This work considers a set of meso-scale weather prediction models using stochastic partial differential equations. These models are simulated by a reconfigurable computing platform involving both general processors and reconfigurable elements based on devices called field programmable gate arrays (FPGAs) which may be able to perform key algorithms in computational fluid dynamics with a speed of 100-1000 times of a serial processor.

We provide an overview of platforms such as the Cray-XD and SRC machines, which have become very affordable, enabling the use of reconfigurable computing based simulation by the aerospace, aviation and airport operations communities at large. Additionally, low cost FPGA hardware acceleration boards can be also used in low-cost cluster based configurations for on-board use. We discuss a family of multi-dimensional partial differential equations solvers based in key formulations such as Poisson and Cauchy-Riemann, which are applied to detailed topographic modeling in weather and turbulence prediction. The probability of occurrence of turbulence in a flight cell or around a local airport is estimated from the forecast by detailed meso-scale weather models.

This work describes how traditional partial differential equation solvers used in weather prediction can be mapped to hardware based digital signal processing (DSP) algorithms, which are mapped to low cost reconfigurable machines. The use of sub-domain techniques to allow parallel processing of the model is discussed. It is assumed that real-time weather data is being continuously provided to the model. The use of high-quality meso-scale forecasting platforms will enable better utilization of airport resources and proper routing of flight traffic to avoid turbulence, key issues for the next generation air transport management (ATM) systems. This work evaluates how the proposed platform can be seamlessly integrated to these systems and the potential economic impact of its deployment.