Welcome to JUMP
JUMP (Japan Uncertainty Modelling Project) is a new project in FRCGC, which links with groups in NIES and CCSR working on similar subjects. Running for 5 years, from April 2007, we aim to investigate and reduce the uncertainty associated with various aspects of climate change ranging over the multidecadal to centennial time scale.
Methods
A range of methods will be used to attack distinct problems. We have used the ensemble Kalman filter to generate ensembles of multivariate parameter sets to drive a number of climate models of varying complexity [2,7,8]. This relatively efficient method enables us to sample the uncertainty arising from the multivariate interactions between parameterisations. We are developing a particle filtering method to improve performance in highly nonlinear applications. We also plan to investigate relationships across the spectrum of model resolution and complexity (see next section) in order to enable us to efficiently estimate uncertainty arising from the full range of processes.
