ICES Journal of Marine Science: Journal du Conseil Advance Access originally published online on May 20, 2008
ICES Journal of Marine Science: Journal du Conseil 2008 65(6):1036-1045; doi:10.1093/icesjms/fsn082
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Correcting for vessel avoidance in acoustic-abundance estimates for herring
Institute of Marine Research, PO Box 1870, Nordnes, 5817 Bergen, Norway
Correspondence to V. Hjellvik: tel: +47 21 07 82 66; fax: +47 21 07 81 46; e-mail: vidar.hjellvik{at}fhi.no
Hjellvik, V., Handegard, N. O., and Ona, E. 2008. Correcting for vessel avoidance in acoustic-abundance estimates for herring. – ICES Journal of Marine Science, 65: 1036–1045.When a research vessel passes over a school or layer of herring (Clupea harengus), the fish may avoid the vessel by swimming downwards and horizontally. If the orientation of the fish is changed in this process, the mean target strength may also be altered. Consequently, the echo abundance measured by the relatively narrow echosounder beam does not always reflect the true density of the school in the undisturbed situation. This avoidance behaviour has been quantified in several experiments where a stationary, submerged transducer has been used to measure the changes in echo abundance during the passage of a survey vessel. Two approaches for correcting the echo abundance for avoidance are presented. The first is to correct in each depth layer separately, but this does not account for diving during vessel passage. The second is to correct the total echo abundance based on the mean depth of the fish at passage. Generalized linear models are fitted to the experimental data in both approaches. The parameters are estimated with uncertainty, which is taken into account when the fitted models are used for correcting standard survey data. The models were fitted to data from various experiments conducted during the period 1996–2004. The avoidance response differed strongly between experiments, indicating that correction factors estimated from one specific experiment should not be used uncritically in a standard correction procedure.
Keywords: acoustic surveys, avoidance correction, generalized linear models, uncertainty
Received 22 April 2007; accepted 3 March 2008; advance access publication 20 May 2008.
 |
Introduction |
|---|
 Top Introduction Material and methodsResultsDiscussionReferences |
|---|
Vessel avoidance is a known source of bias in acoustic abundance-estimation surveys. The problem was raised by Olsen (1971), and experiments were conducted (Olsen, 1979; Olsen et al., 1983). This work was followed up in the early 1990s (e.g. Gerlotto and Fréon, 1992), laying the ground for a comprehensive ICES Cooperative Research Report (Mitson, 1995). That report contains a review of vessel-avoidance work and recommendations for vessel design to minimize the problem. Most new research vessels are built according to these recommendations. Further work was conducted, and for herring (Clupea harengus) recorded in dense schools close to the sea surface, avoidance reactions were reported by comparing sonar observations with the density measured by the research-vessel’s echosounder (Misund et al., 1996; Soria et al., 1996; Gerlotto et al., 2004; see also reviews in Fréon and Misund, 1999). Fernandes et al. (2000a) used an Automated Underwater Vehicle (AUV) running in front of a noise-reduced vessel (RV "Scotia") and reported no avoidance behaviour, and the lack of reaction was attributed to the noise-reduced vessel design (Fernandes et al., 2000b). More recently, Ona et al. (2007b) compared a noise-reduced and conventional research vessel and reported, contrary to expectations, twice as strong a reaction from the noise-reduced vessel. They concluded that the ICES recommendations may be necessary, but not sufficient. De Robertis et al. (2008) confirm this. We are today unable to build true stealth vessels, and a method to correct for this variable bias is necessary and timely.
The assessment of the Norwegian spring-spawning herring includes catch data and eight surveys (acoustic surveys of adults and juveniles, a larval survey, and a 0-group survey), where the acoustic-survey data are used as relative indices of abundance (ICES, 2007). The current study uses vessel-avoidance data collected during the two acoustic winter surveys in the wintering area in Tysfjord, Ofotfjord, and Vestfjord in northern Norway (Figure 1). The importance of these surveys was great during the stock-rebuilding phase from 1983 to 2002, but in recent years, the summer survey on the feeding grounds has been more important for the assessment (ICES, 2007).
|
It has been argued that surveys could provide measures of absolute abundance, independent of commercial catches, but the conversion to biomass from echo abundance may be biased if corrections are not made, e.g. for the shadowing effect (Foote et al., 1992; Zhao and Ona, 2003), depth-dependent target strength (TS) (Ona, 2003), and vessel avoidance (Vabø et al., 2002). If these effects vary between years, they are also relevant when using the survey estimates as relative indices.
