CARICOM Fisheries unit
Kingstown, Saint Vincent and the Grenadines
The first Meeting of the Working group reviewed available information on the biology and assessment of flyingfish fisheries within the Eastern Caribbean, which supported the need to coordinate a collection of statistics, assessment and management of flyingfish at the subregional level. During the first Meeting, therefore, the Group completed a preliminary inventory of catch and effort data and a comparative analysis of available time series of flyingfish landings. This analysis implied some correspondence in interannual variability of landings reported by Barbados and Tobago and hence reinforced the need to manage the Eastern Caribbean flyingfish resource as a single stock. During the first Meeting, the Group also identified certain tasks to be completed, in preparation for a second Meeting of the Working group.
In terms of catch and effort data analysis, it was agreed that available catch and effort data from all participating fisheries should be assembled and evaluated prior to the second Meeting of the Working Group. If possible, countries would try to improve their data in order to facilitate estimation of total catch and abundance trends.
Some data were reviewed prior to this Meeting and country representatives were advised of a suitable format for preparation of their data for analyses. However, most data were not ready before the Meeting. During this Meeting, therefore, participants focused mainly on evaluating the quality of catch and effort data, preparing datasets for catch rate standardizations and developing and interpreting standardized catch rate time series for each participating countrys fishery. The analyses conducted were preliminary, but highlighted certain deficiencies in the data collection systems, which would need to be addressed in the near future.
The aim of the second Meeting of the Working Group was:
- to examine and to determine the quality of available landings and effort data,
- to estimate annual total landings,
- to standardize effort data,
- to develop and examine standardized annual catch rate series for each fleet/country,
- to prepare fish stock management advice based on results of analyses.
3. DATA USED
Barbados - Landings and fishing trip data were available for the period 1959 - 2000. The adoption of the TIP database system in 1995 encouraged collection of additional data on vessel type and length, gear type, area fished and crew size. Data on fishing time (i.e. hours fished) and depth fished were not collected, although the TIP database could accommodate these data. Catch rate was therefore estimated as weight of flyingfish caught per fishing trip. During the period studied, the Barbados flyingfish fishing fleet consisted of two types of vessels: dayboats and iceboats. The dayboats measured 5 - 12 m in length and fishing trips were restricted to one-day trips. The iceboats ranged from 10 m to 17 m in length and fishing trips normally lasted 5 - 8 days.
Tobago - In the present analysis, data were available on monthly landings and fishing trips at three major landing sites (Pigeon Point, Buccoo and Mt. Irvine) during the years 1988 - 1999. Catch rate was therefore estimated as weight of flyingfish caught per fishing trip. The Tobago flyingfish fishing fleet consists mainly of small dayboats and a few iceboats. Tobagos dayboats and iceboats are similar in size and capacity to Barbados dayboats and iceboats respectively. At the time of the present study, the available data did not permit distinction of boat type and this was therefore a significant confounding factor in the analyses.
Dominica - Landings data for the period 1994 - 1999 only were available for the present analysis. Fishing trips were one-day trips. The predominant gear for the fishing trip was also usually recorded. Catch rate was estimated as weight of flyingfish caught per fishing trip. Sampling coverage was estimated to be approximately 70 % during 1994 - 1995 and 60 % during 1996 - 1999.
Grenada - At the time of analysis, landings data only were available for the years 1984 - 1999. During the 1980s, the data collection system did not capture substantial landings, which were not sold within the main markets. Additionally, sampling coverage at the markets was considered to be low. During the 1990s, flyingfish became more important as bait for the large pelagic fishery than as a food fish and so much of the catches were again not recorded. Given the available information, it was not possible to estimate sampling coverage rate, which clearly decreased as the fishery, evolved and hence obtain annual estimates of total landings. This limited the usefulness of exploratory analyses and interpretation of catch trends of the Grenada flyingfish fishery.
Saint Lucia - Landings data for the period 1995 - 1999 only were available in a format suitable for the present analysis. Fishing trips were one-day trips. The predominant gear for the fishing trip was also usually recorded. Catch rate was estimated as weight of flyingfish caught per fishing trip. Flyingfish were normally harvested as part of a multispecies fishery. The field sampling program during the period studied did not permit identification of those fishing trips for which flyingfish were the primary target species. The flyingfish fishery in Saint Lucia has been highly opportunistic, dependent upon conditions such as abundance of flyingfish and market demand which change with time of the year and abundance of other species such as dolphinfish and wahoo.
