- Email: [email protected]
A comparative study of American lobster fishery sea and port sampling programs in Maine: 1998–2000
Fisheries Research 68 (2004) 343–350
A comparative study of American lobster fishery sea and port sampling programs in Maine: 1998–2000 Kevin Scheirer a,c,∗ , Yong Chen a , Carl Wilson b a
b
School of Marine Sciences, University of Maine, 216 Libby Hall, Orono, ME 04469, USA Maine Department of Marine Resources, P.O. Box 8, West Boothbay Harbor, ME 04575, USA c Gulf of Maine Research Institute, P.O. Box 7549, Portland, ME 04112, USA
Received 29 January 2003; received in revised form 22 October 2003; accepted 11 November 2003
Abstract The American lobster (Homarus americanus) supports the most valuable commercial fishery in the northeast United States. The fishery is critical to the Maine economy and society. To better manage this fishery, the Maine Department of Marine Resources has established two fishery-dependent sampling programs: sea sampling and port sampling. This, however, raises a question of consistency in describing the lobster fishery using data collected from the two programs. Using data from 1998 to 2000, we evaluated the consistency in size composition and catch per unit of effort (cpue) between the sea and port sampling programs. The strength of the statistical correlations between the two sampling programs varied depending upon the measure of cpue, the year, and whether time or area was the comparison variable. The overall pattern that emerged was a stronger relationship between sea and port sampling data over time from 1998 to 2000, implying the two sampling programs were consistent in describing temporal variations in cpue. However, mean yearly county cpue estimates showed significant differences between the two programs in all 3 years, suggesting an inconsistency in describing spatial variations in cpue between the two programs. Size composition reported by the two programs was very similar with significant differences in only 3 months out of the 21 tested. This study suggests that either program should be sufficient in monitoring temporal trends of the lobster fishery. © 2003 Elsevier B.V. All rights reserved. Keywords: Sampling program; American lobster; Data consistency; Catch per unit effort; Size composition
1. Introduction Fisheries scientists and managers use fisherydependent sampling programs as a means of monitoring the commercial fishery and collecting fisheries data for stock assessment and management. The benefits of such programs include greater quantities of data and lower costs compared with fisheries-independent sampling programs. The diversity and amount of data ∗ Corresponding author. Tel.: +1-207-772-2321; fax: +1-207-772-6855. E-mail address: [email protected] (K. Scheirer).
are important in ensuring a high quality of stock assessment. Limited data may introduce large uncertainties and biased errors in stock assessment, potentially resulting in the mismanagement of a fishery (Walters, 1998). Fisheries data are often collected by monitoring programs such as port and sea sampling programs and logbook systems (Hilborn and Walters, 1992). Common measurement variables often include catch, measures of fishing effort, length, weight, and fecundity information. Variables measured often differ between sampling programs due to sampling design, the nature of the program (on board a vessel, dockside,
0165-7836/$ – see front matter © 2003 Elsevier B.V. All rights reserved. doi:10.1016/j.fishres.2003.11.003
344
K. Scheirer et al. / Fisheries Research 68 (2004) 343–350
or electronic), or other constraints such as budgets, logistics, and governmental management rules. Multiple sampling programs allow several unique sampling designs that can measure the characteristics of the fishery at the different temporal and spatial scales that the manager wishes to monitor. Comparative study of these sampling programs may also help identify industry fishing behavior. For example, the comparison of sea and port sampling is useful in detecting fleet responses to changes in regulations. Problems, however, may arise when data from the programs characterize the fishery in significantly different ways. In this case, choices may need to be made as to what data source is most reliable and desirable in describing the fishery. This may often depend on sampling design, costs, quality and quantity of data, and temporal and spatial coverage of the sampling programs. The American lobster (Homarus americanus) is distributed throughout the northwest Atlantic from the Straight of Belle Isle, Newfoundland to Cape Hatteras, North Carolina, from mean low water to depths of 700 m (Cooper and Uzmann, 1980; Lawton and Lavalli, 1995). It supports the most valuable commercial fishery in the northeast United States.1 The fishery is critical to the Maine economy and society. Optimal management of the lobster stock requires full understanding of its population dynamics. The quality of stock assessment is thus a central issue in lobster fishery management. Of the factors that may affect the quality of the lobster stock assessment, fisheries data are one of the most important. To have great spatial and temporal coverage of data collection for better management of this important fishery, the Maine Department of Marine Resources (DMR) has established two fisheries-dependent sampling programs. The port sampling program has been in place since 1967, and it has supplied fishery managers with large amounts of data on the lobster fishery including catch, various measures of fishing effort, and biological information of landings. The sampling design is random with the day and location of sampling events selected by a computer program that uses a ta1 Atlantic States of Marine Fisheries Commission (ASMFC), 2000. American lobster stock assessment report. Stock Assessment Report No. 00-01 (supplement), ASMFC American Lobster Stock Assessment Sub-Committee, Washington, DC.
Washington Hancock Waldo
A
Knox Lincoln
B
Sagadahoc
C Cumberland
D E F
York
N
Lobster Zones Coastal Counties
G
W
E S
40
0
40 Miles
Fig. 1. Maine counties and lobster zones.