The calculation of the respective correction factors involves uncertainty, and Løland et al. (2007) estimated the relative contribution of the various sources of uncertainty in the corrected abundance estimate. This they did by describing the shadowing effect, the depth-dependent TS, and vessel avoidance by parametric models, and estimating the parameters (let 
denote all parameters) and the distribution of the parameter estimates (
). The uncertainties involved in the corrections were quantified by a resampling method that involved simulating sets of parameters * from the distribution of (see pp. 1307/8 of Løland et al., 2007, for details). Here, we focus in more detail on the problem of correcting for vessel avoidance, and the data are not corrected for shadowing or depth-dependent TS.
Norwegian spring-spawning herring are mainly distributed at depths between 50 and 400 m. In the upper parts of the water column, a marked avoidance behaviour has been observed in several experiments. Vabø et al. (2002) conducted work in 1996 where vessel avoidance was quantified for each depth layer by means of the vessel-avoidance coefficient (VAC), defined as VAC = sA,pass/sA,ref, where sA,pass is the scrutinized herring sA (Nautical Area Backscattering Coefficient; m2 per nautical mile2; MacLennan et al., 2002) measured during vessel passage, and sA,ref is the sA averaged over a reference period before passage. The measured VAC was <0.15 in several cases for herring layers shallower than 90 m.
For reasons given in the section entitled "The Vabø method" below, the results from Vabø et al. (2002) cannot be used directly for correction purposes. Here, therefore, an alternative approach is presented where generalized linear models are used for estimating depth-dependent correction factors. Correcting the echo abundance in each depth layer separately is often problematic because the fish may dive before the vessel passes. An alternative approach is to correct the total echo abundance, and a method for doing this based on the mean depth of the herring at the time of vessel passage is also presented.
These methods may be applied to historical survey data as long as they have a reasonable vertical resolution, preferably 10-m depth bins, or finer. The correction can be dramatic if most of the fish are close to the sea surface, and insignificant if the herring are distributed deeper than 150 m. Without corrections, surveying the same, enclosed herring population may yield differences in biomass between day and night of one order of magnitude or more (Huse and Korneliussen, 2000), mainly because of changes in the vertical distribution pattern, causing differences in vessel avoidance and depth-dependent TS.
|
Material and methods |
|---|
|
TopIntroductionMaterial and methodsResultsDiscussionReferences |
|---|
Data and experimental setup
We used data from vessel-avoidance investigations carried out in 1996 and from 2001 to 2004 (Figure 1, Table 1). Three experimental setups were used in the investigations. The 1996 experiment was carried out by placing a smaller vessel equipped with a Simrad EK500 38 kHz echosounder in the path of the surveying vessel; this setup is described in detail in Vabø et al. (2002). The Bergen Acoustic Buoy was used for the 2001, 2002, and 2003 experiments; this is a free-floating buoy equipped with a Simrad EK60 38 kHz echosounder (Godø et al., 1999; Godø and Totland, 1999). The method has been successfully used for similar investigations (see Handegard et al., 2003; Jørgensen et al., 2004; Handegard and Tjøstheim, 2005; Skaret et al., 2005, 2006, for a more detailed description). The 2004 experiments were carried out using a bottom-moored, upward-looking EK60 38 kHz echosounder (Ona et al., 2007b). The transducer depths among experiments ranged from 10 to 137 m (Table 1).
|
The Vabø method
In Vabø et al. (2002), the VAC at depth d for a given experiment i was calculated as
|
| (1) |
|
| (2) |
|
| (3) |
The correction factor is the inverse of 
d, and as VACd,i is asymmetric (0<VACd,i <1 if sd,iA,pass <sd,iA,ref, and 1<VACd,i < 
if sd, iA,pass >sd,iA,ref), the distribution of the correction factor is skewed, and it cannot be described by the mean and the s.d. of VACd alone. When correcting the echo abundance for avoidance, the uncertainty of the estimated correction factor should be incorporated to obtain a realistic uncertainty measure in the corrected abundance estimate. This can be done by a bootstrap approach, where the correction factor is drawn at random from its estimated distribution. However, the bootstrap approach is not possible in this situation, because the mean and s.d. are insufficient for describing the skewed distribution. Two alternative approaches are given below.
