Improving predictions of hazardous weather

Stefano Migliorini, Ross Bannister, Mark Dixon

Achieving better forecasts of high-impact weather is currently one of the main challenges for operational meteorological centres. The motivation is that the occurrence of "significant" weather events is expected to increase in the near future due to climate change. The aim of this research is to investigate strategies for implementing an Ensemble Kalman Filter (EnKF) based data assimilation system for a high-resolution version (4 to 1.5 km) of the Unified Model.

At such resolutions (particularly at 1.5 km) it is possible to resolve convection and avoid relying on its sub-grid scale parametrization. High resolution observations, such as radar or geostationary satellite measurements can also be properly modelled and assimilated in the model. This can potentially lead to improvements in forecasts of severe convective storms, which may lead to hazardous events such as flooding.

But why investigate ensemble methods when operational centres (such as the Met Office) have already available variational data assimilation systems? A possible reason is that the balance relations used to model forecast error covariances in variational data assimilation for synoptic scales are not necessarily suited for mesoscale flows. A possible way forward is to adapt the system used for synoptic scale data assimilation and consider balance equations that are more suited for modelling forecast error covariances on the mesoscale (i.e. representing higher order Rossby number descriptions of the flow), by e.g. allowing for gravity waves.

A possible alternative approach for achieving better initial conditions for high-resolution forecasts is to exploit a sequential data assimilation technique such as the EnKF. With this technique forecast errors are directly estimated at each time step, starting from some initial estimate of them, possibly provided by the synoptic model.

The EnKF has advantages wrt variational methods, such as the ability of estimating flow-dependent forecast error covariances, but at the same time has limitations in fitting a large number (i.e. larger than the number of ensemble members) of accurate observations (Lorenc, 2003). However, recent research (e.g. Snyder and Zhang, 2003) has demonstrated that the EnKF approach can achieve successful results when used to assimilate radar data at convective scales.

Lorenc, A. C., 2003: The potential of the ensemble Kalman filter for NWP - a comparison with 4D-Var. Quart. J. Roy. Meteor. Soc., 129, 3183-3204.
Snyder, C., and F. Zhang, 2003: Assimilation of simulated Doppler radar observations with an ensemble Kalman filter. Mon. Wea. Rev., 131, 1663-1677.

You are visitor number [an error occurred while processing this directive] to this page.


Page navigation