The quality of climate data is of extreme relevance, since these data are used in many different contexts. However, few climate time series are free from non-natural irregularities. These inhomogeneities are related to the process of collecting, digitising, processing, transferring, storing and transmitting climate data series. For instance, they can be caused by changes of measuring instrumentation, observing practices or relocation of weather stations. In order to avoid errors and bias in the results of analysis that use those data, it is particularly important to detect and remove those non-natural irregularities prior to their use. Moreover, due to the increase of storage capacity, the recent gathering of massive amounts of weather data implies also a toilsome effort to guarantee its quality. The process of detection and correction of irregularities is named homogenisation. A comprehensive summary and description of the available homogenisation methods is critical to climatologists and other experts, who are looking for a homogenisation method wholly considered as the best. The effectiveness of homogenisation methods depends on the type, temporal resolution and spatial variability of the climatic variable. Several comparison studies have been published so far. However, due to the absence of time series where irregularities are known, only a few of those comparisons indicate the level of success of the homogenisation methods. This article reviews the characteristics of the most important procedures used in the homogenisation of climatic variables based on a thorough literature research. It also summarises many methods applications in order to illustrate their applicability, which may help climatologists and other experts to identify adequate method(s) for their particular needs. This review study also describes comparison studies, which evaluated the efficiency of homogenisation methods, and provides a summary of conclusions and lessons learned regarding good practices for the use of homogenisation methods.
- Data quality
- Methods comparison