Homogeneity adjustments of temperature and precipitation series—Finnish and Nordic data

Authors


Abstract

An analysis is made of the adjustments needed to produce three homogeneous data sets, namely the 1961–1990 mean temperatures in Finland, the North Atlantic Climatolological Dataset (NACD) temperature and precipitation series (1890–1990), and the Finnish daily mean maximum and minimum temperature series (1910–1995), as well as the reasons for making such adjustments. The adjustments in the annual (seasonal) mean temperatures are up to ±1°C (±2°C), and annual precipitation adjustments can be ±40%. In Finland, the homogeneity breaks in the normal period temperatures and in the long-term daily mean maximum and minimum temperatures appear to be random, and thus, do not bias averages based on large numbers of stations. However, both the temperature and precipitation series of the NACD would have been statistically significantly biased without adjustments. Station relocations appear to be the most common cause of homogeneity breaks in the temperature series. In the NACD, the adjustments resulting from relocations are statistically significant and reflect changes to colder observing sites. Also, changes in the formula used for the calculation of mean temperatures and urbanization both cause systematic biases in the data. The installation of improved precipitation gauges has been systematic in the NACD; thus, the original series need to be adjusted upwards in the early years. The applied adjustments are of the same order of magnitude as the observed long-term trends, which stresses the importance of the testing and adjusting of long-term series before analysis of climatic changes. In order to monitor climatic changes in a reliable manner, the observing network should be designed to withstand the common discontinuities (e.g. relocations, observer and environment changes etc.) in observation series, because the number of homogeneity breaks appears to be roughly constant in time. Moreover, the introduction of new technology may cause systematic changes in the observations, and comprehensive comparison measurements are needed. Copyright © 2001 Royal Meteorological Society

Ancillary