Future climate change cannot be supported by evidence from a future that does not exist yet, therefore the evidence comes in the form of model projections in climate-change assessments. However, models are difficult to evaluate as not all parts are possible to measure by observations and, moreover, many sites on Earth are still data sparse. In addition, one model may perform well for some regions/ seasons/ variables but worse for others, where another model may show better performance. The usual way to handle these differences is to use model ensembles, in which results are regarded as well-supported if a sufficiently diverse set of models agrees on the results.
Is there a way to reduce members in the ensemble?
Many impact modellers find difficulties using the large amount of data available from the wide range of combinations of climate models and scenarios. Therefore, often a subset of the data is used. The selections of these subsets are usually based on how the models perform with respect to key variables (such as temperature or precipitation).
There is no general method to reduce the number of members in the ensemble. It is important to think about which variables and processes that are most important for a specific impact study. Models can then be selected to cover the range of a full ensemble for these variables. Disadvantages with reducing the numbers of ensemble members is that the selection may be less general in supporting a variety of user needs. Moreover, the selection becomes more dependent on results from the individual ensemble members.