The western and central waters of Torres Strait north of 11°S were surveyed between the 10th and 13th of November 1996 using two suryey crews on two Partenavia 688 aircraft to minimise the chance of the population estimates being confounded by local movements of dugongs within the survey period. As in the 1987 and 1991 surveys, the area was divided into eight blocks on the basis of sampling intensity and transect placement. Transects were aligned in an east-west direction south of Buro (Tumagain) Island (9° 34'S, 142° 18'E), and north-south along the coast of Papua New Guinea. Transect lines were spaced 5' apart in Blocks 0, 1B, 3 and 4; and at intervals of 2.5' in Blocks 1A, 2A, 2B and 5. The survey design was determined by: (1) the boundariesof Austalian air space, (2) the known distribution of suitable dugong habitat, (3) the enduance of the aircraft from Horn Island, the only site in Austalian territory itr the region where aircraft fuel could be purchased, and (4) the aircraft time available for the survey. The design was the same as that used in the previous surveys except that: (1) of the seven eastern most transects in Block 1A, the two short transects were not flown and the remaining five were truncated at 9°10'S because we were unable to enter Papua New Guinea air space close to Daru, and (2) as in 1991, the survey intensity in Blocks 0 ard 1B was halved from that used in 1987 by increasing the interval between successive transects from 2.5' in 1987 to 5'. A total of 30,568 km² were surveyed in 1996. A global positioning system mounted in the aircraft facilitated prccise and accuste navigation. The aircmfl was frtted with a ladar altimeterfor accurateheight control. In order to increase repeatability,the survey was conducted only when the weather conditions were good (usually Beaufort sea State ≤ 3). Whenever possible, daily schedules were arranged to avoid severe glare associated with a low or midday sun. Block areas were estimated from 1:100,000 digitised topographic coverage (AUSLIG) using the ArcInfo GIS package. The areas of all islands were excluded from the block areas. The length of each transect was also estimated from these digitised maps. Survey methodology As in 1987 and 1991, we used the strip transect aerial survey methodology as detailed by Marsh and Sincair (1989a and b) andMarsh and Saalfeld(1989b) We chose to continue using this methodology rather than the line transect methods now routinely used for dolphin surveys(eg. Barlow et al. 1997) because: (1) consistent methodology is essential to a reliable time series,(2) Marsh and Saalfeld (1990) verified that the strip width used is sufficiently narrow to preclude detectable variation in dugong sightability across the transect, and (3) dugongs are generally more difficult to sight than dolphins. Dugongs are most often seen as solitary individuals or adult female-calf pairs in turbid water and exhibit cryptic surfacing behaviour.We therefore preferred to use a technique in which the observers do not have to take their eyes off the water to read an inclinometer.