Comparison of Trend Detection Approaches in Time Series and Their Application to Identify Temperature Changes in the Valencia Region (Eastern Spain)

Abstract

The identification of systematic small- and intermediate-scale temperature changes (trends) in a time series is of significant importance in the analysis of climate data. This is particularly so in the analysis of local climate change trends and their potential impact on local hydrological cycles. Although many statistical tests have been proposed for detecting these trends their effectiveness is often affected by the presence of serial correlation in the time series. Hence, it is of both interest and necessity to compare the performances of these tests by applying them under a representative range of conditions. In this study, we use Monte Carlo experiments to compare and explore six commonly used tests for detecting trend. For this purpose, we use the confidence level and power to assess the ability to detect trend in two groups of simulated time series with and without serial correlation. The statistical tests are also applied to mean annual temperature measured at 13 weather stations located in the Valencia region (Eastern Spain).