Satellite Remote Sensing: Operational Flood Monitoring: A Reality with RADARSAT?
The recent Red Tiver flood in Canada's prairies provides the chance for RADARSAT to prove its flood extent monitoring capabilities.
By Scott Paterson, Terry Pultz, Ron Saper and Yves Crevier

Flooding in North America causes over $1 billion in damage and is responsible for several dozen deaths every year. Emergency personnel who plan flood relief operations and ensure the safety of the local population often lack reliable information for decision making. Accurate and timely flood extent information can help emergency personnel make better decisions on where to deploy resources, how to plan evacuations and can help provide damage assessments. Making the correct decisions during a flood is critical for minimizing the loss of life and injury, reducing flood damage and reducing operational costs.
      Until the November '95 launch of RADARSAT, Canada's Synthetic Aperture Radar (SAR) satellite, no affordable, reliable data source existed which could provide synoptic information for monitoring flood extent. RADARSAT can provide a cost-effective, birds-eye-view every 2-3 days, in all weather and cloud conditions - a powerful combination for flood monitoring.
      Less than one month after commissioning, RADARSAT proved its value for flood monitoring. A massive three week spring flood in the Canadian prairies provided the opportunity. RADARSAT collected imagery every few days of the flood from space, and illustrated how flood evolution can be monitored.
      Although RADARSAT imagery contains important information on flooded areas, emergency personnel want real-time information, not pretty imagery. To meet the needs of operational users, Vantage Point International Inc. (VPI) of Ottawa, Canada, is developing the RADARSAT Flood Monitoring Workstation (FMWS), a system that will extract, format and deliver flood boundary information within four hours of acquisition.

Case Study: Monitoring Flooding in the Canadian Prairies
RADARSAT's first major test for flood monitoring occurred in the spring of 1996. Flooding started on the U.S. end of the Red River (North Dakota) in mid-April and moved northward into the Canadian province of Manitoba inundating small farming communities and turning farmersÕ fields into lakes. At the peak of flooding, the width of the Red River ballooned from several hundred meters to 15 kilometres. In parts of the Red River Valley, a canoe or a rowboat was the only available mode of transportation.
      Manitoba's Water Resources Branch predicted flooding for most of Manitoba's waterways as early as January, months before flooding. Heavy winter snow accumulation and wet autumn soil conditions provided advance warning of potential flooding. Flood forecasting information enabled planning for RADARSAT image acquisitions to monitor the Red River flood weeks ahead of the actual flooding. Over the three-week flood, RADARSAT imaged the Red River eight times, showing the build up and recession of the flood.
      On March 16 small farming towns along the Red River began building walls of sand bags around their homes to help protect against the rising river. By May 1, the Red River peaked at 19 feet above the winter ice level, inundating about half a million hectares of farm land. Although flooding to the south of the city of Winnipeg was predicted, ice jams that choked the flow of water into Lake Winnipeg, north of Winnipeg were not predicted. Residents along the Red River north of Winnipeg awoke to find their homes surrounded by water and chunks of ice. Ice jams are particularly dangerous because they occur very quickly and without any warning.
      RADARSAT images collected on April 25 and May 9 (Figures 4 and 5 respectively), show excellent contrast between the multi-toned agricultural fields and the dark flooded areas. Even flooded vegetation along the banks of the Red River can be mapped because of its bright appearance on the imagery. Changes in the flood extent are easily seen by comparing the April 25 and May 9 images. The bright rectangular feature in the center of the dark flooded area in the imagery is the town of Morris, protected by a levee. To map the flood stage and inundated areas over this period, a color composite image (Figure 2) was created using the April 25 and May 9 flood data and a winter image from March 23 representative of low flow conditions. The red areas on the color composite (west side of the image) represent flooding on April 25 only, while the blue-green areas (north of Morris) were flooded areas on May 9 only. Dark blue tones on the color composite are areas flooded on both April 25 and May 9.
      Figure 3 shows a prototype FMWS product from the Red River flood. FMWS products have vectorized flood boundaries and are geo-referenced in NAD83 (North American Datum of 1983). The flood extent on April 25 is outlined by pink vectors, whereas the yellow vectors show the flood extent for May 9.

The Flood Monitoring Workstation
RADARSAT can provide imagery to monitor floods, but operational users do not want raster imagery, they want products that their existing systems can use. To provide these friendly products, an end-to-end system is required which can extract and deliver a product to emergency field personnel within a few hours.
      The Flood Monitoring Workstation (FWMS) will monitor river boundaries during periods of seasonal flood risk, producing boundary maps, flood extent statistics, and text advisories every few days. Output flood extent products must be in formats easily interpreted and ingested by the end user, and distributed in a timely and accessible method (e.g., Internet and direct-to-fax output).

Summary
RADARSAT has enormous potential to assist with flood operations. The Red River flood in Canada's Prairies provided the opportunity for RADARSAT to prove its flood extent monitoring capabilities. An automated system is required to extract, format and deliver the flood information to operational field sites where it is needed. The Red River flood data is being used as the baseline dataset to aid with the development of the Flood Monitoring Workstation.

About the Authors:
Terry Pultz is with the Canada Centre for Remote Sensing, Natural Resources Canada, Ottawa, Ontario. Yves Crevier is with Intermap Technologies Ltd., Ottawa, Ontario. Scott Paterson and Ron Saper are with Vantage Point International, Ottawa, Ontario. Vantage Point International Inc. (VPI) is an engineering consulting company dedicated to helping customers use advanced radar technology and data to solve real-world problems.

Back