The presence of positive serial correlation in hydrometeorological data tends to distort the results of traditional trend detection tests by inflating the variance of the test statistic resulting in type I error probabilities larger than the value implied by the nominal significance level of the test. This variance inflation increases the probability of false detection causing problems in the interpretation of the results. Despite the many methodologies developed over the last decades to deal with serial correlation in trend detection analysis, such as Pre-Whitening, Trend-Free-Pre-Whitening and its modified version, Variance Correction Pre-Whitening, and other variance-correction techniques, a consensus on which method should be used in which circumstances has not been reached yet. The aim of this study is to evaluate and characterize the performance of several methodologies when applied to series with different probability distributions, record lengths, coefficient of variation, serial correlation, and trend magnitude. A Monte Carlo study was carried out to evaluate the performance of four methods based upon four measures. The first two are the actual probability of type I error and statistical power. The other two not-so-used measures are the probability of obtaining a significant trend with the wrong sign (type S error) and the bias in the estimate of trend magnitude (type M error). The paper reports preliminary results of an on-going analysis. The unbiased versions of both the Modified Trend-Free-Pre-Whitening and the Variance-Correction Pre-Whitening methods obtained a more balanced performance in terms of the tradeoff between the violation of the type I error and statistical power. Besides, both methods provided also competitive results in terms of type S and type M errors regardless of the degree of serial correlation.
