Prognosis for Weather Forecasting

Abstract

Operational meteorology and its supporting research are based on long-established beliefs that important weather changes are preceded by recognizable meteorological events, and that forecasting accuracy improves along with increased density and frequency of observations and increased computational power. While in recent years these increases have become enormous, weather forecasting has not comparably improved; also, pressure distribution is now predicted no better by ?dynamical? numerical methods than by statistical methods. Turbulence may be to blame for this poor performance. Its role has been obscured by the crude statistical representations used by meteorologists. Recent research in fluid dynamics, however, suggests that true turbulence is so strongly intermittent and anisotropic that it effects most of the atmospheric mixing and transfer processes in only 5 to 10% of the time and that it cannot be properly incorporated in either dynamical or statistical models. The hypothesis is then advanced (and supported by some evidence) that deterministically unpredictable turbulence ?bursts? are the most important agents of atmospheric change on all space and time scales, limiting predictability to the level attainable by sophisticated statistical procedures. If this hypothesis approaches the truth more closely than present beliefs, then efforts to improve forecasting through parameterization and more detailed numerical models will fail. Research should concentrate on the character and distribution of turbulence bursts; in particular, observing systems should be designed to pick up the earliest possible stages of turbulence burst development. Only thus might forecasting be improved.