Chapter 4: Statistical Downscaling

In the early stages of climate modelling, then-called “General Circulation Models” (GCMs, an acronym which is now sometimes taken to stand for “Global Climate Models”) were used primarily for theoretical study of the large-scale dynamics of the atmosphere. As you learned in Chapter 2, these models have been developed into highly complex Earth System Models (ESMs) that include components for additional aspects of the climate system and have extensive parameterizations to represent the effects of sub-gridscale processes. However, these models can still only be trusted at face-value when it comes to large-scale aspects of the climate. The coarse resolution of global models (\(\sim\) 1\(^{\circ}\), or 110 km) and the reliance on empirical parameterizations means that fine-scale features, which can be extremely important for regional and local climate, may not be represented well or even at all.

These systematic model errors result in biases in the output - discrepancies between the simulated and observed climate. Before climate model output can be used for regional climate impact studies, one must make an effort to correct these biases. If your impact study requires information over a contiguous region, then you may also need data on a finer spatial grid than that of the model. These two tasks are the goal of a broad category of methods referred to as downscaling.

In Section 4.1, we will recap some of the basics of downscaling, though readers are encouraged to review Module 2 of the Engineering in a Changing Climate e-Learning resource before continuing. Section 4.2 will introduce commonly used methods of correcting for model biases without altering the spatial resolution of the data, and Section 4.3 will lay out different methods of true spatial downscaling, many of which use one or more bias correction methods from 4.2. We will then start to work with some downscaled climate data by comparing it to observations in Section 4.4 and assessing downscaled future projections in Section 4.5.