Non-zero Catches Research Articles

Comparing predictions of fisheries bycatch using multiple spatiotemporal species distribution model frameworks

Spatiotemporal predictions of bycatch (i.e., catch of nontargeted species) have shown promise as dynamic ocean management tools for reducing bycatch. However, which spatiotemporal model framework to use for generating these predictions is unclear. We evaluated a relatively new method, Gaussian Markov random fields (GMRFs), with two other frameworks, generalized additive models (GAMs) and random forests. We fit geostatistical delta-models to fisheries observer bycatch data for six species with a broad range of movement patterns (e.g., highly migratory sea turtles versus sedentary rockfish) and bycatch rates (percentage of observations with nonzero catch, 0.3%–96.2%). Random forests had better interpolation performance than the GMRF and GAM models for all six species, but random forests performance was more sensitive when predicting data at the edge of the fishery (i.e., spatial extrapolation). Using random forests to identify and remove the 5% highest bycatch risk fishing events reduced the bycatch-to-target species catch ratio by 34% on average. All models considerably reduced the bycatch-to-target ratio, demonstrating the clear potential of species distribution models to support spatial fishery management.

The influence of a severe reservoir drawdown on springtime zooplankton and larval fish assemblages in Red Willow Reservoir, Nebraska

Reservoirs can be dynamic systems, often prone to unpredictable and extreme water-level fluctuations, and can be environments where survival is difficult for zooplankton and larval fish. Although numerous studies have examined the effects of extreme reservoir drawdown on water quality, few have examined extreme drawdown on both abiotic and biotic characteristics. A fissure in the dam at Red Willow Reservoir in southwest Nebraska necessitated an extreme drawdown; the water level was lowered more than 6 m during a two-month period, reducing reservoir volume by 76%. During the subsequent low-water period (i.e., post-drawdown), spring sampling (April–June) showed dissolved oxygen concentration was lower, while turbidity and chlorophyll-a concentration were greater, relative to pre-drawdown conditions. Additionally, there was an overall increase in zooplankton density, although there were differences among taxa, and changes in mean size among taxa, relative to pre-drawdown conditions. Zooplankton assemblage composition had an average dissimilarity of 19.3% from pre-drawdown to post-drawdown. The ratio of zero to non-zero catches was greater post-drawdown for larval common carp and for all larval fishes combined, whereas we observed no difference for larval gizzard shad. Larval fish assemblage composition had an average dissimilarity of 39.7% from pre-drawdown to post-drawdown. Given the likelihood that other dams will need repair or replacement in the near future, it is imperative for effective reservoir management that we anticipate the likely abiotic and biotic responses of reservoir ecosystems as these management actions will continue to alter environmental conditions in reservoirs.

Accounting for space–time interactions in index standardization models

Scientific survey data are used to estimate abundance trends for fish populations worldwide, and are frequently analyzed using delta-generalized linear mixed models (delta-GLMMs). Delta-GLMMs incorporate information about both the probability of catch being non-zero (catch probability) and the expected value for non-zero catches (catch rates). Delta-GLMMs generally incorporate year as a main effect, and frequently account for spatial strata and/or covariates. Many existing delta-GLMMs do not account for random or systematic differences in catch probability or rates in particular combinations of spatial strata and year (i.e., space–time interactions), and do not recognize potential correlation in random space–time interactions between catch probability and catch rates. We therefore develop a Bayesian delta-GLMM that estimates correlations between catch probability and rates, and compare it with either (a) ignoring year–strata interactions, (b) modeling year–strata interactions as fixed effects, or (c) estimating year–strata interactions in catch probability or rates as independent random effects. These four models are fitted to bottom trawl survey data for 28 species off the U.S. West Coast. The posterior median of the correlation is positive for the majority (18) of species, including all five for which the posterior distribution has little overlap with zero. However, estimating this correlation has little impact on resulting abundance indices or credible intervals. We therefore conclude that the correlated random model will have a little impact on index standardization of the West Coast bottom trawl dataset. However, we propose that the correlated model can quickly identify correlations between occupancy probability and density, and provide our code to allow researchers to quickly identify whether such a correlation is likely to be significantly different from zero for their chosen data set.

