Comparison of homogenization methods for daily temperature series against an observation-based benchmark dataset

Abstract

Homogenization of daily temperature series is a fundamental step for climatological analyses. In the last decades, several methods have been developed, presenting different statistical and procedural approaches. In this study, four homogenization methods (together with two variants) have been tested and compared. This has been performed constructing a benchmark dataset, where segments of homogeneous series are replaced with simultaneous measurements from neighboring homogeneous series. This generates inhomogeneous series (the test set) whose homogeneous version (the benchmark set) is known. Two benchmark datasets are created. The first one is based on series from the Czech Republic and has a high quality, high station density, and a large number of reference series. The second one uses stations from all Europe and presents more challenges, such as missing segments, low station density, and scarcity of reference series. The comparison has been performed with pre-defined metrics which check the statistical distance between the homogenized versions and the benchmark. Almost all homogenization methods perform well on the near-ideal benchmark (maximum relative root mean square error (rRMSE): 1.01), while on the European dataset, the homogenization methods diverge and the rRMSE increases up to 1.87. Analyses of the percentages of non-adjusted inhomogeneous data (up to 39%) and substantial differences in the trends among the homogenized versions helped identifying diverging procedural characteristics of the methods. These results add new elements to the debate about homogenization methods for daily values and motivate the use of realistic and challenging datasets in evaluating their robustness and flexibility.

Observed changes in precipitation during recent warming: The Czech Republic, 1961–2019

“…In cases where the detected break‐points were not confirmed by metadata, the limit rose to 10% (the limit values applied resulted from thorough verification simulations for central Europe). The distribute adjustment by percentile (DAP) method (Déqué, 2007) was further employed for the adjustment of daily data based on comparison of percentiles before and after break‐points (Squintu et al ., 2020). …”

Reflections of global warming in trends of temperature characteristics in the Czech Republic, 1961–2019

“…Based on rich existing experience with checking and homogenization of Czech and other climatological series (see e.g., Štěpánek et al ., 2011a; 2011b; 2013; Zahradníček et al ., 2014; Squintu et al ., 2020) and using ProClimDB and AnClim software (http://www.climahom.eu), temperature series from 133 stations were approached in several steps. Once data quality had been checked, three basic approaches were employed: analysis of series of differences between candidate and neighbouring stations (i.e., pair‐wise comparisons); application of limits derived from interquartile ranges (either to individual series, i.e., an absolute test or, preferably, to series of differences between the candidate and the reference series, i.e., a relative test); comparison of the series of tested values with ‘anticipated’ (theoretical) values created by statistical methods for spatial data (e.g., by inverse distance weighting kriging). …”

“…Based on rich existing experience with checking and homogenization of Czech and other climatological series (see e.g., Štěpánek et al, 2011a;2011b;Zahradníček et al, 2014;Squintu et al, 2020) and using ProClimDB and AnClim software (www.climahom.eu), temperature series from 133 stations were approached in several steps. Once data quality had been checked, three basic approaches were employed:…”

Uncertainty evaluation of Climatol's adjustment algorithm applied to daily air temperature time series

“…Evaluation of the efficiency of the detection algorithms has been performed in many works (e.g., Ducré‐Robitaille et al ., 2003; DeGaetano, 2006; Reeves et al ., 2007; Domonkos, 2011; Kuglitsch et al ., 2012; Venema et al ., 2012; Willett et al ., 2014; Killick, 2016; Yozgatligil and Yazici, 2016; Coll et al ., 2020). Assessment of the performance of the adjustment methods has also been considered (e.g., Della‐Marta and Wanner, 2006; Mestre et al ., 2011; Trewin, 2013; Squintu et al, 2020). In both cases, the evaluation was mainly performed in a relative form, that is, several homogenization algorithms are usually compared in order to define which one gives the best output and is most suitable for practical applications.…”