Saint Vincent & the Grenadines - There is no target fishery for flyingfish in Saint Vincent and the Grenadines. Some flyingfish catches were recorded for two months in two years only during the middle to late 1990s. Fishing trips were one-day trips and were the measure of effort recorded. The predominant gear for the fishing trip was also usually recorded. Given the amount of available data, a crude estimation of total annual landings was attempted only.
Martinique - The sampling program for the flyingfish fishery is now being developed, along with the database used for storing the data.
4. STATISTICAL AND ASSESSMENT ANALYSES
Individual country representatives conducted a number of descriptive and exploratory analyses of available catch and effort data.
4.1 Exploratory Analyses
- to examine frequency tables of the variables/factors vessel type, vessel length, crew size, gear type and other available effort data fields
- to examine cross-tabulations of data field combinations believed to be significant. e.g. vessel type*gear type, vessel type*landing site, gear type*site, landing site*month, and month*year
- to obtain basic descriptive statistical measures of the available landings and effort data,
- to examine the distribution of both un-transformed and transformed landings data to determine which type of data would be more appropriate for catch rate standardization trials using General Linear Model (GLM) analysis
- to examine spatial and temporal trends in reported catches and effort.
Prior to commencement of exploratory analyses, most country representatives required time to prepare their datasets in an appropriate format for analysis. The Tobago data had to be re-typed and the dataset from Dominica was incomplete at the start of the Meeting. Participants therefore spent almost 2 days to prepare the datasets for Tobago and Dominica. In the case of Barbados, the representative also spent almost 2 days to incorporate a number of revisions, involving a large part of the dataset.
Standard exploratory analyses were conducted using the statistical package SPSS. Frequency tables were generated to determine the categories of gear, vessel type and size and crew size involved in the fishery, to determine the extent and regularity of sampling through time (months and years) and space (landing sites) and to perform general quality checks on the data. Cross-tabulations of combinations of key factors were constructed to determine potential significance of these interactions to fishing activities.
The Group also examined Stem and Leaf and Box and Whisker plots of untransformed and transformed landings data by the lowest time stratum available, i.e. month. Observed changes in landings with changes in month, year and landing site were examined graphically for each countrys fishery. Total annual effort by country was also estimated, using the formula: total effort = total catch/catch per trip.
4.1.3 Results and Discussion
In all datasets, a number of incorrect gear codes were identified and corrected during the workshop. The Stem and Leaf and Box and Whisker plots of Barbados and Tobago flyingfish data indicated skewed distributions for untransformed catch rates (catch per trip), with large numbers of zero values recorded. Removal of zero data records reduced the skewness of the distributions. Additionally, the non-zero data were transformed by calculating the logarithms of the observed values. Corresponding Stem and Leaf and Box and Whisker plots of the log-transformed non-zero catch data approached normal distributions.
Except for Barbados, it was not possible to distinguish neither among the various vessel types, nor between those vessels, which targeted flyingfish and those vessels for which flyingfish was a bycatch species only. It should be noted that participants concluded that the targeting of flyingfish would have been affected by factors such as time of year, market prices and abundance of other higher-value species. In the case of the Barbados iceboats, it was also noted that flyingfish was not the primary target species during the first 2 - 3 days of a fishing trip. The results of the exploratory analyses were therefore limited by the available data, but were useful for general data quality checks and to obtain basic descriptive statistics such as mean catch by month and by year. Further details are noted in individual country logs. Fig. 1 shows the total reported ladings by country for the period 1989 - 1999. Clearly, the Barbados fishery has been the most important. The decreases in flyingfish landings in Tobago during 1997 - 1999 was believed to be due to problems of reduced access by fishers to a major landing site (Pigeon Point) and lack of guaranteed primary fish processing facilities.
The Group examination of a clustered bar chart showing trends in total annual fishing effort. However, participants queried the accuracy of the results, which indicated that Saint Lucia carried out an unrealistically large number of fishing trips during 1997 - 1999. Additionally, the amount of effort obtained for the different categories of vessels within the Barbadian fleet appeared incorrect and would have to be checked. In view of limited time, it was not possible to redo the analyses of total effort during the Meeting. The estimates of total annual effort for Barbados and Saint Lucia would therefore have to be re-done prior to the next Meeting of the Working Group.
It was decided to use only non-zero catch rate data, both untransformed and log-transformed for input into the GLM standardization analyses. This accommodated the underlying assumption of a normal error distribution.
4.2 Development and Standardization of Catch Rate Time Series
- to develop and standardize time series of catch rate data,
- to determine trends in flyingfish abundance for the time periods investigated.
