Abstract Non-natural irregularities are an inevitable part of long-time climate records. They are originated during the process of measuring and collecting data from weather stations. In order to use those records as an input for environmental projects or climate studies, it is essential to detect and correct the irregularities through the process of homogenisation. The use of geostatistical approaches as homogenisation techniques has already been proven to be successful. The gsimcli homogenisation process is based on a geostatistical simulation method, the direct sequential simulation. This method generates a set of equally probable and independent realisations, and calculates a probability distribution function at the candidate station's location. This probability distribution function is then used in the identification and correction of irregularities. Currently, gsimcli is being developed into an open source software package. During the homogenisation process, gsimcli requires the selection of several parameters in the detection and correction of irregularities. The candidate stations’ order to be homogenised, the value of the probability used in the detection of irregularities, and the statistic value to be used in the correction of the irregularity or in the replacement of missing data, are examples of parameters to be chosen for the homogenisation with gsimcli. This work presents a sensitivity analysis of those parameters, in order to find the most suitable set of values for the homogenisation of monthly precipitation data. A benchmark data set, comprising climate records from an Austrian precipitation network, will be used in this analysis. Performance metrics are calculated to evaluate the efficiency of the homogenisation process. The set of parameters providing the best values of performance metrics will be defined as the default set of homogenisation parameters for precipitation data.
- sensitivity analysis benchmark precipitation homogenization