Peter Düben, University of Oxford, Department of Physics
To reduce numerical precision to achieve higher accuracy in weather and climate modelling
In atmosphere and ocean models, values of relevant physical parameters are often uncertain by more than 100% and weather forecast skill is decreasing significantly after a couple of days. Still, numerical operations are typically calculated with 15 decimal digits of numerical precision for real numbers. If we reduce numerical precision, we can reduce power consumption and increase computational performance significantly. Savings can be reinvested to allow simulations at higher resolution that would hopefully allow improved predictions of future weather and climate at higher accuracy.
We aim to reduce numerical precision to the minimal level that can be justified by information content in the different components of weather and climate models. But how can we identify the optimal precision for a complex model with chaotic dynamics? We found that a comparison between the impact of rounding errors and the influence of sub-grid-scale variability can provide valuable information on the level of precision that should be used. We also found that the influence of rounding errors can actually be beneficial for simulations since variability is increased and that numerical precision can be reduced with forecast lead time as forecast errors are increasing. Our physical understanding of the system also guided us to investigate a scale-selective approach that is using high precision to integrate large-scale dynamics and low precision to integrate small-scale dynamics.
We have performed multiple studies that investigate the use of reduced numerical precision for atmospheric applications of different complexity (from Lorenz'95 to a full global circulation model) and cooperated with computing scientists to study the use of real hardware that allows to trade numerical precision and against performance in atmospheric applications. Results will be presented during the talk.