Response to review of final repor
DATE: 11 October 2017 MEMORANDUM FOR: Cheryl Corbett, Cooperative Programs Specialist SUBJECT: Response to review of final report for NOAA project NA15NMF4540059 ______________________________________________________________________________ Response to questions regarding accomplishing the goals Goals identified in the provided proposal (page 2) did not appear to match the goals fulfilled in the report (see Executive Summary, p. 6 of report and identical language on page 10 (“Objectives”). It is not described how or way these are different from the proposed goals and objectives. The study design itself did not appear to be changed, but objectives and data collected were. Only brief or no rationales were provided. As the research objectives were not a comprehensive approach, the changes did not appear to support them. In summary, all the objectives set out in the research proposal were fulfilled. The difference lies in the form they were written; in the research proposal, all the objectives were described in the method section (not in objectives section), while in the final report all the objectives made during this year project were written in an orderly manner on the correct section. Response to questions regarding the first objective “Quantify groundfish bycatch rates in comparison to scallop meat yield with the goal of optimizing scallop harvest while minimizing impacts to other stocks” Spatial influences are described in a very minimal, broad scale (page 16 of report) or not at all, despite a clear spatial component to the design of the project. Spatial data are presented in a somewhat raw form in the final appendix, G, without interpretation. Interpretation of temporal influences is limited (page 16). Although it can be inferred that multiple years of data of this type have been collected, no synthesis of these data with other years’ is included. A better description of the spatiotemporal distribution has been added to a revised version of the final report (second paragraph on page 17). In addition, a new Table 4 has been added to summarize total catches by area by trip. Synthesis of the all data (from 2011 to 2015) was not a goal for this project. However, we agree synthesizing the data sets from all of the bycatch projects is important. Articles with multi-year data analysis have been published and submitted (Smolowitz et al. 2016; Winton et al. 2017; Leavitt et al.; Siemann et al.), and others are in preparation. In addition, we have compiled data for NEFMC and the Scallop PDT on numerous occasions.
Response to questions regarding the second objective “Compare a modified dredge bag (5row apron), designed to reduce flatfish bycatch, with the standard dredge (7-row apron)” It was difficult to understand exactly what was tested; the text did not appear to match a table, and it was unclear how many comparisons were made. We added the following text to clarify the comparisons. "This planned comparison was completed for seven out of eight trips. But because the control dredge broke during the March trip, the CFF TDDs were redesigned with heavier reinforced center bars, and during the May trip, the old versus new TDD frames were tested with 7-row aprons on both." (page 11) "Results from the May trip (above) were not included in the analysis because that trip did not use the control and experimental dredges used for the rest of the study. Catch data from the remaining seven survey trips were treated as a single data set since the experimental treatment was consistent across trips and the heavier dredge (June 2016) did not appear to fish differently than the original dredge (August 2015 – March 2016)." (page 24) Response to questions regarding the sixth objective “Conduct biological sampling of bycatch crustacean and echinoderm species” Purpose and methodology was not described for this objective. An explanation of the purpose and methodology of this objective was added to the revised version of the final report (third paragraph on page 13). Response to questions regarding figures Many of the figures are not sufficiently legible to be read. There is a general issue with identification of individual trips: in some places, a month designation is used; in others, a cruise number; in others, the dates of the trip. And the month designation is a bit misleading: the trip called “Sep” occurred mostly in August; the trip called “Feb” occurred mostly in March. This naming convention misleads the reader about the spacing of the trips and affects the interpretation of the results. The figures were revised, and the labeling convention was unified. The trip labeled "Feb" in few of the graphs was a typo and should have been labeled "Mar". In August we had two trips, and to avoid confusions in the analysis the second trip, which was the last week of August, was changed to September trip convention. Locations are inadequately described in Figure 1, especially considering the use of CAII and non-CAII as variables in analyses. Figure 1 and Figures G1 to G6: All the maps were modified with the correct locations (CAII and non-CAII; pages: 11, 71-82). In Figure 2, I can see no discernable difference in the examples for “Light Brown” and “Brown”. Examples of “stringiness” which is analyzed but not defined or mentioned in the Methods, could also be provided.
