Diel variations in acoustic recordings of blue whiting (Micromesistius poutassou)

Time series of scientific survey estimates of fisheries resources are used to measurechanges in stock size and composition. Standardization of sampling equipment anddesign is regarded crucial for reliable trawl and acoustic survey estimates, but in spiteof established routines the reliability of the survey estimates are reduced due tovariations in fish behavior and changes in their spatial distribution.The main objectives of this thesis are to investigate the effects of spatial and temporalvariation in fish distribution and behavior on acoustic and bottom trawl estimates, andthe study presents methods to identify and correct for these sources of errors. A largevariety of data are examined ranging from acoustic recordings from the spawninggrounds of blue whiting (Micromesistius poutassou) and Norwegian Spring Spawningherring (Clupea harengus) to bottom trawl survey data and commercial CPUE ofNamibian hake (Merluccius capensis and M. paradoxus) and the results are presentedin six scientific papers.We show that blue whiting is distributed deeper and covers a narrower depth range atdaytime than at night during its spawning west of the British Isles. As a consequencethe tilt angle distribution changes and reduces the average acoustic target strength ofthe measured fish. This explains why we observed higher acoustic densities of bluewhiting at daytime than at night. The effect of the diel variation on the surveyestimates varies between years and therefore reduces the reliability of the survey timeseries (Paper I). Not surprisingly we observed that the diel oscillation in the samplevalues is closely related to rise and set for the sun. However, time-dependent variationin the catch rates and echo density also occur within daylight and the densitymeasures increase from dawn to noon and then subsequently decrease to dusk (PapersI and IV). For the catch rates of Namibian hakes we found that the maximum isreached when the sun is about 20 degrees above the horizon, but the amplitude of thediel oscillation decreases with depth and increases with density of hake. Highestamplitude values are found in shallow waters which are dominated by the M. capensisspecies (Paper IV). The effect of weather and environmental conditions on the dielvariation is not investigated in our studies due to lack of adequate data, but a in situexperiments we carried out in Namibia indicates a between-year difference in theescapement of hake under the fishing line due to annual differences in the oxygenlevels near the seabed (Paper III). This type of escapement is likely to change withtime of the day as hake lifts from the bottom at night and may also explain some ofthe differences in the diel oscillation we found in the different areas of the Namibianwaters.It is often stressed that the assessment surveys should be repeated at the same timeeach year, as many fish stocks move in more or less repeating geographical migrationpatterns. However, the substantial geographical dynamic we identified in theNamibian trawler fleet (Paper V) suggests substantial annual variation in the spatialdistribution of the hakes and changes in survey estimates may therefore be a samplingartifact due to variable environmental conditions and location of fish. Fish populationsare usually patchily distributed and show gradient structures and the individualswithin a shoal or a limited area tend to have more similar characteristics than those ofthe entire population. An example of small scale patchiness is found in the herring trawl samples with a significant relationship between depth distribution and maturitystate of herring where the spawners dominated the trawl catches taken close to thesea-bed (Paper II). In Paper VI, the precision of the abundance estimates by lengthgroup of Namibian hake is increased by including the information of spatialautocorrelation. Using geo-statistic, the geographical structure of the distribution canbe model and thus enhance the quality of the survey estimates.The overall result shows that the catch rates and acoustic density of fish areinfluenced by the fish’s location and its response to the approaching trawl which varywith time of the day. The swimming performance and reaction pattern to catchingdevice are also affected by the density of fish, species and length dependent behavior,bottom depth, level of dissolved oxygen and temperatures. All these parameters canbe included in the generic term “survey condition”, and any change in the surveycondition tend to influence the relationship between measured and true fish density.The survey errors investigated in this thesis are not included in standard estimationprocedures, and obviously it is a difficult task to correct for errors caused by variationin the survey condition. We have applied statistical modeling to identify systematicspatial and diel variability in large survey and commercial fishing data-sets, which canbe used to correct for these errors. However, diel variation and geo-statistical modelsdepend on the correct estimation of several parameters, which negatively affect theprecision of the survey estimates. Therefore, a better approach is to integrate severalsampling methodologies so the effect of variable survey conditions can be directlymeasured. Many places on the globe are also on the doorstep to a complex ecosystemmanagement approach which demands scientific surveillance of several species andtrophic levels. A simultaneous monitoring of inter-specific interaction is a fundamentof this types of surveys and requires advanced survey strategies where many sourcesof data collected with trawls, CTD probes, echo-sounders, transducers and sonars areintegrated and together form the basis for the concurrent measurement of theenvironment, survey condition and fish densities.Furthermore, the allocation of survey effort are occasionally sub-optimize due tovariation in the spatial structure of the target stock between surveys. An adaptivesampling strategy which utilizes for instance satellite tracking data of the commercialfishing fleet to map the stock distribution in advance of the surveys may improve theeffort allocation and hence increase the reliability of the survey estimates