© 2004 by ICES/CIEM International Council for the Exploration of the Sea/Conseil International pour l'Exploration de la Mer
Reliability of a model based on a short fishery statistics survey: application to the Northeast Atlantic monkfish fishery
a ECOBIOMAR, Instituto de Investigaciones Marinas C. Eduardo Cabello 6, 36208 Vigo, Spain
b Economía Aplicada, Facultad de Ciencias Económicas y Empresariales, Universidad de Vigo Spain
*Correspondence to F. Rocha: tel: +34 986231930; fax: +34 986292762. e-mail: frocha{at}iim.csic.es.
A model based on a short fishery statistics survey was applied to estimate catch and catch per unit effort (cpue) of the Galician monkfish (Lophius spp.) trawl fleet during 1998. In all, 35 interviews were conducted with fisheries personnel in ICES Divisions IXa, VIIIc, VIIId, and VIIIe (coastal offshore fishing grounds) and 44 in VIIb, VIIc, VIIj, and VIIk (Grand Sole fishing ground). Reliability of the model estimates was tested using: (i) registered fish market landings at 38 ports; (ii) landings data from sale invoices at six of these ports (93.8% of total landings of the species); (iii) 29 observers' trips made on board coastal offshore trawlers and the logbook of one Grand Sole trawler. Estimated mean total catch from the model was 5110 t (602 t coastal offshore, 4508 t in Grand Sole) and cpue values were 36.2 kg haul–1 vessel–1 coastal offshore and 104.4 kg haul–1 vessel–1 Grand Sole. Differences between the values of cpue estimated by the model and those determined directly in both fisheries were not significant. However, there were differences between the total catch estimated by the model and the total landings deduced from sale invoices and Galician fish market information.
Keywords: catch estimates, fishery statistics, Lophius, monkfish, survey methods
Received 12 June 2003; accepted 5 October 2003.
 |
1 Introduction |
|---|
 Top 1 Introduction 2 Material and methods3 Results4 DiscussionAppendixReferences |
|---|
Some countries have made good progress in installing input controls on their fisheries, notably control of fleet size, but such measures rarely include specific controls on total fishing effort (Caddy, 1999). Therefore, fisheries management advice is still based largely on output controls such as catches and quotas, and corresponding effort data. However, when fisheries statistics are incomplete or inaccurate, scientists have to resort to alternative methods of generating the data on which to base their advice. For example, alternative methodologies and models have been used variously to estimate catch and catch per unit effort (cpue) in small-scale fisheries (e.g. Gómez-Muñoz, 1990; Pollock et al., 1994, 1997; Simón et al., 1996; Hoenig et al., 1997; Kirchner and Beyer, 1999; Neis et al., 1999a, b). The reason why such model-based estimates have had to be made is generally that the availability of accurate catch statistics is confounded by the particular characteristics of such fisheries (seasonality, variability, dispersion), as well as the peculiarities of the markets.
Galicia is the home of Spain's and Europe's biggest fishing fleet. In addition to fishing vessels operating in distant-water fisheries, some 5300 vessels fish in the Northeast Atlantic (Figure 1). Most of these (>2800, all <10 m long) operate inshore. In addition, 2000 vessels (10–30 m long) fish in coastal offshore waters, using trawls, gillnets, longlines, traps, and seines. Another 150 Galician trawlers and longliners (25–40 m long) fish in Grand Sole waters off southwest Ireland (ICES areas VIIb, VIIc, VIIj, and VIIk) (Figure 1). Both coastal offshore and Grand Sole trawl fisheries target hake (Merluccius merluccius), blue whiting (Micromesistius poutassou), Norwegian lobster (Nephrops norvegicus), horse mackerel (Trachurus trachurus), and monkfish (Lophius piscatorius and Lophius budegassa). The two species of monkfish are not separated in official statistics and are sold as one product.
|
Although a huge amount of effort has been made by several European countries to improve the reliability of their fishery statistics, unreported and misreported catches are still common, notably for some commercially important species regulated through imposition of a total allowable catch (TAC). Therefore, and in order to achieve sustainable utilization of such marine resources, it is vital that the extent and quantity of unreported and misreported catches is minimized, and that the cpue be estimated accurately. This information is crucial for those investigating multispecies, multigear fisheries, the case for most of the fisheries prosecuted by Galician fishers.
