Created 2020-08-07
Updated 2022-01-02 with adjusted Uppsala series up to 2021
Updated 2022-02-24 with new comparison station Siljansfors (climate data from The reference climate station in the Siljansfors Experimental Forest, run by the Unit for Field-based Forest Research at The Swedish University of Agricultural Sciences, Dalarna County)
Updated 2023-01-13 with adjusted Uppsala series up to 2022
Updated 2023-02-16 with better description of algorithm for creating yearly averages
Updated 2024-01-05 with adjusted Uppsala series up to 2023
Updated 2024-06-01 with major update regarding the new adjustments to the University Uppsala series. No change to the Cederlöf series.
The temperature record from Uppsala in Sweden starts already in 1722. This is one of the world's longest temperature series available. A great job has been made to adjust and homogenize this data to be consistent over time, measurement locations and changed equipment. However, during the 20th and 21th century this temperature series has diverged substantially from surrounding rural stations. This is probably due to urbanisations effects from the growing city of Uppsala, which has more than doubled its inhabitants from the 1950th to now. The location of the measurements has also been moved, not further away from the city, but instead into the city. This gives the incorrect impression that the temperature lately has increased much more than it actually has.
There is however a well maintained reference measurement station, located some 5km north of the city center, at the Uppsala airport (N 59.895 E 17.594, 13 m above sea level). This is a former military airbase which has records of temperature 8 times per day from mid 1952 and every hour from 1962. The Uppsala Airport station was moved during 1995 from a place near the air traffic control tower to the beginning of the runway. But there is no apparent jump in the temperature records at that time, when comparing with surrounding stations.
The long temperature series of Uppsala is maintained by Uppsala university and more information is available on their homepage here. The Uppsala series is described in an article:
The homogenisation method is described in:
The Uppsala series can be downloaded from Swedish Meteorological and Hydrological Institute (SMHI) here.
To make the Uppsala series better reflect rural conditions, I have created a new series, called Cederlöf Uppsala series, which uses the station at Uppsala airport to adjust the Uppsala University series from 1953 and onwards. The resulted data, methodology and validation is shown in the sections below. Hourly data from Uppsala airport has been downloaded from SMHIs site for open data.
Between the publications of the University series containing the years up to 1999 and the next one containing years up to 2000, Uppsala university has made significant changes to the temperature data after year 2000. They have lowered the temperature with about 0,3 to 0,4 degrees. See below a comparison with the 2019 edition and the latest published edition:
The motivation for this homogenization is unclear. The Uppsala series definately needs adjustments but this drastic change in the data after 2000 makes the difference to peak around year 2000 compared to surrounding stations.
My improved Uppsala series, which has been adjusted by use of the Uppsala airport data, can be downloaded in several formats:
- Daily data in the format "year month date temperature". One line per day (updated 2024-01-05):
- Monthly and yearly data in csv format that can be opened with spreadsheet programs (with decimal comma or decimal dot) (updated 2024-01-05):
uppsala_mcederlof_monthly_yearly_v1_0_decimal_comma.csv
uppsala_mcederlof_monthly_yearly_v1_0_decimal_dot.csv
Please note that the yearly average is a simple average of the months of the year. By doing it that way, the impact of possible missing days will be small.
The yearly average of the Cederlöf Uppsala series is plotted below (updated 2024-01-05):
The difference between the original Uppsala series from University and the Cederlöf series is plotted below per month (10 year moving average smoothing).
To create the Cederlöf Uppsala Series, the University Uppsala series has been adjusted with a simple algorithm described below. It should be easy to reproduce this calculation for someone else.
1. Calculate daily means of Uppsala airport using 24 hours per day if available, otherwize 8 hours per day
2. Use monthly means for the years 1953 to 1955 to adjust (homogenize) the daily means from Uppsala airport so that these have the same monthly means as the Uppsala series for these years.
3. Replace the Uppsala daily means with the Uppsala airport adjusted daily means for the year 1953 and onwards.
4. For those days that miss values from Uppsala airport, use instead the Uppsala values but homogenized so that the monthly mean is the same as for Uppsala airport for this particular month. If no data is available from Uppsala for a particular day, this day will be skipped.
For validation of the Cederlöf Uppsala series, I have used 8 different rural stations (7 nearby and 1 further away) except for Uppsala airport itself. The 8 stations has the following coordinates, altitude over sea level and distance to Uppsala ("A" means automatic station):
Film A: N 60.2358 E 17.9043, 33 m a.s.l, 53 km
Films kyrkby: N 60.2357 E 17.9072, 39 m a.s.l, 53 km
Kerstinbo A: N 60.2683 E 16.9738, 55 m a.s.l, 56 km
Risinge: N 60.1748 E 18.2240, 17 m a.s.l, 59 km
Sala: N 59.9045 E 16.6607, 58 m a.s.l, 44 km (moved 1990)
Sala A: N 59.9093 E 16.6843, 57 m a.s.l, 43 km
Svanberga: N 59.8317 E 18.6313, 15 m a.s.l, 67 km
Siljansfors: N 60.8846 E 14.3786 260 m a.s.l, 210 km
All data is downloaded from SMHI's site for open data except for Siljansfors from 1990 and onwards. That part comes from "The reference climate station in the Siljansfors Experimental Forest, run by the Unit for Field-based Forest Research at The Swedish University of Agricultural Sciences, Dalarna County".
The plots below show the yearly difference between the Uppsala University/Cederlöf data and the reference rural station.
The trend lines shown are divided up into two parts, one before 2000 and one from 2000 onwards. The reason for this is the new adjustments applied by Uppsala university affecting data from 2000 onwards.
A positive trend means that Uppsala has become warmer than the rural station. The yearly difference has been calculated as a mean of monthly difference for years with 11 or 12 months of data. The monthly difference has been calculated as a mean of daily difference for months with less than 15 missing days.
Please note that the stations above have a daily mean calculated using the Ekholm-Modén formula (based on daily max/min) except for Siljansfors from 1990 and onwards. That is the reason why the Uppsala airport doesn't have a straight line for the Cederlöf data.
The University Uppsala series seems to have a much higher trend than most of the nearby rural stations before 2000. On the other hand, the Cederlöf Uppsala series seems to have a slightly lower trend compared to the nearby rural stations. However, for the last comparison station, Siljansfors, which is located further away from Uppsala (210km), has an almost flat trend compared to the Cederlöf Uppsala series.
Siljansfors is located in the woods with very little human constructions. The station was moved 1990 to a place 500m further into the woods when SMHI stopped the operation at that site and Sveriges Lantbruksuniversitet took over.
After 2000 the situation is different, the University Uppsala series has a much lower trend than the nearby stations. This means that the changes made by Uppsala university must have been done in an incorrect way. For the complete time period, the Uppsala university version still has a much higher trend than surrounding stations.
The reason why the nearby rural stations shows a slightly higher trend than the Cederlöf data is not clear. One reason for this discrepancy could be newly built houses, roads or grown up vegetation around those rural stations. The history and placement of the Uppsala airport station and Siljansfors are known, but that is not the case for the nearby rural stations. Therefore I have chosen to only rely on Uppsala airport for the adjustment. The correlation with the nearby rural stations has improved significantly in the Cederlöf series, but the remaining discrepancy remains to be explained.