Previous PageTable Of Contents

Assessing flood damage using SPOT and NOAA AVHRR data

Peter Worsley1 and John Bowler2

1,2NSW Agriculture, 161 Kite Street (Locked Bag 21), Orange NSW 2800
Corresponding author – Phone: 02-63913606, Fax: 02-63913767


The February 2001 floods in the Moree, Narrabri, Collarenebri area of northern NSW were particularly serious given the investment in both irrigated and dryland summer crops, and in particular cotton. Levees had been constructed around many areas to protect cotton crops from flood water. These levees have been very effective for protecting targeted areas, however flooding has been exacerbated in non-cropping areas through their construction. In order to assess both agricultural losses and the effect of levees, a combination of pre-dawn thermal AVHRR data and SPOT 2 data were used. Daily AVHRR data were acquired to determine the date that the flood peak moved through the major cotton areas. Two SPOT 2 super scenes were purchased for this date. The SPOT data proved suitable for mapping the areas affected by flood waters. The effects of levees were clearly visible and these data are now available for future planning of levee systems.


Severe flooding and droughts are a feature of the Australian landscape. Major flooding intermittently affects all Australian rivers to a greater or lesser degree. Depending on the timing and duration, major floods can cause millions of dollars of losses to agricultural produce and resources. The NSW Agriculture, Agricultural Protection Program regularly requires mapping and quantification of areas inundated by water. The only cost effective, accurate method of mapping broadscale flooding is through the use of remotely sensed data. Satellite data is cheaper per unit area than data acquired from aircraft. The cost and time required for processing of data collected from an aircraft precludes the use of anything except satellite data. A range of satellite data sources are suitable to varying degrees, including; Landsat (TM and MSS), SPOT (XS and Pan), Radarsat (SAR), MODIS, IKONOS and NOAA (AVHRR). When choosing a particular data source, a trade off between cost, spatial, spectral and temporal resolution has to be accepted.

The Department of Land and Water Conservation (DLWC) operate a network of river height gauging stations that offer a predictive capacity for down stream flood levels. In the Moree, Narrabri, Collarenebri districts, significant amounts of overland flow can occur. Gauging stations may not pick up this overland flow and thus down stream flooding predictive capacity may be compromised. In this very flat country, water often spreads for tens of kilometres, however it may only be 200mm to 300mm deep. It is important for flood level prediction and damage assessment to quantify these broad sheets of shallow water. Given that this water may move on to and off an area in a reasonably short period, it is necessary to map inundated areas as frequently as possible. Data are needed very regularly, preferably at least every few days. This means that either a satellite with a very high temporal resolution is necessary, or a multi-sensor approach was needed using a range of satellite data types. A multisensor approach using SPOT, Landsat, IKONOS and Radarsat data would produce a satisfactory temporal picture. However the high purchase cost and high demand placed on human and computer resources for pre-processing precluded the implementation of a multisensor approach. The NSW Agriculture, Resource Information Unit uses NOAA to map land inundated by water at a regional scale. The technique using pre-dawn thermal imagery has proven successful during past floods (Worsley and Pradhan, 1999), (Worsley, et al 1999a), (Worsley, et al 1999b), (Worsley, et al 2000). The 1.1km spatial resolution is the only drawback of using this data source. NOAA data were analysed each morning to determine the extent and location of overland flow and the timing of the main flood front moving through the major cotton growing areas. Once the date of the worst flooding in the cotton areas was determined, high resolution (SPOT) imagery was acquired for that date to determine the extent of crop damage and the effect of levee placement.

Daily overland flow monitoring

NSW Agriculture has a formal arrangement with the Queensland Department of Natural Resources (QDNR) to acquire NOAA data at short notice during emergency situations. QDNR routinely acquire data direct from the NOAA satellites via their receiver. When data are required, a live feed script is activated to electronically transfer the data to NSW Agriculture as soon as they are available. Data are supplied in a calibrated signed 16-bit georeferenced format. Bands 3, 4 and 5 are calibrated to absolute temperatures in 1/100ths of a degree, while bands 1 and 2 are calibrated reflectance values.

The contrast between water and land is greatest in images recorded shortly before dawn. At this point, the soil surface has cooled to its lowest point in the diurnal cycle. Water surface temperature remains relatively stable throughout the diurnal cycle as convection currents maintain a relatively stable surface temperature in water bodies (Sabins, 1996). Previous studies have shown overpasses close to dawn provide greatest discrimination between land and water in the thermal bands 3, 4, and 5 (Worsley and Pradhan, 1999).

The main flooding occurred during the first week of February 2001. Four cloud free pre-dawn images were acquired during this period. The area of interest was subsetted to improve classification accuracy. The 4 subsets were completely free of cloud, therefore an isodata classification was run on each to output a 50 class image.

The classified images were overlaid upon a false colour composite displayed as RGB 3,4,5. Classes were remapped to produce a binary image of water and non-water. The binary image was overlaid upon a false colour composite and rigorously checked for omission and commission errors. Hand masking was used to remove commission errors, which generally occurred as a result of high cirrus cloud.