Given that flyingfish is considered to be an annual species and the well-documented seasonality of fish availability and abundance, the time series of catch rates were developed for fishing seasons rather than calendar years. A fishing season was defined as the time period commencing in October of one year and finishing in September of the following year.
Weight of flyingfish landed per fishing trip was used as the measure of catch per unit of effort (CPUE) or catch rate in all instances. Except for the iceboats, the length of a fishing trip for other fishing vessels normally ranged from 8 to 11 hours. Notwithstanding and given the possibility of country differences in fishing techniques and areas fished, a separate annual CPUE series was developed for each country using available monthly data on landings and fishing trips. As noted previously, the data currently available from Tobago did not distinguish between dayboats and iceboats and so the time series developed for Tobago was mixed. In the case of Barbados, two annual catch rate series were developed: a dayboat catch rate series for the seasons 1995 - 2000 and an iceboat catch rate series for the seasons of 1995 - 2000. The dayboat catch rate data for the seasons of 1995 - 2000 were also used to update a historical time series of CPUE data (26 years) which had been developed from manual examination and extraction of data records kept by the main flyingfish markets operating during that time (Mahon, 1989).
Available CPUE time series were standardized using the technique of General Linear Modelling (GLM) available within SPSS (see also McCullagh and Nelder, 1989). The consultant briefly presented and reviewed the methodology:
GLMs are of the form: Y = SbiXi + e, where the variable Y is modelled assuming a multilinear regression relationship with several
significant factors or variables (Xi) and the error (e) follows an exponential distribution.
In the case of the present study, season was included as a factor in order to obtain standardized CPUE estimates through time. There was a general and lengthy discussion about patterns of changes in catchability and the possible ways in which these patterns may be relevant to the Eastern Caribbean flyingfish fishery. It was noted that catch rates varied on a monthly basis and were also affected by moon phase and time of day. Participants discussed the need to include any other factors or variables believed to influence catchability of flyingfish. Fishing area may affect catchability, but no detailed data on area were available. In addition, it was assumed that fishing techniques have not changed over time for the different types of boats. This assumption was probably not valid and data should be gathered in future to facilitate some measure of increases in fishing efficiency.
In the case of Barbados, the factors/variables included in the GLMs were: season, month and vessel length. In other instances, the following factors were examined: season, month and landing site. All 2-way interaction effects were also investigated for all CPUE series. The terms included in the final models adopted depended mainly on their contribution to the sum of squares explained, the significance of the effect (F - test) and the reduction in R - squared value due to their omission.
4.2.3 Results and Discussion
Barbados - The final GLM for the Barbados dayboat and CPUE data collected during the seasons of 1995 - 2000 included the factors season and month and the variable vessel length. The GLM of the historical dayboat CPUE series included the factors season and month only, as data on vessel length were not available. The shorter time series for the Barbados dayboats and iceboats showed increases from 1996 to 1997 onwards, but these increases appear to be either levelling off or decreasing to previous levels during the most recent years (see figs. 6 and 7 of Barbados log). Given the short time period investigated, the observed trends must be interpreted with caution. The updated historical CPUE series showed a clear increase from 1979 to 1988, after which the CPUE values dropped to pre-1979 levels (see fig. 8 of Barbados log). It was suggested that the upgrade of several dayboats to iceboats during the early 1990s could have caused the apparent decline, as the more successful dayboats were probably the first dayboats to upgrade to iceboats. The establishment of the Bridgetown Fisheries Complex encouraged boats from smaller landing sites to land their flyingfish at the larger market. It is possible, therefore, that the increased sampling coverage of the less successful vessels from smaller landing sites could also have contributed to the decline observed after 1988.
Tobago - The standardized CPUE series is shown in fig. 2. The series showed little fluctuation during 1989 to around 1993, after which mean CPUE generally increased, although the fluctuations between seasons also became larger. The highest CPUE was observed in 1999. The apparent increase in CPUE in the later years was likely due to the gradual introduction of iceboats into the Tobago fishing fleet during this time and the lack of separation of these boats within the present dataset. Changing market facilities during the middle to late 1990s could also have affected fishing practices and so contributed to the larger fluctuations observed at that time.
Dominica and Saint Lucia - These data showed such wide interseasonal variation, that the GLMs did not produce any meaningful results. This was probably due to the fact that flyingfish was not a primary target species for most vessels and CPUEs appeared to be influenced by different factors or combinations of factors and these also varied randomly within season and among seasons.
4.3 Standardization of Fishing Effort
- To develop a standardized measure of fishing effort by country and if possible for the entire fishery
- To estimate total fishing effort in standard units.