Improving fishery-independent indices of abundance for a migratory walleye population

The primary goal of many biological surveys is to provide an unbiased representation of trends of population abundance. However, there are often factors other than abundance that vary over time and influence catch rates and thus inferences about population trends from surveys. This is particularly true for highly mobile species because of interannual variation in the timing, extent, and duration of movements, and for surveys that are not randomized with respect to space and time. We developed general and generalized linear mixed models to standardize Canadian and United States fishery-independent surveys that provide an index of basin-level walleye (Sander vitreus) population abundance trends in Lake Erie (1983–2008). In Canadian waters, the probability of a non-zero catch was associated with the type of gill net set (canned>bottom sets), the presence of hypoxia (negative trend; −), and secchi depth (−). Positive catch rates were associated with the set type (canned>bottom) and water depth (+). In United States waters, survey catch rates were associated with secchi depth (−) and surface water temperature (+). For each case, the best model included random effects (interactions between year, week, basin, sub-basin) which accounted for a modest amount of the total variation. General abundance trends were similar between the standardized and nominal indices, but substantial annual variation in the direction and magnitude of the difference between indices was observed. We recommend the use of standardized indices for walleye population assessments because these account for factors influencing catch rates other than changes in abundance.

Survey estimates of king crab (Paralithodes camtschaticus) abundance off Northern Norway using GLMs within a mixed generalized gamma-binomial model and Bayesian inference

Abstract Hvingel, C., Kingsley, M.C.S., and Sundet, J.H. 2012. Survey estimates of king crab (Paralithodes camtschaticus) abundance off Northern Norway using GLMs within a mixed generalized gamma-binomial model and Bayesian inference. – ICES Journal of Marine Science, 69: . A trawl survey provides information on number and biomass of introduced king crab (Paralithodes camtschaticus) to the management of a fishery off the coast of Northern Norway; the annual catch quotas are largely set as a percentage of the survey estimate. A specially built sledge trawl was designed for the survey. It needs only small areas of trawlable bottom, performs well on a wide range of bottoms, and appears to have good catchability for benthic organisms. Many survey hauls catch no crabs and the non-zero catches have a highly skewed distribution. Data were therefore analysed with a compound model, in which separate predictors were fitted for the proportion of zero catches and for the catch size of the non-zero catches. The compound model was fitted by Bayesian methods using WinBUGS. The distribution of non-zero catches fitted well to a generalized gamma distribution, but with parameter values that made it approximate a lognormal distribution. Numbers of fishable crabs peaked in 2003, and total numbers in 2010 were about two-fifths of the 2003 maximum.

Integrated risk analysis for rare marine species impacted by fishing: sustainability assessment and population trend modelling

Abstract Zhou, S., Milton, D. A., and Fry, G. C. 2012. Integrated risk analysis for rare marine species impacted by fishing: sustainability assessment and population trend modelling. – ICES Journal of Marine Science, 69: 271–280. Risk assessment and conservation management of rare species are challenging due to a lack of data. We developed an integrated risk assessment approach to assess human impact on population sustainability of rare species. The approach involved two components: a quantitative sustainability assessment coupled with modelling trends in relative abundance. Both components took nil catches into account through zero-inflated statistical distributions that simultaneously modelled the zero and non-zero catches separately in submodels. The sustainability assessment used detection–non-detection data for population estimation and linked sustainability to easily collected life-history traits. This component provides an assessment of population sustainability at one point in time. The trend modelling applied zero-inflated negative binomial models to temporal trends in density and dispersion of species. It provided a complement to the static sustainability assessment. We applied this integrated approach to assess the risk to 14 species of rare, protected sea snakes incidentally caught in the Australian Northern Prawn Fishery. This approach can be applicable for risk assessment of many species with limited abundance data, a large number of absences and some presence–absence information only.

Assessment of landed and non-landed by-catch of walleye, yellow perch and white perch from the commercial gillnet fisheries of Lake Erie, 1994–2007

Theoretical by-catch (including landed and non-landed bycatch) of walleye ( S. vitreus), yellow perch ( Perca flavescens), and white perch ( Morone americana) from the Lake Erie commercial gillnet fisheries during 1994–2007, was predicted by a delta model developed on the fishery-independent survey data (Lake Erie Partnership Index Fishing Survey). The delta model consisted of one generalized additive model and one AdaBoost model. The generalized additive model was used to predict non-zero catches of the by-catch species, and the AdaBoost model was used to predict the probability of obtaining non-zero catches. Non-landed by-catch was estimated as the difference between the theoretical by-catch predicted from the delta model and the landed by-catch recorded in the commercial fishery data. The theoretical by-catch of walleye was relatively higher in the west basin in October. A higher theoretical by-catch of yellow perch occurred in the west central basin in November, and a higher theoretical by-catch of white perch occurred in the west central basin in October. We observed higher levels of non-landed by-catch of walleye in the west basin during August to September, higher levels of non-landed by-catch of yellow perch in the west central and east central basins in November, and higher levels of non-landed by-catch of white perch in the west basin in August and November. The combination of the AdaBoost model with the delta model provided an alternative model in by-catch analyses when the percentage of zero observations was high.

