Homogenization of monthly and daily temperature and precipitation data in Sweden Erik Engström and Thomas Carlund Swedish Meteorological and Hydrological Institute EMS 6 September 2017
Digitalization of historical archives Procedure continuously improved to increase digitalization speed. Current focus is on precipitation and snow depth data from 1945 to 1960. A new database for observational data facilitate access to data and quality control. N stations Almost all station data digitalized from 1960 and forward. Temperature Precipitation N stations Monthly Daily 15 min 2 Earlier estimation
Digitalization of historical archives Station meta data from inspection protocols is scanned and digitalized to enable processing by scripts. 3
Historical precipitaiton gauge Parallel tests of historical precipitation gauge at two stations in Sweden. Primarily for two whole years. 4
Introduction of wind screen Parallel tests of historical precipitation gauge at two stations in Sweden. Primarily for two whole years. 5
Homogenization in Sweden Monthly mean temperature: 81 temperature series (red circles) with quality controlled values from 1860 to 2016. Monthly precipitation sum: 94 precipitation series (blue points) with quality controlled values from 1860 to 2013. COST-HOM software HOMER run in both automatic and interactive mode Meta data available from station inspections. 6
Homogenization in Sweden Monthly homogenization of north region of Sweden. QC (red), interactive/erik (green), interactive/lisa (blue), automatic (black) 7
Homogenization in Sweden Monthly homogenization of north region of Sweden. Automatic mode removed the warming trend! QC (red), interactive/erik (green), interactive/lisa (blue), automatic (black) 8
Daily homogenization in Sweden Daily homogenization of 1952-2016 for Tx and Tn 8 stations in network 2 stations without breaks HOM and SPLIDHOM homogenized 5 stations for Tx (2 for Tn) Climatol homogenized all 6 stations with breaks. 9
Daily homogenization in Sweden Daily homogenization of 1952-2016 for Tx and Tn 8 stations in network 2 stations without breaks HOM and SPLIDHOM homogenized 5 stations for Tx (2 for Tn) Climatol homogenized all 6 stations with breaks. 10
Daily homogenization in Sweden Days (%) Tn10p, cold nights, percentage of days when TN < 10th percentile of 1961-1990. Comparison of indices based on QC-data (red), Splidhom (blue), Hom (black) and Climatol (green). Vertical line = Break. 11
Daily homogenization in Sweden Days (%) Tn10p, cold nights, percentage of days when TN < 10th percentile of 1961-1990. Percentage of cold nights is increasing in QC-data. Comparison of indices based on QC-data (red), Splidhom (blue), Hom (black) and Climatol (green). Vertical line = Break. 12
Daily homogenization in Sweden Days (%) Tn10p, cold nights, percentage of days when TN < 10th percentile of 1961-1990. Percentage of cold nights is increasing in QC-data. Percentage of cold nights is decreasing in homogenized data. Comparison of indices based on QC-data (red), Splidhom (blue), Hom (black) and Climatol (green). Vertical line = Break. 13
Daily homogenization in Sweden Days (%) Tn90p, warm nights, percentage of days when TN > 90th percentile of 1961-1990. Comparison of indices based on QC-data (red), Splidhom (blue), Hom (black) and Climatol (green). Vertical line = Break. 14
Daily homogenization in Sweden Days (%) Tn90p, warm nights, percentage of days when TN > 90th percentile of 1961-1990. Percentage of warm nights is increasing in QC-data. Comparison of indices based on QC-data (red), Splidhom (blue), Hom (black) and Climatol (green). Vertical line = Break. 15
Daily homogenization in Sweden Days (%) Tn90p, warm nights, percentage of days when TN > 90th percentile of 1961-1990. Percentage of cold nights is increasing in QC-data. Percentage of warm nights is increasing in homogenized data. Comparison of indices based on QC-data (red), Splidhom (blue), Hom (black) and Climatol (green). Vertical line = Break. 16
NordHom Daily homogenization methods evaluation Preliminary conclusions: Climatol and Splidhom/Hom are relatively fast and easy to use when you have learn how to run them. But it is quite a threshold to get started. All tested methods are sensitive to network/data density. Splidhom/Hom more strict compared to Climatol (due to settnings?). All tested methods usually shift the data in the same direction and magnitude, but not always. Larger station networks are needed to make more robust conclusions. The work will continue with the rest of the available stations in Sweden. 17
Thank you for your attention! Erik Engström, SMHI erik.engstrom@smhi.se 18