Nt t be cited withut prir reference t the authrs ICES Annual Science Cnference Bruges, Belgium, 27-30 September 2000 Incrpratin f External Factrs in Marine Resurce Surveys ICES CM 2OOO/K:37 Measurements f Distance Fished During the Trawl Retrieval Perid by Charles W. West and Jhn R. Wallace Natinal Marine Fisheries Service, FRAM Div., NWFSC 2725 Mntlake Blvd. E., Seattle, Washingtn 98112 U.S.A. Abstract Observatins f sampling trawl perfrmance made during a multi-vessel grundfish trawl survey cnducted during 1998 and again in 1999 raised cncerns that the trawls might be cntinuing t fish during the retrieval perid, after the end f the sampling perid but befre cming ff bttm. Fllwing the 1998 survey, a simple gemetric analysis f times and psitins recrded at critical mments during and fllwing each sampling tw was develped t estimate the fllwing parameters: 1) the distance alng the bttm that the gear swept during the retrieval perid, and 2) the speed at which the trawl mved ver the seabed. This analysis suggested that the distances swept were substantial, and systematically increased with the depth f the tw. The effective trawl speed apprached r even exceeded the twing speed specified by the sampling prtcls, and this varied systematically amng the participating vessels. The same analysis was perfrmed fr sampling tws cnducted during the 1999 survey and cmpared against trawl psitins recrded during the same perid by an ultra-shrt baseline acustic psitining system. Bth techniques yielded similar results, and were in accrd with the findings frm the 1998 data: distances swept by the trawls during the retrieval perid were substantial and the trawls were mving at speeds cmparable t fishing speed, and these effects varied systematically frm depth t depth and vessel t vessel. Neglect f these effects culd increase the impact f depth-related bias and inter-vessel variability n survey results while knwledge f them culd help explain the vessel effect cmmnly bserved when cmparing the fishing perfrmance f tw r mre vessels. Intrductin One f the cre issues in bttm trawl survey methdlgy and in ther experimental trawling situatins is measuring and/r standardizing the fishing effrt. Whether survey results will be used t prepare area swept estimates f abslute abundance, t cmpute indices f relative abundance, r t calculate ther fishing efficiency statistics, cntrlling and quantifying tw duratin and/r tw distance are typically cnsidered quite imprtant (Gundersn, 1993; Engfis,
1994; Gd@, 1994; Parsns and Sandeman, 1981; Grsslein, 1971; and Byrne et al. 1981). The significance f these factrs is cmparable t such thers as the cnstructin and rigging f the gear, adherence t a specified twing speed and ther peratinal prtcls, and emplyment f a standardized vessel. Fr many surveys tw duratin is defined as the perid between the time the trawl is determined t be n the bttm and in a stable fishing cnfiguratin until the mment when the tw is declared ver (usually at the end f sme fixed, predetermined sampling perid) and the trawl winches are started up t retrieve the gear. In such cases twing distance is similarly defined as equal t the distance transited by the vessel between the starting pint and end pints f the tw as defined abve. These are versimplificatins f the true situatin, as is shwn frm bservatins cnducted with trawl instruments (e.g. Wathne, 1977). It is quite cmmn fr a trawl t remain n the bttm fr cnsiderable perids after haulback has begun (see example in Figure 1). Even if the twing vessel s thrust is reduced (as is typically dne) it is pssible that the trawl cntinues t advance acrss the bttm during this perid and may cntinue t catch fish if it remains in its nrmal fishing cnfiguratin and its speed f advance is sufficiently high. Starting in 1998, every summer the Fishery Resurce Analysis and Mnitring Divisin f the Nrthwest Fisheries Science Center, U.S. Natinal Marine Fisheries Service, charters fur cmmercial fishing vessels t cnduct a bttm trawl survey assessing the distributin and abundance f cmmercial grundfish resurces inhabiting the slpe zne (100 t 700 fathms, r 183 t 1280 m) ff the casts f Washingtn, Oregn, and Califrnia. Each f the fur vessels ccupies its wn unique set f sampling statins distributed alng the full length f the survey area, essentially cnducting its wn cmplete, independent, castwide mini-survey. The fur sets f statins are interleaved such that all statins cmbined cnstitute a full survey at relatively high sampling density. N separate fishing pwer calibratins are cnducted amng the fur vessels since cmparisns amng the fur mini-surveys can, with the crrect analytical apprach, cntribute infrmatin abut the relative fishing perfrmance f the fur vessels. Fr this apprach t succeed it is necessary