Use of generalized linear Bayesian model to mitigate the impact of spatial contraction in fishing pattern on the estimation of relative abundance

We developed a hierarchical Bayesian generalized linear model (Delta model) for CPUE standardization and tested this model with Japanese longline fisheries data from 1975 to 2006 for south Pacific Ocean albacore tuna ( Thunnus alalunga). The Delta model consists of binomial and lognormal sub-models and was developed to predict catch rates for areas bypassed by the fishery in some years. Incorporation of these predicted catch rates into the standardization process mitigates the impact of spatial contractions in fishing pattern on the estimation of abundance indices. The relative abundance of albacore as measured by the standardized and nominal CPUEs are similar for the early years when the spatial pattern of the fishing did not change. However, when the spatial coverage of fishing was decreasing, the standardized CPUEs without predicted catch rates and nominal CPUEs were similar in the first half of this period, but the standardized CPUE was lower than nominal CPUE in the remaining years of this period. In contrast, standardized CPUEs with the inclusion of predicted catch rates indicated significantly lower abundances than the nominal CPUEs for this entire period. Explanatory variables considered in this study were Year, Month, Area, Year–Area interaction, and Depth using hooks per basket (HPB) as a proxy variable. All variables except for HPB were important in explaining variations in catch rates. Comparison of model predicted probabilities and observed proportions of non-zero catches showed that the binomial sub-model fit was adequate, and calculated Bayesian p-values indicated that the lognormal sub-model fit was acceptable. We also show that model fitting can be improved by adding random process error into the relationship between the explanatory variables and the function of the expected values of the response variables.

Modeling environmental, spatial, temporal, and operational effects on blue shark by-catches in the Mediterranean long-line fishery

A study was conducted from 1998 to 2001 on blue shark (Prionace glauca) by-catch of the Italian and Greek surface-drifting swordfish long-line fisheries in the Mediterranean Sea. The focus was on examining whether catches are related to some environmental, spatial, temporal or operational parameters and to what extent, applying generalized linear model (GLM) approaches. Analyses indicated that most appropriate for the dataset was the Delta-lognormal model, which is a binomial error distribution for the probability of a non-zero catch and lognormal error for positive catch rates. Spatial and temporal factors were the most influential regarding blue shark distribution and abundance, with a considerable interaction between them; the modeled environmental factors were of minor importance. Spatial distribution revealed a strong longitudinal gradient whereby blue shark occurrences increased in an east to west direction, whereas catches by latitude were higher in southern- and northern-most regions. Blue sharks were more frequently encountered in autumn and in distant open waters; however, the likelihood of making a larger catch peaked in late spring–early summer and in the vicinity of land. Catch rates differed significantly depending on the fishing gear configuration. Deeper settings (>20 m), more resilient lines and use of fish attractants increased the probability of P. glauca capture.

Alternative error distribution models for standardization of catch rates of non-target species from a pelagic longline fishery: billfish species in the Venezuelan tuna longline fishery

Alternative error distributions were evaluated for calculating indices of relative abundance for non-target species using catch and effort data from commercial fisheries. A general procedure is presented for testing the underlying assumptions of different error distributions. Catch rates, from an observer program, of billfish caught mainly as bycatch in a pelagic tuna longline fishery in the Western Central Atlantic were standardized. Although catches of billfishes are not common in pelagic tuna longline fisheries, these fisheries are one of the main sources of fishing mortality for these stocks in the central Atlantic due to the magnitude and spatial extent of longline fishing effort. Billfish CPUE data are highly skewed with a large proportion of zero observations. Delta distribution models can accommodate this type of data, and involve modeling the probability of a non-zero observation and the catch rate given that the catch is non-zero separately. Three different Delta models were compared against other error distributions, including the lognormal, log-gamma, and Poisson. Diagnostic checks and deviance table analyses were performed to identify the best error distribution and the set of factors and interactions that most adequately explained the observed variability. The results indicated that the Delta-lognormal model (a binomial error distribution for the probability of a non-zero catch and lognormal error for the positive catch rates) complied best with the underlying characteristics of the data set. Analyses of catch rates for blue marlin, white marlin and sailfish confirmed the spatio-temporal nature of their distribution in the central Atlantic and Caribbean Sea. Also, the analyses indicated that catch rates of billfish differed among fishing vessel types; larger vessels had a higher probability of catching blue marlin, the more oceanic-oriented species, and lower probabilities of catching the more coastal-oriented species white marlin and sailfish. Standardized catch rates indicated in general a lower relative abundance for blue and white marlin in the most recent years, although estimated confidence intervals overlap through the years especially for white marlin.

