Graphical-statistical method to explore variability of hydrological time series

• Main Author: Charles Onyutha
• Format: Article
• Language: English
• Published: IWA Publishing 2021-02-01
• Series: Hydrology Research
• Subjects: climate variability; hydrological change attribution; mann–kendall test; river nile; spearman's rho test; sub-trend analysis
• Online Access: http://hr.iwaponline.com/content/52/1/266

Due to increasing concern on developing measures for predictive adaptation to climate change impacts on hydrology, several studies have tended to be conducted on trends in climatic data. Conventionally, trend analysis comprises testing the null hypothesis H0 (no trend) by applying the Mann–Kendall or Spearman's rho test to the entire time series. This leads to lack of information about hidden short-durational increasing or decreasing trends (hereinafter called sub-trends) in the data. Furthermore, common trend tests are purely statistical in nature and their results can be meaningless sometimes, especially when not supported by graphical exploration of changes in the data. This paper presents a graphical-statistical methodology to identify and separately analyze sub-trends for supporting attribution of hydrological changes. The method is based on cumulative sum of differences between exceedance and non-exceedance counts of data points. Through the method, it is possible to appreciate that climate variability comprises large-scale random fluctuations in terms of rising and falling hydro-climatic sub-trends which can be associated with certain attributes. Illustration on how to apply the introduced methodology was made using data over the White Nile region in Africa. Links for downloading a tool called CSD-VAT to implement the presented methodology were provided. HIGHLIGHTS Common trend tests are purely statistical. They can yield meaningless results in some cases.; Conventional testing of the null hypothesis (no trend) using entire data leads to lack of information about sub-trends.; Analysis of sub-trends maximizes understanding on how changes in the data can be linked to certain driving factors.; This paper presents a methodology for graphical-statistical analyses of sub-trends.;