Abstract
Between 17 and 22 April 1990 widespread flooding occurred throughout Queensland, New South Wales, and Victoria, with several record rainfalls recorded. This paper investigates the forcing of the rainfall, assesses the quality of the operational numerical weather prediction model guidance during this period, and presents some numerical sensitivity experiments both to diagnose forcing mechanisms and to identify reasons for model weaknesses. All diagnostics used in this study could be made available to forecasters in real time, and one aim is to show what information could be used to enhance the forecaster's understanding, and thus confidence in, the numerical forecast guidance.
The event can be divided into three phases, all associated with an evolving upper-tropospheric cutoff low. In the first phase, the Queensland rainfall was focused by a broadly divergent upper-level regime and a lower-tropospheric trough, which is shown to be partly topographically induced. A plentiful moisture supply was present in the northeasterly flow around a higher-latitude anticyclone. This phase of the event was well predicted by the numerical forecasts.
The second phase of the event was the development of a surface cyclone over New South Wales. This was formed below an area of increasing upper-tropospheric divergence associated with the negative tilting of an upper closed low. On the scale of the forecast model, this was also well predicted; however, it is also shown that mesoscale detail identified in subjective postanalyses could be forecast using a higher-resolution NWP model.
The third stage of the event was the extremely heavy rainfall over southeastern Victoria. This was poorly forecast by the operational model; however, it is shown that higher-resolution forecasts that include enhanced topography and information from a high-resolution sea surface temperature analysis produced a dramatic improvement in forecast quality.