Gómez-Muñoz (1990) developed a model to estimate catch and cpue in small-scale multispecies fisheries, and it has been applied successfully in Mexico (Gómez-Muñoz, pers. comm.) and in the squid and clam fisheries of Galicia (Simón et al., 1996, unpublished data). The method involves surveying fishing sector personnel to obtain the basic model parameter data. However, this model has to be broadened if it is to be applied to larger scale fisheries. The main objective of the current study is therefore to develop the Gómez-Muñoz method of questionnaire and interview so as to be able to apply it to the estimation of catch and cpue in large-scale commercial fisheries. We also document some of the statistical assumptions necessary to determine the model's precision and the errors of its estimates, and test the model's reliability by comparing its outputs with landing statistics and data obtained by shipboard observers in the monkfish fishery.
|
2 Material and methods |
|---|
|
Top1 Introduction2 Material and methods3 Results4 DiscussionAppendixReferences |
|---|
2.1 The model
Figure 2 is a flowchart of the model. There are two phases: phase 1, in which the main parameters for the model are determined through port interviews with fishers; and phase 2, when the values of these main parameters are used to estimate the catches for each vessel type and for the whole fleet.
|
2.1.1 Phase 1: variables and parameters
The main parameters of the model are listed in Table 1. For all data obtained through interview, M, I, S, and L are defined as the mode, and Cmax and Cmin as the mean catch per haul. Assuming that these parameters are normally distributed, the mean or the mode of the interview data are used for estimation.
|
A set of catch curves based on secondary variables and parameters was estimated from these main parameters. To ensure that the month of maximum catch coincides with the origin (x(M)=0), the data set was time-transformed (x(t)). The curve showing the rate of decrease was calculated in such a way that the rate was always between zero and +1. The type of curve was determined from the interview data and defined by their degree of asymmetry (Figure 3), measured by the parameter TE in the model. TE is the relationship between the minimum and the maximum times to or from Cmax.
|
If the decrease curve is type 1 (a slow decrease), then I=1 and
|
| (1) |
|
|
|
| (2) |
|
|
Transformation is different for each type of curve. If the decrease curve is type 1 (slow), then
|
| (3) |
If the decrease curve is type 2 or 3 (intermediate or rapid), then
|
| (4) |
The term Month is used to calculate x(t) in all cases. Later, x will substitute x(t) to simplify the equations, except where the context is not clear. Then the weighting function
|
| (5) |
2.1.2 Phase 2: estimation of monthly catch per haul, per trip, and per vessel
The steps to generate the monthly catch per haul per vessel are shown in the Appendix. Mean monthly catch per haul per vessel (Ct) was estimated as the mean of a uniform distribution from
|
| (6) |
The mean monthly catch per trip per vessel (CLt) was estimated from
|
| (7) |
|
| (8) |
2.2 Total catch of the fishing fleet per fishing season
The first step in determining this is to estimate the average catch per trip per vessel during the fishing season (Cmean). If the average catch per haul per month is used, then
|
| (9) |
|
| (10) |
The total catch of the fishing fleet per fishing season (CT) can then be estimated under the assumption that all fishing units operating with each type of fishing gear would land the same catch. CT can be estimated in three ways:
- If the total number of trips undertaken for the whole fleet (V) is known, then
(11)
- If V is unknown, then
where v(t) is the average number of monthly trips per vessel and B is the number of vessels.
(12)
- Conversely,
(13)
2.3 Sample size
Given the model definition above, it may be assumed that the means of the catches conform to a normal distribution when the sample size (n) is sufficiently large. If random samples of size n are drawn for each combination (of species, fishing gear, and fishing ground), the standard errors (s.e.) of the variable catches (sc) must be known from previous samples. With sc known and a confidence level of 95%, the sampling error for each combination would be
|
| (14) |
|
| (15) |
Assuming a percentage error (ep) of 10%, we can recalculate the sample size as a function of this error, i.e.
|
| (16) |
the sample size, would be given by
|
| (17) |
If we take a size nt for each month, the total sampling size is
|
| (18) |
|
| (19) |
|
| (20) |
2.4 Outliers
Before estimation of the main parameters in the model (Table 1), the "outliers" specified by the person interviewed were discarded. In particular, it is not possible for M to be before S or after its end (S+L–1); M must therefore lie during the fishing season. In other words, all observations that fulfil the requirement M<S if S+L<12 or M+12>S+L, were discarded.