Figure 2 was changed. A better picture of gray meats was added (page 12). Text was added to describe stringiness (page 12). Why is there no measure of variation per cruise (error bars, SD, SE, boxplot) for the bycatch rates in Figure 3? It should be made explicit whether the “scallop LB” in the axis labels is shell weight or meat weight. Figure 3 does not have error bars because bycatch rates were calculated from the entire catch per trip per each species, generating only one catch weight per month per species (page 1819). In Figure 4, the more important information is seasonal variation in SHMW ratio. A figure showing this ratio over time would help emphasize the importance of adjusting management and exploitation to maximize efficiency. I cannot see the marks within the boxes of the boxplots, and they are not defined in the legend – presumably, mean and median. The lengths of the whiskers should be also explained. The information contained in this graph relates to the variability in the raw data that was collected over the course of the study. These data are simply the observed shell heights and meat weights collected by sub-area over the eight sampling cruises. Text has been added to the caption to define the whisker plots (page 20). Figure 5 shows seasonal variation in meat weight for a 120-mm scallop (page 21). Figure 6a and 6b: Should the axes both be “MEAT WEIGHT (GRAMS)”? Or is one shell height, as the caption suggests? Figure 6 has been modified in the revised version of the final report (page 22). The axes labels have been corrected. Figure 9: “Multiple” is misspelled. This has been corrected in the revised version of the final report (page 27). Figure 14: Meanings of dots and of the lengths of the box whiskers need to be added Text has been added to the caption to define the whisker plots (page 32). Figure D1: These figures are all hard to read (similarly, the rest of the figures in the Appendices). The legends do not need to be repeated. The second y axis is labeled “FREQUENCY” but appears to be “Count” or “Number”. All graphs have been replaced with better resolution graphs in panel format which allows for the use of common X, Y and Y2 axes (pages: 59-62). The labels have also been changed. The second y-axis is the proportion of the catch in each size bin. Also, it appears that the modeling technique only fits linear relationships. In other gear comparison studies using GLMMs to analyze length effects, inflection points are allowed, and provide better fits. Figure D3 illustrates a problem with linear fits; the Cruise 221 panel shows a significant difference for sizes above 35 cm, with the experimental dredge catching less – but the only data points show retention in the experimental dredge alone – a very poor fit, and misleading as well. Most of the panels in D3 show similar confusing fits. Trimming the data so that only lengths with a certain level of observations (numbers at length) could produce more useable and believable figures. (Similarly for all model fits).
The cumulative proportion at length is not an indication of goodness of fit and only provides a visual reference as to the observed proportion because proportions can be misleading at lengths where the overall numbers are low. Comparing predicted proportions to the observed is useful when catches at length are high and there is some stability to the proportions. To minimize confusion, we trimmed the plots to include only these higher catch number regions as suggested. Figure D5 lacks units on the labels. I suspect counts, as the data do not appear to be continuous. The caption now specifies the catch as numbers (page 63). Figure D6 illustrates the confusion caused by inconsistent labels, as it is not easy to find the “January” cruise. Labels are now consistent throughout (page 64). Response to questions regarding tables Table 1: Also, there is insufficient information in the table or text to identify when the headbale was changed, and whether the “new CFF TDD” is very different from the “CFF TDD”. I could not tell whether there were actually 3-4 different gear comparisons instead of 2. In addition to adding clarifying text, we added months and more descriptions of the dredges below Table 1 (page 10). Table 4 caption: Isn’t intercept-only the most parsimonious model? How could any other model be simpler? The choice of the most parsimonious model depends on both the number of variables and how the best fitting model is selected. For this analysis, the most parsimonious model was chosen based on AIC value, indicating the model with the greatest explanatory power. We decided to compare the observed data with the “intercept only” model even though other, more complex models did provide better fits to the data for some species. Note that Table 4 is now Table 5 (page 26). Table 8 is an example of where the odd naming convention for the trips causes the trip to trip changes in data to be falsely interpreted. There is no intent to mislead. The dates specific to each month label are clearly defined in Table 1. Numbers of individuals captured by each dredge are shown in the table, while the text points out how many females and males were examined from the subsample of 10 individuals per tow (now Table 9, page 32). Table 14: It is false to insert zeros where data was not collected. 213 lobsters cannot weigh zero lbs. This has been fixed (now Table 15, page 37). Table 15 does not mention whether the data are in numbers or lbs. The caption now specifies that the catch is in numbers (now Table 16, page 39). Table E1 is an extremely simplistic assessment of the economic impact. The landed value of the discarded fish is not included; possible fuel cost or effort-linked costs are not included. Seasonal
pricing is not accounted for. Also, trip was a significant variable in the best fit model – why not show how the change in value changes by trip? An economic analysis was not one of the objectives of the project. The intent of this information was present a preliminary economic analysis, and suggest the value of more in-depth economic analysis (taking into account costs, landed value of fish, scallop price, etc.) that allows decision makers and fishermen to recognize that the different modifications of the dredges can not only can diminish the environmental impact, but also lead to economic gains (page 67). This point is emphasized in the final paragraph of the conclusion (page 41). Table F1 omits counts for “stringiness” That is correct. This table was put together by Susan Inglis, our collaborator at SMAST. She did not consider stringiness in her analysis.
