Notes about school detection

Corrections

When using a vertical echosounder, the horizontal and vertical dimensions of schools suffer distortion in the acoustic image due to the effects of pulse length and beam width: similar images may be produced by quite different schools, and conversely, the same school may produce widely different images at different depths.

The true thickness of a school is inflated when observed with the echosounder by the time taken for the trailing edge of the pulse to pass through the lower bound of the school, corresponding to a distance of one half pulse length. Echoview corrects for this effect by subtracting one half pulse length (ct/2) from the uncorrected measurement of school thickness (Reid and Simmonds, 1993; Barange 1994). This correction assumes that the ping data are high resolution data that fully defines the echo envelope. The School Detection module can be used to detect schools where data are of lower resolution (for example, using Simrad Q telegram data), but any corrections to thickness will have little effect because the error in thickness due to the binning of the data may be greater than the correction. Data in Sonic, BioSonics, Simrad sample Sv and Simrad sample power formats would in general have sufficient resolution.

The true length of a school is inflated by an amount determined by the echosounder's directivity function in relation to school echo energy and depth. School length is calculated from the point at which a portion of the school enters the acoustic beam where the number of fish is sufficiently large to generate an echo exceeding the processing threshold. At this point in time, the school will only be at the edge of the beam, but length is calculated from the center of the beam, resulting in over-estimation. Diner (1998) has empirically derived algorithms based on analysis of simulated schools (of ellipsoid shape and uniform density) to correct for this effect, based on the concept of the 'attack angle' (f): the angle from the beam's axis to the edge of the school where school detection begins; that is, where measured acoustic energy exceeds the threshold. The length in the acoustic image by which true length is increased at the start of the school is given by D x tan(f/2), and thereby the correction becomes 2 x D x tan(f/2) for error at both the start and end of the school. The equations given in School Detection module algorithms, allow the calculation of the attack angle from the nominal beam angle, school depth, and the difference between the school's mean acoustic energy and the processing threshold. See Diner (1998) for the derivation of these algorithms, since their explanation is beyond the scope of the present document.

Measurements of average back-scattered energy are also decreased by the effect of beam width, due to pings at the start and end of school insonification which are not entirely filled by school. Diner's (1998) algorithms additionally allow correction for this under-estimation of acoustic energy.

School length is underestimated by the fact that the vessel is unlikely to have passed over the center of the school. If the school is assumed to be circular in plan view, school width can be corrected for this problem by multiplying by a 4/p factor (a mathematical derivation can be found in Simmonds and MacLennan, 2005) This additional correction is not incorporated in the corrected length analysis variable in Echoview.

Warnings

It is advised that schools regions shouldn't extend beyond the data. This may occur when you detect schools then remove data files and add new data such that schools regions extend beyond the new data. In this case, the calculation of inconsistent Uncorrected_length and Uncorrected_thickness may have undesired effects on other schools analysis variables.

Echoview allows a school detection when GPS data is missing for some pings. A GPS extrapolation is used and is limited to 30 seconds beyond the last known position.

Cautions concerning school corrections

Diner's (1998) corrections can only be applied when the uncorrected length of the school, normalized in terms of numbers of beam-widths (see calculation given in School Detection module algorithms) exceeds 1.5. As length decreases below this threshold towards one, the correction for school length (2xDxtan(f/2) asymptotes to negative infinity.

The value Sv (mean energy - processing threshold) should be in the range of 10 to 30 dB re m-1 to reduce angular errors.

In assessing his corrections to school acoustic energy, Diner (1998) notes:

"For this parameter, the proposed corrections do not appear to be very efficient. In all cases, it does not appear to be necessary to correct the image energy value."

You are thus cautioned that the corrections to energy parameters performed by Echoview may not be necessary, and that the uncorrected measurements (also exportable) should perhaps be used, at least until these corrections are subjected to further testing and refinement.

You are referred to the acoustic literature to follow developments in this field and requested to keep Echoview Software informed of any new developments so that we can continue the enhancement of Echoview.

Additional possible calculations

Variables not available in Echoview

The following variables are not available in Echoview but they can be calculated from variables available in the School detection module:

  • Fractal dimension (Nero and Magnuson, 1989):

[ln(P/4)x2]/lnA.

An index of shape complexity, where a value of 1 represents a smooth or square outline and 2 a highly complex outline.

  • Elongation (Weill et al., 1993):

L/H.

An index of the length to thickness (height) relationship.

  • Unevenness1 (Weill et al., 1993):

[P-2(L+H)]/N.

An index of school shape unevenness, calculated as the difference between the observed perimeter and the perimeter of the rectangle that would be formed by the observed school thickness (height) and length, normalized to the number of pings in the school.

  • Unevenness2 (Weill et al., 1993):

P/[2(L+H)].

An index of the relationship between school perimeter and the perimeter of the rectangle that would be formed by the observed school thickness (height) and length.

  • Rectangularity (Scalabrin and Massé, 1993):

(L.H)/A.

An index of the ratio between the area of the rectangle that would be formed by the observed school thickness (height) and length, to the school's observed area.

Other school/region descriptors in Echoview

The Analysis domain optional export variables Region_top_altitude max, mean and min, and Region_bottom_altitude max, mean and min are similar to the descriptors for school bathymetric altitude above the bottom as discussed by Haralabous and Georgakarakos (1996).

The Integration results optional export variables Center_of_mass, Inertia, Proportion_occupied, Equivalent_area and Aggregation_index.

See also

School Detection module references
School Detection module algorithms
Using the School Detection module
Schools analysis variables
School detection while live viewing algorithm