Centre for Resource and Environment Studies Machine CM
Co-Investigators David P Hansen, David A Post, Sergei Schreider and Wei Ye
Centre for Resource and Environment Studies
Dynamic Separation of Catchment Water Balance.
The aim of this project is to perform dynamic separation of catchment water balance using rainfall, temperature and stream flow time series. The computer program, IHACRES, will be applied to a large number of catchments (of different sizes and in different hydrodynamic regimes) world wide. The outcome will be a better understanding of how catchments respond to climatic and land use changes.
What are the basic questions addressed?
IHACRES has proven to be a relatively simple but effective model, incorporating the basic physics and having around six parameters, that has been shown to be applicable in a wide variety of hydroclimatologies. The accuracy of IHACRES compares extremely favourably with all other, including more complex models, which are in common use for calculating streamflow runoff from rainfall data. A number of catchments in different climates, and with different vegetation cover, are being studied. It is hoped that this will lead to improvements in IHACRES and that the model can then be incorporated into global climate models as well as being used to assess the effects of prescribed vegetation changes on the water balance in a prescribed catchment.
What are the results to date and the future of the work?
The program IHACRES is currently being ported to the CM-5. A `cloning' code, using the CMMD libraries, has initially been written. The processing is divided between the processors allowing for the parallel processing of 32 different sets of input parameters. Currently the program has been partially rewritten in CMFortran, however it is envisaged that a further increase in speed of the code over the workstation version can be achieved once more efficient algorithms are used for the major processing subroutines. This will allow for a greater level of parallelisation. Currently this program is being used to search the parameter space on a number of different catchments, at a speed and ease much greater than what could be achieved on workstations. Catchments being modelled include various sub catchments in the Murray-Darling basin, relatively dry catchments in Western Australia, as well as benchmark catchments in Australia and the United States.
What computational techniques are used and why is a supercomputer required?
The bulk of the calculation involves parameter estimation using an instrumental variable algorithm. A large parameter space needs to be searched and this can be efficiently carried out in parallel.
The CM-5 provides a perfect platform for the program IHACRES. The programming language CMFortran will allow an efficient version of the program to be written for the CM-5 and the CMMD library functions then allow the program to be run in parallel on the 32 nodes thus allowing the parameter space to be more efficiently searched.