t standardize r accurately measure as many as pssible f the ptential cntributrs t differences in between-vessel fishing pwer. Given the knwn differences amng the vessels in such areas as engine hrsepwer, twing capabilities, and winch characteristics, we had serius cncerns abut ptential differences in fishing perfrmance due t such phenmena as prlnged fishing after the nminal end f the sampling tw as described abve. These cncerns were heightened by ur chice f a relatively shrt 1%minute standard tw duratin, which wuld increase the relative imprtance f a few minutes delay in the gear actually cming ff bttm. Fllwing the 1998 survey, a simple gemetric analysis f times and vessel psitins recrded at critical mments during and fllwing each sampling tw was develped t estimate the fllwing parameters: 1) the distance alng the bttm that the gear swept during the retrieval perid, and 2) the speed at which the trawl mved. This analysis (West et al., 1999) suggested that the distances swept were substantial and systematically increased with the depth f the tw, that the effective trawl speed apprached r even exceeded the twing speed specified by the 2
sampling prtcls, and that these perfrmance parameters varied systematically amng the participating vessels. During the 1999 survey the same data were recrded as in the previus year, making pssible the same srt f analysis. In additin, a new data-lgging system was emplyed t autmatically and cntinuusly recrd many significant parameters thrughut the tw including trawl psitins expressed in latitude and lngitude as calculated by ne f the instrument systems. Thus tw independent means were available fr examining trawl perfrmance during the retrieval perid, and their results cmpared. Methds The 1999 West Cast slpe survey was cnducted in the same survey area with the same sampling trawls and accrding t the same sampling prtcls as were used in 1998. Significant peratinal prtcls included a standard target twing speed f 2.2 knts ver the grund and a nminal tw duratin f 15 minutes between the time the net was prperly cnfigured n bttm and fishing, and the time at which heaving was initiated. The fur vessels chartered fr the 1999 survey were the Miss Lena, Blue Hrizn, Captain Jack, and Sea Eagle. Simrad ITI trawl instrumentatin systems and self-lgging bttm cntact sensrs attached t the ftrpe were used t recrd varius aspects f the trawls fishing perfrmance including their vertical and hrizntal dimensins at the muth, the time at which the gear cntacted the bttm at the beginning f each sampling tw, and the time at which it left the bttm at the end f the tw. Using ultra-shrt baseline acustic psitining technlgy, the ITI systems als calculated and reprted the trawls psitins in latitude and lngitude thrughut each tw. During each tw scientists in the wheelhuse recrded the time and ship s psitin (using a highly accurate differential GPS system) assciated with such critical events as the trawl reaching the bttm, the beginning and nminal end f the tw, the mment at which the trawl actually lifted ff bttm during retrieval, and the time at which the trawl drs had been fully heaved. In additin, every few secnds an autmated data-lgging prgram linked t the ITI and the GPS recrded the time, vessel psitin, trawl psitin, and ther parameters thrughut each sampling tw. Tw methds were emplyed t assess the amunt f bttm swept while the trawls were still lingering n bttm during retrieval peratins and the trawl s speed f advance ver the seabed during this perid, which we have dubbed liftff lag. All tws cnducted during the survey were cnsidered unless instrument r datalgger malfunctins made it impssible t use the data, r if the tw had t be abrted due t the trawl hanging up r ther peratinal prblems. We use the term ITI methd t describe the technique first emplyed with the 1999 data since it relies n the trawl psitins determined by the ITI system. Linking the cntinuusly-lgged trawl psitin data t the time at which retrieval began versus the time at which the IT1 shwed the trawls lifting ff made it pssible t directly calculate the distance the trawl cvered and the speed at which it mved. Because the GPS system s ability t crrectly cmpute vessel heading was degraded when vessel speeds were less than half a knt, such as during haulback, the IT1 was unable t crrectly determine the trawl s gegraphic psitin during sme f these perids. 3
