Carbon-cycle modelling and Phanerozoic climate change

Greenhouse gases trap heat in the Earth’s atmosphere and warm our planet and on geological time-scales CO₂ is the most influential greenhouse gas in modulating atmospheric temperature. During most of the Phanerozoic (past 540 million years), our planet was warmer than today, and a greenhouse-dominated climate (80%) was only interrupted by three periods of cold glacial conditions during the end-Ordovician (Hirnantian) glaciation, the Permo-Carboniferous (~330-260 Ma) and the second half of the Cenozoic (34-0 Ma). Icehouses are characterized by lower CO₂ concentrations and temperatures, and a modern CO₂ threshold for continental-scale glacial inception is estimated to 500 ppm. But with a fainter sun, the glacial inception threshold during the Hirnantian (445 Ma) glaciation was probably closer to 1000 ppm.

CO₂ concentrations cannot be measured in deep time, and we therefore must rely on proxies, or models. For the past 450 million years, CO₂ proxies during greenhouse climates average ~1100 ppm whilst the Phanerozoic icehouse intervals average ~480 ppm. But a proxy-based picture of CO₂ concentrations before 450 Ma is lacking and thus CO₂ levels for most of Earth’s history must be estimated from carbon-cycle models. Models are also important for capturing the processes (sources and sinks) that can explain shifting greenhouse and icehouse climates and can loosely be classified as inverse or forward models, pending on whether isotopic proxy data are parametrized or predicted from the model, respectively. Both model types, however, incorporate several biological and geological/tectonic forcing parameters that should be similar in all models.

Carbon-cycle models predict very different atmospheric CO₂ levels for large of the Phanerozoic, differing by more than 4000 ppm and model-proxy differences can exceed 5000 ppm. Many of the relatively large, modelled differences in atmospheric CO₂ are arguable caused by differences in time-dependent parametrization of plate tectonic degassing and silicate weathering, and benchmarking of carbon-cycle models are urgently required. In this contribution we focus on carbon-cycle modelling with GEOCARB_NET — a user-friendly version of the GEOCARB model. In GEOCARB_NET input parameters can easily be changed, tested, and compared with other models (e.g., COPSE, SCION and GEOCLIM). The system also contains databases for CO₂ proxies and temperatures that be visualized together with CO₂ predictions. We highlight how key input parameters can seriously affect reconstructed CO₂ levels but also how models and proxies can better be reconciled.