The choice of estimation of deterministic effect method in climate series is usually the starting point for most studies on climate variability. This decision is more sensitive when the presence of nonlinear phenomena is explored. Therefore, this paper analyzes the robustness and stability of different methods to estimate the annual cycle in time series of daily temperature. Robust estimators are compared and show that the calculation of the annual cycle using the median is the more robust estimate for reporting periods of less than 50 years. However, for longer periods of information no differences between the various estimators were observed. Regarding the persistence of the temperature series, there is a dependency between the time of independence (To) and the period under review (there is a variation of To which can vary between 3 and 7 days). In addition, the non-lineal component of the anomalies tend to represent higher values of To if compared with the linear departures, i.e. the analysis of this component leads to estimate long lasting effects that are affected by the slope of the annual cycle. Asymmetries in the annual temperature cycle in some cases are the result of nonlinear interactions between the annual wave and disturbances. Finally, the processes that show higher dependence in the annual cycle are characterized by greater persistence of warm anomalies. This is associated with slow advective processes where warm air masses remain for several days. This effect is better represented for minimum temperature. Therefore it is proposed that in general, the estimation or design of the anomalies with respect to the annual wave should be considered as a series expansion that considers the independent effects (linear) and dependent (nonlinear) of the annual cycle.
Temperature; Time of independence; Daily time scale
