Relative homogenization: Special problems

Problems related to signal-to-noise ratio, synchronous breaks, annual cycle of biases, temporal resolution of bias correction terms, and several other issues are thematically discussed. In time series of monthly or daily resolution, considerably high autocorrelations occur. The autocorrelation structure is different for relative time series in comparison with that of the raw data, and either of them is similar to any red noise. The correct detection and correction of the annual cycle of inhomogeneities is a challenge for the lower signal-to-noise ratio for sub-annual values. An additional problem of daily homogenization is that biases of daily values are often weather dependent. Sometimes, synchronous or semi-synchronous breaks occur for a network, and this problem can be treated better with pairwise comparisons than with other statistical tools. However, sometimes statistical tools are generally ineffective for removing concerted inhomogeneities; therefore, metadata and parallel measurement data are particularly valuable in such homogenization tasks.