A method of correction of climatological temperature records for the effect of relocation of meteorological stations

We introduce a method of removing the influence from relocation of meteorological stations on monthly mean temperature data. The long term changes in temperature recorded at meteorological stations are important in understanding global warming and regional climate change. However, for the stations which experienced relocation in the past, time-series data are not homogeneous over a long period of time. In order to create homogenized datasets, we devised a method of correction to remove the impact of the relocation on the monthly mean temperature data. In the method of correction, we calculate a correction value for a relocated station by applying the principal component analysis to monthly time-series data of surrounding stations which have not been relocated. Period of the data used to calculate the correction value is 16 years preceding and succeeding the relocation. We made a correction for the 64 relocations operated in Japan that have been relocated up to 2004. As a result, a measure of inhomogeneity of the data has been improved from about 20% to about 6%.