ble of randomly assorted digits. Lobster dealers who buy from five or more boats are included in the sample set (over 120 dealers), and 10 of these dealers on 10 separate days per month are selected as sampling locations. These samples are representative of the distribution of dealers in the seven coastal counties in Maine (Fig. 1). DMR biologists obtain catch and effort information from each lobsterman that lands at the dealer location by using an interview format. Random samples of 10 lobsters are drawn from each lobsterman’s catch for biological information. Usually more than one boat is sampled per sampling trip. The sample design remained largely unchanged until 2000, when sampling time was expanded to the entire year while previous port samplings covered only from April to December. The sea sampling program, which places DMR biologists on commercial fishing boats to record lobstermen’s catch and sample for biological information, has been in place since 1985. It is a directed sampling program with a certain number of sampling events planned for specific areas of the coast. The Maine coast is divided into seven lobster management zones (Fig. 1). From 1998 to 2000, sampling efforts were greatly increased to cover more boats and more fishing time. Currently three sampling trips per month are planned for each zone, totaling a possible 21 trips per month from May to November. One boat per trip is sampled, and because fishermen voluntarily allow
K. Scheirer et al. / Fisheries Research 68 (2004) 343–350
sampling on their boat, most boats are sampled more than once in a season. It is more efficient to make return trips with a cooperative fisherman than to convince another fisherman to allow state biologists on board. There is no law requiring fishermen to allow biologists or observers on their boats. The increase in effort in the sea sampling program since 1998 has provided a more comprehensive and detailed coverage of the Maine lobster fishery. As a result of the expansion of effort and increases in the costs of the sampling program, a comparative analysis was needed for evaluating differences in the data collected from the two sampling programs. Of key interest were catch per unit of effort (cpue) and size composition estimates as well as the overall scale of sampling and data collected, which are essential in assessing and monitoring the lobster stock and developing management plans for the lobster fishery in the state of Maine. Such a study will indicate if the data collected from the two sampling programs were consistent in describing the lobster fishery. A consistent pattern would allow us to combine the two programs and use limited financial resources to have greater spatial and temporal coverage of the fishery in fishery-dependent sampling. An inconsistent pattern in describing the lobster fishery, however, would require us to identify factors that result in the differences in the two sampling programs. Using data from 1998 to 2000, we evaluated the consistency in size composition and cpue between the sea and port sampling programs. Because there are many measures of fishing effort in each sampling program, we also evaluated the difference in calculating cpue using different measures of catch and effort for each sampling program.
2. Materials and methods The port and sea sampling programs were compared using data from 1998 to 2000, because the sea sampling effort was initially expanded in 1998 and continued expanding through 2000. Also, the analysis was limited to the months from May to November by the length of the sea sampling program season. Five different measures of cpue were calculated for each sea and port sampling trip: pounds per trap haul
345
Table 1 Example calculation of five measures of catch per unit of effort CPUE measure
Numerator
Denominator
Number/trap haul Number/trap haul set-over-day Number/boat hour Pounds/trap haul Pounds/trap haul set-over-day
303 212
885 640
122 255 569
9.75 765 406
CPUE 0.34 0.06 12.51 0.60 0.07
(lbs/th), pounds per trap haul set-over-day (lbs/thsod), number per trap haul (num/th), number per trap haul set-over-day (num/thsod), and number per boat hour (num/bh). Pounds (lbs) are the pounds of legal lobsters landed, and numbers (num) are the number of legal lobsters landed. Trap hauls (th) are the number of traps that a lobsterman pulls out of the water in one trip. Set-over-days (sod) are the number of days that a trap has been fished without being checked (generally 1–10 days). Trap haul set-over-days (thsod) are calculated by multiplying the number of trap hauls by the number of set-over-days for that group of traps (i.e. 300 trap hauls multiplied by five set-over-days equals 1500 thsod). The sum of the catch for each sampling trip was divided by the sum of the effort for each sampling trip (example in Table 1). The mean, median, and coefficient of variance (standard deviation/mean) were calculated for each year from the sampling trip cpue’s (Table 2). It was unclear as to whether pounds or numbers was a more appropriate measure of catch when being used to compare two different data sets. To answer this question, cpue calculated by using pounds and numbers needed to be compared within sea and port sampling data sets. Because the total number of pounds of lobster and the total number of lobsters sampled each trip are different (e.g. 300 lobsters weighing a total of 450 pounds) the five measures of cpue were standardized (Z score) for both port and sea sampling data sets from 1998 to 2000. The standardization (Z score) consisted of subtracting the mean cpue (calculated from the sampling trip cpues) from each sampling trip cpue and dividing that number by the standard deviation (calculated from the sampling trip cpues). This standardization gave the two sets of cpue the same scale (amount of variation around the mean), making them comparable.
346
K. Scheirer et al. / Fisheries Research 68 (2004) 343–350
Table 2 Statistics for five measures of CPUE from port and sea sampling (1998–2000) CPUE measure
Summary statistic
1998 Port
1999 Sea
Port
2000 Sea
Port
Sea
Number/trap haul
Mean Median CV
0.87 0.87 0.54
1.18 1.01 0.74
0.86 0.81 0.52
1.12 0.88 0.70
1.02 1.00 0.55
1.45 1.30 0.64
Number/trap haul set-over-day
Mean Median CV
0.22 0.20 0.91
0.25 0.23 0.65
0.19 0.19 0.53
0.21 0.19 0.68
0.24 0.20 0.65
0.29 0.24 0.75
Number/boat hour
Mean Median CV
21.21 21.05 0.56
30.84 27.27 0.67
21.31 19.77 0.54
28.57 22.62 0.72
24.73 21.69 0.60
33.78 29.90 0.68
Pounds/trap haul
Mean Median CV
1.08 1.06 0.53
1.49 1.27 0.76
1.07 1.01 0.52
1.44 1.16 0.69
1.30 1.25 0.56
1.88 1.65 0.68
Pounds/trap haul set-over-day
Mean Median CV
0.27 0.24 0.94
0.32 0.28 0.65
0.23 0.24 0.53
0.28 0.25 0.68
0.30 0.25 0.66
0.39 0.31 0.77
sampling trips, but port sampling collected data for more boats than sea sampling (Figs. 2 and 3). The mean monthly cpues were calculated using the sampling trip cpues (mean of sampling trip cpues in each month). A regression analysis was conducted with the port sampling cpue as the independent or X variable and the sea sampling cpue as the dependent or Y variable. The measures of cpue used in the regression analysis were num/th, num/thsod, num/bh, lbs/th, and lbs/thsod. The monthly means of the sampling trip cpues were plotted in one regression per year per measure of cpue, totaling 15 regression analyses (five regressions for 1998–2000).
