Projecting the Response of Climate to Orbital Variations in the Pliocene Using an Emulator

Abstract:

The mid-Pliocene occurred between 3.264 and 3.025 Ma BP, and was characterized by several periods of relatively warm climatic conditions. It has been researched extensively, in part due to the availability of proxy data, allowing for data/climate model comparisons, but also because it is thought that it can give insights into the characteristics of future climate, due to the fact that periods of the mid-Pliocene were characterised by CO₂ levels similar to present. However, it is now being realised that the orbital forcing during the mid-Pliocene is also important for properly understanding its relevance to future climate. Computational resource limitations associated with running General Circulation Model (GCM) experiments often restrict the number of simulations that can realistically be carried out, inhibiting the study of the long-term evolution of climate on orbital timescales (>10s kyr). Here, we model the evolution of climate over the Marine Isotope Stage K1 interglacial period in the mid-Pliocene warm period in response to variations in orbital forcing, using an emulator approach. An emulator is a fast statistical model based on the output of a more complex model, in this case the Hadley Centre climate model (HadCM3); the premise of this approach being that once the emulator is built no additional GCM runs are required. A 40-member ensemble of simulations are run using HadCM3, with atmospheric CO₂ and the three orbital parameters varying within ranges appropriate for the Pliocene. Latin hypercube sampling is used to optimally and efficiently sample the 4-D parameter space, and the model results are used to develop and calibrate an emulator. Combined with various CO₂ scenarios, this is then used to simulate climate at 1 kyr intervals over the period 3.08 to 3.04 Ma, which includes MIS K1. The emulator projections are then compared with existing model simulations [Prescott et al., 2014], as well as proxy data. It is demonstrated that climate emulators may provide a useful tool for rapidly simulating the long-term evolution of climate, both past and future, with relatively small computational cost.