ML-TOMCAT: machine-learning-based satellite-corrected global stratospheric ozone profile data set from a chemical transport model

Abstract. High-quality stratospheric ozone profile data sets are a key requirement for accurate quantification and attribution of long-term changes. Satellite instruments provide measurements over typical mission durations 5–15 years. Various methodologies have then been applied to merge homogenise the different satellite in order create observation-based with minimal gaps. However, individual use measurement methods, sampling patterns retrieval algorithms which complicate merging these sets. In contrast, atmospheric chemical models can produce chemically consistent simulations based on specified changes external forcings, but they subject deficiencies associated incomplete understanding complex processes uncertain photochemical parameters. Here, we self-consistent output from TOMCAT 3-D transport model (CTM) random-forest (RF) ensemble learning method merged 42-year (1979–2020) set (ML-TOMCAT V1.0). The underlying CTM simulation was forced by meteorological reanalyses, trends long-lived source gases, solar flux aerosol variations. RF is trained using Stratospheric Water OzOne Homogenized (SWOOSH) time periods Microwave Limb Sounder (MLS) Upper Atmosphere Research (UARS) (1991–1998) Aura (2005–2016) missions. We find that ML-TOMCAT shows excellent agreement available independent satellite-based pressure as vertical coordinate (e.g. GOZCARDS, SWOOSH non-MLS periods) weaker altitude-based SAGE-CCI-OMPS, SCIAMACHY-OMPS). at almost all levels concentrations well within uncertainties observational (V1.0) ideally suited evaluation profiles tropopause 0.1 hPa freely via https://doi.org/10.5281/zenodo.5651194 (Dhomse et al., 2021).