Owing to the uniformity of the catch distribution and because the distribution of the maximum catch departs from normality (see Appendix), the statistical method of Tukey (1977) was used to exclude outliers from the interview data. Consequently, any Cmax per haul, vessel, or trip was considered to be an outlier if it did not fall within the interval Me±1.5H, where Me is the median of the maximum catches and H the distance between the third (Q3) and the first (Q1) quartile of those catches
|
| (21) |
2.5 Area of study and interview data
The area of study was the fishing grounds on which the Galician fleet operates in the Northeast Atlantic. These fishing grounds (see Figure 1) were:
- coastal offshore, comprising the continental shelf and slope waters of ICES Divisions IXa, VIIIc, VIIId, VIIIe;
- Grand Sole, fishing grounds in ICES Divisions VIIb, VIIc, VIIj, VIIk.
A preliminary visit to the 82 fishing ports of Galicia was made to establish contacts with the personnel at each port and to determine which of them recorded landings of monkfish. In all, 38 ports were selected for this study, and weekly visits were made to each during 1998 and 1999 to obtain general fisheries data on each, and on the number and characteristics of the fleets operating in 1998. A sampling network was established, and observers regularly visited the six most important ports for monkfish landings (Vigo, Marín, Ribeira, Coruña, Celeiro, Burela), the regularity of visit determined by the volume of monkfish sold at the market in each port. Observers checked each sale invoice to monitor the landings for each vessel during 1998. More data were obtained from 29 observers who accompanied coastal offshore vessels targeting monkfish during 1998 and 1999. The 1998 logbook of a trawler operating in both the coastal offshore and Grand Sole areas was also obtained. Finally, focused interviews with fishers, skippers, and fishing sector personnel were carried out according to the same protocol in the 38 ports selected. Interviews covered each target species, type of gear, and fishing ground. The interview protocol focused on obtaining the information required for the model: (a) name and profile of the person interviewed; (b) name and technical data of the vessel; (c) number of trips per month; (d) gear/s normally used; (e) fishing ground/s usually visited; (f) target species; (g) number of hauls per trip; (h) duration and depth of hauls; (i) schedule of fishing activities (average amount of time spent steaming to and from the fishing ground, and on operations other than fishing, and the amount of fishing effort directed at the target species per trip); (j) trip duration; (k) actual time spent fishing (time between shooting and recovering the net); (l) catch of each target species per haul, specifically the maximum and the minimum catch; (m) the month when the fishing season starts; (n) the months of maximum and minimum catch during the fishing season; and (o) the rate of decrease of catches during a fishing season (slow, intermediate, or rapid). Symbols used in the interviews and their meaning are summarized in Table 1. Although interviews gave information on a large number of different gear types, the current analysis is restricted to trawlers only. This was because only the interviews for trawlers attained the minimum number required by the model (see section Sample size).
2.6 Reliability of the model
To test the reliability of the model and to estimate its precision and bias, a cross-check between the catch and cpue of monkfish obtained from shipboard observers, logbook, sale invoice sampling network data, and fishing market statistics, and the cpue estimated by the model was undertaken, applying a t-test and single-factor ANOVA (Zar, 1999).
|
3 Results |
|---|
|
Top1 Introduction2 Material and methods3 Results4 DiscussionAppendixReferences |
|---|
3.1 Model estimates of monkfish catch and cpue
In all, 79 focused interviews on monkfish were carried out in the 38 ports selected. The interviews were distributed as follows: 35 with fishers operating on coastal offshore grounds and 44 with fishers operating on the Grand Sole grounds. The minimum sample sizes to obtain catch estimates with an error of 10% were 33 and 41 interviews for coastal offshore and Grand Sole, respectively. None of the 79 interviews was totally rejected. However, some incoherent or outlier data for each of the ten variables obtained from each interview were rejected. The maximum number of rejected data for each variable was <10%.