Correcting each depth layer separately using generalized linear models
The first approach is to model log (sd,iA,ref) as a function of depth and log (sd,iA,pass) as
|
| (4) |
d,i is random noise. To find the predicted value 
dA,ref corresponding to a given observed value sdA,pass at a given depth d, we back-transform formulation (4) to get
|
| (5) |
is a bias-correcting term. This method of bias correction is known as smearing (Duan, 1983). Confidence intervals for dA,ref can be found by bootstrapping or analytically given that log(dA,ref) is unbiased and normally distributed, with
|
| (6) |
|
| (7) |
We can now simulate (cf. Løland et al., 2007) a correction vector *A,ref=[d1*A,ref, d2*A,ref,... dp*A,ref]T for the depth layers d1, d2, ..., dp as
|
| (8) |
|
| (9) |
, and 
Several modifications of formulation (4) are possible. For example, one could model an individual slope bd for each depth, or one could let the intercept a and the slope b have some functional dependence on depth (e.g. a = a0 + a1di + a2di2 + a3di3).
Formulation (4) does not account for fish diving. For example, the model will fail in a situation where all the fish are situated in one depth layer before passage and move down to the next layer during passage. The correction factor will then tend to infinity for the upper layer and to zero for the lower layer. A model taking this into account could be
|
| (10) |
Correcting the total sA using mean depth at passage in a generalized linear model
An approach that deals with fish migration among the depth layers in a more robust and general way is to model the vessel correction factor VCFi = siA,ref/siA,pass as
|
| (11) |
iA,pass is the mean fish depth during passage. An observed sA,pass value with a corresponding mean depth pass could then be corrected by multiplying sA,pass with |
| (12) |
Confidence intervals for V
Fi can be found by bootstrapping or analytically assuming that log (VFi) is unbiased and normally distributed with variance:
|
| (13) |
In simulations (cf. Løland et al., 2007), we can let
|
| (14) |
Fi), var {log(VFi)}).
A downwards vertical movement will incur a reduction in swimbladder volume, and consequently, a reduced TS (and hence siA,pass < siA,ref and VFi >1) without necessarily reducing the abundance in numbers of fish within the echobeam. This effect could be interpreted as an "avoidance reaction", because there is a reduction in echo abundance. One solution would be to correct for depth-dependent TS before correcting for avoidance, i.e. to base the avoidance correction on fish density rather than on sA. Another solution is to estimate the mean vertical displacement, 
d, and use this information to obtain the correct depth-dependent TS. The fish density at location i and depth d is estimated as [cf. formulation (9) in Løland et al., 2007]:
|
| (15) |
0 and 
are parameters, l is fish length, and d is depth. By replacing d in the formula above with d–
d, we have accounted for the TS reduction caused by vertical displacement in the experiment situation. In simulations, we may draw *d from N(d, var (d)). |
Results |
|---|
|
TopIntroductionMaterial and methodsResultsDiscussionReferences |
|---|
The vessel avoidance differed strongly between the various experiments conducted during 1996 and 2001–2004 and analysed here (Table 1). Echograms for two of the experiments are shown in Figure 2. In both experiments there is a relatively deep herring layer at
200–300 m. In the second experiment there is a marked diving reaction (Figure 2a), which is also seen in the detailed echogram from the fourth passage (Figure 2b and c), but the reaction occurs slightly after the transducer passage. In fact sA,pass is higher than sA,ref in 9 of the 11 passages, and on average 19% higher. In the third experiment there is no visible reaction (Figure 2d), but variations that could be mistaken for vessel passages are seen. Echograms for all except the 1996 experiment are shown in Hjellvik et al. (2006).
|
In 1996, there was a marked reduction in echo energy at passage in the upper layers, but there was no corresponding increase in the lower layers, which would have indicated a vertical displacement of the biomass through diving (Figure 3; Vabø et al., 2002). In 2004 the situation was quite different, with a strong diving response, particularly for the "G. O. Sars" (Figure 3; Ona et al. 2007b). Therefore, it would be inappropriate to fit formulation (4) to the 2004 experiment, but for the 1996 experiment the formulation would be suitable. In fact, it described the data reasonably well (Figure 4), giving very high correction factor estimates in shallow waters (Figure 5).
|
|
|
For the 2004 data, where a strong diving response was detected, the alternative approach is to use formulation (11) and therefore to correct the total sA by letting the correction depend on the mean fish depth at passage. We fitted this model to all the experiments, and in general the avoidance reaction decreased with increasing fish depth, but the variation between experiments was quite large (Figure 6, left panels). The 2004 experiment was atypical, with almost no reduction in echo energy at passage despite the strong diving response and the shallow mean fish depth (Table 1). Based on data from all years (Figure 6, lower left panel), back-transformation gave a correction factor of about 4 at 50 m depth, decreasing to 1 at 300 m depth (Figure 6, lower middle panel). The r2 was in this case just 0.25.