The method described by Robson (1966) was reviewed and discussed. This method essentially compares catch rates of vessels operating within the same fishing area and at the same time, in order to estimate fishing power relative to a specified standard. It accommodates for comparisons between different types of vessels at different times. For instance, vessel type m can be compared with vessel type n at one time and then vessel type n could be compared with the specified standard vessel s at another time.
Additionally, the consultant also reviewed and discussed the possibility of developing a simple linear regression equation which would explain fishing power in terms of some characteristic of effort, say engine horse power or vessel gross tonnage (Gulland, 1983). The use of GLMs to standardize effort was also noted. Assuming that all vessels in the subregion are fishing on one stock and the subregion could be treated as an uniform fishing area, comparison of the mean CPUEs of Groups of vessel types could be used to develop estimates of fishing power relative to a specified standard. However, flyingfish are believed to be rather patchily distributed and hence the assumption of an uniform distribution of the stock and hence an uniform fishing area within the subregion would probably not be valid.
While the importance of effort standardization was noted, there was insufficient time to conduct these analyses during the Meeting.
5.1 Statistics and Research
1. The standardized historical CPUE series for Barbados showed a steady substantial increase during 1978 to around 1984. The reasons for this marked increase were unclear, but the Group believed that it was probably associated with an increase in fishing efficiency. The Group therefore recommended that an interview study be carried out to determine and quantify changes in the dayboat fishery (gear and vessel characteristics and fishing strategies) for as many years as possible but particularly for the period of concern.
2. The interview study noted in 1 should be carried out on the Tobago fishery.
3. The standardized CPUE series for Tobago showed a notable increase in recent years, with larger interseasonal fluctuations. The reasons for the increase was unclear, but was probably largely due to the introduction of some iceboats into the fleet, as well as marketing changes during the period of concern. In future catch rate standardization analyses, it would be necessary to distinguish among the different types of vessels in the Tobago fleet.
4. In other countries where flyingfish is harvested by a multispecies and multigear fishery, the sampling program should be modified to ensure greater sampling coverage of the main fishing sites and to permit identification of fishing trips for which flyingfish is a main target species.
5. Field sampling programs also need to ensure that the species of flyingfish is correctly identified during sampling. This recommendation is particularly important for countries like Saint Lucia, which noted that at least 3 species of flyingfish are usually harvested by the local fishery.
6. Following discussions about the expected effect of increases in fishing efficiency on catchability, the Group recognized the importance of collecting more detailed effort data and to maintain updated Licensing and Registration systems which would include general information on changes in fishery characteristics. By this means, increases in fishing efficiency, as well as shifts in fishing areas, could be monitored closely and quantified.
7. Given the potential influence of environmental conditions on catchability and recruitment of flyingfish, the acquisition of environmental data such as sea surface temperature was recommended. These data would also be useful for refining the stock - recruit analysis and risk assessment conducted by Mahon (1989).
8. The Group also recommended that studies should be conducted to identify and determine the existence of preferred spawning grounds for flyingfish. This information would contribute to a better understanding of flyingfish distribution and vulnerability to the fishing gear, as well as to management.
Standardized time series of catch rates were developed and examined for both the Barbados and Tobago fishery. There was a marked increase in the Barbados dayboat CPUE during the late 1970s to 1988, which may be associated with an increase in dayboat fishing efficiency at that time. After 1988, CPUE declined to levels similar to those observed during the late 1960s - early 1970s. The reason for this decline was not clear, but market changes in the early 1980s may have biased reporting in later years, as well as affected sampling coverage. The shorter and more recent time series of Barbados dayboat and iceboat CPUEs also showed some increase during 1996 - 1998, but CPUEs tended to level off by 1999.
The Tobago fisherys CPUEs were fairly stable during the late 1980s - early 1990s, with increases observed from 1993 onwards. These increases may have been due to the introduction of a small iceboat fleet during this time period. Given logical explanations for the observed notable increases and decreases of CPUE over short time periods, the Group concluded that flyingfish CPUE trends have probably remained more or less constant over the time periods studied.
1) Given the uncertainties in the quality of current data, the Group recommended that future expansion of the fishery should be approached with caution.
2) Additionally, the Group recognized the need for further thought and action in implementing the institutional arrangement for handling shared management and for strengthening the current Working Group in analysing and interpreting data analyses.
Figure 1. Total annual report landings (t) by country for the period 1989-1999.
Figure 2. Standardized CPUE series for Tobago fleet.