RIASSUNTO
The availability and ease of collection of fish data is crucial for the quantification of this marine resource. For example, it can be used to determine breeding grounds and other key sensitive areas in need of protection. It also provides critical information for fisheries management and conservation. Visual census techniques have been used for many years to assess reef fish populations mainly because of their relative accuracy and cost effectiveness. The methodology is ideally suited to monitoring the abundance of coral reef fish as it allows for the collection of community level data without the disturbance inherent in other more destructive sampling techniques. In its simplest form the 3D belt transect method for visual census of fish populations involves a trained diver, equipped with SCUBA equipment, estimating the abundance of fish within a given volume of marine area, usually along a transect. A multitude of factors, including fish mobility and habitat complexity, have been shown to affect the precision of this survey technique. Furthermore, the results obtained using this method are influenced by differences in diver experience; ability to recognize fish species, estimating size, and actually counting large numbers of moving fish. Additional errors are likely to be introduced through observer bias and situations like diver decompression times and cold or rough weather conditions. Belt transect video surveys are also a standard procedure for collecting data on sessile and slow-moving marine benthos. They have also been shown to be effective for fish surveys using closed-circuit SCUBA, suspended cameras or tow fish. Sonar is possible from the surface, but deep swimming fish may be masked by fish swimming above them. In addition, the video transect as a sampling unit is easily incorporated into the overall design of a monitoring program since it encompasses established benthic data collection techniques. One powerful advantage is the possibility of archiving the raw video data. This allows for checking results as well as the ability to revisit the raw data to extract other information via newer or revised techniques. In 2005, studies were carried out in the North-east of Hong Kong, at the Hoi Ha Wan Marine Park. These studies included using an ROV to record video belt transect surveys to collect coral data. The results showed that similar techniques could be used to determine fish numbers in areas, particularly areas immediately above a coral reef. The coral data extraction method, via selection of random video frames and a grid analysis of each frame also lent itself well for extracting fish data. This paper explores and expands on the concept of using ROV's and AUV's to carry out 3D video belt transects for the collection of fish data in coral reef areas. It shows the technique allows for the standardized fish data collected to be compared both spatially and temporally. The paper also discusses an analysis technique, based upon the existing standard used for other video surveys. This method is also compared with other established fish data collection methodologies to confirm the suitability and to consider the limitations of this methodology.