160 Sea
140
NUMBER OF TRIPS
A regression analysis was performed for mean lbs/th versus mean num/th for each year within both sea and port sampling data sets. We would consider there to be no significant difference in using pounds or numbers as a measure of catch if all the regression models had a slope estimate not significantly different from 1, an intercept estimate not significantly different from 0, and an r2 greater than 90%. The comparison of port and sea sample cpues was done on a monthly time scale, using May to November for each sampling program. The sea sampling program ran from May to November until 2000 when it was expanded further to include all 12 months. As a result, the analysis was constrained to this time period even though port sampling has run from April to December except for 2000 when sampling occurred in all 12 months. Even though both the sea and port sampling programs sampled the full 12 months of 2000, the 7-month time period was used in this year for ease of comparison with 1998 and 1999. Sampling trips in the winter and early spring were weather-dependent resulting in sporadic data collection. The time frame of a month was used because it averaged out the differences in sampling techniques (both in number of boats sampled per trip and number of trips per month) and would preserve a certain amount of variation over time. There were more sea sampling trips than port
Port
120 100 80 60 40 20 0 1998
1999
2000
Fig. 2. Comparison of yearly sampling effort between sea and port sampling in number of sampling trips.
Sea
500 450 400 350 300 250 200 150 100 50 0
Port
Percent (%) of Total Measured
NUMBER OF BOATS
K. Scheirer et al. / Fisheries Research 68 (2004) 343–350
347
100 80 Sea
Port
60 40 20 0
1998
1999
83-94mm
2000
Fig. 3. Comparison of yearly sampling effort between sea and port sampling in number of boats sampled.
95-108mm
109-124mm
125-127mm
14% molt groupings
Fig. 4. Example size composition data from July 2000.
3. Results Another regression analysis compared mean num/thsod and num/th for each county from sea and port sampling data. Sea sampling locations were categorized by lobster management area, whereas port sampling locations were categorized by county (Fig. 1). A table that lists each sea sampling location with the county was used to organize all locations according to county. Lincoln and Sagadahoc counties were combined because there were no sea sampling locations in Sagadahoc in 1998 and 2000. They are geographically adjacent and have smaller sample sizes on average than Cumberland or Knox counties (west and east of Lincoln and Sagadahoc; Fig. 1). The mean sample trip cpue per county was calculated for 1998–2000. The regression analysis compared the two sampling programs by county for each year. Size composition of the lobster catch between the sea and port sampling programs was compared using a non-parametric test, the Kolmogorov–Smirnov (KS) method (Zar, 1984). The test compares two independently sampled distributions to determine if the samples have been drawn from the same population. The size categories followed the 14% carapace length (molt) groupings used since 1989: 83–94, 95–108, 109–124, and 125–127 mm.2 The frequency of each grouping was calculated for each month from May to November for 1998–2000 (Fig. 4). The frequencies were then compared by the KS test.
The regression analysis of the two measures of catch (i.e. the standardized num/th and lbs/th) within the sea and port sampling data sets showed that slopes were not significantly different from 1 (the obtained P > 0.50), intercepts not significantly different from 0 (the obtained P > 0.50) and all coefficients of determinant r2 were greater than 0.90. Therefore, we concluded that there was no difference in using numbers or pounds to compare cpue for both the sampling programs. The analysis of cpue data from the two sampling programs, conducted on a monthly time scale, revealed a trend in all five measures of cpue data. Results for the regression analysis of cpue by month include slope, regression P-value, r2 adjusted by sample size, and number of data pairs (months) for the regression analysis (Tables 3 and 4). A trend of improved relationship from 1998 to 2000 was seen in all cpue measures (increased adjusted r2 and smaller model P-values), but only num/th and num/thsod are shown due to page limitation. Cpue is used as a relative measure of fish stock abundance, so while the comparison of the cpue values is important, time series plots of cpue help give a more Table 3 Port sampling versus sea sampling monthly cpue comparison regression results (number/trap haul): May to November 1998–2000 Regression statistic
1998
1999
2000
Slope Model P value Adjusted r2 N (number of months)
0.97 0.079 0.39 7
1.75 0.011 0.71 7
1.38 0.005 0.78 7
2
Thomas, J.C., 1971. An analysis of the commercial lobster (Homarus americanus) fishery along the coast of Maine, August 1966 to December 1970. State of Maine Department of Sea and Shore Fisheries, Fisheries Research Station, West Boothbay Harbor, ME.
K. Scheirer et al. / Fisheries Research 68 (2004) 343–350
Regression statistic
1998
1999
2000
Slope Model P value Adjusted r2 N (number of months)
0.51 0.144 0.36 7
1.11 0.010 0.71 7
1.23 0.001 0.91 7
complete picture on differences in temporal variations of stock abundance implied by different measures of cpue. Two measures of standardized cpue (num/th and num/thsod) were plotted from port and sea sampling data on a monthly time scale for 1998–2000 (Figs. 5 and 6). The variation above or below the mean was 2.00 Sea
1.00 0.50
-0.50
June
July
Aug.
Sept.
Oct.
0.00 May
1998 Month
Port
1.50
Sea
1.00 0.50
Aug.
Sept.
Oct.
Nov.
1998
-1.00 Month
-1.50
1.50 Port Sea
1.00 0.50 0.00 May
June
July
Aug.
Sept.
Oct.
Nov.
-0.50
1999
-1.00 Month
1.50
May
June
July
Aug.
Sept.
Oct.
Nov.
-1.00 -1.50
1999
-2.00
Sea
1.00 0.50 0.00 May
June
July
Aug.
Sept.
Oct.
Nov.
-0.50 -1.00
2000
-1.50
0.00
-2.00
Month
Fig. 6. Standardized number per trap haul set-over-day (1998–2000).
Month
1.50 Port 1.00
Sea
0.50 0.00 May
June
July
Aug.
Sept.
Oct.
Nov.
-1.00 -1.50 -2.00
July
Port
-1.50
-0.50
June
-0.50
Nov.