Table 2 lists the parameter values estimated for coastal offshore and Grand Sole trawl fisheries for monkfish. Table 3 lists the cpue, the monthly and total catches per trip, the mean catch (kg) per haul and vessel (Ct), and the range and s.e. of the mean estimated.
|
|
In all, 231 vessels registered in 36 Galician ports prosecute the monkfish coastal offshore trawl fishery. Interviews revealed that those vessels catch monkfish in one of every five hauls per trip (n=1) and in six of every 18 trips per month (v=6). Observers on board such trawlers noted that monkfish were caught in 11 of every 29 trips. The Grand Sole trawl fishery targeting monkfish utilizes 75 vessels registered in six Galician ports. These vessels made only one trip per month (v=1) and carried out an average of 48 hauls per trip (n=48).
Based on model estimations and considering the number of vessels in each port and the total catches by port, the mean, the maximum, and the minimum total catch for the two fisheries were estimated (Table 4). Total Galician catches for coastal offshore and Grand Sole trawl fisheries were 602 t (566–638 t) and 4508 t (4248–4768 t), respectively.
|
3.2 Cross-check between estimated and observed data
The average monkfish catch per haul from the four different sources of data for the coastal offshore fishery is summarized in Table 5. In comparing the values in Table 5 with model estimates of cpue, it must be noted that:
- vessels with observers on board were only six times (v=6) on the commercially viable monkfish fishing grounds;
- the sale invoice and fish market data used for the analysis of catch per haul do not give temporal or geographic information about the monkfish catch.
|
This limitation of the information implies that cpue would be underestimated, because trips undertaken without monkfish as the target species were included in the calculations. The differences (p>0.05) between catch per haul determined from interview and catch per haul obtained from other sources (Tables 3 and 5) were not significant, likely because of the very high standard deviation.
In the Grand Sole trawl fishery for monkfish, data on the number of hauls per trip and on the monkfish catches per haul were obtained from the logbook of a single vessel for the period 1997–1998. The same trawler was inspected several times by EU observers. During the period, the vessel carried out 46.9 (s.e. 12.0) hauls per month and monthly caught 77.8 kg (s.e. 22.4 kg) per haul and 3476 kg (s.e. 1123 kg) per trip. There were no significant differences (p>0.05) between the mean monthly catches per haul estimated from the model and from the logbook data. The total annual catch of monkfish by the same trawler was 60 147 kg, just 0.1% different from the amount estimated by the model (60 108 kg in Table 3).
Table 6 lists the total annual landings (kg) of monkfish in the six main fishing ports during 1998, and for comparison, the model estimates of total annual catch. The same table also gives a comparison of the total catches and landings of the whole Galician fleet obtained from three different sources.
|
|
4 Discussion |
|---|
|
Top1 Introduction2 Material and methods3 Results4 DiscussionAppendixReferences |
|---|
Previous studies using interview data have revealed that a large amount of information from fishers, useful for assessment purposes, can be collected from fishers themselves and from landing points (Simón et al., 1996; Pollock et al., 1997; Kirchner and Beyer, 1999; Neis et al., 1999a). Personal interviews provide much detailed information that will increase the likelihood of precision and reliability in model estimates (Neis et al., 1999a). For example, the interviews revealed that monkfish catches by coastal offshore trawlers are occasional, vessels only catching monkfish in one of every five hauls per trip and in only six of every 18 trips per month. Such information allows us to improve the accuracy of the model parameters and avoid overestimating catches. However, because the information is based on the appraisals and memory of fishers, a minimum number of interviews by species and fisheries are required. As fishers have different interpretations of the variables used in the model, the interviewer must be able to interpret the data fairly, though there should be no prior expectation of the result of the interview (Simón et al., 1996; Neis et al., 1999a, b). Notwithstanding, there must still be some reliable means of rejecting outliers. The results given herein lend some confidence to the means of rejecting outliers from the current analysis.
There are two main requirements for the model to be used optimally. First, variables or parameters must be estimated accurately. From this analysis, the most reliable estimates seem to be the average values of Cmax and Cmin, and the month when Cmax is most frequently attained (Gómez-Muñoz, 1990; Simón et al., 1996). Second, for total catch in any port or for the total fleet to be estimated accurately, the true number of vessels fishing in a given area or landing at a specific port has to be known precisely. If vessels vary their port of landing, target species change, or preferred fishing ground varies, then the discrepancies between estimated and true catches will be large. In order to avoid invoking this bias, we were careful to determine the true number of vessels landing in the six main Galician ports during our period of study. Ideally, we should have done this for all ports, but owing to the complexity of the Galician fishery, such an exercise would have been extremely expensive and time-consuming. Therefore, we feel that, given that those six ports account for 94% of the total Galician landings, we did the best we could with the resources at our disposal.
