To assess the performance of current state of the art LSMs, we use the signature based metric of the phase lag of heat fluxes to solar radiation following Renner et al. (2019b). This method has important advantages to standard model observation evaluations. First, as a signature approach it does not evaluate the variability over time, but the response to a common reference variable such as solar radiation. Second, the approach effectively decomposes diurnal variability into two metrics whose statistical significance can be tested and which can be directly interpreted. We will employ the PLUMBER dataset, which was established by the Land Surface Model Benchmarking Evaluation Project published by Best et al. (2015). The PLUMBER dataset includes 13 different land surface models at 20 different FLUXNET sites. This dataset thus represents a wide range of observations and includes key land surface models used in climate and weather models.