The Coongie Lakes is an area of freshwater wetlands located in the Strzlecki desert of north-eastern South Australia. The annual flood pulse from the unregulated Cooper Creek is the principal supply of water to this area of wetlands and of fundamental importance to the ecosystems of the wetlands. Understanding the hydrological factors that affect the flood patterns in arid zone wetlands is vital in the study of the ecology of wetlands and when assessing the likely ecological effects of any upstream diversion of water from dryland rivers, such as Cooper Creek. The variability of the flood regime of dryland rivers requires that the flooding patterns from a range of flood pulses, with widely differing characteristics, be examined in order to best understand the subtle and complex interactions that are occurring. The use of the AVHRR data provides the only viable synoptic and temporal coverage that can map the floodwater patterns during current and recent historic, large-scale flood events.

the spatial and temporal distribution of floodwaters in the Coongie Lakes was mapped, using NOAA-AVHRR satellite imagery, for the period 1988-1990. Variations between images, not associated with the inherent albedo of the surface, were minimised by applying coefficients to the satellite data that corrected for the different path length of each image. These were measured for the centre of the image and this also minimised in-scene variations due to the wide view angle of the AVHRR. The images were not corrected for changes in atmospheric conditions between the different scenes. The effects of increased atmospheric haze levels and/or highly off-nadir view anlges on some images was found to be quite significant and compromoised the utility of the affected images in mapping out the distribution and movement of floodwaters.

The floodwaters in the wetlands were identified using the ratio of NIR/Red<1 ('Ratio criterion') and also a NIR Threshold technique ('Threshold criterion'). The NIR class ranges were selected following the single channel unsupervised classification of a representative subset of images. Three spectral classes were defined (0-16%, 16-23%, 23-30%) and were applied to all images. The floodwaters classes principally corresponded to the percentage of water surface within the pixel. A relationship was established between the spectral classes and depths of the floodwaters by calibrating the measured surface area of lakes against their known volumes. The accuracy of the AVHRR data in measuring the inundation was also evaluated by comparing selected images with Landsat MSS images and aerial photographs. The AVHRR data underestimated the surface area of flooding by 20-30% using the Ratio criterion and 15-20% range using the Threshold criterion. However, the latter criterion could grossly overestimate the area of flooding in regions of scattered water bodies and dark albedos. Comparing the volumes within the wetlands calculated using the satellite data and those measured by Cullyamurra gauge station in each of the ghree years of flooding (1988-1990) gave an average underestimation by the Ratio criterion of 51% and by the Threshold criterion of 35%. These underestimations also were caused by water losses other than by evaporation and so overstate the error range for the volume calculations.

The range of flood pulse sizes observed during the study period were divided into five classes based on ranges for the peak daily flow amplitude and total volume of the flood pulse. The extent of flooding for each of these classes was defined. The monitoring of the movement of a range of flood pulses through the wetlands also allowed the identification of the main controls on the patterns of inundation. The pattern of flooding within the wetlands was most influenced by the amplitude (peak and at flood head), shape and total volume of the flood pulse. Other critical factors included the pre-existing water volume in the wetlands, the flow rate, catchment source of the flooding and the floodplain volume capacity upstream of important geomorphological, flow-regulating features.

Four Landsat TM images covering a period of seven years from 1987 to 1994 were used to analyse changes in land cover over time of an area of coastal wetlands in Thailand. Each of the images was classified according to the NOAA Coastal Change Analysis Program (C-CAP) classification scheme. Post classification change detection was then carried out on these images, and changes compared both quantitatively and qualitatively. In addition, changes in NDVI over time were also compared for each of the images.

The post classification change detection revealed a dramatic drop in area for the marsh. These changes were the result of either encroachment of other land cover types into the marsh or of the drying out of the marsh. Within the marsh, palustine short emergent plants also experienced a significant drop in the area, while tall emergent plants increased in area. Other noteworthy changes were increases in area for both shrimp farms and mangroves, and a large drop in area for croplands.

NDVI change detection showed an increase in NDVI values for the marsh and orchards. Streams north-west of the marsh had even greater increases in vlaues. Decreases in NDVI values occurred in forested mountainous areas, areas where shrimp farms have intruded into the marsh, areas of palustrine submerged plants, and upland areas surrounding the streams north of the marsh. Most of the increases in NDVI took place in the 1992-1994 image. This could have been the result of much lower than average rainfall figures for 1992 and 1993 affecting the moisture levels in the area.