Improving seasonal matching in the STARS model by adaptation of the resampling technique

The STatistical Analogue Resampling Scheme (STARS) has already been used for several climate projection studies in different parts of the world and climate projections obtained by STARS play an important role in several impact studies. Thus, it is crucial to provide results that are reasonable in terms of physical consistency. This also includes the annual cycle of the different variables. In this paper, we address the seasonal inconsistency that appears in the results if a demanding setting is used. This is shown on the example of applying STARS to the entire European continent, where the size and the climatological variety of the region determine the demanding setting. In this setting, the current model is not able to provide results with a realistic annual cycle, as it replaces days in autumn with days from spring. This problem is solved by an adaptation of the resampling method, resulting in a new model version of STARS which yields physically reasonable annual cycles. This model adaptation and its effects on the model results are presented in this work. While the old model results show a large overestimation of shortwave radiation in autumn, the annual cycles provided by the new model version are in agreement with observations.