Stefano Migliorini, Met Office

Improving the impact of EO data on moisture and cloud forecasts

William Bell (Met Office), Andrew Lorenc (Met Office)


Satellite data are the most important observation type used for improving numerical weather prediction, both in terms of number of assimilated observations and of impact on short-range forecast error reduction as measured by adjoint-based indicators. It is, however, still very challenging to improve satellite data use in the presence of cloud, where total moisture measurements are most important to constrain the model’s representation of the evolution of weather systems and the large-scale circulation effects of latent-heat release. The difficulties are related to the realism of cloud and precipitation (large-scale and convective) parametrizations and their linearizations, which may lead to large discrepancies between observations and predictions; the nonlinear dependence of observation operators on cloud, temperature and humidity; the univariate/multivariate and non-Gaussian nature of forecast errors depending on relative humidity values; shortcomings in radiative transfer modelling. Despite these difficulties, over the last few years the major numerical weather prediction centres have been making significant progress in model parametrizations, data assimilation and radiative transfer modelling to make the assimilation of microwave remote sensing data in the presence of cloud either a reality or a closer goal. At the Met Office, the introduction of a nonlinear humidity transform in the minimization used in data assimilation has proved beneficial in improving the predictions of humidity-sensitive observations. Still, more work is needed to better constrain the partition between humidity and cloud data assimilation increments to be used as input to observation operators. In this talk, the strategy and the first attempts towards the assimilation of clear-sky and cloud-affected (or all-sky) microwave radiance data are discussed.

File available here