• 06 Oct 17
  • Posted by admin
  • 0 Comments

Climate predictions and downscaling to extreme weather

BINGO WP2 outcomes reveal that the space of possible climate states in uninitialized climate projections (i.e. the uncertainty) is larger than the variability in the recent observations! Initializing these simulations (i.e. turning projections into predictions) reduces this variation.

Thus for near term planning, skillful decadal predictions are preferable. On continuing a climate prediction until the end of the century, it will tend to forget initial conditions and just depend on the RCP assumed. Long term planning requires choosing an RCP.

Climate projections are used to explore the range of possible climate developments under a given scenario of greenhouse gas emissions, a so-called RCP (representative concentration pathway). RCPs are based on assumptions of greenhouse gas emissions in the future which are difficult to estimate, as future global climate policies are impossible to foresee. A number of numerical climate model runs are performed to explore the range of possible climate developments under a given RCP. They are commonly started from an arbitrary climate state under the greenhouse gas and solar forcing of the late 19th century. From this state, each individual numerical simulation develops under the observed change of greenhouse gas forcing prescribed until the end of the 20th century, and under the chosen RCP afterwards. With this approach, the model runs are designed to represent the range of possible climate developments. They are typically considered for the time range until the end of the 21st century.

Knowing the present day climate state, the range of variability can be reduced for a number of years ahead. Simulations are then initialized from states close to the present day situations. Ocean and land surface states are particularly important in this respect. We speak of (initialized) predictions instead of (uninitialized) projections. As the influence of the specific RCP is comparatively small for short time scales of typically less than 10 years, these predictions do not depend much on unknown climate policies. As the influence of the initial state decreases with increasing simulation time, initialization would become unimportant again when these simulations are run for a long-term period into the future (i.e. to the end of the century), while the chosen RCP would be the more important factor again.

These decadal (near-term) climate predictions are thus valuable for near-term planning. With hydrological impact studies (starting with the hydrological modelling done in WP3) fed by downscaled decadal predictions, the work done in BINGO complements existing (and ongoing) hydrological impact studies based on climate projections. Climate projections studies, on the other hand, explore the potential range of climate variability until the end of the century and under various RCPs. Best- and worst-case scenarios can be deduced, but as the range of variability (uncertainty) inherent to the unknown future developments of greenhouse gas concentrations (RCPs) is large, the uncertainty for planning arising from long term climate projections is typically larger.