2.00 Standardized CPUE (num/th)
0.50
-1.50 May
-1.00
-0.50
Sea
1.00
0.00
-2.00
Standardized CPUE (num/th)
Port 1.50
Standardized CPUE (num/thsod)
Standardized CPUE (num/th)
Port 1.50
2.00 Standardized CPUE (num/thsod)
Table 4 Port sampling versus sea sampling monthly cpue comparison regression results (number/trap haul set-over-day): May to November 1998–2000
Standardized CPUE (num/thsod)
348
2000 Month
Fig. 5. Standardized number per trap haul (1998–2000).
consistent between the two sampling programs from 1998 to 2000. The degree and pattern of variation differed between the two measures of cpue. Num/thsod showed greater variation from the mean than num/th. The two measures also depicted increases or decreases in cpue differently. What appeared to be a dramatic change in num/thsod would not appear as dramatic in num/th, or an increase in num/th for 1 month would be a decrease in num/thsod in the same month (i.e. Port sampling, November 2000; Figs. 5 and 6).
K. Scheirer et al. / Fisheries Research 68 (2004) 343–350 Table 5 Difference in size composition between sea and port sampling by month and year Month
1998
1999
2000
May June July August September October November
True True True True True True True
True True True True True True False
True True True False True True False
Percent true
100.00
85.71
71.43
Kolmogorov–Smirnov test: no difference = true, difference = false.
Counties were used to compare how the two sampling programs represented areas of the coast. Two measures of cpue (num/thsod and num/th) were used in the regression analysis. The relationship between the two programs was not significant in any of the 3 years (P > 0.05, Adj. R2 < 0.5). The Kolmogorov–Smirnov test of size compositions estimated from the two sampling programs showed that they are very similar (Table 5). Over the 3 years they became less similar with the size composition in November being different in 1999 and 2000, and different in August 2000.
4. Discussion The comparison of the sea and port sampling program data was conducted on an absolute scale and a relative scale. The absolute scale was concerned with statistical differences in the absolute cpue values of the two sampling programs, while the relative scale looked at patterns over time between the two data sets. The strength of the statistical relationships between the two sampling programs varied depending on the measure of cpue, the year, and whether time or area was the classification variable. The overall pattern that emerged on both the absolute and relative scales was that there was a strong correlation between sea and port sampling data over time from 1998 to 2000. The per trip cpue data seem to be normally distributed for both programs with little difference between mean and median values (Table 2), an in-
349
dication of symmetric distribution of the data. Sea sampling tends to report higher values for monthly and area cpue (Table 2). Also the data show that a majority of the time, variation (CV) is higher in sea sampling (Table 2). This may be caused by the differences in the choice of fishermen sampled in the two programs. Sea sampling may select more successful fishermen, while port sampling selects from a range of fishermen. Port sampling also samples from a larger number of boats, which may explain why there is less variation (CV) in cpue estimates (Table 2). The standardized monthly cpue shows that the two data sets vary similarly around their means (Figs. 5 and 6). The cpue data grouped by county do not show this similarity and appear to vary differently around the yearly mean. This may be caused by a factor of time scale (year versus month). Also, port sampling randomly selects dealers and may not sample in the same county more than once a month or not at all. Sea sampling makes trips three times a month in each of the seven zones. This may result in a difference in sample size depending on the county, contributing to variations in cpue. The cpue estimates for each county can vary widely from year to year in either sampling program. The regression analysis of monthly cpue showed slope values closer to 1, smaller P values, and higher adjusted r2 values from 1998 to 2000 (Tables 3 and 4). The overall pattern of increasing relationship with time was evident, when standardized cpue data were plotted in a monthly time series for the 3 years (Figs. 5 and 6). The degree and direction (positive or negative) of variation around the mean increased in similarity. Differences among measures of cpue were noted in terms of variation around the mean. Most likely the measure of effort or changes in sample size introduced different amounts of variation over time. Size composition of the legal catch recorded by each program was statistically the same for the majority of the months from May to November. The differences in 1999 and 2000 may have been caused by the large increase in number of measured lobsters in sea sampling, particularly in 2000. Also of interest is that in November 1998–2000, over 90% of the lobsters measured by port sampling were in the first 14% grouping (83–94 mm). In contrast, 84%
350
K. Scheirer et al. / Fisheries Research 68 (2004) 343–350
or less of the measured catch in sea sampling was in the first 14% grouping. Again, sample size may be the cause of this difference as well as the technique by which port sampling selects lobsters to be measured. Multiple fisheries sampling programs are common in collecting fishery-dependent data (Hilborn and Walters, 1992). These programs are often established by different fisheries management agencies (e.g. for stocks distributed across different states) and in different time periods, resulting in different spatial and temporal coverage of the fishery. This raises an important question in using information collected from multiple sampling programs in describing the fishery. Because different sampling programs are often created for different purposes and have different designs and different spatial and temporal coverage, the information derived from them may be inconsistent in indicating the status of the fishery, which may have negative impacts on stock assessment and management. A comparative study should be done to compare the consistency of data collected in different sampling programs, identify the causes of any inconsistencies, and subsequently reconcile the differences of multiple sampling programs. This paper demonstrates the importance of such a study.
Acknowledgements Gary Robinson, Jason Bartlett, Glenn Nutting, Norris Bowie, Robert Russell, and others collected all the data used in this paper and were generous in sharing both the data and background knowledge associated with it. Dr. Linda Mercer directed financial support (as DMR Marine Studies Fellowship) and focus to this project. Financial support also came from the Thistle Marine, LLC located in Ellsworth, Maine, the University of Maine, and the Maine Sea Grant Program. References Cooper, R.A., Uzmann, J.R., 1980. Ecology of juvenile and adult Homarus. In: Cobb, J.S., Phillips, R. (Eds.), The Biology and Management of Lobsters, vol. 2. Academic Press, New York, pp. 97–141. Hilborn, R., Walters, C., 1992. Quantitative Fisheries Stock Assessment: Choice, Dynamics, and Uncertainty. Chapman & Hall, New York. Lawton, P., Lavalli, K.L., 1995. Postlarval, juvenile, adolescent and adult ecology. In: Factor, J.F. (Ed.), Biology of the Lobster Homarus americanus. Academic Press, New York, pp. 47–88. Walters, C.J., 1998. Evaluation of quota management policies for developing fisheries. Can. J. Fish. Aquat. Sci. 55, 2691–2705. Zar, J.H., 1984. Biostatistical Analysis, 2nd ed. Prentice-Hall, Englewood Cliffs, NJ.
