Influence of serial correlation on climatic variables trends analysis in Sicily

Abstract

The intense climatic anomalies occurred in the last decades, led the scientific community to officially admit the existence of a climate change, mainly due to human activities. Researchers have studied trends and variations in temperatures and rainfall in several areas of the world. The analysis of precipitation and temperature trends has become a topic of particular interest because the knowledge of rainfall and temperatures evolution represents a support tool in forecasting future climatic scenarios and, consequently, for water resources management. In several studies aimed at identifying trends in climatic data, the nonparametric test of Mann-Kendall is frequently used with the assumption that the observed data are characterized by serial independence. However, many hydrological series, such as minimum and average annual runoff, often show a statistically significant serial correlation. This study is based on previous analysis of time series of precipitation and temperature recorded in Sicily in the past century, which have confirmed the existence of trend in precipitation and temperature time series, using the Mann-Kendall test. The aim of this work is to deepen the analysis of trends in rainfall and air temperature using techniques able to remove the serial correlation in order to obtain results not influenced by it. Whit this purpose, two different procedures have been used, namely the pre-whitening and the trend-free pre-whitening. With reference to the monthly precipitation data, which are self-correlated, the prewhitening process identifies a number of decreasing trends less than those identified in the previous studies while the trend-free pre-whitening procedure doubles the number. When data are aggregated at seasonal and annual scale, the trend-free pre-whitening procedure individuates the same number of decreasing trends observed in the previous studies, while the alternative pre-whitening procedure, slightly reduces this number. The temperature series are highly auto-correlated and for this reason, the pre-whitening procedure completely eliminates the trend presence. The trend-free pre-whitening procedure individuates more stations with positive trends than the previous studies ones. If the temperature data are aggregated to seasonal and annual time scale, the trend-free prewhitening procedure results are consistent with previous studies. In fact positive trends are observed at all levels of confidence and time aggregation.