Stochastic simulation of surface variables (most often precipitation) conditional on synoptic states is one approach to weather generation. As a limited number of conditioning states will be identified, many events are grouped together under the category of a typical weather type, potentially limiting the amount of variability that will be represented in the simulation. The problem can be addressed by improving synoptic resolution, i.e., increasing the number of classified states, but this decreases the sample size within events, increases computational demands during simulation, and increases the risk of the weather typing (clustering) algorithm inventing spurious states.

 

In this project researchers apply a Bayesian approach to describing local response to synoptic conditions, in order to determine whether improving representation of within state variation can accomplish the same result as increasing the number of conditioning states.

 

Time frames: Research commenced in September 2014, time frames not confined

Funder: Climate Systems Analysis Group (CSAG)

Partners:

For further details: Contact Dr Piotr Wolski