Tim Palmer

Abstract:

In the 2013/14 winter, the eastern USA was exceptionally cold, the Bering Strait region was exceptionally warm, California was in the midst of drought and the UK suffered severe flooding. It has been suggested that elevated SSTs in the tropical West Pacific (TWPAC) were partly to blame due to their producing a Rossby wavetrain that propagated into the extratropics. We find that seasonal forecasts with the tropical atmosphere relaxed towards a reanalysis give 2013/14 winter-mean anomalies with strong similarities to those observed in the Northern Hemisphere, indicating that low-latitude anomalies had a role in the development of the extremes. Relaxing just the TWPAC produces a strong wavetrain over the North Pacific and North America in January, but not in the winter-mean. This suggests that anomalies in this region alone had a large influence, but cannot explain the extremes through the whole winter. We also examine the response to applying the observed TWPAC SST anomalies in two atmospheric general circulation models. We find that this does produce winter-mean anomalies in the North Pacific and North America resembling those observed, but that the tropical forcing of Rossby waves due to the applied SST anomalies appears stronger than that in reanalysis, except in January. Therefore both experiments indicate that the TWPAC influence was important, but the true strength of the TWPAC influence is uncertain. None of the experiments indicate a strong systematic impact of the TWPAC anomalies on Europe.

Abstract:

Programmable hardware, in particular Field Programmable Gate Arrays (FPGAs), promises a significant increase in computational performance for simulations in geophysical fluid dynamics compared with CPUs of similar power consumption. FPGAs allow adjusting the representation of floating-point numbers to specific application needs. We analyze the performance-precision trade-off on FPGA hardware for the two-scale Lorenz '95 model. We scale the size of this toy model to that of a high-performance computing application in order to make meaningful performance tests. We identify the minimal level of precision at which changes in model results are not significant compared with a maximal precision version of the model and find that this level is very similar for cases where the model is integrated for very short or long intervals. It is therefore a useful approach to investigate model errors due to rounding errors for very short simulations (e.g., 50 time steps) to obtain a range for the level of precision that can be used in expensive long-term simulations. We also show that an approach to reduce precision with increasing forecast time, when model errors are already accumulated, is very promising. We show that a speed-up of 1.9 times is possible in comparison to FPGA simulations in single precision if precision is reduced with no strong change in model error. The single-precision FPGA setup shows a speed-up of 2.8 times in comparison to our model implementation on two 6-core CPUs for large model setups.