资源环境科技发展态势分析平台

For some applications, long-term coverage and climate consistency are important requirements in gridded climate data sets. This is often challenged by the inhomogeneity of measurements and by the variation of station networks over time. We describe the development of new monthly grid data sets of temperature and precipitation for Switzerland that range back as far as 1864 and satisfy high standards in long-term consistency. To this end, data are incorporated from stations only with complete and homogenized series available. To compensate for the limited spatial coverage, our analysis integrates statistical information from high-resolution grid data sets based on a dense network but over a limited period. For Switzerland, we demonstrate that the statistical reconstruction recovers mesoscale features not resolved by the stations alone. The method is fairly robust against representativity biases during calibration. The resulting trend maps are more plausible than those from common interpolation with more extensive but variable station networks. Typical errors in the analyses are 0.2 degrees C for monthly temperature and 10% for precipitation, but they can be larger in regions with low station density. In 75% of all months, the reconstructions explain more than 90% of the spatial variance for temperature and more than 60% for precipitation. From the new data sets, we infer a warming trend of around 1.3 degrees C per century in the period 1864-2017, increasing to more than 0.35 degrees C per decade for 1961-2017. Precipitation trends are not statistically significant, except in winter, when an increase of 1-3% per decade is found in the north and west of the country.