References Leavitt J.S., C.J. Huntsberger, R.J. Smolowitz, and L.A. Siemann. (in review). The Seasonal distribution and abundance of barndoor skate on Georges Bank based on scallop dredge surveys. submitted to Fisheries Research. Siemann L.A., C.J. Huntsberger, J.S. Leavitt, and R.J. Smolowitz. (in review). Summering on the bank: seasonal distribution and abundance of monkfish on Georges Bank. submitted to Fisheries Oceanography. Smolowitz R.J., L.A. Siemann, C. Huntsberger and D. Boelke. 2016. Application of seasonal closures to reduce flatfish bycatch in the U.S. Atlantic sea scallop fishery. Journal of Shellfish Research. 35(2): 475-480. Winton M., C. Huntsberger, D. Rudders, G. DeCelles, K. Thompson, K. Goetting, and R. Smolowitz. 2017. Spatiotemporal patterns of flatfish bycatch in two scallop access areas on Georges Bank. Journal of the Northwest Atlantic Fishery Science 49: 23–37.
Response to questions regarding the second objective “Compare a modified dredge bag (5row apron), designed to reduce flatfish bycatch, with the standard dredge (7-row apron)” It was difficult to understand exactly what was tested; the text did not appear to match a table, and it was unclear how many comparisons were made. We added the following text to clarify the comparisons. "This planned comparison was completed for seven out of eight trips. But because the control dredge broke during the March trip, the CFF TDDs were redesigned with heavier reinforced center bars, and during the May trip, the old versus new TDD frames were tested with 7-row aprons on both." (page 11) "Results from the May trip (above) were not included in the analysis because that trip did not use the control and experimental dredges used for the rest of the study. Catch data from the remaining seven survey trips were treated as a single data set since the experimental treatment was consistent across trips and the heavier dredge (June 2016) did not appear to fish differently than the original dredge (August 2015 – March 2016)." (page 24) Response to questions regarding the sixth objective “Conduct biological sampling of bycatch crustacean and echinoderm species” Purpose and methodology was not described for this objective. An explanation of the purpose and methodology of this objective was added to the revised version of the final report (third paragraph on page 13). Response to questions regarding figures Many of the figures are not sufficiently legible to be read. There is a general issue with identification of individual trips: in some places, a month designation is used; in others, a cruise number; in others, the dates of the trip. And the month designation is a bit misleading: the trip called “Sep” occurred mostly in August; the trip called “Feb” occurred mostly in March. This naming convention misleads the reader about the spacing of the trips and affects the interpretation of the results. The figures were revised, and the labeling convention was unified. The trip labeled "Feb" in few of the graphs was a typo and should have been labeled "Mar". In August we had two trips, and to avoid confusions in the analysis the second trip, which was the last week of August, was changed to September trip convention. Locations are inadequately described in Figure 1, especially considering the use of CAII and non-CAII as variables in analyses. Figure 1 and Figures G1 to G6: All the maps were modified with the correct locations (CAII and non-CAII; pages: 11, 71-82). In Figure 2, I can see no discernable difference in the examples for “Light Brown” and “Brown”. Examples of “stringiness” which is analyzed but not defined or mentioned in the Methods, could also be provided.
Figure 2 was changed. A better picture of gray meats was added (page 12). Text was added to describe stringiness (page 12). Why is there no measure of variation per cruise (error bars, SD, SE, boxplot) for the bycatch rates in Figure 3? It should be made explicit whether the “scallop LB” in the axis labels is shell weight or meat weight. Figure 3 does not have error bars because bycatch rates were calculated from the entire catch per trip per each species, generating only one catch weight per month per species (page 1819). In Figure 4, the more important information is seasonal variation in SHMW ratio. A figure showing this ratio over time would help emphasize the importance of adjusting management and exploitation to maximize efficiency. I cannot see the marks within the boxes of the boxplots, and they are not defined in the legend – presumably, mean and median. The lengths of the whiskers should be also explained. The information contained in this graph relates to the variability in the raw data that was collected over the course of the study. These data are simply the observed shell heights and meat weights collected by sub-area over the eight sampling cruises. Text has been added to the caption to define the whisker plots (page 20). Figure 5 shows seasonal variation in meat weight for a 120-mm scallop (page 21). Figure 6a and 6b: Should the axes both be “MEAT WEIGHT (GRAMS)”? Or is one shell height, as the caption suggests? Figure 6 has been modified in the revised version of the final report (page 22). The axes labels have been corrected. Figure 9: “Multiple” is misspelled. This has been corrected in the revised version of the final report (page 27). Figure 14: Meanings of dots and of the lengths of the box whiskers need to be added Text has been added to the caption to define the whisker plots (page 32). Figure D1: These figures are all hard to read (similarly, the rest of the figures in the Appendices). The legends do not need to be repeated. The second y axis is labeled “FREQUENCY” but appears to be “Count” or “Number”. All graphs have been replaced with better resolution graphs in panel format which allows for the use of common X, Y and Y2 axes (pages: 59-62). The labels have also been changed. The second y-axis is the proportion of the catch in each size bin. Also, it appears that the modeling technique only fits linear relationships. In other gear comparison studies using GLMMs to analyze length effects, inflection points are allowed, and provide better fits. Figure D3 illustrates a problem with linear fits; the Cruise 221 panel shows a significant difference for sizes above 35 cm, with the experimental dredge catching less – but the only data points show retention in the experimental dredge alone – a very poor fit, and misleading as well. Most of the panels in D3 show similar confusing fits. Trimming the data so that only lengths with a certain level of observations (numbers at length) could produce more useable and believable figures. (Similarly for all model fits).