Hwever, such phenmena did nt affect the ITI s ability t cmpute the trawl s range and bearing relative t the vessel s by applying the last knwn gd headings t the recrded range and bearing data, then smthing the results, we were able t btain sund estimates f the trawl s actual curse ver the seabed during the liftff interval. The ther apprach, which we have dubbed the 1998 methd since it was riginally develped and emplyed fr the 1998 data when trawl psitins were nt recrded, used varius knwn and interplated factrs t estimate the distance fished and the trawl s speed during the liftff lag perid (Figure 2). Knwn factrs: Time and ship s psitin at the beginning f haulback Time and ship s psitin at the mment f liftff The scpe (length f twing warp deplyed) and the fishing depth fr the tw Duratin f the perid between the beginning f haulback and the mment f liftff Time required t recver varius lengths f twing warp (Figure 3) The first step was t estimate the trawl s psitin (actually the psitin f the trawl drs in this analysis) at the beginning f haulback. This was dne by treating the scpe as the hyptenuse f a right triangle and the depth as ne f the sides, then slving t find the length f the remaining side (Fig.2). This crrespnds t the trawl s hrizntal distance behind the vessel and this ffset distance can then be applied t the vessel s bserved psitin t estimate the trawl s psitin. Fr this and subsequent steps sme simplifying assumptins were made: 1) that the trawl warps described a straight line between the trawl and the ship withut significant sagging r ther deflectin; 2) that the utward deflectin f the warps due t dr spread was insignificant relative t the scpe; 3) that the trawl was directly behind the ship thrughut the twing and haulback perids; 4) that the trawl s fishing depth was the same as the bttm depth displayed by the ship s echsunder; and 5) that the depth did nt change substantially during haulback. The next step was t estimate the amunt f warp still deplyed at the mment f liftff. The absence f suitable instrumentatin n these vessels made it impssible t directly measure this value. Hwever, the time required t recver varying scpes had been recrded n each vessel and frm this it was pssible t estimate the recvery speed fr each vessel s winches using the data depicted in Figure 3, then use this t predict hw much wire had been recvered by the time. liftff ccurred. Subtracting this estimate frm the initial scpe yielded an estimate f the length f the warps still deplyed at the mment f liftff. Using this resulting estimate f the scpe at liftff as the straight-line distance (the hyptenuse f the imaginary triangle) between the ship and the trawl, the same apprach as described abve was used t find the gegraphic psitin f the trawl at the mment f liftff. With estimates f these tw psitins it was pssible t estimate the hrizntal distance cvered by the trawl, and dividing this distance by the duratin f the liftff lag perid yielded an estimate f the trawl s speed f advance.
Fr each bat, estimates f the liftff lag distance btained via the 1998 methd were pltted against thse btained with the ITI methd. Linear regressins were perfrmed n these relatinships t determine hw well the tw techniques crrespnded.. Results During the 1999 West Cast slpe survey ver 340 sampling tws were carried ut. As shwn in Table 1 belw, valid data frm 238 tws were available t analyze by the ITI methd and frm 236 tws by the 1998 methd. The trawl perfrmance bservatins revealed patterns similar t thse seen in 1998: there frequently were prlnged intervals during retrieval when the trawls lingered n bttm; the length f such delays tended t increase with depth, and there were systematic differences amng the fur vessels in this delay/depth relatinship (Fig. 1). Recrds f the trawls vertical and hrizntal penings, cnfirmed by readings frm the bttm cntact sensrs, shwed that during the liftff lag perids the trawls seemed t remain in rughly the same cnfiguratin they had assumed during the actual tw. Vessel Miss Lena Blue Hrizn Captain Jack Sea Eagle Number f Tws Analyzed Length Overall Rated Hrsepwer ITI Methd 1998 Methd 87 ft 850 65 64 91 ft 675 66 63 75ft 425 52 51 88 ft 673 55 58 Table 1. Characteristics f the vessels emplyed during the survey and number f tws frm each vessel s cruise that culd be analyzed by the tw methds. Results are graphically depicted in Figures 4,5, and 6. Figure 4 shws that the distances cvered by the trawl during the liftff lag perid were substantial and increased with depth t meet r exceed ne kilmeter, the apprximate target tw distance that wuld be cvered in 15 minutes at 2.2 knts, and als shws that this effect varied frm vessel t vessel. A cmparisn f the distance estimates btained with the IT1 methd versus the 1998 methd (Fig. 4) shws that the tw techniques yielded similar results. This similarity is further reflected in the plt in Figure 5 and the regressin and crrelatin statistics presented in Table 2 belw. Testing the null hypthesis that the tw methds wuld yield exactly the same estimates, (I&: Slpe = 1.O) it can be seen that they did nt exactly agree, althugh agreement was clse. 5
