Presentation Information
[ACG64-06]Uncertainty Quantification of an Earth System Model for risk assessment of Climate Tipping via Non-stationary Extreme Events
*Amane Kubo1, Yohei Sawada1 (1.University of Tokyo)
Predicting climate tipping points, such as the collapse of the Atlantic Meridional Overturning Circulation (AMOC), remains a formidable challenge due to the absence of direct observation records of such events and large parametric and structural uncertainties in Earth System Models (ESMs). It is a grand challenge to explore whether climate tipping or its risk is foreseeable using ESMs and observation without any direct records of climate tipping in the past. To contribute to the practical predictability of climate tipping, we quantified parametric uncertainty in an Earth System model and demonstrated how the parametric uncertainty resulted in the uncertainty in the projection of AMOC.
Climate tipping is driven by factors such as anthropogenic carbon dioxide emission and freshwater forcing. Thus, capturing the non-stationarity in observation is crucial for reliable uncertainty quantification. In this context, non-stationarity refers to the temporal shifts in the underlying climate states caused by the external forcing, particularly the changing frequency and intensity of extreme events. We proposed a new parameter uncertainty quantification method specialized for climate tipping risk assessment, which captures the non-stationarity appearing in the trends of mean climate and extreme events of the observation better than the existing method. We performed Observation System Simulation Experiments (OSSE) using the LOVECLIM intermediate resolution Earth system model. We employed a surrogate model-based uncertainty quantification approach to estimate five uncertain parameters related to atmospheric and oceanic physics and quantified the predictability of AMOC tipping under a freshwater hosing scenario.
We successfully quantified parameter uncertainty assuming that sea surface salinity, sea surface temperature, relative humidity or geopotential height can be globally observed. We found that whether we successfully reduce the uncertainty of the precipitation adjustment parameters, which are essential to the AMOC strength, depends on the observation variable selection. Our results also demonstrated that while the conventional methods exhibit significant estimation errors when applied to non-stationary data, the proposed method maintains high robustness and estimation accuracy even with the nonstationary observation.
Climate tipping is driven by factors such as anthropogenic carbon dioxide emission and freshwater forcing. Thus, capturing the non-stationarity in observation is crucial for reliable uncertainty quantification. In this context, non-stationarity refers to the temporal shifts in the underlying climate states caused by the external forcing, particularly the changing frequency and intensity of extreme events. We proposed a new parameter uncertainty quantification method specialized for climate tipping risk assessment, which captures the non-stationarity appearing in the trends of mean climate and extreme events of the observation better than the existing method. We performed Observation System Simulation Experiments (OSSE) using the LOVECLIM intermediate resolution Earth system model. We employed a surrogate model-based uncertainty quantification approach to estimate five uncertain parameters related to atmospheric and oceanic physics and quantified the predictability of AMOC tipping under a freshwater hosing scenario.
We successfully quantified parameter uncertainty assuming that sea surface salinity, sea surface temperature, relative humidity or geopotential height can be globally observed. We found that whether we successfully reduce the uncertainty of the precipitation adjustment parameters, which are essential to the AMOC strength, depends on the observation variable selection. Our results also demonstrated that while the conventional methods exhibit significant estimation errors when applied to non-stationary data, the proposed method maintains high robustness and estimation accuracy even with the nonstationary observation.