The discrepancies in the landings by port using the different sets of data ranged from 9.6 to 44.8% (Table 6), part no doubt resulting from some vessels changing their port of landing during the year of study, and part because landings at some ports were trucked to other ports for economic reasons, and there registered as having been landed there. Finally, a small portion of the landings does not pass through the official market. For the current analysis, however, the effect of these biases was minimized, because the total catch was estimated for the whole fleet rather than the fleet representing each port.
Values of catch per haul and per trip estimated for both fishing areas were similar to the true data obtained from shipboard observers and logbook data, respectively. Considering the variability in monkfish catches, the minimum number of interviews needed to apply the model adequately for each combination of gear and fishing ground was relatively low. This requirement was complied with for the interviews carried out in both areas.
For both coastal offshore and Grand Sole monkfish fisheries carried out by Galician trawlers, it is of note that the total catches estimated by the model are generally higher (12.3%) than those reflected by the fish market statistics. Furthermore, the values obtained from sale invoices are not within the confidence intervals estimated from interview parameters (Table 6). Such differences can be explained by the characteristics of Galician fisheries. As indicated by Simón et al. (1996) and through data collected during interviews, at least some portion of the landings does not appear in official statistics. Depending on socio-economic conditions and local control at each port, the amount of unregistered landings varies. Thus, data obtained from sale invoices at each port would seem to offer more accurate information than the official data from fish markets. On the basis of the information collected during the interviews, we conclude that the discrepancy cannot be attributed to discarding at sea, so the 12.3% difference between model and fish market data would represent the unreported catch in the official statistics.
Our analysis has demonstrated that the model can estimate catches and cpue in both small-scale (Simón et al., 1996) and large-scale fisheries with accuracy and reliability. The model-based estimates were significantly larger than those based on fish market receipts and invoices (Table 6), implying that this sampling approach yields better estimates of total catch than those based on official statistics.
In conclusion, the methodology developed herein provided statistical validation of our model which, on the basis of interview data, can be used as an independent means of estimating catch and effort in a fishery and to test the reliability of landing statistics. However, while some fishers continue to ignore TAC and other regulations, better knowledge of catch will still not lead to better fisheries management.
|
Appendix |
|---|
|
Top1 Introduction2 Material and methods3 Results4 DiscussionAppendixReferences |
|---|
A.1 Monthly catch determination
Let us assume a given period of fishing starting at month S and a fishing season with a duration of L months. Let M, the month during which the maximum catch is taken, be known. M must occur between S and S+L. In some cases, S+L can be higher than 12 so, to avoid negative values, periods of 24 months were considered. In this case, if M is lower than S, then M* = M + 12. Therefore, the parameter representing the month of maximum catch would be
|
|
|
|
Let Ct be the cpue in month t, t=S,...,S+L–1. Note that Ct occurs in an "ideal" vessel, applying the average of the interview data normal distribution. Owing to the absence of prior information about catches in a given month, it is assumed that Ct is a sequence of random, uniform, and independent variables, so the information obtained on the catches in month t is independent of previous catches but dependent on some particular exogenous parameters for each fishing period (decrease curve followed by the catches, I; month when the maximum catch occurs, M; actual month, t).
The parameters of the uniform distribution are variable, so
|
|
1), corrected by a factor that is a function of the time until the month of peak catch and of the type of decrease curve. Hence, if |
|
|
|
|
|
S2: the maximum theoretical catch in period t depends on the total maximum catch (2), corrected by a factor that is a function of the time that remains until the month of peak catch and of the type of decrease curve. Hence, if
|
|
|
|
|
|
According to this model, the global maximum will be obtained in M, and the global minimum in S+L–1 in the first case (I=1), and in S in the other two cases (I=2 or 3). Consequently, the weighting factors must be modified to avoid coincidence of maximum and minimum.
A.2 Distribution of Ct
Let (u1,..., un) be the observations of one uniform variable in a specified month, where u
(1,2) with unknown parameters.
