On the relative contribution of wind and mass observations in numerical weather forecasting

Abstract

The impacts of wind and mass observations on analyses and forecasts are studied. Idealized perfect and complete observations are examined. A realistic three-dimensional data assimilation technique is considered. Spatial structures of analysis errors resulting from assimilation of wind or mass observations are theoretically investigated. Forecast impacts are assessed using the geostrophic adjustment theory and spatial probabilistic error models. Numerical predictability experiments are conducted to verify the theoretic conclusions. The main result is that, on average, wind observation are more important than mass observations, especially in the tropics and in the upper troposphere and stratosphere. In the mid and high latitude lower troposphere, the wind and mass observational data appear to be of comparable utility.