A natural phenomenon in the hydrological cycle is flooding. Flooding is necessary to replenish soil fertility by periodically adding nutrients and fine grained sediment; however, it can also cause loss of life, temporary destruction of animal habitat and permanent damage to urban and rural infrastructure. Inland floods can result from disruption to natural or man-made dams, catastrophic melting of ice and snow (jökulhlaups in Iceland), rain, river ice jams and / or excessive runoff in the spring.
Why remote sensing?
Remote sensing techniques are used to measure and monitor the areal extent of the flooded areas , to efficiently target rescue efforts and to provide quantifiable estimates of the amount of land and infrastructure affected. Incorporating remotely sensed data into a GIS allows for quick calculations and assessments of water levels, damage, and areas facing potential flood danger. Users of this type of data include flood forecast agencies, hydropower companies, conservation authorities, city planning and emergency response departments, and insurance companies (for flood compensation). The identification and mapping of floodplains, abandoned river channels, and meanders are important for planning and transportation routing.
Many of these users of remotely sensed data need the information during a crisis and therefore require "near-real time turnaround". Turnaround time is less demanding for those involved in hydrologic modelling, calibration/validation studies, damage assessment and the planning of flood mitigation. Flooding conditions are relatively short term and generally occur during inclement weather, so optical sensors, although typically having high information content for this purpose, can not penetrate through the cloud cover to view the flooded region below. For these reasons, active SAR sensors are particularly valuable for flood monitoring. RADARSAT in particular offers a high turnaround interval, from when the data is acquired by the sensor, to when the image is delivered to the user on the ground. The land / water interface is quite easily discriminated with SAR data, allowing the flood extent to be delineated and mapped. The SAR data is most useful when integrated with a pre-flood image, to highlight the flood-affected areas, and then presented in a GIS with cadastral and road network information.
Canada vs. International
Requirements for this application are similar the world over. Flooding can affect many areas of the world, whether coastal or inland, and many of the conditions for imaging are the same. Radar provides excellent water/land discrimination and is reliable for imaging despite most atmospheric limitations.
Case study (example)
Radarsat maps the Manitoba Sea:
The Floods of 1997:
In 1997, the worst Canadian flood of the 20th century inundated prairie fields and towns in the states of Minnesota, North Dakota, and the Canadian province of Manitoba. By May 5th, 25,000 residents of Manitoba had been evacuated from their homes, with 10,000 more on alert. The watershed of the Red River, flowing north from the United States into Canada, received unusually high winter snowfalls and heavy precipitation in April. These factors, combined with the northward flow into colder ground areas and very flat terrain beyond the immediate floodplain, caused record flooding conditions, with tremendous damage to homes and property, in addition to wildlife and livestock casualties. For weeks emergency response teams, area residents, and the media monitored the extent of the flood, with some input from remote sensing techniques. It is impossible to imagine the scale of flooding from a ground perspective, and even video and photographs from aircraft are unable to show the full extent. Spectacular satellite images however, have shown the river expand from a 200 m wide ribbon, to a body of water measuring more than 40 km across. Towns protected by sand-bag dikes, were dry islands in the midst of what was described as the "Red Sea". Many other towns weren't as fortunate, and home and business owners were financially devastated by their losses.
Insurance agents faced their own flood of claims for property, businesses, and crops ruined or damaged by the Red River flood. To quickly assess who is eligible for compensation, the insurance companies can rely on remotely sensed data to delineate the flood extent, and GIS databases to immediately identify whose land was directly affected. City and town planners could also use the images to study potential locations for future dike reinforcement and construction, as well as residential planning.
Both NOAA-AVHRR and RADARSAT images captured the scale and extent of the flood. The AVHRR sensors onboard the NOAA satellites provided small-scale views of the entire flood area from Lakes Manitoba and Winnipeg south to the North Dakota - South Dakota border. Some of the best images are those taken at night in the thermal infrared wavelengths, where the cooler land appears dark and the warmer water (A) appears white. Manmade dikes, such as the Brunkild Dike (B), were quickly built to prevent the flow of water into southern Winnipeg. Dikes are apparent on the image as very regular straight boundaries between the land and floodwater. Although the city of Winnipeg (C) is not clearly defined, the Winnipeg floodway (D) immediately to the east, paralleling the Red River at the northeast end of the flood waters, is visible since it is full of water. The floodway was designed to divert excess water flow from the Red River outside of the city limits. In this case, the volume of water was simply too great for the floodway to carry it all, and much of the flow backed up and spread across the prairie.
RADARSAT provided some excellent views of the flood, because of its ability to image in darkness or cloudy weather conditions, and its sensitivity to the land/water differences. In this image, the flood water (A) completely surrounds the town of Morris (B), visible as a bright patch within the dark flood water. The flooded areas appear dark on radar imagery because very little of the incident microwave energy directed toward the smooth water surface returns back to the sensor. The town however, has many angular (corner) reflectors primarily in the form of buildings, which cause the incident energy to "bounce" back to the sensor.
Transportation routes can still be observed. A railroad, on its raised bed, can be seen amidst the water just above (C), trending southwest - northeast. Farmland relatively unaffected by the flood (D) is quite variable in its backscatter response. This is due to differences in each field's soil moisture and surface roughness.
Did you know?
It is worth your while to pay attention to the polarization characteristics of the radar imagery that you are collecting. If your target is to map flooded versus dry land, then HH (horizontal transmit, horizontal receive) is a much better choice than (say) VV (vertical transmit, vertical receive) polarization. The HH imagery will produce a noticeably stronger contrast between these two types of surfaces, allowing greater accuracy in the mapped result.