[M-GI26] Data-driven approaches for weather and hydrological predictions

[MGI26-02] Predictability of climate tipping: data assimilation approach

*Amane Kubo¹ (1.University of Tokyo)

Climate tipping is an irreversible change in the Earth system associated with climate change and is feared to have a significant impact on the entire planet and our society. To assess and mitigate these impacts, it is of paramount importance to predict whether and when climate tipping will occur under current global warming scenarios. However, it is difficult to detect the signs of climate tipping since many systems that will exhibit climate tipping are currently in a steady state. Despite a lot of efforts to model and observe climate tipping, predictability of climate tipping is poorly understood. Here we examined whether climate tipping can be predicted by data assimilation which combines physical models and observations. We specifically focused on how accurate observations should be to improve the predictability of climate tipping through idealized experiments of the large-scale mortality of the Amazon rainforest and the cessation of the North-South Atlantic thermohaline circulation (AMOC). We found that the successful reproduction of the internal variability of the system from observations is a key to realizing the prediction of climate tipping. When signal-to-noise ratio (S/N ratio) is defined as the magnitude of the internal variability of the system relative to the observation error, sufficiently high S/N ratio of observations is necessary to recover model parameters which can predict climate tipping. Current real observation networks of Amazon and AMOC are found to be sufficiently accurate in terms of the S/N ratio necessary to predict climate tipping according to our idealized experiment, although it should be noted that our idealized experiment substantially simplified the real-world assessment of climate tipping.