Single Echo Detection (SED)

Measures of TS of the organisms present are useful for scaling echo integration data to absolute abundance estimates and for interpreting the observed echoes.  However, these measures are biased if fish are not sufficiently separated to be observed as single targets.  The analysis software includes algorithms for filtering out single echoes.  The most commonly used is derived from Soule et al. (1997).  Single targets are detected following these general steps:

1. peak amplitudes are selected above a threshold (Single echo detection threshold- SEDT);
2. the echo width (either time- or range-based) is measured, and;
3. the echo width is compared to the pulse duration.
4. phase jitter (angle standard deviation) should be smaller than a certain value
5. the calculated beam compensation should be smaller than a certain value

The threshold in step 1 should be small enough to observe the lower range of the TS distribution. Examination of this distribution is the basis for setting analysis thresholds in both the S_v and TS domains (see below).  The chosen threshold for analysis will be higher than the initial SEDT.  In addition, Sonar5 provides a different approach for detecting traces (cross-filter detector, Balk and Lindem 2000) which can be very useful in echo.

For the Great Lakes, we recommend using an initial SEDT of -75 dB, a maximum beam compensation of 6 dB (two way, 3dB one-way), an echo duration of 0.5 to 1.5 times the pulse duration, and a maximum standard deviation of both angles of 0.6.  These are similar to software default settings except that the default lower TS threshold is often set at -50 dB, which is too large for Great Lakes applications.  The limits for echo duration are also slightly wider than default suggestions which is only important when using short pulse durations.  The choice of acceptable echo duration for single targets is partly dependent on the shape of the pulse and it may be possible to make those limits more stringent with newer transducer and longer pulse durations.  In shallow lakes or in lakes with few targets (e.g. Lakes Superior and Huron), we recommend using a larger beam compensation to increase the number of detected single fish echoes.  For example, we got a 1 dB difference when beam compensation was increased from 3 to 12 dB with our 70kHz split beam sounder (Table 9).  This potential bias has to be weighed against the increase in the precision of in situ TS measures due to more identified targets.

Different echosounder manufacturers and post-processing packages may apply single target detection methods differently.  It is important to understand the specific methods used to ensure that they are consistent and comparable with other studies.  Echoview presents two methods (method 1 and 2) and a Simrad and a Biosonics single fish filter.  They give slightly different TS values although the difference is small.  Even so, use the same method for analysis as used during calibration.  Check the software for recommended method with different echosounders.

A survey from Onondaga Lake collected at 70 kHz with a Simrad EY500 split-beam unit (0.2 msec pulse duration, 11.4^o beam width, Table 9) provides an example of single-fish detection settings.  The open-water fish population was dominated by one age-group of alewife with an average length of 148 mm.  Echoview suggests using single-target detection method 1 for EY500 data.  When method 1 and 2 were compared, differences were small—in the order of 0.1 dB.  Since this difference is present also in calibration data, the two methods are essentially identical after correcting for differences in calibration.  We present data only for method 1  in Table 9.  This survey had relatively high noise levels and some noise spikes were present in water deeper than 10 m.  These spikes were often accepted as single-fish echoes when the angle standard deviation was high resulting in a decrease in in situ TS of more than 1 dB.  When the analysis was restricted to water depth of 2 to 10 m, the effect was less.  Mean TS increased from -42.55 dB for an angle standard deviation of 0.6 to -41.73 dB for an angle standard deviation of 5 (both at 6 dB beam compensation, Fig. 28).  This difference of 0.82 dB is equivalent to a 20% difference in fish abundance.  Mean in situ TS also increased with higher beam compensation (Table 9), but this increase was small.  For this survey, the mean TS was -42.58 dB with 3 dB beam compensation and -42.41 dB with 12 dB beam compensation (at a angle standard deviation of 0.6); a difference of 0.17 dB (and a 4% difference in estimated fish density).  The effect of changing an acceptable lower limit for the normalized pulse length from 0.6 to 0.8 times the initial pulse length was a 0.3 dB decline in in situ TS and a six-fold decrease in number of accepted targets (from 3000 to 500).  Decreasing the upper pulse length limit from 1.5 to 1.2 had no effect.  Although a difference of less than 0.8 dB in mean TS may be considered relatively small, such differences do result in an up to 20% change in estimated fish density, which is of similar magnitude to several other components of uncertainty associated with acoustic surveys (Simmonds et al. 1992).   

Another difference between EV SED methods 1 and 2 is the application of the TS minimum threshold.  EV SED method 1 applies the minimum threshold to uncompensated targets, resulting in the loss of small targets that are above threshold after compensation for position in the beam.  Conversely, method 2 accepts any targets are above the minimum threshold after compensation.   If SED is performed in EV using method 1, we recommend using a TS minimum threshold 6-10 dB lower than the desired TS minimum threshold and then applying a data threshold to your compensated targets.   If method 2 is used for SED, the desired compensated target strength threshold may be used directly in SED without a data threshold.

Table 9.  Mean TS calculated for targets larger than -60 dB in the 2-10 m depth layer using different single-echo detection (SED) settings.  The data are based on a survey with at 70 kHz (11.4o, 0.2 msec pulse duration) in Onondaga Lake, NY, May 2005, when an age-3 alewife year class dominated and constituted over 95% of the catch).  Data analysis is with Echoview, method 1.  ΔTS is the difference in mean TS for targets > -60dB compared to standard recommended settings (row 2), which had a mean TS of -42.55 dB. Angle variance is given in mechanical degrees, echo length is a multiplier of pulse length.

Beam compensation (dB)	Angle variance	Minimum echo length	Maximum echo length	Number of targets detected	Mean TS > -60dB (dB)	ΔTS (dB)
3	0.6	0.6	1.5	1567	-42.58	-0.03
6	0.6	0.6	1.5	2976	-42.55	0
9	0.6	0.6	1.5	4177	-42.44	0.11
12	0.6	0.6	1.5	5235	-42.41	0.14
6	2	0.6	1.5	4815	-41.97	0.58
6	5	0.6	1.5	6172	-41.73	0.82
6	0.6	0.8	1.5	504	-42.86	-0.31
6	0.6	0.8	1.2	466	-42.82	-0.27
6	0.6	0.6	1.2	2950	-42.55	0

Figure 28.  TS distribution from Onondaga Lake, May 2005, in water 2 to 10 m deep.  The three curves represent different angle standard deviations (0.6, 2, and 5) Analysis with EchoView method 1.  The shape of the distributions with method 2 is almost identical to method 1. With this distribution we consider -60 dB to be an appropriate minimum TS value to consider alewife, although -54dB may also be appropriate.  The lake was dominated by one age group of alewife (age 3 in 2005).