Improving simple climate models through a traceable and process-based analysis of ocean heat uptake in AOGCMs and observations

Future climate change predictions provide essential guidance for the global efforts to curb the global warming trend caused by human emissions of greenhouse gases. The main tool to provide these projections are coupled atmosphere-ocean general circulation models (AOGCMs). However, these models are computationally still very expensive to run. Therefore, simple climate models (SCMs) have been developed, which are able to mimic the climate response seen in the AOGCMs, but at a much reduced computational cost. SCMs are being used for several purposes, e.g. simulating how the projections depend on key climate parameters, or for the interpretation of the AOGCM projections. SCMs are often used for policy advice.

To ascertain the usefulness and accuracy of SCMs, it is essential to establish their traceability to comprehensive AOGCMs and ultimately to reality. The link and hence the traceability between SCMs, AOGCMs and observations is established through the process of calibration, which is the key step whereby the values of the control parameters of a given model are set up. There appear to be two main kinds of calibration: physical and behavioural, which operate very differently. A physical calibration is one that specifies the control parameters of a given model by invoking physical arguments or observational constraints on the physical processes involved. In contrast, a behavioural calibration is one that specifies the same control parameters so that the model reproduces various emergent properties of the actual or simulated climate system, under past, present or future conditions.

Because climate change is primarily controlled by ocean heat uptake (OHU), the accuracy of climate change projections depends on the validity of the representation of the physical processes controlling OHU in SCMs and AOGCMs. How these processes are represented varies widely across SCMs, and ranges from explicit to entirely implicit representations. In this project, three specific SCMs are considered. In MAGICC, ocean heat uptake is explicitly represented through two vertical advection/diffusion equations for each hemisphere. In Gregory (2000)'s two-layer model, it is represented via two simple ordinary differential equations with two-time scales. In Good et al. (2011,2012)'s step-response approach, ocean heat uptake is entirely implicit.

In this proposal, our first objective will be to assess the degree of generality of each SCM by testing their relative performances on the same range of climate change scenarios, in order to identify which simplified representation of OHU performs better. Our second objective will be to investigate the link between physical and behavioural calibration, by implementing a physical calibration of MAGICC, and testing whether it improves or deteriorates its performances and why. This will provide key new insights on which aspects of ocean heat uptake are robust, and which are in need of further study, which will enhance the credibility of SCMs. The first and second objectives address Goal 2 of the Call. Our third objective will be to develop a versatile and flexible approach to constraining ocean heat uptake processes by using NEMO and its adjoint NEMOTAM, allowing for the optimal calibration of mixing parameters from both physical and behavioural constraints. This will endow the UK with a key capability for systematically re-calibrating ocean models by incorporating observational and theoretical advances on ocean mixing processes, and testing the implications for climate change projections. Moreover, the optimisation set-up for NEMO is potentially useful for many other types of studies in the future. This addresses Goal 1 of the Call. This will make a key contribution to the joint NERC-Met Office strategy for the development of UKESM1.

The simple climate models, with improved parameters and a clarified traceability, will be of great benefit for the interface between climate science and climate policy. MAGICC is widely used by the IPCC as a tool to interpolate AOGCM scenarios for impact assessment as well as adaptation and mitigation studies (IPCC AR4). The proposed research will establish alternative simple climate models for these purposes, and will improve MAGICC. The fact that this is based on physical processes and constraints will lend more credibility and reliability to these models and the scenarios they produce. As a highly relevant impact, this affects policymakers and the wider public alike. The project results will be of great use for the IPCC, and might increase the acceptance of climate models and their scenarios in the public debate. These impacts should materialize already during the project lifetime, as soon as the first publications have appeared.

The optimization routine we will set up for NEMO and its adjoint has a large and far-reaching benefit for the UK climate modelling community as a whole, and for those modelling centres using NEMO in particular. Once the NEMO optimization is up and running, and has been shown to work for the ocean heat uptake processes, it can be used for many different other ocean processes, such as freshwater transports, specific mixed layer processes like the Langmuir circulation, or specific parameterizations like bottom-enhanced mixing. Processes like these can be used as control variables in order to constrain them (or their parameterizations) on a physical basis. This is an important and very useful impact for the development of NEMO as undertaken at the Met Office (HadGEM3), at the ECMFW as part of their forecasting activities, and for the development of UKESM1 at the Met Office and NERC centres. This impact should be realised towards the end of the project.

The career development of the PDRA will benefit from this project. The PDRA will gain knowledge of unusually many different aspects of climate modelling, namely full-fledged, three-dimensional AOGCMs; adjoint techniques involving AOGCMs; simple one-dimensional climate models; and a simple climate model based on the impulse-response approach. With climate science becoming more and more interdisciplinary, having worked on this project will equip the PDRA with broad skills for their future career.