Objective Time-scale-dependent Homogenization of Early Instrumental Temperature Series

An objective method for homogenization of early (pre $1850) instrumental near surface air temperature records is developed. The method is grounded on the same variational principle as used in the objective analysis of meteorological ®elds; i.e. the method consists of a statistically optimal spatial-coherence-based adjustment of nearby station temperature records. The adjustments are made for several different ranges of time scales, where the wavelet transform is used for the decomposition. The method takes into account that early instrumental temperature records are supposed to contain observational disturbances which are more or less smooth functions of time as a result of summations of numerous sources of biases. The method differs from traditional homogenization techniques in that corrections are not only made for a discrete number of abrupt or linear changes, but for continuously changing errors. The power of the method is illustrated with an example of homogenization of three very long temperature records from Sweden.