A generalized linear mixed model analysis of a multi-vessel fishery resource survey

A generalized linear mixed model (GLMM) that treats year and spatial cell as fixed effects while treating vessel as a random effect is used to examine fishing power among chartered industry-based vessels and a research trawler, the FRV Miller Freeman, for bottom trawl surveys on the upper continental slope of U.S. West coast. A Bernoulli distribution is used to model the probability of a non-zero haul and the gamma distribution to model the non-zero catch rates of four groundfish species. The use of vessel as a random effect allows the data for the various vessels to be combined and a single continuous time-series of biomass indices to be developed for stock assessment purposes. The GLMMs fit the data reasonably well. Among the different models examined, the GLMM incorporating a random vessel × year effect had the smallest ΔAIC and was thus chosen as the best model. Also, estimated random effects coefficients associated with the industry-based vessels and the FRV Miller Freeman for each year suggests that these vessels can be assumed to be from a common random effects distribution. These results suggest that combining data from the chartered industry-based vessels and from the research trawler may be appropriate to develop indices of abundance for stock assessment purposes. Finally, an evaluation of variances associated with abundance indices from the different models indicate that analyzing these data as a fixed effect GLM may underestimate the level of variability due to ignoring the grouped nature of tows within vessels. As such, use of a mixed model approach with vessel as a random effect is a reasonable approach to developing abundance indices and their variances.

Use of binary and truncated negative binomial modelling in the analysis of recreational catch data

Data from surveys of recreational anglers fishing on three estuaries in eastern Australia reveal highly skewed distributions of catches with many zeros. Such data may be analyzed using a two-component approach involving a binary (zero/non-zero catch) response and the conditionally distributed non-zero catches. A truncated negative binomial model was effective in analyzing the non-zero catches. Covariates were incorporated in the modelling, and their critical assessment has led to improved measures of fishing effort for this recreational fishery.

In search of a better unit of effort in the coastal liftnet fishery with lights for small pelagics in Indonesia

Despite major criticism, catch per unit effort (CPUE) is still widely used as a measure for the size of the exploited stock, but its indicative value is affected by selection of a proper unit of effort. The unit of effort used in the Indonesian fisheries statistical system is poorly standardised with one trip made by a fishing unit of any type. Therefore, changes in catchability, induced by technical innovations cannot be accounted for in official statistics and thus bias the perception of fisheries authorities on the state of stocks. This study shows that differences in technical characteristics in a coastal liftnet fishery on small pelagics around Ambon, such as unit size and type of lamps used to attract fish, influence both effort allocation and catch per trip. Larger fishing units (more windlasses) are more commercially orientated. They have a larger nonzero catch per trip—1.5 times larger nonzero catch per added windlass—and ignore potentially small catches, as shows from their higher proportion of zero-catches. The shift from kerosene to electric lamps does not lead to a higher catch per trip, but the use of electric lamps is less laborious, inviting fishermen to fish also under more unfavourable light conditions such as during full moon. Model simulations of average catch per trip of two fishing areas show that technical innovations in the last 20 years have increased the mean catch per trip by a factor of 4.1 in the fishing area with the largest technical improvements. Technical improvements explain 80% of the change in catch per trip in the official statistics over the last 15 years. Furthermore, the variance in catch per trip among individual fishing units has increased and so has the uncertainty in the average CPUE, as only part of the fishermen implemented technical improvements. Both developments stress the poor indicator value of catch per trip for developments in fish stock biomass and the importance to standardise the unit of effort. The characteristic high variability in catch in this fishery will always obscure effects of technical improvements in more local situations and smaller timeframes. Because of the inconsistent effects of the type of lamps used on the catch rate, it is questionable whether standardising the unit of effort by the type of lamp used will enlarge the precision in the CPUE. Nevertheless, standardising the unit of effort by the size of the unit will enlarge the precision in, and the indicator value of the CPUE in more aggregated catch data as used on higher levels in the fisheries administration of Indonesia.