A comparative study of American lobster fishery sea and port sampling programs in Maine: 1998–2000 Kevin Scheirer a,c,∗ , Yong Chen a , Carl Wilson b a
b
School of Marine Sciences, University of Maine, 216 Libby Hall, Orono, ME 04469, USA Maine Department of Marine Resources, P.O. Box 8, West Boothbay Harbor, ME 04575, USA c Gulf of Maine Research Institute, P.O. Box 7549, Portland, ME 04112, USA
Received 29 January 2003; received in revised form 22 October 2003; accepted 11 November 2003
Abstract The American lobster (Homarus americanus) supports the most valuable commercial fishery in the northeast United States. The fishery is critical to the Maine economy and society. To better manage this fishery, the Maine Department of Marine Resources has established two fishery-dependent sampling programs: sea sampling and port sampling. This, however, raises a question of consistency in describing the lobster fishery using data collected from the two programs. Using data from 1998 to 2000, we evaluated the consistency in size composition and catch per unit of effort (cpue) between the sea and port sampling programs. The strength of the statistical correlations between the two sampling programs varied depending upon the measure of cpue, the year, and whether time or area was the comparison variable. The overall pattern that emerged was a stronger relationship between sea and port sampling data over time from 1998 to 2000, implying the two sampling programs were consistent in describing temporal variations in cpue. However, mean yearly county cpue estimates showed significant differences between the two programs in all 3 years, suggesting an inconsistency in describing spatial variations in cpue between the two programs. Size composition reported by the two programs was very similar with significant differences in only 3 months out of the 21 tested. This study suggests that either program should be sufficient in monitoring temporal trends of the lobster fishery. © 2003 Elsevier B.V. All rights reserved. Keywords: Sampling program; American lobster; Data consistency; Catch per unit effort; Size composition
1. Introduction Fisheries scientists and managers use fisherydependent sampling programs as a means of monitoring the commercial fishery and collecting fisheries data for stock assessment and management. The benefits of such programs include greater quantities of data and lower costs compared with fisheries-independent sampling programs. The diversity and amount of data ∗ Corresponding author. Tel.: +1-207-772-2321; fax: +1-207-772-6855. E-mail address: [email protected] (K. Scheirer).
are important in ensuring a high quality of stock assessment. Limited data may introduce large uncertainties and biased errors in stock assessment, potentially resulting in the mismanagement of a fishery (Walters, 1998). Fisheries data are often collected by monitoring programs such as port and sea sampling programs and logbook systems (Hilborn and Walters, 1992). Common measurement variables often include catch, measures of fishing effort, length, weight, and fecundity information. Variables measured often differ between sampling programs due to sampling design, the nature of the program (on board a vessel, dockside,
0165-7836/$ – see front matter © 2003 Elsevier B.V. All rights reserved. doi:10.1016/j.fishres.2003.11.003
344
K. Scheirer et al. / Fisheries Research 68 (2004) 343–350
or electronic), or other constraints such as budgets, logistics, and governmental management rules. Multiple sampling programs allow several unique sampling designs that can measure the characteristics of the fishery at the different temporal and spatial scales that the manager wishes to monitor. Comparative study of these sampling programs may also help identify industry fishing behavior. For example, the comparison of sea and port sampling is useful in detecting fleet responses to changes in regulations. Problems, however, may arise when data from the programs characterize the fishery in significantly different ways. In this case, choices may need to be made as to what data source is most reliable and desirable in describing the fishery. This may often depend on sampling design, costs, quality and quantity of data, and temporal and spatial coverage of the sampling programs. The American lobster (Homarus americanus) is distributed throughout the northwest Atlantic from the Straight of Belle Isle, Newfoundland to Cape Hatteras, North Carolina, from mean low water to depths of 700 m (Cooper and Uzmann, 1980; Lawton and Lavalli, 1995). It supports the most valuable commercial fishery in the northeast United States.1 The fishery is critical to the Maine economy and society. Optimal management of the lobster stock requires full understanding of its population dynamics. The quality of stock assessment is thus a central issue in lobster fishery management. Of the factors that may affect the quality of the lobster stock assessment, fisheries data are one of the most important. To have great spatial and temporal coverage of data collection for better management of this important fishery, the Maine Department of Marine Resources (DMR) has established two fisheries-dependent sampling programs. The port sampling program has been in place since 1967, and it has supplied fishery managers with large amounts of data on the lobster fishery including catch, various measures of fishing effort, and biological information of landings. The sampling design is random with the day and location of sampling events selected by a computer program that uses a ta1 Atlantic States of Marine Fisheries Commission (ASMFC), 2000. American lobster stock assessment report. Stock Assessment Report No. 00-01 (supplement), ASMFC American Lobster Stock Assessment Sub-Committee, Washington, DC.
Washington Hancock Waldo
A
Knox Lincoln
B
Sagadahoc
C Cumberland
D E F
York
N
Lobster Zones Coastal Counties
G
W
E S
40
0
40 Miles
Fig. 1. Maine counties and lobster zones.
