Improving numerical weather prediction: error growth at the convective scale and speed
Date
2009
Journal Title
Journal ISSN
Volume Title
Abstract
Despite the continuous growth of the available computational power, it is undoubtedly beneficial, for both the research and operational communities, to increase the efficiency of Numerical Weather Prediction. Because parameterizations often occupy a significant portion of the total execution time the first focus of this work is to provide a methodology to transform parameterizations into algorithms that provide the same output at a fraction of the computational cost (i.e., transfer schemes). Several transfer schemes are developed for the Harrington radiation parameterization, in the clear sky case and implemented in the Regional Atmospheric Modeling System. The best one requires roughly 5% of the computational expense of the parent scheme. Accuracy is generally preserved and an analysis of the main meteorological fields after two days of simulations does not show significant differences. The differences for the 2 m temperature are larger than for the other fields, but still smaller than the differences introduced by a second common parameterization. A second area where NWP is in need of improvements is convective-scale forecasting. The advantages of more accurate forecasting derive from the high societal impact of convective events, which can be severe and lead to loss of life and property. Ensemble forecasting is an ideal tool to handle uncertainties in forecasts and the second aim of this study is to identify the processes that lead to error growth at the convective scale, for a case study over the United Kingdom using the Met Office Unified Model. The perturbation was applied to the potential temperature at a specific model level within the boundary layer, either sequentially (every 30 minutes) or at specific times. It was determined that acoustic waves are generated and can affect the background state. Vertical stability is also altered and occasionally lids can be set or removed. The unique boundary-layer scheme also contributes to error growth, by triggering different parameterizations as a response to the perturbation. Finally there are qualitative differences between high amplitude perturbations (1 K) and the smaller ones (0.01 and 0.1 K), but the root mean square error reaches similar values at saturation.
Description
Rights Access
Subject
convective scale
error growth
weather prediction
atmospheric sciences