Vessel Miss Lena Blue Hrizn Captain Jack Sea Eagle Regressin cefficients Slpe Intercept 0.7950 0.0266 0.9289-0.0366 1.1444-0.1226 0.7641 0.0261 P-value, H: Slpe =T 1.0 Crrelatin 0.0001 0.8825 0.3172 0.8564 0.0101 0.9458 0.0011 0.8171 Table 2. Results f regressin analysis f the distance estimates btained by the 1998 methd versus thse btained by the IT1 methd, and the cefficient f crrelatin between the tw techniques. Figure 6 shws that the speed f advance f the trawl during the liftff lag perid was substantial and ften exceeded the specified 2.2~knt twing speed. It als demnstrates that these speeds varied systematically frm vessel t vessel but that there was n depth-related effect. Discussin At the utset it must be nted that these bservatins were made n a specific set f similar vessels, twing a particular type f trawl under a unique set f circumstances. Different trawl systems perated frm ther types f vessels under different cnditins will perfrm in a different manner. Nnetheless, while the details f gear perfrmance will dubtless vary it is likely that similar patterns can be bserved, r shuld at least be suspected. Because the time elapsed during liftff lag may be a significant prtin f the time the net is n bttm, it is imprtant t understand the physical perfrmance f the gear during this interval t facilitate understanding its likely catching perfrmance. If the gear is n bttm, is cnfigured prperly, and is mving ahead at r near its nrmal fishing speed, it is likely t be catching fish. If the net is indeed fishing during this interval then the distance fished during a sampling tw must be adjusted t include this additinal time n bttm. Neglecting these effects culd lead t serius and systematic verestimates f fish abundance, especially at increasing depths, and the relative impact will increase as the duratin f the sampling tws decreases. This culd have cntributed t the catch by surprise effect pstulated by Gd et al. (1990). They describe experiments in which it was seen that catch rates during sampling tws fr gadids were at least as high fr shrt tws as fr lng, and were even higher fr the shrtest tws. T explain this they prpsed that due t fish behaviral respnses a trawl s catching efficiency might be highest during the first few minutes f a tw, then drp ff t a mre steady state as the tw cntinues. They assumed, based n snar bservatins, that the trawl came ff bttm immediately at the beginning f haulback, but if this was nt the case then the liftff lag phenmenn described here culd have cntributed t their bserved higher catch rates fr shrter tws. Vessel-t-vessel differences f the magnitude bserved here culd cntribute t an therwise undetected, but substantial, vessel effect which culd reduce the accuracy and precisin f any survey that ever emplys mre than ne vessel. Alternatively, being able t bserve and quantify 6
vessel-specific liftff lag patterns may prve useful in crrecting actual vs. nminal fishing effrt and thus help eliminate the impact f vessel effects. As pinted ut abve, the impact f liftff lag n catch quantity and cmpsitin depends n the extent t which the trawl is fishing, r is fishing in the same way, during this perid. The bserved tendency fr the gear t traverse the bttm at speeds higher than the standard twing speed has trubling implicatins fr its catching perfrmance. It is pssible that the gear s bttm tending characteristics culd be affected by these higher speeds, r that its size- and species-specific catching efficiency culd therwise be affected. If any f these pssible impacts d actually ccur then the impact n survey results will g beynd the difference between actual versus nminal tw duratin and/r distance fished. In future effrts we intend t undertake studies f the catching perfrmance f the gear during liftff lag under varius circumstances typical f ur survey. These will include mre detailed examinatins f gear perfrmance during this perid as well as in situ bservatins f fish behavir and interactins with the gear. If such phenmena are bserved and prve t be methdlgically r analytically intractable, then it may be necessary t cnsider ther technical measures such as the use f a remtely-perated cdend clsure system like the MultiSampler (Eng%s et al., 1997) t ensure that the entire sample is captured during a knwn, discrete sampling perid. There were interesting vessel-t-vessel differences in the relatinship between depth and liftff lag distance. As ne might expect, there was a steady increase in distance with increasing depth fr the Miss Lena and Captain Jack, but there was a mre cmplex relatinship fr the Blue Hrizn and Sea Eagle. On these vessels, distances increased with increasing depth up t a pint, then declined. This can be explained by the dynamics f the twing