Climate System Dynamics and Modelling

OUTLINE

Models are essential tools in climatology to understand the processes responsible for observed changes and to make predictions. After defining what is traditionally meant by climate modelling, this chapter presents the different types of models. Then the chapter briefly reviews the main characteristics of the components of those models. Finally, a discussion is provided of the methodology applied to test the performance of the models and to interpret their results in conjunction with observations.

Introduction

3.1.1 What Is a Climate Model?

In general terms, a ‘mathematical model’ of the climate system is defined as a mathematical representation of the system based on physical, biological and chemical principles or laws. The equations derived from these laws are so complex that they must be solved numerically, leading to the formulation of a numerical or computer model (Figure 3.1). Such a model is used to perform simulations corresponding to realistic past or future conditions and idealised experiments. This is the reason why the term ‘climate simulator’ has been proposed to make a distinction between various types of models, such as physical, statistical or conceptual models (e.g., Crucifix 2012a). Nevertheless, we will keep here the classical denomination of climate model, which includes the following model equations (mathematical model) and the way these are solved (numerical model). This approach is largely practised in the field, keeping in mind that the word ‘model’ can have different meanings in different communities.

Numerical resolution of the equations provides an approximation of the true solution, which is, in practise, discrete in space and time. This can be interpreted as if the numerical models give estimates of climate variables averaged over regions, whose size depends on model spatial resolution, and over specific time intervals, which are related to the time step. For instance, some models provide globally or zonally averaged values, whilst others have a numerical grid, whose spatial resolution could be less than 100 km (Figure 3.2).