Sensitivity of trend estimation to subsampling and estimation algorithms in radiosounding historical time series

In the frame of the Copernicus Climate Change Service (C3S), a study has been carried out with the aim to discriminate among competing estimation methods of decadal trends and to quantify the effect of spatial and temporal subsampling. Trend estimation methods used fall into two main categories: parametric and non-parametric. Given also their quite common use in the climate community, decadal trends as well as performances have been evaluated for the following four fitting methods: simple linear fitting, calculating the linear trend(slope) based in statistical significance; LADFIT robust linear fitting, a robust least absolute deviation method; LANZANTE robust linear fitting, a resistant and non-parametric regression based on the median of pairwise slopes and LMROB robust linear fitting, based on a fast MM-type estimator linear regression models.