situatin. All fur vessels were able t carry just enugh twing warp t successfully execute tws at the deepest statins; i.e. the maximum scpe the bats culd deply was barely adequate fr these depths. During haulback, any increase in the cmbinatin f frward thrust applied by the vessel and/r higher trawl winch recvery speed will result in the gear lifting ff relatively quickly at these deeper depths. This appears t have taken place with the Blue Hrizn and Sea Eagle. On the ther hand, the ther tw bats may nt have been applying as much frward thrust and/r their winches were perhaps nt as pwerful. There was very gd agreement between the estimates f liftff lag distance btained by the ITI methd and the 1998 methd. This is particularly remarkable in view f the substantial differences in technlgy and sphisticatin and the many assumptins and apprximatins that had t be emplyed with the 1998 methd. The similarity f the results btained with these tw techniques suggests that a reasnable apprximatin f the extent f liftff lag can be btained withut sphisticated trawl psitining systems, s lng as a netsunder r ther means is available fr detecting the mment the trawl cmes ff bttm. Mving in the ther directin, twards increased sphisticatin and hpefully increased accuracy, we intend in future effrts t fit ur ITI system with a flux-gate rate-sensing cmpass s as t stabilize the heading parameters used by the ITI t calculate trawl psitin, and thereby btain better trawl psitin slutins during haulback and ther slw-speed peratins. Liftff lag is nly ne f the gear perfrmance issues that can be a ptential surce f bias r variability in sampling effrt. As nted by Wathne (1977) and many thers, the gear can arrive
n the bttm and ptentially start fishing quicker than expected, r alternatively it can fail t settle nt the bttm and int its fishing cnfiguratin fr substantial perids after it shuld have. Our survey prtcls, which relied n real-time bservatins f the gear s perfrmance t determine the beginning and endpints f each haul, ffered substantial prtectin frm such types f errr. Hwever, survey prgrams that d nt emply trawl instrumentatin are vulnerable tthese effects, and the degree f vulnerability increases as the survey s standard nminal tw duratin decreases. Acknwledgments While many have cntributed t the wrk described here, the authrs wish t ffer their special thanks t: the captains and crews f the fishing vessels Miss Lena, Blue Hrizn, Captain Jack, and Sea Eagle; Keith Bsley, Alisn Bailey, and Janelle Zimmerman fr their valuable assistance analyzing the data; and the scientific staff (especially the varius scientific party chiefs) wh participated in the survey activities.
References Byrne, C.J., T.R. Azarvitz, and M.P. Sissenwine. 1981. Factrs affecting variability f research vessel trawl surveys. h W.G. Dubleday and D. Rivard (Eds.), Bttm trawl surveys/relevq au chalut de fnd. Can. Spec. Publ. Fish. Aquat. Sci./Publ. spec. can. sci. halieut. aquat. 58. Chambers, J. M., W.S. Cleveland, B. Kleiner, and P.A. Tukey. 1983. Graphical methds fr data Analysis. Wadswrth, Belmnt, Califrnia. Cleveland, W.S. 1979. Rbust lcally weighted regressin and smthing scatterplts. Jum. Am. Stat. Assn., 74:829-836. Engas, A., R. Skeide, and C.W. West. 1997. The MultiSampler : a system fr remtely pening and clsing multiple cdends n a sampling trawl. Fisheries Research, 29(1997) 295-298. EngAs, A. 1994. The effects f trawl perfrmance and fish behaviur n the catching efficiency f demersal sampling trawls. b A. Femci and S. Olsen (Eds.) Marine fish behaviur in capture and abundance estimatin. Fishing News Bks, Oxfrd. Gd, O.R. 1994. Factrs affecting the reliability f grundfish abundance estimates frm bttm trawl surveys. b A. Femii and S. Olsen (Eds.) Marine fish behaviur in capture and abundance estimatin. Fishing News Bks, Oxfrd. Gd, O.R., M. Penningtn, and J.H. Vlstad. 1990. Effect f tw duratin n length cmpsitin f trawl catches. Fisheries Research 9: 165-179. Grsslein, M.D. 1971. Sme bservatins n the accuracy f abundance indices derived frm research vessel surveys. ICNAF Redbk 1971, Part III: 249-266. Gundersn, D.R. 1993. Surveys f fisheries resurces. Jhn Wiley & Sns, Inc. NY. Parsns, D.G. and E.J. Sandernan. 1981. Grundfish survey techniques as applied t abundance surveys fr shrimp. h W.G. Dubleday and D. Rivard (Eds.), Bttm trawl surveys/relevcs au chalut de fnd. Can. Spec. Publ. Fish. Aquat. Sci./Publ. spec. can. sci. halieut. aquat. 58. Wathne, F. 1977. Perfrmance f trawls used in resurce assessment. Mar. Fish. Rev., 39(6): 16-23. West, C.W., J.R. Wallace, and T.A. Turk. 1999. Mdeling trawl perfrmance during the retrieval perid. Presented at the meeting f the ICES Wrking Grup n Fishing Technlgy and Fish Behaviur, St. Jhn s, Newfundland.
