Analyzing bias in prominent climatic data sets

The majority of documented climatic data set biases can be divided into two categories: physical biases and calendar biases. Factors affecting the consistency and accuracy of the variable measurement (physical bias) throughout the data record include systematic measurement error and changes in instrumentation, time of observation, measurement site location, observation methods, and surrounding land use/land cover. Factors influencing data aggregated from shorter temporal scales to longer temporal scales (i.e. days to years or days to months) are calendar biases and have been studied to a much lesser extent than physical biases. Previous homogenization efforts for prominent temperature and precipitation data sets are detailed in this review. Numerous physical biases are accounted for in these homogenization efforts and a debate exists in the literature regarding the effectiveness of these bias adjustment methods, leading some investigators to question the suitability of these adjusted data sets for the identification of large-scale climate change signals. Calendar biases are not addressed in previous homogenization efforts. After a brief analysis, a calendar bias known to exist in homogenized data sets (the leap year bias) is identified in the CRUTEM3v, an important global monthly near-surface air temperature data set.