A FLOOD NOWCASTING SYSTEM FOR THE eTHEKWINI METRO.
Volume 1: Umgeni Nowcasting using Radar - An Integrated Pilot Study
Nov 2003
Report No 1217/1/04

Two studies were undertaken on Flood Nowcasting for the eThekwini Metro. The first which focussed on the Umgeni River was started in early 2001 (contract K5/1217), the second was started in early 2002 (contract K8/456) and focussed on the Mlazi River. Both were completed in mid 2003. The study comprises the two projects. They appear in two volumes describing the two parts.

Part 1: UMGENI FLOOD NOWCASTING USING RADAR - AN INTEGRATED PILOT STUDY

EXECUTIVE SUMMARY

The original Aims of the project (K5/1217) as stated in the project proposal were:

1. The primary aim of this study is to provide decision makers in Umgeni Water and the Drainage and Disaster Management Departments of Durban Metro (and eventually the Umgeni Catchment Management Authority) with the tools to become proactive rather reactive in the context of flood warning.
2. To pull together the outcome of previous research funded by the Water Research Commission in the areas of Radar Estimation of Rainfall, Space Time Modelling and Forecasting of Rainfall, and Integrated Catchment and Rainfall Modelling, for Flood Forecasting in a Real World Application.

In the year 2000, the Durban Metro Disaster Management Centre did not have any facility for anticipating floods except from emergency weather reports and forecasts. They typically found themselves reacting to information phoned in by people who had either experienced damage or who had noticed that flooding was occurring. In the year 2003 they have, in the Disaster Management Centre, a GIS display overlain by real time images of rainfall measured by radar at 5 minute intervals showing them where the rain is and has fallen. They also have information coming from the Shongweni and Inanda Dams indicating the river flows, in near real time, in the two major rivers affecting the Metro.

The people living near rivers have now got the potential for some warning about impending floods and the knowledge that the Disaster Management Group is working towards mitigating floods in their area in a proactive rather than reactive way. Major industrial developments have been established in Mgeni Park and in the Mlazi Basin. Some of these are strategic industries. With the flood forecasting capability in the Durban Metro Disaster Management Centre, 6 to 12 hour warning of an impending flood will enable industry to evacuate staff and perform controlled shut downs or take steps to reduce the damage to the sensitive plants. This is a far cry from September 1987 when the SAPREF Refinery and Mondi paper mill were closed for 10 days with serious economic consequences.

This report describes how the system which assists the Disaster Managers has been put in place. The components are meteorological, hydrological and hydraulic. The meteorological component comprises telemetering rain gauges and a radar feed from the SA Weather Service METSYS branch in Bethlehem. The hydrologic component comprises the modelling of the rainfall-runoff response of river basins from the small to the large, interpreted in a transfer function framework using a linear model developed in a previous WRC contract K5/1050: A Linear Catchment Model for Real Time Flood Forecasting. The hydraulic component comprises the work done in a subsidiary contract K8/456: Mlazi River Nowcasting to include Levels of Inundation, performed by Nokuphumulu Mkwananzi under the guidance of the Project Leader of both these contracts. This last will be reported separately in Volume 2.

Chapter 1 outlines in detail the tasks undertaken during the 2-year project which include:

• Accessing Rainfall Data: Ensure the reliability of the Durban radar. Understand and write software to handle MDV format data. Obtain the MDV data stream directly in real-time. Generate catchment masks.
• Forecasting rainfall fields and inclusion of the String of Beads Model: Set up a forecasting version of the rainfall-runoff model. Set up a conditional forecasting implementation of SBM.
• Validation and testing: Validation on “unseen” historical data (offline). Validation and testing in real-time (online). (This task was not completed because the real-time data were not available until after completion of the project.)

In addition to the original aims of the project (all of which were achieved except those relating to real-time data hydrological data acquisition system and its consequences, which has now been put in place in August 2003) useful work was done on rainfall forecasting and estimation. The forecasting of rainfields in real time gives good forecasts up to an hour ahead over a large area; the improved estimation comes from the merging of information from raingauges with that from radar. This work was done in place of the few undeliverable tasks.

Chapter 2 outlines the Flood Forecasting System and its components in some detail. Some of the innovations detailed there include descriptions of:

• Development of algorithms for merging of raingauge and weather radar estimates of rainfall over large areas - this provides good information in detail about where it is raining and is essential input to the decision making system
• Employment of GIS to capture real-time rainfall over sub-catchments as input to the rainfall/runoff model by integrating the optimal spatial rainfields
• Derivation of a short term rainfield nowcasting method to advect possible future rainfields in real time in order to anticipate where it will rain up to an hour ahead. This makes use of a space-time model of rainfields developed under a previously completed WRC project K5/1010 (Clothier and Pegram, 2002)
• Exploitation of the speed and efficiency of a linear transfer function rainfall/runoff model developed under a previous WRC project K5/1005 (Pegram and Sinclair, 2002) to make flood nowcasts - this was exploited in the parallel Mlazi study
• Establishment of a GIS-based information base in the eThekwini Metro Disaster Management Centre, giving instantaneous visual display of real-time information on whereabouts of storms (and their nowcasts) relative to suburbs, townships, rivers, roads etc

Chapter 3 summarises the deliverables and tasks achieved in the project,

In conclusion, the pilot project has been successful enough in its application in Durban to persuade the WRC to fund a new project K5/1429 which started in 2003: National Flood Nowcasting Initiative: Towards an Integrated Mitigation Strategy. This is being undertaken by personnel in SAWS:METSYS, DWAF & NDMC under the leadership of the Civil Engineering Programme of the University of Natal, DURBAN.