|
The mean vertical displacement of fish from reference time to passage time, averaged over all experiments from 2001 through 2004, was
d = 12.5 m, with a s.e. of 0.12. This corresponds to a 2–3% decrease in sA atributable to swimbladder compression at 100 m depth. The largest displacements were in experiments where the mean fish depth was <250 m and sA at passage was >1000 m2 nautical mile2 (Figure 7).
|
The time intervals used in formulations (4) and (11) for averaging the reference- and passage densities, sA,ref and sA,pass were the same as in Vabø et al. (2002). To check the effect of changing these intervals, we averaged the reference density over the 70-s interval ending 158 s before passage, and the passage density over the 3-s interval starting 1 s before passage, and the results were similar to those obtained using the original intervals, except in 2001 and 2004 where a changed reference interval led to a change in the slope (Figure 6, right panels). The range in mean depth was, however, small in 2001 and 2004. The 1996 experiment was not included in this sensitivity analysis, because data for the new reference period were not available for this experiment.
|
Discussion |
|---|
|
TopIntroductionMaterial and methodsResultsDiscussionReferences |
|---|
Two methods for correcting the observed echo abundance for avoidance have been presented. In the first method the correction is made at each depth layer separately, and in the second the total echo abundance is corrected based on the mean depth of the fish during passage. The first method assumes no vertical displacement of fish during vessel passage, whereas no such assumption is needed for the second method.
If the assumption of no vertical displacement is met, the first method makes it possible to estimate the vertical distribution of echo energy as it was before vessel passage, and not only the total echo energy, as is the case for the second method. The first method also provides more data points to analyse than the second (one per depth layer per experiment vs. one per experiment), but more parameters need to be estimated (slope and one intercept per depth layer vs. slope and one intercept).
If the assumption of no vertical displacement is not fulfilled, which is often the case, the first method may lead to erroneous corrections, whereas the second method is much more robust. Still, the decrease in total echo abundance at passage may differ with circumstances where the mean depth is the same. For example, one could expect different behaviour in a layer reaching from 20 to 180 m deep than in one concentrated between 90 and 110 m. In the first situation, it is likely that the fish in the upper layers react to the vessel, and if fish farther down in the water column react to the fish above, a cascading effect may result (see Gerlotto et al., 2006).
In some cases a correction factor of <1 was estimated. For formulation (4) this can happen if fish dive from one layer to another. Then fish "avoid" the upper layer and "are attracted to" the lower layer. For formulation (11), correction factors <1 can occur at great depths if the assumption of a linear relationship between siA,ref/siA,pass and depth does not hold. Looking at the 2003 data in Figure 6, where the span in mean depth is large, the decrease in log(VCFi) seems to stop at 200 m, and the threshold model
|
| (16) |
d0. Alternatively, the truncated correction factor |
| (17) |
A more serious concern is the great variability in vessel avoidance between seasons (or experiments). All experiments were performed on overwintering fish, and some similarities were expected. One reason for the variability may be a potential bias in the method used to observe fish avoidance. Røstad et al. (2006) showed an increase in echo abundance under a stationary research vessel. If the platform acts as a fish-attracting device (FAD), the density under the platform will be greater during the experiment than it would have been in the undisturbed situation, and they argued that this may bias avoidance observations. A small vessel was used as an observation platform in the 1996 experiments, and a FAD effect may therefore explain the stronger fish reaction observed in these experiments. Other floating platforms such as buoys can also act as FADs, and the bottom-mounted equipment used in the 2004 experiments are preferred over floating platforms in consequence.
Røstad et al. (2006) also argued that the approaching vessel itself could attract fish and hence cause the echo abundance to be overestimated. However, based on observed swimming velocities of fish (2004 experiments) and interpretation of the echograms (2001–2004 experiments), this hypothesis is not supported for these experiments.
Varying fish depth explains some of the variation in avoidance reaction in experiments 2–9 (Table 1). In experiments 4, 7, and 9, mean fish depth was <100 m, and in experiments 4 and 7 there was a marked reduction in echo energy at passage, whereas in experiment 9 there was virtually no reduction even though a strong diving response was observed (Figure 3). One possible explanation is that there was a stronger horizontal fish movement in experiments 4 and 7 than in experiment 9, and that changes in tilt angle did not have any significant influence on the echo energy. It is also possible that the motivation for reaction may be locally altered by, for instance, recent fishing activity and variable predation pressure, but as long as the motivation for reacting is unknown it is not possible to incorporate it in the avoidance correction model. The model should therefore be fitted to experiments performed on each survey to obtain parameter estimates with a reasonable accuracy and uncertainty. The experiments should cover the fish depths that are encountered on the survey and at which avoidance is expected, i.e. down to 200 m for herring.