ble of randomly assorted digits. Lobster dealers who buy from five or more boats are included in the sample set (over 120 dealers), and 10 of these dealers on 10 separate days per month are selected as sampling locations. These samples are representative of the distribution of dealers in the seven coastal counties in Maine (Fig. 1). DMR biologists obtain catch and effort information from each lobsterman that lands at the dealer location by using an interview format. Random samples of 10 lobsters are drawn from each lobsterman’s catch for biological information. Usually more than one boat is sampled per sampling trip. The sample design remained largely unchanged until 2000, when sampling time was expanded to the entire year while previous port samplings covered only from April to December. The sea sampling program, which places DMR biologists on commercial fishing boats to record lobstermen’s catch and sample for biological information, has been in place since 1985. It is a directed sampling program with a certain number of sampling events planned for specific areas of the coast. The Maine coast is divided into seven lobster management zones (Fig. 1). From 1998 to 2000, sampling efforts were greatly increased to cover more boats and more fishing time. Currently three sampling trips per month are planned for each zone, totaling a possible 21 trips per month from May to November. One boat per trip is sampled, and because fishermen voluntarily allow
K. Scheirer et al. / Fisheries Research 68 (2004) 343–350
sampling on their boat, most boats are sampled more than once in a season. It is more efficient to make return trips with a cooperative fisherman than to convince another fisherman to allow state biologists on board. There is no law requiring fishermen to allow biologists or observers on their boats. The increase in effort in the sea sampling program since 1998 has provided a more comprehensive and detailed coverage of the Maine lobster fishery. As a result of the expansion of effort and increases in the costs of the sampling program, a comparative analysis was needed for evaluating differences in the data collected from the two sampling programs. Of key interest were catch per unit of effort (cpue) and size composition estimates as well as the overall scale of sampling and data collected, which are essential in assessing and monitoring the lobster stock and developing management plans for the lobster fishery in the state of Maine. Such a study will indicate if the data collected from the two sampling programs were consistent in describing the lobster fishery. A consistent pattern would allow us to combine the two programs and use limited financial resources to have greater spatial and temporal coverage of the fishery in fishery-dependent sampling. An inconsistent pattern in describing the lobster fishery, however, would require us to identify factors that result in the differences in the two sampling programs. Using data from 1998 to 2000, we evaluated the consistency in size composition and cpue between the sea and port sampling programs. Because there are many measures of fishing effort in each sampling program, we also evaluated the difference in calculating cpue using different measures of catch and effort for each sampling program.
2. Materials and methods The port and sea sampling programs were compared using data from 1998 to 2000, because the sea sampling effort was initially expanded in 1998 and continued expanding through 2000. Also, the analysis was limited to the months from May to November by the length of the sea sampling program season. Five different measures of cpue were calculated for each sea and port sampling trip: pounds per trap haul
345
Table 1 Example calculation of five measures of catch per unit of effort CPUE measure
Numerator
Denominator
Number/trap haul Number/trap haul set-over-day Number/boat hour Pounds/trap haul Pounds/trap haul set-over-day
303 212
885 640
122 255 569
9.75 765 406
CPUE 0.34 0.06 12.51 0.60 0.07
(lbs/th), pounds per trap haul set-over-day (lbs/thsod), number per trap haul (num/th), number per trap haul set-over-day (num/thsod), and number per boat hour (num/bh). Pounds (lbs) are the pounds of legal lobsters landed, and numbers (num) are the number of legal lobsters landed. Trap hauls (th) are the number of traps that a lobsterman pulls out of the water in one trip. Set-over-days (sod) are the number of days that a trap has been fished without being checked (generally 1–10 days). Trap haul set-over-days (thsod) are calculated by multiplying the number of trap hauls by the number of set-over-days for that group of traps (i.e. 300 trap hauls multiplied by five set-over-days equals 1500 thsod). The sum of the catch for each sampling trip was divided by the sum of the effort for each sampling trip (example in Table 1). The mean, median, and coefficient of variance (standard deviation/mean) were calculated for each year from the sampling trip cpue’s (Table 2). It was unclear as to whether pounds or numbers was a more appropriate measure of catch when being used to compare two different data sets. To answer this question, cpue calculated by using pounds and numbers needed to be compared within sea and port sampling data sets. Because the total number of pounds of lobster and the total number of lobsters sampled each trip are different (e.g. 300 lobsters weighing a total of 450 pounds) the five measures of cpue were standardized (Z score) for both port and sea sampling data sets from 1998 to 2000. The standardization (Z score) consisted of subtracting the mean cpue (calculated from the sampling trip cpues) from each sampling trip cpue and dividing that number by the standard deviation (calculated from the sampling trip cpues). This standardization gave the two sets of cpue the same scale (amount of variation around the mean), making them comparable.
346
K. Scheirer et al. / Fisheries Research 68 (2004) 343–350
Table 2 Statistics for five measures of CPUE from port and sea sampling (1998–2000) CPUE measure
Summary statistic
1998 Port
1999 Sea
Port
2000 Sea
Port
Sea
Number/trap haul
Mean Median CV
0.87 0.87 0.54
1.18 1.01 0.74
0.86 0.81 0.52
1.12 0.88 0.70
1.02 1.00 0.55
1.45 1.30 0.64
Number/trap haul set-over-day
Mean Median CV
0.22 0.20 0.91
0.25 0.23 0.65
0.19 0.19 0.53
0.21 0.19 0.68
0.24 0.20 0.65
0.29 0.24 0.75
Number/boat hour
Mean Median CV
21.21 21.05 0.56
30.84 27.27 0.67
21.31 19.77 0.54
28.57 22.62 0.72
24.73 21.69 0.60
33.78 29.90 0.68
Pounds/trap haul
Mean Median CV
1.08 1.06 0.53
1.49 1.27 0.76
1.07 1.01 0.52
1.44 1.16 0.69
1.30 1.25 0.56
1.88 1.65 0.68
Pounds/trap haul set-over-day
Mean Median CV
0.27 0.24 0.94
0.32 0.28 0.65
0.23 0.24 0.53
0.28 0.25 0.68
0.30 0.25 0.66
0.39 0.31 0.77
sampling trips, but port sampling collected data for more boats than sea sampling (Figs. 2 and 3). The mean monthly cpues were calculated using the sampling trip cpues (mean of sampling trip cpues in each month). A regression analysis was conducted with the port sampling cpue as the independent or X variable and the sea sampling cpue as the dependent or Y variable. The measures of cpue used in the regression analysis were num/th, num/thsod, num/bh, lbs/th, and lbs/thsod. The monthly means of the sampling trip cpues were plotted in one regression per year per measure of cpue, totaling 15 regression analyses (five regressions for 1998–2000).