ṿ - (\J.- ọ - c.q thru 365 366 thru 731 732 thru 1096 1097 thru 1236 Depth intervals (meters) Figure 1. Depth-related mean delay between the initiatin f haulback and the mment at which the trawl came ff bttm fr the tur 1999 NWFSC West Cast Slpe Survey vessels. VPH = Vessel's psitin a t beginning f haulback, VPL = Vessel's psitin at MOMent f gear liftff, Scpe-H = Scpe a t beginning f haulbgck, and Scpe-L = EstiMGted scpe Gt MMent f gear liftff, tgether with Depth are used t cglcula te: Offset-H = GeGr's hrizntal ffset behind vessel Gt h<lulbacl< and Offset-L = GeGr's.hrizntQl ffset behind vessel Qt Ilftff Tgether with VPH and VPL these are used t estimate: GPH = GeGr's, psitin a t beginning f hqulbacl<, Gnd GPL = GeGr's psitin at MMent f liftff, rnstgnce fished during liftff lag is the difference between GPH Gnd GPL VPL GPH ~ Offset-H.{ Offset-L ~ I Distance fish~d I r" during tiftff lag "I Figure 2. Schematic f the critical gemetry and variables used in the "1998 methd" (see text) t e sti mate the trawl's psitin at haulback and liftff. 10
50 TV O + D Miss L Blue Hrizn ø' - O> :3.~ E -30 C:- O> > u O> a: 40 20 Ir 10 v. u u 500 1000 Scpe (meters) 1500 2000 Figure 3. The time required t fully recver the twing warps at varying scpes. 11
200 400 600 800 1000 1200 (J > 1.5 1.0 B 0.5 0,0 E ~ "'O Q) cn '- Q) > 1.5 1.0 0.5 0.0 ~ ~~~~~-t~~0 ~ 00 O ~'-Q) () c ~-cn (5 80 ~O -~~ -'.-Y ~ 1.5 1.0 0.5 0.0 1.5 1.0 0.5 8 CO 00 9> (f!j ~ ' ~-~--;;I~ "9 00 ~ "~~-~ ~ :1:>0 08 0.0 I I I,, I I I, I I I 200 400 600 800 1000 1200 Depth (m) Figure 4. Estimated distance the net mved acrss the bttm during the liftff lag perid fr the tur vessels used fr the 1999 NWFSC West Cast Slpe Survey, calculated by bth the "m methd" and the "1998 methd." Lines thrugh the scatterplts were generated by a lcally-weighted regressin and smthing prcedure (Cleveland, 1979, and Chambers ~t al., 1983). 12
~ "" //0 00 O 6' '" 0.0 0.5 1.0 1.5 /' 1.5 /',/ /' ~.p-.,8'" ". /--: 1.0 0.5 00-0.0,/ ~/ '" /' /'./ /' /' 0.-" 0.0 0.0 0.5 10 1.5 Liftff lag (km) calculated using the "JTI Methd" Figure 5. Relatinship between the liftff lag distances estimated with the "m methd" and the "1998 methd." Slid lines represent the linear regressin flt t the data while the dashed line represents exact agreement. 13
~ I 200 400 600 800 1(00 1200 "(TI MØlhnil'! 00 II#-!-~~ O c9 O O O t)...& O O... '~-::;. O O 8 O O d' O 4 3 2.- Cl> +-- c 6 --== ~w~ 5 3 2 ::::.~~~~ --- ~ii~~~" 0<9 O O Q:> O ~ O @ O O O O O.. - Q) Q) D- U) 1"11'1 MølhM" - 'b -~QR MøthM" 4 3 2 -"'TI M;;lh;',i" O "1QQA M"thnn"~ ~ ~ Aj~~~-~-~ r-:~ do 0000 I I 200 400 600 800 1000 1200 Depth (m) Figure 6. Estimated speed f the trawl' s advance acrss the bttm during the liftff lag perid fr the tur vessels used fr the 1999 NWFSC West Cast Slpe Survey, calculated by bth the "m methd" and the "1998 methd." The nminal standardized twing speed was 2.2 knts. Lines thrugh the scatterplts were generated by a lcally-weighted regressin and smthing prcedure (Cleveland, 1979, and Chambers et al., 1983). 14