Significant bodies of water were identified between the main drainage lines and away from gauging stations (Figure 1). The identification of these water bodies was an invaluable aid in the prediction of down stream flood levels. Analysis of these flood images revealed the most serious flooding in the cotton areas had occurred on the 4th February. A search of suitable high resolution imagery sources was conducted to suit this date. A moderately cloud free overpass was captured by SPOT 2 for the 4th of February, 2001.

Figure 1. Classified NOAA Layers for 4 Dates showing progression of the Flood Front.

Crop Loss and Levee Effect Monitoring

Given the high value of summer crops in the area, mainly cotton, a more thorough study of losses was required. As mentioned above, SPOT 2 acquired a cloud free overpass for the 4th February, 2001. Data capture for this overpass was performed in dual mode with the 2 sensors operating in multispectral mode (XS) and angled at -21.16 and -17.24 degrees. Two super scenes were acquired (approx. 270km long) one from each sensor. These were orthorectified prior to delivery by Geoimage Pty Ltd, Brisbane.

The high spatial (20m) and appropriate spectral resolution, Figure 2 (G,R,NIR) make SPOT XS a useful data source for mapping damage to individual cotton bays. Each super scene was classified using an isodata routine to produce a 200 class image. Each classified image was analysed with cloud classes amalgamated to create a cloud mask. This cloud mask was then applied to the original image. An isodata routine was run on this masked image, again 200 classes were output. The classified image was analysed and classes were reassigned to water, irrigated crop or land. The water, crop, land, cloud image, provides the capability to calculate areas of inundation on a property by property basis using the digital cadastre database in a cross tabulation analysis, Figure 3. Automatic recognition of inundated irrigated crops proved unsuccessful, even though these crops were reasonably easy to identify by eye. Visual interpretation was required to identify these inundated crops. Radarsat data may have been a valuable tool for the identification of crops standing in water.

The effects of levees can be seen quite clearly in the unprocessed imagery. Visual analysis of this imagery by field officers familiar with the area proved valuable. This peak flood imagery with a drainage line and levee overlay gave field staff a considerably better understanding of the hydrological processes operating within the flood prone farming district between Narrabri, Moree and Collarenebri. The imagery now serves as a valuable archive for use during future levee siting and design projects. In the future, as more accurate Digital Elevation Models (DEMs) become available, modelling the flow of flood waters without the imposition of levees, and comparing the result with the actual flow as created through the construction of levees, should provide valuable information on land use planning and design of levees and irrigation infrastructure.

Figure 2. SPOT XS Image showing the Flood Front as it moves west across the Narrabri Flood Plain

Figure 3. Classified SPOT XS Image showing Inundated Land, Irrigated Crops and Dry Land with a Property Boundary Overlay


NOAA data was a useful tool providing field staff with a rapid assessment of potential down stream flood levels. Significant bodies of water were identified away from the main drainage lines. This water between the DLWC gauging stations has the potential to significantly increase downstream flood levels over and above the levels predicted from gauging stations data. NOAA data used for this purpose is fast, cheap and effective. Mapping flood waters every 1 or 2 days during the flood using NOAA provided researchers with a dynamic picture of floodwater movement through this area, thus further enhancing their understanding of the hydrological processes operating within the subcatchment.

The mapping achieved from NOAA provided field staff with a good overall picture of flooding within their region. It was then necessary to use a high resolution data source to assess flood damage at a farm level. Analysis of flood levels on the NOAA imagery revealed the most suitable day (highest water level within the main summer cropping area) to acquire high resolution data.

The high resolution data source (SPOT XS) proved valuable for assessment of flood damage at a paddock scale. The effect of levees was easily seen, however a more thorough study using a suitable DEM to undertake some flood modelling would be valuable. A comparison of modelled flooding without levees against the actual flooding measured by SPOT would provide a valuable tool for the future design and construction of levees in this area.


Sabins, F.F. (1997) Thermal Infrared Images, In Remote Sensing : Principles and Interpretation. 3rd ed. , P161

Worsley, P.M, and Pradhan, U.C. (1999) Northern and Western NSW Flood Mapping – Spring 1998. Final Report. New South Wales Agriculture, Orange.

Worsley, P.M, Pradhan, U.C. and Bowler, J.K. (1999a) Mapping floodwater at a regional scale using NOAA Satellite data. In Proceedings of The International Rangeland Congress, Townsville Australia July 1999.

Worsley P.M., Bowler J.K., Freckelton D.A., McGowen I.J., Pradhan U.C., Roger R.E. and Worsley M.A, (1999b) The Use of Remote Sensing and GIS Technology by NSW Agriculture for Emergency Management. In Proceedings of The Australian Disaster Conference, Canberra, November 1999, p215-220

Worsley, P.M., Pradhan, U.C., Bowler, J.K. and Roger, R.E. (2000) Regional scale flood mapping using AVHRR data.. In Proceedings of 10th Australasian Remote Sensing and Photogrammetry Conference, Adelaide, August 2000.

Previous PageTop Of Page