Inflated uncertainty in multi-model based regional climate projections

Date: 26.05.2016

Time: 11:00 - 12:00

Place: Nordlys, DMI

Contact: Cathrine Fox Maule

Climate scenarios are widely used to explore likely impacts of climate change. As risk based impact analyses are often required, there is an increasing demand for probabilistic climate change information based on multi-model approaches. However, information from multi-model ensembles may be extracted in several ways.

The uncertainty in multi-model solutions is often (e.g. in IPCC) portrayed by collecting all available model information and evaluating the appropriate statistics for a given climate variable (e.g. mean annual temperature) grid point by grid point using for example certain percentiles to express likely lower and upper bounds around a mean value. Here, we use CMIP5 data for temperature and precipitation changes to illustrate that this approach has some inherent inconsistencies when it comes to regional climate projections.  An alternative approach is suggested where dominant patterns of climate change from multi-model ensembles are identified from EOF analysis and used to construct globally constrained and physically more consistent maps of climate change uncertainty.