The effect of homogenization when constructing long-term gridded monthly precipitation and temperature data

When assessing long-term climate trends and variability, it is important to analyse data that are not disturbed by external factors that might lead to misleading trends. It is also urgent to analyse consistent series that cover the entire time period. Since observation networks are continuously changing not too many complete series are available for a centennial long analysis. In this paper an approach to construct long-term homogenized climate series to be used to derive a long-term gridded dataset is presented. A gridded dataset of monthly climate anomalies based on all available observations was used to construct 165 raw temperature series and 323 raw precipitation series covering the entire period 1901-2020. These series are homogenized by using the automatic procedure in the R-package climatol. The gridded data based on the homogenized data show the same regional trends and variability as the data derived from non-homogenized data. The local spatial variance is smaller in the gridded data based on homogenized series than those based on the raw data series. Both the construction and homogenization of data series to provide spatial and temporal consistent data for gridding lead to better and more reliable gridded datasets for analysing climate trends and variability.

Automated summary

When analyzing long-term climate trends and variability, it is important to use undisturbed data and consistent data series. This paper presents an approach to construct long-term homogenized climate series, which leads to better and more reliable gridded datasets for analyzing climate trends and variability.