A tool for objective detection of abrupt transitions in CMIP6 models

We present here a tool for the detection of abrupt transitions in CMIP6 model outputs, that is aimed to update and extend the catalog of tipping points presented in Drijfhout et al. 2015, based on the evaluation of CMIP5 intercomparison. The tool consists of three fundamental steps:  Data manipulation: model outputs are sampled according to the user’s preferences, aggregated along the integration period and interpolated to a common grid for the whole multi-model ensemble. A 10-years moving average is also applied; Criteria for abrupt transitions: Criteria for the detection of abrupt transitions are computed and combined. These are: exceedance of the preindustrial 99-percentile standard deviation, exceedance of the preindustrial 99-percentile jump over 10 years period, exceedance of the preindustriak 99-percentile yearly anomaly for each year in the last 30 years of the simulation, p-value of a Kolmogorov-Smirnov hypothesis test for normality of the distribution; Masking and clustering: grid points for which the time series of anomalies with respect to preindustrial conditions that satisfy at least 3 out of 4 of the criteria illustrated above are selected. Successively, grid points are clustered in order to exclude sparse points and highlight significant regions affected by widespread abrupt transitions; We present a preliminary analysis demonstrating the usage of this tool on a set of ocean-sea-ice-related quantities for a number of models participating in CMIP6 project under disparate SSP scenarios.