Warming air temperatures induce ground warming and can therefore lead to permafrost degradation, which in turn may produce rock fall, debris flows, or cause serious problems for infrastructure stability. In the last two decades, many boreholes for thermal permafrost monitoring have been drilled in Alpine regions. Factors such as snow or scree cover, hydrology, or topography strongly affect the ground thermal regime and make trend estimation very difficult. Furthermore, ground temperature series exhibit a pronounced annual cycle and are strongly temporally correlated. For the sake of simplicity, often only annual values are analyzed for trend estimation, which is usually not robust for short measurement periods. This study intends to model time series with annual cycles and a daily database. A family of nine different models for trend estimation is proposed. The models are able to fit both overall trends and trends in the annual cycle amplitudes and therefore allow more robust and accurate trend estimation than when annual values only are used. Evaluation and comparison of the models is demonstrated using short‐range (less than two decades, 1996–2008) permafrost temperature series from two adjacent boreholes in the Muot da Barba Peider ridge (Eastern Swiss Alps). The results suggest an increase in cycle amplitudes for near‐surface coarse‐blocky ground layers and an overall warming trend for deeper bedrock layers where the annual cycles are less pronounced.