160 Sea
140
NUMBER OF TRIPS
A regression analysis was performed for mean lbs/th versus mean num/th for each year within both sea and port sampling data sets. We would consider there to be no significant difference in using pounds or numbers as a measure of catch if all the regression models had a slope estimate not significantly different from 1, an intercept estimate not significantly different from 0, and an r2 greater than 90%. The comparison of port and sea sample cpues was done on a monthly time scale, using May to November for each sampling program. The sea sampling program ran from May to November until 2000 when it was expanded further to include all 12 months. As a result, the analysis was constrained to this time period even though port sampling has run from April to December except for 2000 when sampling occurred in all 12 months. Even though both the sea and port sampling programs sampled the full 12 months of 2000, the 7-month time period was used in this year for ease of comparison with 1998 and 1999. Sampling trips in the winter and early spring were weather-dependent resulting in sporadic data collection. The time frame of a month was used because it averaged out the differences in sampling techniques (both in number of boats sampled per trip and number of trips per month) and would preserve a certain amount of variation over time. There were more sea sampling trips than port
Port
120 100 80 60 40 20 0 1998
1999
2000
Fig. 2. Comparison of yearly sampling effort between sea and port sampling in number of sampling trips.
Sea
500 450 400 350 300 250 200 150 100 50 0
Port
Percent (%) of Total Measured
NUMBER OF BOATS
K. Scheirer et al. / Fisheries Research 68 (2004) 343–350
347
100 80 Sea
Port
60 40 20 0
1998
1999
83-94mm
2000
Fig. 3. Comparison of yearly sampling effort between sea and port sampling in number of boats sampled.
95-108mm
109-124mm
125-127mm
14% molt groupings
Fig. 4. Example size composition data from July 2000.
3. Results Another regression analysis compared mean num/thsod and num/th for each county from sea and port sampling data. Sea sampling locations were categorized by lobster management area, whereas port sampling locations were categorized by county (Fig. 1). A table that lists each sea sampling location with the county was used to organize all locations according to county. Lincoln and Sagadahoc counties were combined because there were no sea sampling locations in Sagadahoc in 1998 and 2000. They are geographically adjacent and have smaller sample sizes on average than Cumberland or Knox counties (west and east of Lincoln and Sagadahoc; Fig. 1). The mean sample trip cpue per county was calculated for 1998–2000. The regression analysis compared the two sampling programs by county for each year. Size composition of the lobster catch between the sea and port sampling programs was compared using a non-parametric test, the Kolmogorov–Smirnov (KS) method (Zar, 1984). The test compares two independently sampled distributions to determine if the samples have been drawn from the same population. The size categories followed the 14% carapace length (molt) groupings used since 1989: 83–94, 95–108, 109–124, and 125–127 mm.2 The frequency of each grouping was calculated for each month from May to November for 1998–2000 (Fig. 4). The frequencies were then compared by the KS test.
The regression analysis of the two measures of catch (i.e. the standardized num/th and lbs/th) within the sea and port sampling data sets showed that slopes were not significantly different from 1 (the obtained P > 0.50), intercepts not significantly different from 0 (the obtained P > 0.50) and all coefficients of determinant r2 were greater than 0.90. Therefore, we concluded that there was no difference in using numbers or pounds to compare cpue for both the sampling programs. The analysis of cpue data from the two sampling programs, conducted on a monthly time scale, revealed a trend in all five measures of cpue data. Results for the regression analysis of cpue by month include slope, regression P-value, r2 adjusted by sample size, and number of data pairs (months) for the regression analysis (Tables 3 and 4). A trend of improved relationship from 1998 to 2000 was seen in all cpue measures (increased adjusted r2 and smaller model P-values), but only num/th and num/thsod are shown due to page limitation. Cpue is used as a relative measure of fish stock abundance, so while the comparison of the cpue values is important, time series plots of cpue help give a more Table 3 Port sampling versus sea sampling monthly cpue comparison regression results (number/trap haul): May to November 1998–2000 Regression statistic
1998
1999
2000
Slope Model P value Adjusted r2 N (number of months)
0.97 0.079 0.39 7
1.75 0.011 0.71 7
1.38 0.005 0.78 7
2
Thomas, J.C., 1971. An analysis of the commercial lobster (Homarus americanus) fishery along the coast of Maine, August 1966 to December 1970. State of Maine Department of Sea and Shore Fisheries, Fisheries Research Station, West Boothbay Harbor, ME.
K. Scheirer et al. / Fisheries Research 68 (2004) 343–350
Regression statistic
1998
1999
2000
Slope Model P value Adjusted r2 N (number of months)
0.51 0.144 0.36 7
1.11 0.010 0.71 7
1.23 0.001 0.91 7
complete picture on differences in temporal variations of stock abundance implied by different measures of cpue. Two measures of standardized cpue (num/th and num/thsod) were plotted from port and sea sampling data on a monthly time scale for 1998–2000 (Figs. 5 and 6). The variation above or below the mean was 2.00 Sea
1.00 0.50
-0.50
June
July
Aug.
Sept.
Oct.
0.00 May
1998 Month
Port
1.50
Sea
1.00 0.50
Aug.
Sept.
Oct.
Nov.
1998
-1.00 Month
-1.50
1.50 Port Sea
1.00 0.50 0.00 May
June
July
Aug.
Sept.
Oct.
Nov.
-0.50
1999
-1.00 Month
1.50
May
June
July
Aug.
Sept.
Oct.
Nov.
-1.00 -1.50
1999
-2.00
Sea
1.00 0.50 0.00 May
June
July
Aug.
Sept.
Oct.
Nov.
-0.50 -1.00
2000
-1.50
0.00
-2.00
Month
Fig. 6. Standardized number per trap haul set-over-day (1998–2000).
Month
1.50 Port 1.00
Sea
0.50 0.00 May
June
July
Aug.
Sept.
Oct.
Nov.
-1.00 -1.50 -2.00
July
Port
-1.50
-0.50
June
-0.50
Nov.
