Estimating the current climate mean state at regional to local scales

Determining the current climate mean state (CCMS) on the regional and local scale is an important task of climate monitoring. The CCMS and the long-term climate change signal derived from it are relevant for a wide range of users. For climate variables with strong, possibly non-linear trends, accelerating climate change more and more disqualifies the use of traditional normals and long-term linear trends. Although several alternatives are in use, there are few comprehensive assessments of different approaches to estimate the CCMS, let alone a consensus on a new widely applicable standard. Here, we identify approaches based on historical data that allow accurate estimates, mainly using the example of the strongly changing Swiss mean temperature. The performance is assessed for the past and future combining long-term observations and climate projections with the centred 30-year mean (15 years observations, 15 years predictions) as CCMS benchmark. Several approaches, e.g. short-term linear trends, cubic splines and weighted local linear regression (LOESS) provide unbiased CCMS estimates for a broad range of climate scenarios and independent of trend magnitudes. Additional requirements such as the applicability to a wide range of variables, simplicity and the straightforward availability of uncertainty information are used to identify the most-suitable approaches. LOESS emerged as the most promising method in the overall assessment. KNMI already uses LOESS and MeteoSwiss plans to implement and use LOESS operationally in the near future. It will become MeteoSwiss’ new standard for determining long-term climate change signals and will also replace the Gaussian smoother currently used to visualise the evolution of climate variables.