Creighton, Glenn, Air Force Weather Agency
The Air Force Weather
Agency (AFWA) now has the ability to run numerical weather prediction models at
higher resolutions and over larger areas than ever. However, as processors have become faster, data
accessibility has lagged sorely behind.
The expected increase in demand for large, high spatial and temporal
resolution simulations carries with it an implicit demand for more efficient
data handling, especially in an environment like AFWA where bandwidth and I/O
are such precious commodities.
With the advent of Mesoscale Ensemble Prediction Suite (MEPS) at AFWA,
ignoring efficient data handling procedures during development would be a
potentially crippling oversight.
Currently
post-processing speed is still largely bottlenecked by the need to handle
numerous large three-dimensional datasets. It is proposed that this problem can be alleviated by
implementing the strong message-passing interface architecture available within
WRF. In exchange for a marginal
decrease in model processing speed, many of the three-dimensional datasets currently
required by the MEPS post-processor, can be eliminated altogether and replaced
with a comparatively lean set of two-dimensional diagnostics output from WRF
itself. This talk will outline
some of the steps AFWA has taken to develop and insert diagnostics directly
into WRF.