Abstract
Temperatures in the Arctic are changing faster than in other areas of the world due to Arctic amplification. The Greenland weather station network managed by DMI is the longest temperature record available from Greenland. This paper investigates the calculation of daily average near surface air temperatures, which are used to calculate monthly and annual averages, climate normals and in global temperature products and data banks. The occurrence of days with uneven sampling times is currently not accounted for when calculating daily averages. There are long periods of time (up to 25 years) for 13 stations in the network where the sunlit hours of the day are oversampled, leading to warm biases in daily average temperatures of magnitudes up to 1.3°C on average for a single station depending on the hours of observation. In addition, all stations have varying amounts of days with missing values leading to an uneven sampling of the diurnal cycle. A correction method is presented and applied that reduces the highest average warm bias to 0.5°C. A simple trend analysis is performed by comparing the difference between period averages before and after the correction. The trend analysis indicates that trends calculated for stations with long periods where the sunlit hours are undersampled are underestimated, as the earlier parts of the record with manual weather stations (the first automatic weather stations were introduced in this station network in the 1980s) in these cases are colder than previously estimated. Corrections on Qasigiannguit have the largest effect on trends at a monthly level with a difference of 1.21°C. The corrected and uncorrected temperatures are compared with ERA5. The corrections decrease the difference between the weather station temperatures and ERA5, but not always the difference in trends. The correction method can be applied to all weather station networks and has no requirements for network density as each station is corrected separately.