It can be a challenge to manage the complex flooding risks in urban areas. With New Zealand’s topography and climate, many densely developed and vulnerable communities are often located in small, steep flood prone catchments. Those attempting to minimise damage to property and prevent injury or loss of life are increasingly looking for alternative solutions to expensive infrastructure.
Recent short-duration / high-intensity rainfall events in both Auckland and Wellington have led to notable flooding events resulting in inundation of habitable floors and disruption to the community. Weather forecasts based on global numerical models provide valuable advanced warning but have lacked the resolution in time and location to facilitate targeted operational responses to flooding. On the other hand, the telemetered rain gauge network that provides real time intelligence does not provide sufficient warning in small catchments with short time-to-concentration.
Rainfall radar is the best available technology for measurement of the spatial distribution and evolution of the short-duration / high-intensity rainfall events which have been a major contributor to flooding damages. In order to make better use of the available national rain radar observations from the NZ MetService network, the authors established a collaboration which has already generated a shared real-time GIS platform for regional radar-derived Quantitative Precipitation Estimation (QPE).
To extend the operational usefulness of the radar data and provide enhanced warnings in catchments with short response times, a Quantitative Precipitation Forecast (QPF) based on a radar echo extrapolation nowcasting method has also been implemented. The radar nowcast QPF is generated by the Short Term Ensemble Precipitation System or "STEPS" and provides ensemble estimates of possible rainfall distributions up to 2 hours into the future, every 7.5 minutes.
The existing Auckland Council / Wellington Water radar QPE product, which includes clutter suppression, attenuation corrections, advection-interpolation and gauge scaling provides the input data for STEPS. The millions of data points which comprise the radar QPE and radar nowcast QPF are combined on-the-fly in a web-based GIS portal, allowing for treatment of antecedent condition and intensity/duration accumulation-based alarming in target catchments. The system provides a platform to enable location specific operational response to the event as it occurs and ultimately alerting to at risk properties.
In order to characterise the operational performance of the catchment alarming system, the QPE analysis and radar nowcast QPF have been run in hindcast mode covering the last 2 years. The radar nowcast QPF is validated against the radar QPE product at stormwater-catchment scale, achieving the highest skill at shorter lead times. This is the expected result given the chaotic nature and limits of prediction of the evolution of convective systems. Case studies of the (hindcast) performance of STEPS for the recent flooding events are also presented and implications for operational hydrology and hydraulic modelling are discussed.