Let random variables
|
|
Then the distribution of X is
|
|
|
|
|
|
|
|
The distribution of Y would be
|
|
|
|
|
|
|
|
|
|
The combined distribution x and y is
|
|
|
|
|
|
|
|
=2–1 the support interval |
|
|
|
Hence, the combined density function is
|
|
The covariance of x and y will be given by
|
|
|
|
|
|
Introducing this change, different values of error will be obtained. The mean will be obtained directly as
|
|
|
|
|
|
A.3 Standard errors
The standard errors (s.e.) of the estimates of catch per trip (CLt) and total catch per month (Ctot) are readily derived from the equations
|
|
|
|
Then
|
|
|
|
|
|
|
Acknowledgements |
|---|
We thank V. Verdugo and I. Cal, University of Vigo, for their valuable assistance in statistical development of the model and an anonymous reviewer and A. I. L. Payne for their corrections to and helpful comments on the manuscript. We are also grateful to J. Castro, M. T. Fernández, E. Ledo, and M. C. Rodríguez for technical support, and all the fishers, skippers, and other members of the fishing fraternity who allowed themselves to be interviewed during the study. The research was funded by EU Study Project (CE 97/0107) entitled "Development of software to estimate unreported and misreported catch and effort data and to apply fishery management models".
|
References |
|---|
|
Top1 Introduction2 Material and methods3 Results4 DiscussionAppendixReferences |
|---|
-
Caddy J.F. (1999) Fisheries management in the twenty-first century: will new paradigms apply? Reviews in Fish Biology and Fisheries 9:1–43.[CrossRef][Web of Science]
Gómez-Muñoz V.M. (1990) A model to estimate catches from a short fishery statistics survey. Bulletin of Marine Science 46:719–722.[Web of Science]
Hoenig J.M., Jones C.M., Pollock K.H., Robson D.S., Wade D.L. (1997) Calculation of catch rate and total catch in roving surveys of anglers. Biometrics 53:306–317.[CrossRef][Web of Science]
Kirchner C.H. and Beyer J.E. (1999) Estimation of total catch of silver kob Argyrosomus inodorus by recreational shore-anglers in Namibia using a roving–roving creel survey. South African Journal of Marine Science 21:191–199.
Neis B., Felt L., Haedrich R.L., Schneider D.C. (1999) An interdisciplinary methodology for collecting and integrating fishers' ecological knowledge into resource management. In Newell D. and Ommer R. (Eds.). Fishing Places, Fishing People: Issues and Traditions in Canadian Small-Scale FisheriesUniversity of Toronto Press pp. 217–238 412 pp.
Neis B., Schneider D.C., Felt L., Haedrich R.L., Fischer J., Hutchings J.A. (1999) Fisheries assessment: what can be learned from interviewing resource users? Canadian Journal of Fisheries and Aquatic Sciences 56:1949–1963.
Pollock K.H., Hoenig J.M., Jones C.M., Robson D.S., Greene C.J. (1997) Catch rate estimation for roving and access point surveys. North American Journal of Fisheries Management 17:11–19.[CrossRef]
Pollock K.H., Jones C.M., Brown T.L. (1994) Angler survey methods and their applications in fisheries management. American Fisheries Society Special Publication 25: 371 pp.
Simón F., Rocha F., Guerra A. (1996) The small-scale squid hand-jig fishery off the northwestern Iberian Peninsula: application of a model based on a short survey of fishery statistics. Fisheries Research 25:253–263.[CrossRef][Web of Science]
Tukey J.W. (1977) Exploratory Data Analysis. Addison-Wesley Series in Behavioural Science, Quantitative Methods, 16(Addison Wesley Publishing Company, Massachusetts, USA) 506 pp.
Zar J.H. (1999) Biostatistical Analysis 4th edition (Prentice Hall, New Jersey, USA) 928 pp.
CiteULike 
Connotea 
Del.icio.us What's this?
This article has been cited by other articles:
|
|
 |
|
  A. Forcada, C. Valle, J. L. Sanchez-Lizaso, J. T. Bayle-Sempere, and F. Corsi Structure and spatio-temporal dynamics of artisanal fisheries around a Mediterranean marine protected area ICES J. Mar. Sci., October 26, 2009; (2009) fsp234v1. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||