Nasby Lucas et al.: Use of submersible transect data and multibeam sonar imagery for habitat assessment 



749 



for p-eenstriped rockfish, rosethorn rockfish, sliarpchin 

 rockfish, shortspino tliornyhoad. and Dover solo (between 

 3.7'r and 6'; ) and slightly higher for pygmy rockfish, yel- 

 lowtail rockfish, lingcod, sablefish, and rex sole (between 



7.8' r and 11.9%). 



Discussion 



A primary finding of our study was that distinct bottom 

 types found on Heceta Bank are distinguishable through 

 the use of sonar data and that these interpreted habitats 

 correlate with direct submersible observations of bottom 

 type. The determination of habitat information from sonar 

 data is significant in that it provides a broad-scale view 

 of the seafloor habitat, previously unavailable, and allows 

 a habitat-based groundfish assessment. Although seafloor 

 imaging and GIS techniques have previously been used in 

 the study of marine habitats (e.g. Meaden, 1999; McRea et 

 al., 1999;Sherin, 1999; Yoklavich et al., 2000), ours is one of 

 the first published studies where GIS technique was used 

 to combine a detailed analysis offish and habitat transect 

 data with broad-area high-resolution sonar seafloor imag- 

 ery and where total fish abundances were calculated for 

 large areas of the seafloor (see also O'Connell et al.^). 



Habitat type could not be determined by bathymetry or 

 backscatter data alone, but the information provided by 

 both data sets, in addition to groundtruthing by direct sub- 

 mersible observation, gives a clearer picture of the overall 

 habitat environment. The use of the backscatter data 

 combined with the bathymetric data has the advantage of 

 providing an indication of substrate type, which is clearly 

 important in fish-habitat associations. In general, however, 

 backscatter provides a better indication of habitat for fish 

 association purposes. Bathymetric data can provide geo- 

 logical structure on the resolution of five to ten meters, 

 whereas backscatter data give an indication of structural 

 variation on a smaller scale, which is of ecological impor- 

 tance in influencing the distribution of groundfish. For ex- 

 ample, the physical properties of a substrate influence the 

 types of invertebrates that colonize the seafloor, and local 

 relief can provide microhabitats for some fish. 



The extent to which a groundfish habitat can be ef- 

 fectively mapped by remote sensing is determined by 

 the resolution of the system used. In general, sonars are 

 optimized for specific operational depth ranges. A sys- 

 tem designed for very shallow water can have sufficient 

 resolution to provide contours of features or objects that 

 deeper water (lower frequency and longer range) systems 

 will only "see" as backscatter changes. As more sites are 

 studied by combining visual seafloor transects, high-reso- 

 lution sonar, and GIS techniques, it is likely the geologic 

 indices most relevant to groundfish habitats will become 



s O'Connell, V., and C. Carlile, and C. Brylinsky. 2001. De- 

 mersal shelf rockfish stock assessment and fishery evaluation 

 report for 2002. Regional Information Report lJOl-35, 43 p. 

 Southeast Region, Division of Commercial Fisheries, Alaska 

 Department of Fish and Game, P.O. Box 25526. Juneau. AK 

 99802. 



apparent. These methods should lead to a more coherent 

 approach to habitat-based stock assessments. 



One of the limitations of this habitat-based approach 

 to stock assessments has to do with strong reliance upon 

 the fish-substratum association. The distribution and 

 abundance of groundfish has been shown to be strongly 

 correlated with substrate type (see introduction), but fish 

 distributions and densities may vary with other factors as 

 well, such as depth, currents, nutrients, and food avail- 

 ability. In this study there was an attempt to address this 

 potential problem of over emphasizing the fish substra- 

 tum relationship through the grouping of habitats into 

 patches. The designation of habitat patches allowed the 

 grouping of areas of potentially similar biotic and physi- 

 cal characteristics. Thus, the use of patches as areas for 

 fish density estimates allowed for increased accuracy in 

 forming abundance estimates from the habitat-groundfish 

 association information. For example, this advantage was 

 apparent by high variance in fish density estimates among 

 patches of similar bottom type, such as the high density of 

 juveniles in one of the three rock ridge patches. Variations 

 in density in similar patch types of our study were also ob- 

 served for Dover sole, rex sole, and pygmy rockfish. These 

 patterns may be due to differences in depth, food avail- 

 ability or variations in percent composition of substrate 

 type in separate areas of the bank. The other benefit of 

 using habitat patches was that it allowed the testing of 

 new groundfish assessment methods without making pre- 

 dictions for areas of high uncertainty where submersible 

 transects were not performed. 



Hixon et al.'s study* is one of only a few comprehensive 

 habitat-groundfish studies available and has provided a 

 foundation for testing this new approach. Hixon et als 

 dataset provided invaluable habitat information, but had 

 several shortcomings. One problem was that of the incon- 

 sistency in positional data because of the use of Loran-C 

 ( GPS was not yet available ). Another problem, characteris- 

 tic of all submersible studies, was the overall limited spa- 

 tial sampling provided by the survey. The stations for the 

 study were chosen as representative habitats for Heceta 

 Bank from exploratory submersible dives conducted by 

 Pearcy et al. (1989) in 1987. However, not all of the rep- 

 resentative habitat areas were sampled because maps 

 of high-resolution bathymetry and backscatter were not 

 available at that time. Lack of complete habitat data made 

 it difficult to extrapolate bottom-type and fish-density data 

 to the entire bank. 



The use of observational data from submersibles for de- 

 termination of fish density, and extrapolation from these 

 data to total abundance within regional habitat patches, 

 were based on several assumptions. First, we assumed 

 that the areas sampled were representative of the entire 

 regional patch to which each transect belonged. Whatever 

 error was associated with this assumption, our approach 

 is certainly more accurate that any method that ignores 

 habitat variation. Second, we assumed that all fish within 

 each submersible transect were accurately identified 

 and counted. Identification of similar species of rockfish 

 can be problematic, and counts clearly become estimates 

 when dense schools are encountered. Third, we assumed 



