Ensemble global warming integrations with a coupled ocean-atmosphere model

Anthropogenic global warming, with both short- and far-reaching consequences, is a fact which has been supported by multiple independent studies. However, observed regional and global climatic changes are due to both human-induced changes and those generated naturally in the climate system over a broad range of timescales. Differentiating anthropogenic climate change from background climate noise is a big challenge for the climate science community. This thesis provides an improved understanding of the impacts of long-term internal climate variability on 20th and 21st century climate simulations, which is of great importance for socio-economic planners. Here, a quantitative assessment of the uncertainty associated with internally generated variability in 21st century climate projections of dynamic sea level (DSL), defined as the local departure from globally averaged sea level, is presented. Furthermore, the contribution of internal climate variability to the observed multidecadal changes in the tropical Pacific, a key component of the global climate system, is investigated. To this end, a series of unforced and forced climate model simulations, including a unique set of initial-condition experiments with the Kiel Climate Model (KCM), as well as several observational data sets, are analysed. Special emphasis is placed on the role of climate variability in designing and analysing a novel experimental setup, in which all realizations undergo identical external forcing but differ in their initial conditions covering a wide range of climate states. It is shown in chapter 2 of the thesis that the uncertainty in the DSL centennial trend over large parts of the globe, in particular the mid- and high-latitudes, is of the same order of magnitude as globally averaged steric sea level rise. Thus, the impacts of long-term internal variability cannot be neglected in centennial projections of the DSL. However, the uncertainty is substantially reduced when perfect knowledge of oceanic initial condition is assumed and only the atmospheric component is perturbed. This suggests that if climate model integrations start with an initial state which is reasonably consistent with the current oceanic state and its past radiative forcing, the uncertainty in centennial climate projections can be significantly reduced. Analysis of model results shows that despite the presence of considerable uncertainty, a large signal-to-noise ratio over some oceanic sectors is simulated, indicating the robustness of the anthropogenic global warming signal in these regions. These results are further confirmed by several multi-model ensembles of global warming provided in the Coupled Model Intercomparison Project Phase 5 (CMIP5) archive. The contribution of long-term internal variability to the observed multidecadal changes in the tropical Pacific climate, taken as a showcase, is the focus of chapter 3 of the thesis. In this context, model unforced simulations indicate the high level of internal variability over the tropical Pacific sector, which can potentially obscure anthropogenic climate change. It is shown that the observed decadal and multidecadal changes over this region are still within the range of unforced internal variability simulated by models. Moreover, it is found that the spatial structure of extreme decadal trends in Sea Surface Temperature (SST), Sea Level Pressure (SLP) and wind stress obtained from control runs are in a good agreement with the observed trend patterns. Findings of this chapter suggest that the most recent decadal changes in the tropical Pacific surface climate can be largely attributed to internal variability. Using multiple large ensembles of global warming experiments conducted with different models, the uncertainty in tropical Pacific climate projections due to internal variability is investigated in the fourth chapter. Results of this assessment indicate large irreducible uncertainty associated with internal variability. It is found that the anthropogenic climate changes in decadal timescale can be readily obscured by internal variability. Further, each ensemble provides some members displaying decadal timescale trends in SST, SLP and wind stress which are fairly similar to the observed decadal trends. This feature is poorly represented in CMIP5 experiments forced by historical radiative forcing. The findings of this chapter suggest using the probabilistic approach to study tropical Pacific climate trends which would provide new insights for the environmental planners