2.00 Standardized CPUE (num/th)
0.50
-1.50 May
-1.00
-0.50
Sea
1.00
0.00
-2.00
Standardized CPUE (num/th)
Port 1.50
Standardized CPUE (num/thsod)
Standardized CPUE (num/th)
Port 1.50
2.00 Standardized CPUE (num/thsod)
Table 4 Port sampling versus sea sampling monthly cpue comparison regression results (number/trap haul set-over-day): May to November 1998–2000
Standardized CPUE (num/thsod)
348
2000 Month
Fig. 5. Standardized number per trap haul (1998–2000).
consistent between the two sampling programs from 1998 to 2000. The degree and pattern of variation differed between the two measures of cpue. Num/thsod showed greater variation from the mean than num/th. The two measures also depicted increases or decreases in cpue differently. What appeared to be a dramatic change in num/thsod would not appear as dramatic in num/th, or an increase in num/th for 1 month would be a decrease in num/thsod in the same month (i.e. Port sampling, November 2000; Figs. 5 and 6).
K. Scheirer et al. / Fisheries Research 68 (2004) 343–350 Table 5 Difference in size composition between sea and port sampling by month and year Month
1998
1999
2000
May June July August September October November
True True True True True True True
True True True True True True False
True True True False True True False
Percent true
100.00
85.71
71.43
Kolmogorov–Smirnov test: no difference = true, difference = false.
Counties were used to compare how the two sampling programs represented areas of the coast. Two measures of cpue (num/thsod and num/th) were used in the regression analysis. The relationship between the two programs was not significant in any of the 3 years (P > 0.05, Adj. R2 < 0.5). The Kolmogorov–Smirnov test of size compositions estimated from the two sampling programs showed that they are very similar (Table 5). Over the 3 years they became less similar with the size composition in November being different in 1999 and 2000, and different in August 2000.
4. Discussion The comparison of the sea and port sampling program data was conducted on an absolute scale and a relative scale. The absolute scale was concerned with statistical differences in the absolute cpue values of the two sampling programs, while the relative scale looked at patterns over time between the two data sets. The strength of the statistical relationships between the two sampling programs varied depending on the measure of cpue, the year, and whether time or area was the classification variable. The overall pattern that emerged on both the absolute and relative scales was that there was a strong correlation between sea and port sampling data over time from 1998 to 2000. The per trip cpue data seem to be normally distributed for both programs with little difference between mean and median values (Table 2), an in-
349
dication of symmetric distribution of the data. Sea sampling tends to report higher values for monthly and area cpue (Table 2). Also the data show that a majority of the time, variation (CV) is higher in sea sampling (Table 2). This may be caused by the differences in the choice of fishermen sampled in the two programs. Sea sampling may select more successful fishermen, while port sampling selects from a range of fishermen. Port sampling also samples from a larger number of boats, which may explain why there is less variation (CV) in cpue estimates (Table 2). The standardized monthly cpue shows that the two data sets vary similarly around their means (Figs. 5 and 6). The cpue data grouped by county do not show this similarity and appear to vary differently around the yearly mean. This may be caused by a factor of time scale (year versus month). Also, port sampling randomly selects dealers and may not sample in the same county more than once a month or not at all. Sea sampling makes trips three times a month in each of the seven zones. This may result in a difference in sample size depending on the county, contributing to variations in cpue. The cpue estimates for each county can vary widely from year to year in either sampling program. The regression analysis of monthly cpue showed slope values closer to 1, smaller P values, and higher adjusted r2 values from 1998 to 2000 (Tables 3 and 4). The overall pattern of increasing relationship with time was evident, when standardized cpue data were plotted in a monthly time series for the 3 years (Figs. 5 and 6). The degree and direction (positive or negative) of variation around the mean increased in similarity. Differences among measures of cpue were noted in terms of variation around the mean. Most likely the measure of effort or changes in sample size introduced different amounts of variation over time. Size composition of the legal catch recorded by each program was statistically the same for the majority of the months from May to November. The differences in 1999 and 2000 may have been caused by the large increase in number of measured lobsters in sea sampling, particularly in 2000. Also of interest is that in November 1998–2000, over 90% of the lobsters measured by port sampling were in the first 14% grouping (83–94 mm). In contrast, 84%
350
K. Scheirer et al. / Fisheries Research 68 (2004) 343–350
or less of the measured catch in sea sampling was in the first 14% grouping. Again, sample size may be the cause of this difference as well as the technique by which port sampling selects lobsters to be measured. Multiple fisheries sampling programs are common in collecting fishery-dependent data (Hilborn and Walters, 1992). These programs are often established by different fisheries management agencies (e.g. for stocks distributed across different states) and in different time periods, resulting in different spatial and temporal coverage of the fishery. This raises an important question in using information collected from multiple sampling programs in describing the fishery. Because different sampling programs are often created for different purposes and have different designs and different spatial and temporal coverage, the information derived from them may be inconsistent in indicating the status of the fishery, which may have negative impacts on stock assessment and management. A comparative study should be done to compare the consistency of data collected in different sampling programs, identify the causes of any inconsistencies, and subsequently reconcile the differences of multiple sampling programs. This paper demonstrates the importance of such a study.
Acknowledgements Gary Robinson, Jason Bartlett, Glenn Nutting, Norris Bowie, Robert Russell, and others collected all the data used in this paper and were generous in sharing both the data and background knowledge associated with it. Dr. Linda Mercer directed financial support (as DMR Marine Studies Fellowship) and focus to this project. Financial support also came from the Thistle Marine, LLC located in Ellsworth, Maine, the University of Maine, and the Maine Sea Grant Program. References Cooper, R.A., Uzmann, J.R., 1980. Ecology of juvenile and adult Homarus. In: Cobb, J.S., Phillips, R. (Eds.), The Biology and Management of Lobsters, vol. 2. Academic Press, New York, pp. 97–141. Hilborn, R., Walters, C., 1992. Quantitative Fisheries Stock Assessment: Choice, Dynamics, and Uncertainty. Chapman & Hall, New York. Lawton, P., Lavalli, K.L., 1995. Postlarval, juvenile, adolescent and adult ecology. In: Factor, J.F. (Ed.), Biology of the Lobster Homarus americanus. Academic Press, New York, pp. 47–88. Walters, C.J., 1998. Evaluation of quota management policies for developing fisheries. Can. J. Fish. Aquat. Sci. 55, 2691–2705. Zar, J.H., 1984. Biostatistical Analysis, 2nd ed. Prentice-Hall, Englewood Cliffs, NJ.