Climatological studies require sufficiently long homogeneous time series. Long observational records are often subject to non-climatological changes, for example changes in location, measurement equipment or technique, or changes in the surrounding environment. In order to serve as basis for climatological studies, the homogeneity of observational data therefore must be tested, and if required, homogenised.
At the Swedish Meteorological and Hydrological Institute (SMHI) 35 key homogenised monthly average 2 m temperature data series has served as the basis for an indicator of climate change. The time series were homogenised manually with the Standard Normal Homogeneity Test. SMHI recently adopted the new climatological standard normal period 1991–2020, which called for an updated homogenised temperature data set.
In order to use a larger part of the observational data set and to enable systematic updating the monthly average temperature data set is homogenised with a new automated version of the homogenisation tool HOMER. Data from 836 individual time series (1860–2021) are merged into 456 time series with a novel automatic merging method. The expansion of the homogenised data set from 35 to 456 time series enables studies of regional climate. Merging limits the need of interpolation of data and increase the number of long time series without a net loss of data.
22 of the merged time series were found to be homogeneous. For the other time series, the median time per homogeneity break is 17 years which correspond well to the typical homogeneity break frequency of European temperature data sets. 40 % of the detected homogeneity breaks are supported in meta data. 27 % of the data points are corrected by ±0.5 °C or less, 2 % by ±1 °C or more. The average correction is negative, larger in the early periods, and larger in the summertime.
The average trend 1860–2021 in the resulting merged and homogenised data set is (0.13 ± 3) °C / 10 a, which does not significantly differ from that of the raw observational data or the previous homogenised data set. Extremely warm months defined as being outside of three times the standard deviation from the average of the full time series are most frequent and extremely cold months least frequent in the most recent 30-year period (1991–2020). In the homogenised data set, extremely warm months are even more frequent and extremely cold month even less frequent in 1991–2020, than in the raw observational data set.
In the presentation, the automation of HOMER and the novel automatic merging method is described. Results from the homogenised data set is presented in more detail and compared with the previous homogenised data set.
How to cite: Joelsson, M., Engström, E., and Kjellström, E.: Homogenisation of Swedish mean monthly temperature series 1860–2021, EMS Annual Meeting 2022, Bonn, Germany, 5–9 Sep 2022, EMS2022-111, https://doi.org/10.5194/ems2022-111, 2022.
