Data Analysis in Nonstationary State

Abstract

Climate change impact must be taken into account in any water resources planning management studies, because it does not allow future occurrences to be repeated as a replicate of the past. The stationarity is no longer valid, because the climate change plays significant role on ascending or descending trend components in any hydro-meteorological events such as temperature, precipitation, evaporation, runoff and discharge. The identification of trends can be represented by well-known methodologies, such as the Mann-Kendall and sequential Mann-Kendall. These methodologies require restrictive assumptions such as data length, serial independence and Gaussian (normal) probability distribution function (PDF). On the other hand, Innovative trend analysis (ITA) method proposed by Şen is helpful to identify even visually with direct interpretations without restrictive assumptions. The PDF or cumulative distribution function (CDF) is effective tool for risk level determination but it cannot tell anything about the trend in a given hydro-meteorological data. The PDF (CDF) does not yield any clue about the trend possibility, and hence, it’s alone use in any water resources structure design may lead to erroneous planning studies. In this study, nonstationary nature of a given monthly hydro-meteorological data is examined by trend determination procedures. For the application, monthly averages of maximum daily temperatures are used on Oxford station, UK. It is observed that the temperature values of each month have a positive trend and the nonstationary empirical cumulative frequency curves on first half group match better all data group than the stationary state. © 2018, Springer Science+Business Media B.V., part of Springer Nature.