Principal Investigator John A Taylor Project s56

Centre for Resource and Environmental Studies Machine VP

Co-Investigators Jay W Larson, David Hansen and Gary Bates

Centre for Research and Environmental Studies and National Center for Atmospheric Research, USA

Mesoscale Meteorological Modelling

Global climate change due to increasing concentrations of greenhouse gases is anticipated to have substantial impacts on natural ecosystems and the agricultural systems and, in combination with the urban heat island effect, impact on urban areas. Unfortunately global general circulation models (GCMs) are not sufficiently high in spatial resolution to estimate the impacts at the regional to the local scale. For this task mesoscale meteorological models (RegCM2 and MM5-NCAR/PSU models) are used in conjunction with global meteorological observations and GCMs to assess the impacts at the local scale. This project involves porting existing models (RegCM2/MM5) and applying them to the study of climate and land use change, initially in urban areas.

What are the basic questions addressed?

The research seeks to improve our understanding of the physics and chemistry of the urban atmosphere and investigate the potential impact of global climate change at the regional scale.

What are the results to date and the future of the work?

Two models RegCM2, a regional climate model, and MM5 the current US National Center for Atmospheric Research and Pennsylvania State University mesoscale meteorological model have been ported to the VP2200. Initial model runs using RegCM2 with a model domain over the USA and Japan have commenced. Future model runs will focus on the Murray-Darling Basin.

What computational techniques are used and why is a supercomputer required?

The model code is highly vectorised (~85% on the VP2200 for RegCM2) with a one-month simulation requiring ~2 hours CPU time. MM5 and RegCM2 are grid-point models with finite differences centred in space and time. Second-order finite differences are used for the advection terms and an Asselin time filter is applied to all prognostic variables. MM5 can be either hydrostatic or non-hydrostatic. The hydrostatic solution uses split semi-explicit time integration for efficient treatment of the fast gravity modes and the non-hydrostatic solution uses semi-implicit integration.