A cycled sensitivity observing system experiment on simulated Doppler wind lidar data during the 1999 Christmas storm ‘Martin’

Abstract

In a companion paper in this issue Sensitivity Observing System Experiment (SOSE) has been introduced as a new method to assess the potential added value of future observing systems for Numerical Weather Prediction (NWP). There, SOSE was introduced as a single cycle experiment, meaning that additional synthetic observations, to extend the existing global observing system (GOS), are applied in only one assimilation cycle. In this study SOSE has been extended to run over three subsequent days to enable impact assessment of additional prospective observations over a prolonged period prior to an event. This is achieved by a cycled implementation of the SOSE method where analysis adaptations from previous cycles evolve progressively in subsequent cycles. This implementation of a cycled SOSE results in a sequence of pseudo-true atmospheric states that are subsequently used for the simulation of prospective extensions of the existing GOS. A cycled SOSE has two attractive properties as compared to the single-cycle implementation (i) the resulting pseudo-true atmospheric state at the end of the cycling period, that is, at forecast initial time, provides a better forecast and (ii) the cycling implementation makes the SOSE method more suitable for absolute impact assessment of continuously operated observing systems such as from polar satellite platforms. The NWP case investigated concerns the 1999 Christmas storm ‘Martin’ that caused much havoc in Western Europe. We show that additional observations from a spaceborne Doppler wind lidar over a 3-d period would have improved the 2-d forecast of ‘Martin’ substantially. This is substantiated by a 50-member ensemble forecast run.