Objectives of and Potential Approaches to Harmonizing Regional Information

Chair: David McCollum

Draft prepared by Elmar Kriegler, Kate Calvin and David McCollum

Why should we harmonize regional information?

Global energy-economy and integrated assessment models (IAMs) have made different choices of native model regions, including both the number and the configuration of regions (i.e., which countries are included in which model regions). Model input and output will be specified on the level of these native regions, or aggregates of those. As a consequence, some incomparability of regional information from different models will arise. Since it cannot be expected that IAMs will strive to choose identical regional breakdowns in the future, nor that most of them will be capable of regional flexibility that would allow them to adapt their regional breakdown to individual study settings, intermediate solutions will be needed to move towards a more harmonized treatment of regional information. The benefit of such a more harmonized treatment is fourfold:

  • The comparability of regional model output is improved, if it refers to an identical or very similar region.
  • Likewise, model input assumptions that are mostly specified on the level of model regions can be better compared.
  • Models can be better harmonized in comparative studies of regionally fragmented policies. As real world climate policies will likely continue to differ across countries/regions for some time yet, such studies will play an increasingly important role in the future.
  • More generally, any regional specific external influence to be imposed on the model – including climate impact and climate policy information – will be more broadly applicable to a large set of models, if those models offer a larger degree of regional harmonization.

The objective of the subgroup on regional information of the IAMC SWG on data protocols and management is to facilitate the better harmonization of regional information across IAMs in the future.

How to achieve a more harmonized treatment of regional information in the future?

Greater harmonization of regional information will be a long term and voluntary process. The general concept is to establish good practice standards and useful regional breakdowns that can serve as benchmarks for future choices on the regional resolution of models that will be taken by modelling teams individually as they continue to develop and re-program their models. Since modeling teams will continue to make different choices about the most appropriate number/configuration of native model regions for their application context, it would be useful to establish benchmark regional breakdowns at different resolutions, e.g. at the level of approx. 5, 10, 20, 30, and 40 world regions (see Figure 1). Modeling teams could then choose a set of native model regions that

  1. would derive from a benchmark regional breakdown with high regional resolution (blue circles in Figure 1)
  2. could be aggregated to a benchmark regional breakdown with low regional resolution (green circles in Figure 1).

As a result of (1), model input information, e.g. relating to policies, technologies, sectoral climate impacts etc. could be specified on the level of a benchmark regional configuration with high resolution and then used consistently by all models. Of course, country level information constitutes a benchmark with very high regional resolution that is already drawn upon by models today. But for many applications, it will be impractical to specify input information at this level of regional detail (e.g. reference climate policy scenarios), and a benchmark regional configuration with only 30 or 40 world regions would greatly alleviate the problem.

figure1

Figure 1: Basic approach to regional benchmarking of native model regions

As a result of (2), model output can be aggregated to some benchmark regional configuration (with lower regional resolution) that is applicable to all models. This has a direct benefit for the comparison of output from different models since it refers to the identical regions on the level of this benchmark configuration. The more harmonized the regional breakdown of individual models will become, the higher the resolution of the harmonized output regions (2; green circles), and the lower the resolution of the harmonized input regions (1; blue circles) that is needed.

We propose that the identification of the benchmark regional configurations should take into account the following four criteria:

  • Current regional breakdowns used by IAMs (which would need to be assessed and collated in a spreadsheet to identify differences and commonalities)
  • Regional resolution of key data sources used by IAMs (e.g. PWT, IEA, GTAP, policy and technology databases etc.)
  • Relevance of regions for current climate policy considerations   (e.g., major economies, largest emitters and regional blocks may be resolved, etc.)
  • Similarity of countries within a region with regard to pertinent characteristics for climate policy analysis (e.g., growth dynamics, income, fossil resource endowment, institutional and cultural factors, etc.)

The application of these criteria will be a function of the number of regions that a benchmark regional configuration includes. Of course, the higher the regional resolution, the better these criteria will be fulfilled. However, computational limitations usually limit the number of regions that IAMs can represent. The optimal number of regions based on these criteria and computational tractability will vary between different modeling frameworks, and be determined by teams individually. The activities and output of this subgroup will offer guidance to teams when doing analyses and undertaking model development in the future.