The effect of data size on detecting trend: A case study of Filyos River (Turkey)

Scaling effect on detecting trend from different segments or different extends from the same time series is a well known problem. A highly significant increasing trend may be found in a given segment, while a highly significant decreasing trend may be found in a different segment. The reconstructed series obtaining from tree ring data which is extended from 1657 to 1997 for a length of 341 years is considered as well as observed data from 1964 to 1997 to discuss this effect. Mann-Kendall trend analysis is used to detect trend on both data series. The main objective of this study is to illustrate the effect of scaling on detecting trend. The present study clearly reveals that both observed and reconstructed data nearly throughout exhibit persistence behaviors. Similar decreasing trend patterns are observed at the different time scales although the trend result for observed time series is not statistically significant. As a result, it can be concluded that trend results can be changed under the effect of scaling and it affects making stationarity and climate change interpretations which are very important to time series model. Keywords: Trend analysis, scaling effect, tree – ring, Mann – Kendall