The cumulative proportion at length is not an indication of goodness of fit and only provides a visual reference as to the observed proportion because proportions can be misleading at lengths where the overall numbers are low. Comparing predicted proportions to the observed is useful when catches at length are high and there is some stability to the proportions. To minimize confusion, we trimmed the plots to include only these higher catch number regions as suggested. Figure D5 lacks units on the labels. I suspect counts, as the data do not appear to be continuous. The caption now specifies the catch as numbers (page 63). Figure D6 illustrates the confusion caused by inconsistent labels, as it is not easy to find the “January” cruise. Labels are now consistent throughout (page 64). Response to questions regarding tables Table 1: Also, there is insufficient information in the table or text to identify when the headbale was changed, and whether the “new CFF TDD” is very different from the “CFF TDD”. I could not tell whether there were actually 3-4 different gear comparisons instead of 2. In addition to adding clarifying text, we added months and more descriptions of the dredges below Table 1 (page 10). Table 4 caption: Isn’t intercept-only the most parsimonious model? How could any other model be simpler? The choice of the most parsimonious model depends on both the number of variables and how the best fitting model is selected. For this analysis, the most parsimonious model was chosen based on AIC value, indicating the model with the greatest explanatory power. We decided to compare the observed data with the “intercept only” model even though other, more complex models did provide better fits to the data for some species. Note that Table 4 is now Table 5 (page 26). Table 8 is an example of where the odd naming convention for the trips causes the trip to trip changes in data to be falsely interpreted. There is no intent to mislead. The dates specific to each month label are clearly defined in Table 1. Numbers of individuals captured by each dredge are shown in the table, while the text points out how many females and males were examined from the subsample of 10 individuals per tow (now Table 9, page 32). Table 14: It is false to insert zeros where data was not collected. 213 lobsters cannot weigh zero lbs. This has been fixed (now Table 15, page 37). Table 15 does not mention whether the data are in numbers or lbs. The caption now specifies that the catch is in numbers (now Table 16, page 39). Table E1 is an extremely simplistic assessment of the economic impact. The landed value of the discarded fish is not included; possible fuel cost or effort-linked costs are not included. Seasonal
pricing is not accounted for. Also, trip was a significant variable in the best fit model – why not show how the change in value changes by trip? An economic analysis was not one of the objectives of the project. The intent of this information was present a preliminary economic analysis, and suggest the value of more in-depth economic analysis (taking into account costs, landed value of fish, scallop price, etc.) that allows decision makers and fishermen to recognize that the different modifications of the dredges can not only can diminish the environmental impact, but also lead to economic gains (page 67). This point is emphasized in the final paragraph of the conclusion (page 41). Table F1 omits counts for “stringiness” That is correct. This table was put together by Susan Inglis, our collaborator at SMAST. She did not consider stringiness in her analysis.
References Leavitt J.S., C.J. Huntsberger, R.J. Smolowitz, and L.A. Siemann. (in review). The Seasonal distribution and abundance of barndoor skate on Georges Bank based on scallop dredge surveys. submitted to Fisheries Research. Siemann L.A., C.J. Huntsberger, J.S. Leavitt, and R.J. Smolowitz. (in review). Summering on the bank: seasonal distribution and abundance of monkfish on Georges Bank. submitted to Fisheries Oceanography. Smolowitz R.J., L.A. Siemann, C. Huntsberger and D. Boelke. 2016. Application of seasonal closures to reduce flatfish bycatch in the U.S. Atlantic sea scallop fishery. Journal of Shellfish Research. 35(2): 475-480. Winton M., C. Huntsberger, D. Rudders, G. DeCelles, K. Thompson, K. Goetting, and R. Smolowitz. 2017. Spatiotemporal patterns of flatfish bycatch in two scallop access areas on Georges Bank. Journal of the Northwest Atlantic Fishery Science 49: 23–37.
