Division of Archaeology and Natural History, Machine VP
Research School of Pacific and Asian Studies
Co-Investigator Jozef Syktus
Division of Atmospheric Research, CSIRO
Palaeoclimate and Climate Modelling with the New CSIRO Coupled Atmosphere-Ocean-Landsurface-Sea-Ice Scheme
Modelling the global climatic system is recognised as a computational science `grand challenge'. In general, palaeoclimate modelling has so far used global climate models (GCMs) that are restricted in their representation of ocean-atmosphere-sea ice-land interactions than those now entering use for modelling present-day and future climates. In Australia, CSIRO has developed an advanced family of GCMs (that include ocean dynamics) which have been shown to simulate present climate as accurately as the best overseas GCMs.
Our project is aimed at verifying that the CSIRO models are also world-leading palaeoclimatic models by simulating past climates which are extensively documented from geological and palaeoclimatic data, such as the initiation and the climax of the last ice age. The project has three strands - testing model sensitivity to boundary conditions such as CO2 and seasonal solar radiation, simulation of selected past climates, and international intercomparison with other models.
We are using two model schemes, a lower resolution scheme known as CSIRO4 which is coupled to a dynamic upper ocean model, and CSIRO9 which has higher resolution and is optionally coupled to a full ocean circulation model. We run CSIRO4 on the CSIRO Division of Atmospheric Research (DAR) Cray supercomputer. CSIRO9 is now established on the ANU VP2200 supercomputer. In 1994 we completed a major series of sensitivity experiments with CSIRO4, including intercomparison with the US National Centre for Atmospheric Research CCMl model. This successfully simulated present climate and we are completing a simulation of climate of 6000 years ago.
What are the basic questions addressed?
We address the general objective of testing the CSIRO GCMs against palaeoclimatic data through a set of questions that are of high international interest and have not been resolved by modelling groups elsewhere.
Can ice age initiation be simulated with appropriate orbital forcing? Does the phase of climate change in the southern hemisphere relative to the northern hemisphere change with intensity of orbital forcing and of CO2 forcing? Can deep ocean-climate interactions that occurred during the ice ages be simulated?
Are results from the above experiments robust or sensitive to internal and external model parameters?
What are the results to date and the future of the work?
We have developed a coupled atmosphere (CSIRO4)/dynamical upper ocean climate model, the first of its kind, with comprehensive sea ice - soil water - cloud and other submodels, for palaeoclimate studies. In the course of this work we also examined the sensitivity behaviour of the coupled model with variations of solar constant, orbital forcing and a wide range of CO2, and compared the sensitivity and signal/noise behaviour of our model with the widely used North American CCMl climate model.
During the last six months the CSIRO9 climate model with vertical resolutions of 9 levels and R42 resolution in the horizontal domain (approx. 1.6x2.8 deg) was integrated on the VP2200 for 11 years of present day climate conditions. This represents the control case against which a range of perturbation experiments will be compared. A second 11-year run now being executed on VP2200 represents the case of the Holocene climatic optimum of 6000 years ago. (This is part of our contribution to the international Palaeoclimate Modelling Intercomparison Project (PMIP) -- Presentation of our simulations and comparisons with other model results will take place in early October in France, during the second PMIP workshop.) Both of these runs used fixed sea surface temperatures. At the same time, the modelling group at CSIRO DAR completed highly successful integration of the full atmosphere-ocean-sea ice-biosphere model for transient CO2 doubling and tripling.
We now propose a series of experiments, shared between the VP2200 and the DAR Cray. Because of the large compute time required for the coupled CSIRO9/full ocean scheme, we propose only one, carefully selected experiment to be run on the VP2200 that will require 520 hours of cpu time. This 21, 000 year experiment (climax of last ice age) will be the first to simulate full ice age climate with fully coupled and tested model scheme. By comparison with well established palaeo data, the predictive ability and performance of the deep ocean circulation model under highly altered climatic conditions can be ascertained. Complementary control experiments of shorter duration will be run on the Cray.
In the longer term, we foreshadow using the fully coupled model with transient boundary conditions to simulate initiation of the last ice age (116,000 years ago). Our previous modelling with the lower resolution CSIRO4 with coupled dynamic ocean has indicated icesheet initiation and monsoon changes compatible with palaeo data; this has not been well simulated by teams overseas and is a major test for climate models.
What computational techniques are used and why is a supercomputer required?
A combination of spectral and finite difference techniques are used to solve both three-dimensional atmospheric dynamics and ocean circulation as well as coupling. The computational complexity of the coupled model, the essential high resolution and the time-scales of global climatic behaviour all conspire to place this problem beyond the reach of all but the largest supercomputers.
Sensitivity of a coupled atmosphere - dynamic upper ocean GCM to variations of CO2, solar constant and orbital forcing, J Syktus, H Gordon and J Chappell, Geophysical Research Letters, 21, 1599-1602 (1994).
Signal-noise patterns from two GCMs with CO2 forcing: implications for recognition of enhanced greenhouse, J Syktus, J Chappell, R Oglesby, J Larson, S Marshall and B Saltzman, Nature, submitted.
Simulation of orbitally forced glacial initiation using a coupled GCM, J Syktus and J Chappell, Geophysical Research Letters, submitted.
Palaeoclimate modelling: A Pacific perspective. In: Coupled Climate System Modelling: a Southern Hemisphere Perspective, J Chappell and J Syktus, J Wiley, A Henderson Sellers and T Giambelucca (eds.), in press.