhomogeneities are also manifested more strongly in extremes than in the mean (e.g., Trewin, ) and are thus important for studies of changes in climatic extremes. State‐of‐the art homogenization methods can only make modest improvements in the variability around the mean of daily temperature (Killick, ) and humidity data (Chimani et al , ).…”