Fig 1.
Flow diagram illustrating the workflow of FISH software showing the different steps in the processing phase (steps 1–3) and the review phase (step 4).
Table 1.
School parameters and their descriptions for Fish-finder Image Segmentation Helper (FISH) outputs.
Fig 2.
Transducer placements on catamaran ski-boat (profile view of stern) for the recreation fish-finder (RFF) and the scientific echo-sounder (SES) showing athwartship displacement (2 m), transducer depths (0.5 m), beam angles and area of beam overlap (BO), not to scale.
Table 2.
Specifications of the scientific echo-sounder and the recreational fish-finder.
Fig 3.
Map of study area showing regular pelagic survey transect routes (dashed lines) around the African Penguin colonies on Bird and St Croix islands and calibration survey route (solid line).
Shaded areas denote the 95% kernel density foraging range of African Penguins provisioning small chicks on St Croix and Bird islands between 2008 and 2011 (extracted from [20]).
Fig 4.
Linear regressions showing the relationships between school desciptors, mean depth (top graph) and log transformed school area (bottom graph), for 36 matched schools from the recreational fish-finder (RFF) and the scientifc echo-sounder (SES).
Shaded areas denote 95% confidence intervals.
Fig 5.
Plots for three scenarios, 1–3, representing three schools selected from the recreational fish-finder (RFF) outputs with different pixel skewness values: a) 2D scatter plots showing the influence of different combinations of mean volume backscattering strength (Sv) starting values and colour step values on the difference in Sv values (Sv diff) between the RFF and the scientific echosounder (SES) outputs (colour scale bar—low values indicate optimal estimates), cross-hatch denotes optimal colour step values at starting values of -65 dB; b) least-squares regressions between volume backscattering coefficients () of the SES (transformed) and RFF outputs of the 36 matched schools using starting and colour step values for the RFF outputs as determined by the 2D scatter plot analyses, coefficients are given at the top of each plot and shaded areas denote the 95% confidence intervals; c) histograms showing the frequency of schools in 4 dB Sv bins for the SES-derived schools, the RFF-derived schools before application of a correction factor as predicted by the linear regression models (b), and the RFF-derived schools with this correction factor applied (RFF corr).
Table 3.
Summary statistics of mean volume backscattering strength (Sv) estimates for 36 matched schools as insonified by the scientific echosounder (SES) and the recreational fish-finder (RFF) using outputs for three scenarios representing different pixel derived outputs for the RFF system: IQR—interquartile range, p—Wilcoxon signed rank statistic probability estimates between the pairs of schools derived from different system outputs, i.e. between all RFF scenarios and the SES outputs.
Fig 6.
Least squares linear regressions showing the relationship between volume backscattering coefficients () and mean school depth for the 36 matched schools insonified by the recreational fish-finder (RFF) and the scientific echosounder (SES).
Shaded areas denote 95% confidence intervals.
Fig 7.
Coefficient of determination (R2) values of least squares regressions for 18 models comparing relative abundance estimates using the nautical area scattering coefficients (sA) (m2 nmi-2) between schools from the scientific echo-sounder (SES) and the recreational fish-finder (RFF) at different scales: 0.5 km–18 km, hatched vertical line denotes asymptote (12.5 km).
Table 4.
Frequency of fish school encounter scenarios of the recreational fish-finder (RFF) and the scientific echo-sounder (SES) quantified by 500 m Elementary Distance Sampling Units (EDSU).