If there is still reason to believe that the experiments are not representative of the survey, alternative methods are needed to obtain information of vessel avoidance. For example, an AUV can be run in front of the surveying vessel to obtain estimates of vessel avoidance (Fernandes et al., 2000a), but speed and mission time are still low compared with the usual requirements for a standard survey. Other methods include using sonars to estimate the avoidance (Soria et al., 1996). Recent trials with scientific, multibeam sonars, which can observe and measure schools close to the sea surface at a range away from the surveying vessels, are also promising tools (Ona et al., 2007a). The advantage with these methods is that the avoidance is measured continuously from the surveying vessel, reducing the problem of validity, as mentioned above.
Designing a non-invasive survey method should be the long-term goal. However, designing true stealth vessels still appears to be unrealistic (Ona et al., 2007b), and if the acoustic-biomass estimate is to be made from vertical-echosounder data, the only realistic option currently available is to correct the measured echo energy for vessel avoidance.
|
Acknowledgements |
|---|
We thank the Norwegian Research Council for financial support under contact 143249/I10, and Chris Wilson and an anonymous referee for very useful comments.
|
References |
|---|
|
TopIntroductionMaterial and methodsResultsDiscussionReferences |
|---|
-
Duan N. Smearing estimate—a nonparametric retransformation method. Journal of the American Statistical Association (1983) 78:605–610.[CrossRef][Web of Science]
De Robertis A., Hjellvik V., Williamson N. J., Wilson C. D. Silent ships do not always encounter more fish: comparison of acoustic backscatter recorded by a noise-reduced and a conventional research vessel. ICES Journal of Marine Science (2008) 65:623–635.
Fernandes P., Brierley A., Simmonds E., Millard N., McPhail S., Armstrong F., Stevenson P., et al. Fish do not avoid survey vessels. Nature (2000) a 404:35–36.[CrossRef][Medline]
Fernandes P., Brierley A., Simmonds E., Millard N., McPhail S., Armstrong F., Stevenson P., et al. Fish do not avoid survey vessels. Nature (2000) b 407:152.[CrossRef]
Foote K. G., Ona E., Toresen R. Determining the extinction cross section of aggregating fish. Journal of the Acoustical Society of America (1992) 91:1983–1989.[CrossRef][Web of Science]
Fréon P., Misund O. A. Dynamics of Pelagic Fish Distribution and Behaviour: Effects on Fisheries and Stock Assessment. (1999) Oxford: Fishing News Books. 348.
Gerlotto F., Fréon P. Some elements of vertical avoidance of fish schools to a vessel during acoustic surveys. Fisheries Research (1992) 14:251–259.[CrossRef][Web of Science]
Gerlotto F., Castillo J., Saavedra A., Barbiere M. A., Espejo M., Cotel P. Three-dimensional structure and avoidance behaviour of anchovy and common sardine schools in central southern Chile. ICES Journal of Marine Science (2004) 61:1120–1126.
Gerlotto F., Bertrand S., Bez N., Gutierrez M. Waves of agitation inside anchovy schools observed with multibeam sonar: a way to transmit information in response to predation. ICES Journal of Marine Science (2006) 63:1405–1417.
Godø O. R., Somerton D., Totland A. Fish behaviour during sampling as observed from free-floating buoys—application for bottom-trawl survey assessment. ICES Document CM 1999/J (1999) 10:10.
Godø O. R., Totland A. Bergen acoustic buoy (BAB)—a tool for the remote monitoring of marine resources. In: Joint IASA/EASA meeting (1999) Berlin. Paper 2pAO10.
Handegard N. O., Tjøstheim D. When fish meets a trawling vessel: examining the behaviour of gadoids using a free-floating buoy and acoustic, split-beam tracking. Canadian Journal of Fisheries and Aquatic Sciences (2005) 62:2409–2422.
Handegard N. O., Michalsen K., Tjøstheim D. Avoidance behaviour in cod (Gadus morhua) to a bottom-trawling vessel. Aquatic Living Resources (2003) 16:265–270.[CrossRef][Web of Science]
Hjellvik V., Handegard N. O., Ona E. Correcting for avoidance in acoustic abundance estimates for herring using a generalized linear model. ICES Document CM 2006/I (2006) 22:24.
