The media often lends more credence to dramatic predictions from individual climate models than does the climate science community as a whole. Models are metaphors of reality; they should be taken seriously, not literally. Predictions made by simple climate models need to be confirmed using more comprehensive climate models. Deep uncertainty surrounds estimates of climate metrics like climate sensitivity, meaning that the error bars presented with these metrics may have their own unquantifiable error bars. It is important to make the distinction between precision and accuracy when evaluating uncertainty estimates of climate parameters. Overconfidence in numerical predictions of extreme climate scenarios may lead to a “doomist” belief that a climate catastrophe is inevitable and that there is nothing we can do to prevent it.

The concept of Occam’s Razor, also known as the principle of parsimony, is a motivating force in science. Galileo’s experiment of dropping objects of different weights from atop the Leaning Tower of Pisa to show that they fall at the same rate is an illustration of this principle. But the application of Occam’s Razor to climate modeling is less straightforward. Simple models of the kind used by Manabe provide qualitative insights, but they are not well-suited for quantitative predictions. To understand this, we can make an analogy between the hierarchy of climate models and the hierarchy of biological models, from fruit fly to mouse. Simple models are used to explore “climate tipping points,” where amplifying feedbacks lead to abrupt climate change, but they may not consider all the stabilizing feedbacks. It is therefore important to use a hierarchy of models, with varying degrees of complexity, to study climate phenomena.

A reservoir simulator consists of a large set of models and parameters to describe the geology, fluid behavior, wells and surface facilities, and rules and conditions that describe and control the production process. On the numerical side, you have a combination of discretization methods, nonlinear solvers, linearizations, linear solvers, preconditioners, stability and convergence checks, and algorithms for automated time-step selection. The AD-OO framework in MRST is developed to encapsulate all these details, such that on one side it offers industry-grade simulator capabilities, and on the other side a flexible framework for rapid development of new proof-of-concept implementations. This chapter outlines the design philosophy behind AD-OO and explains many of its ingredients in detail. By reading this chapter, you will get an in-depth introduction to all the details you need to make a full-fledge reservoir simulator. The chapter ends with a discussion of three cases: a simple pressure depletion, multisegment representation of instrumented wells, and the full SPE 9 benchmark.