Forecast sensitivity to the assimilation of observational data - two case studies for Australia

Australian weather forecasts use Numerical Weather Prediction (NWP) model output. Forecast accuracy is improved by assimilating a range of observational data which includes Australian Bureau of Meteorology station data. The significant investment by the Bureau of Meteorology in the national observing network, and the constant evolution of observational technologies, requires an ongoing assessment of the scientific value of the network components. Examining an objective measure of the impact of each assimilated observing system on the quality of short-term NWP forecasts can potentially guide planning and investment decisions related to network efficiency and effectiveness. 

Traditional techniques for assessing the impact of observations in NWP are inflexible (i.e. they require dedicated trials) and computationally expensive, but a widely used technique, known as adjoint-based Forecast Sensitivity to Observations (FSO), can provide forecast impact information continuously, flexibly, and in near real-time. We use archived FSO data to assess the relative forecast impact of in-situ data for different instruments and variables. We use two case studies to examine the impact of 1) three upper-air measurement instruments - radiosondes, aircraft, and a wind profiler - through the atmosphere at Sydney Airport, and 2) Automatic Weather Station surface observations along the Great Barrier Reef. These studies aim to provide network planners with information that can guide observations rationalisation decisions.