Huse I., Korneliussen R. Diel variation in acoustic-density measurements of overwintering herring (Clupea harengus L.). ICES Journal of Marine Science (2000) 57:903–910.
ICES. Report of the Northern Pelagic and Blue Whiting Fisheries Working Group, 27 August–1 September 2007. ICES Document CM 2007/ACFM (2007) 29:226.
Jørgensen R., Handegard N. O., Gjøsæter H., Slotte A. Possible vessel-avoidance behaviour of capelin in a feeding area and on a spawning ground. Fisheries Research (2004) 69:251–261.[CrossRef][Web of Science]
Løland A., Aldrin M., Ona E., Hjellvik V., Holst J. C. Estimating and decomposing total uncertainty for survey-based abundance estimates of Norwegian spring-spawning herring. ICES Journal of Marine Science (2007) 64:1302–1312.
MacLennan D. N., Fernandes P. G., Dalen J. A consistent approach to definitions and symbols in fisheries acoustics. ICES Journal of Marine Science (2002) 59:365–369.
Misund O. A., Aglen A., Hamre J., Ona E., Røttingen I., Skagen D., Valdemarsen J. W. Improved mapping of schooling fish near surface: comparison of abundance estimates obtained by sonar and echo integration. ICES Journal of Marine Science (1996) 53:383–388.
Mitson R. B. Underwater noise of research vessels: review and recommendations. ICES Cooperative Research Report, 209. (1995) 61 pp.
Olsen K. Influence of vessel noise on the behaviour of herring. In: Modern Fishing Gear of the World—Kristjonsson H., ed. (1971) London: Fishing News Books. 291–293.
Olsen K. Observed avoidance behaviour in herring in relation to passage of an echosurvey vessel. ICES Document CM 1979/B (1979) 18:21.
Olsen K., Angell J., Pettersen F., Løvik A. Observed fish reactions to a surveying vessel with special reference to herring, cod, capelin and polar cod. FAO Fisheries Report (1983) 300:131–138.
Ona E. An expanded target-strength relationship for herring. ICES Journal of Marine Science (2003) 60:493–499.
Ona E., Andersen L. N., Knudsen H. P., Berg S. Calibrating multibeam, wideband sonar with reference targets. (2007) a. Proceedings of OCEANS 2007, 1-5: Aberdeen, Scotland. [doi 10.1109/OCEANSE.2007.4302473].
Ona E., Godø O. R., Handegard N. O., Hjellvik V., Patel R., Pedersen G. Silent research vessels are not quiet. Journal of the Acoustical Society of America (2007) b 121:EL145–EL150.[CrossRef][Web of Science][Medline]
Røstad A., Kaartvedt S., Klevjer T. A., Melle W. Fish are attracted to vessels. ICES Journal of Marine Science (2006) 63:1431–1437.
Skaret G., Axelsen B. E., Nøttestad L., Fernö A., Johannessen A. The behaviour of spawning herring in relation to a survey vessel. ICES Journal of Marine Science (2005) 62:1061–1064.
Skaret G., Slotte A., Handegard N. O., Axelsen B. E., Jørgensen R. Pre-spawning herring in a protected area showed only moderate reaction to a surveying vessel. Fisheries Research (2006) 78:359–367.[CrossRef][Web of Science]
Soria M., Fréon P., Gerlotto F. Analysis of vessel influence on spatial behaviour of fish schools using a multibeam sonar and consequences for biomass estimates by echosounder. ICES Journal of Marine Science (1996) 53:453–458.
Vabø R., Olsen K., Huse I. The effect of vessel avoidance of wintering Norwegian spring-spawning herring. Fisheries Research (2002) 58:59–77.[CrossRef][Web of Science]
Zhao X., Ona E. Estimation and compensation models for the shadowing effect in dense fish aggregations. ICES Journal of Marine Science (2003) 60:155–163.
CiteULike 
Connotea 
Del.icio.us What's this?
This article has been cited by other articles:
|
|
 |
|
  J. A. Koslow The role of acoustics in ecosystem-based fishery management ICES J. Mar. Sci., July 1, 2009; 66(6): 966 - 973. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
R. Patel and E. Ona Measuring herring densities with one real and several phantom research vessels ICES J. Mar. Sci., July 1, 2009; 66(6): 1264 - 1269. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||