Use of generalized linear models to analyze catch rates having zero values: the Kuwait driftnet fishery

Fishery survey data often contain zero values. This paper uses a gamma-based generalized linear model for non-zero values and a Bernoulli-based model for the proportion of non-zero catches. Yearly variation in catch rates is then estimated by integrating the estimates from both models. The results for the Kuwait driftnet fishery show that the catch rate of silver pomfret ( Pampus argenteus) decreased from 115 kg per 1000 m clearances in 1984 to about 10 kg per 1000 m clearances in 1999, but its alternative target species, hilsa shad ( Tenualosa ilisha), exhibited no clear pattern with great fluctuation over time. The proportion of non-zero catches for silver pomfret decreased with time during the study period. In contrast, the proportion for hilsa shad increased, indicating fishermen shifted their targeting from silver pomfret to hilsa shad with the decreasing catch rate of silver pomfret. Other effect factors like month, boat type and landing port were also included in the models, and their effects on catch rates were estimated and tested.

Standardization of catch and effort data in a spatially-structured shark fishery

The methods used to develop catch rate based indices of relative abundance for the school shark Galeorhinus galeus resource off southern Australia are outlined. These methods are based on fitting generalized linear models to catch and effort data for several regions in this fishery. This is to take account of the multi-gear nature of the fishery and the spatial structure of the trends in catch rate. The data on whether or not the catch rate is zero and the catch rate given that it is non-zero are analysed separately and then combined to provide indices of abundance. The former analysis is based on assuming the data are Bernoulli random variables. Given the uncertainty about the appropriate error-model to assume when fitting generalized linear models to catch and effort data, four alternative error-models — log-normal, log-gamma, Poisson, and negative binomial — were explored when modelling the non-zero catch rates.

Catch, effort and sampling strategies in the highly variable sardine fisheries around East Java, Indonesia

Temporal and spatial patterns in the fishery for Sardinella spp. around East Java, Indonesia, were studied in an attempt to develop an efficient catch and effort sampling strategy for this highly variable fishery. The inter-annual and monthly variation in catch, effort and catch per unit of effort (cpue) are quantified and discussed, using official catch reports and reports from local village cooperatives. In 1993 the total catch of Sardinella spp. in East Java was 56 000 tonnes and the total effort was equivalent to 5000 units of purse seine. A maximum sustainable yield (MSY) of 40 000 tonnes of sardines is estimated, at an optimum effort of 3000 units of purse seine. The present study focuses on the purse seine fishery since this fishery accounts for more than 50% of all sardines landed in East Java. Purse seiners use lamps to attract fish and peak catches are landed around the new moon. Catches are smaller towards the full moon, when the fishery is inactive during a period of 10 days. Considerable differences exist between monthly catches from peak seasons and monthly catches from off-seasons. The fishery is further characterized by a relatively large amount of zero catches. A catch and effort data recording system (CEDRS) is developed based on the analysis of temporal variation in cpue and effort in the purse seine fishery for sardines. A purse seine fleet, with a variable size of 40 to 70 vessels, was sampled in Probolinggo harbour at the Madura Strait, during 4 to 6 days per month from March 1990 until March 1991. The overall mean effort of the fleet was 36.1 trips per day, of which on average 12.6 trips per day resulted in non-zero catches. The geometric mean cpue over the non-zero catches was 404 kg per trip. The proposed CEDRS has an accuracy of 30% for the estimated annual catch, 21% for the cpue and 9% for the effort (non-zero catches). A minimum required number of 4 sampling days per month is calculated for the temporally stratified CEDRS, with months used as strata within years, and lunar phases as strata within months. For this purpose, the lunar cycle is divided in three lunar phases, one around the full moon, one around half moon and one around the new moon. Within the proposed CEDRS, no sampling takes place in the lunar phase around full moon, when the fishery is inactive. The sample of 4 days per month is proportionally allocated over the other two lunar phases.

A Model Expressing the Relationship between Catch and Soak Time for Trap Fisheries

In trap fisheries, catch per trap haul changes with soak time. Catch rates should therefore be standardized for differences in soak time when used to measure abundance. Several models have been used to describe catch per trap haul while adjusting for soak time. Some of these models are limited with respect to time spent fishing or have difficulties in parameter estimation. In this paper, we describe a model without limitations on soak time. The model is Ct = ab + a(t – b)e–ct; Ct is the catch per trap haul at soak time t; a, b, and c are parameters to be estimated. When sufficient data are used, the term ab denotes asymptotic catch after an infinite soak time, and a and c depend on local animal density, entry rate, and escape probability. These parameters can be readily estimated by many popular computer programs. This model is equivalent, or superior, to other models and is the only suitable model for long soak times with nonzero catch.