Scrutinizing variability in full and partial rainfall time series by different approaches

Abstract

The main goal of this study was established on the effect of splitting full data in disclosing time-related internal variability in a given time series, as well as classical trend analysis procedure. For this purpose, Mann–Kendall (MK), Spearman's Rho (SR), linear regression (LR), and innovative trend analysis (ITA) approaches were applied to the full part and the split data of the seasonal and annual rainfall data provided from the Kizilirmak Basin in Turkey. In the study, it was determined that the ITA method was more effective than other methods in terms of finding out the statistically significant trend in a significant portion of the rainfall data (approximately 87%), whereas this rate varied between 7 and 13% in other methods. Similar results have been experienced when the full data were divided into three groups. Among the trend analysis approaches applied to the split data groups formed for the analysis of the internal variability that occurs during the observation periods in the rainfall time series, the ITA method was quite effective in detecting the mentioned change in the time series. The ITA procedure revealed that there was a statistically significant decrease in rainfall amounts in the seasons with the highest rainfall of the region. This result could not be presented sufficiently when the data were divided into two groups. The remaining three methods failed to produce a remarkable result in determining the internal variability experienced in the data. In this sense, the consideration of full data in the analysis produces information about the average trend, it is insufficient to reveal the hidden information in the data.