Flood forecasts for New Zealand communities

 

Parking beside the River Leith proved precarious as floods tore through Dunedin in 2006. (Photo: NIWA)

Precipitation forecasts from NZLAM for the 6-hour period ending at 1800 NZST on 29 November 2006.

Quantitative streamflow forecasts for the Buller River at Te Kuha for the high-flow event on 29–30 November 2006.

Year after year, people across New Zealand face the threat of damaging floods. Martyn Clark, Ross Woods, and Richard Ibbitt are working on an early warning system that could help lessen the impact on lives and property.

Floods are New Zealand’s most frequent and costly hazard. The country is hammered by a diverse range of severe weather systems, and flood waters rage through communities somewhere in New Zealand every one or two years. Damaging storms and floods in the last decade alone include: November 1999 in Queenstown, Alexandra, and Balclutha ($51 million in insurance payouts); June 2002 North Island ‘weather bomb’ ($24 million, one person drowned); February 2004 in Manawatu, Wanganui, Taranaki, and Wellington ($116 million); July 2004 in Bay of Plenty ($18 million, 156 homes destroyed); May 2005 in Matata ($29 million); March and July 2007 in Northland ($12.5 million and counting). (Figures from Insurance Council of New Zealand website: www.icnz.org.nz/current/weather/)

NIWA scientists are developing and testing a nationwide flood early-warning system to give people time to prepare for floods and, if necessary, evacuate threatened areas. Given forecasts of heavy rainfall, NIWA’s flood early-warning system predicts what will happen to the rain after it falls. It predicts how much rain will be soaked up by the soil, how long it will take flood waters to rise, the maximum level of flood waters, and the duration of the flood.

Predicting heavy rainfall

The first component of NIWA’s flood-forecasting system is NZLAM (New Zealand Limited Area Model), a sophisticated numerical weather-prediction model that provides forecasts of heavy rain throughout New Zealand (see ‘From weather prediction to hazard forecasting’). The series of weather maps at right shows four independent precipitation forecasts for the 6-hour period ending at 1800 NZST on 29 November 2006. Each of the forecasts was produced at a different time, ranging from 48 hours (top map) to 12 hours (bottom map) before the period illustrated. There is strong consistency between the different forecasts; for example, note how all four NZLAM forecasts predict heavy precipitation (shown in red) on the western side of the Southern Alps.

Estimating soil moisture

The second component of NIWA’s flood early warning system calculates the water stored in the soil before the start of the flood. The hydrological model simulates evaporation and drainage from the soil and the fraction of precipitation that runs off directly to the streams. It provides an initial estimate of the water stored in the soil by adding the precipitation that falls on the land surface and subtracting the sum of evaporation, drainage, and runoff.

Then the model’s initial estimate of soil water storage is updated using all available observations of streamflow and soil moisture in a process known as data assimilation. Drainage from the soil, and the seepage of water from the soil to the streams, is much higher when the river basin is wet. The difference between the streamflow simulated by the hydrological model and the measured streamflow provides information for the computer model on how to improve the model’s initial estimate of soil water storage.

Translating the predictions: from rainfall to streamflow

The final component of NIWA’s flood early-warning system predicts how long it takes flood waters to rise, the maximum level of flood waters, and the duration of the flood. In this component the rainfall predicted by the localscale numerical weather prediction model is fed into the hydrological model to transform the forecast rainfall into forecasts of river flow.

The diagram illustrates four example forecasts for the Buller River for a highflow event on 29–30 November 2006 together with the actual, observed flow (black line). The forecast generated at 1800 on 27 November (blue line) predicts that flow will rise early on 29 November. The next forecast (0600 28 November; green line) predicts a flood peak, but it is well above what actually occurred. The streamflow forecast from 1800 on 28 November (orange line) is closer to observations, in terms of both magnitude and timing. Finally, the streamflow forecast generated at 0600 on 29 November (red line) correctly predicts the timing of the flood peak in the next several hours but still over-estimates the flood magnitude. These forecasts were all computed before the river began to rise at all. We will still see differences between the forecast and the flood – partly because of uncertainties in the precipitation forecasts, and partly because all models are simplified replicas of hydrological processes. In future, we will produce probabilistic streamflow forecasts that provide information both on the uncertainty in the forecast and the probability that flood waters will rise above critical levels.

Getting flood warnings out to the communities

Flood forecasts are important because they provide people with time to protect their property and to move away from dangerous flood waters. NIWA’s floodforecasting system can be applied anywhere in the country. It can give advance warning of floods in regions where no flood-forecasting systems are currently available, and can provide additional guidance on impending floods to regional flood and emergency managers in regions that have established floodforecasting capabilities. Speedy development and implementation of NIWA’s nationwide flood-forecasting system can save money and lives.

Pack up your car and head for the hills

  • Exceptional flood events can overwhelm our flood defences.
  • Damage and losses can be minimised through accurate forecasts of flood flows and levels.
  • A flood forecasting system could help emergency managers decide on responses such as sandbagging and evacuation.

Teachers’ resource for NCEA Achievement Standards or Unit Standards:
Geography Level 1 US5083, Level 3 US5095, US5097