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



743 



bathymetry and backscatter and the submersible transect 

 data included bottom type characteristics and fish density 

 data. In order to represent the dive transects in GIS as 

 linear features displaying changes in habitat and fish 

 density, a dynamic segmentation data structure was used 

 (ESRL 1994). 



Bottom type and fish density data to be displayed using 

 dynamic segmentation were derived from transect observa- 

 tion data. Transects were divided into segments by uniform 

 bottom type. Fish density was calculated along each seg- 

 ment of habitat tjT)e by using the data for the most common 

 species observed, accounting for 90"^ of the total, plus a few 

 rare species of commercial importance (i.e. lingcod, sablefish, 

 Dover sole and rex sole). This complex of species consisted of 

 a mixture of demersal and benthopelagic species. The follow- 

 ing species were assessed: juvenile Sebastes sp. (unknown 

 juvenile rockfish), Sebastes chlorostictiis (greenstriped rock- 

 fish), Sebastes wilsojii (pygmy rockfish), Sebastes helvomac- 

 ulatus (rosethorn rockfish), Sebastes zacentrus (sharpchin 

 rockfish), Sebastes flavidus (yellowtail rockfish), Ophiodon 

 elongatus (lingcod), Sebastolobus alascanus (shortspine 

 thomyhead), A?iop/opo/nn fimbria (sablefish). Microstomus 

 paclficus (Dover sole), and Errex zachirus (rex sole). Tlie 

 density of fish (number per hectare) was calculated by 

 taking the number of fish sighted in that habitat segment, 

 dividing by the area of the habitat segment in meters, and 

 multiplying by 10,000 square meters per hectare. 



The use of dynamic segmentation data structure al- 

 lowed for the display of changes in bottom type and fish 

 density data within the transect lines. This was done by 

 creating a "route" system in Arclnfo from the dive transect 

 data and associating it with an "event table." The event 

 table consisted of bottom type and fish density data, and 

 their corresponding locations along the transect, and a 

 route-identifier number to link the information to the 

 corresponding transects in the route system. For visual 

 display, bottom type segments were combined into three 

 major habitat groups: 1) mud, which consisted of "MM" 

 observations, 2) rock ridge, which consisted of "RR" obser- 

 vations, and 3) mixed substrate habitat, which consisted of 

 combinations of all other bottom type observations. 



In order to combine the sonar and submersible data 

 sets, all segmented dive transect data were then re- 

 projected with a 500-meter offset to the east. This was 

 determined to be the best correction for discrepancies 

 between the transect position data which were acquired 

 by Loran-C and the sonar data which came from GPS posi- 

 tions. It was determined that this offset was necessary by 

 comparing the two data sets and matching depth contours 

 and borders of well-defined habitat, specifically interfaces 

 between the mud and rock features of the bank. There 

 did not appear to be a significant north-south offset, al- 

 though this effect was more difficult to determine because 

 the submersible transects did not cross any well-defined 

 north-south boundaries. Transects segmented by both 

 bottom type and fish density were overlaid on the sonar 

 data in AxcView (Fig. 4, A-C). 



Assessments of fish abundance within large habitat ar- 

 eas were performed by selecting patches of relative habi- 



tat homogeneity on the sonar map around the location of 

 each submersible transect. These patches were chosen by 

 examining both patterns in the backscatter values and 

 topographic features indicated by the backscatter and 

 bathymetric data. In areas of mud off the bank, borders 

 were chosen by maintaining constant depth as well as 

 equal distance from the bank. In selecting patches to 

 represent areas of similar habitat, the boundaries were 

 relatively well defined in areas of rock and mud, but for 

 mixtures of sand, cobble, pebble, and boulder, it was more 

 difficult to distinguish distinct boundaries and therefore 

 these patch borders were drawn conservatively. 



Using the observational data from the transects from 

 all three years, we were able to characterize each habitat 

 patch by percent bottom type, density of fish, and esti- 

 mated abundance of fish as extrapolated from the dive 

 transect data contained within that patch. The grand 

 mean density and standard error for each species was de- 

 termined by using a weighted density for each habitat seg- 

 ment based on a proportion of the length of that segment 

 to the overall transect distance within that habitat patch. 

 The grand mean density was calculated as 



* = X'^'^" 



(1) 



where x = x; 



d= density offish within a segment of continuous 



bottom type; and 

 p = bottom type segment length/total transect 

 length within the patch. 



This calculation used the associations of the fish species 

 with substrate type and weighted its contribution to the 

 overall density by the comparative length of that segment. 

 Total fish abundance for each habitat patch was deter- 

 mined by multiplying the area of the patch and the grand 

 mean density and standard error of each species. The total 

 abundance for each species for all habitat patches was 

 determined by adding the abundance for that species for 

 all eight habitat patches. The standard error for the total 

 abundance for each species was determined by calculating 

 a grand mean standard error weighted by using the stan- 

 dard error of each habitat patch multiplied by the propor- 

 tion of the abundance of that species for that habitat patch 

 to the overall abundance for that species. Total abundance 

 standard error was calculated by using Equation 1 where 

 X = SE; d = standard error offish abundance within a habi- 

 tat patch; and p = abundance within that habitat patch/ 

 total abundance in all eight patches. 



Results 



Comparing the submersible data with the sonar data, 

 we found that there was high correlation between the 

 direct observations of bottom type and the habitat type 

 indicated by the sonar data. Side-lit bathymetry revealed 

 areas of outcropping substrata and the backscatter data 



