- Split View
-
Views
-
Cite
Cite
Jochen Depestele, Sofie Vandemaele, Willy Vanhee, Hans Polet, Els Torreele, Herwig Leirs, Magda Vincx, Quantifying causes of discard variability: an indispensable assistance to discard estimation and a paramount need for policy measures, ICES Journal of Marine Science, Volume 68, Issue 8, September 2011, Pages 1719–1725, https://doi.org/10.1093/icesjms/fsr030
- Share Icon Share
Abstract
Fishery-dependent data underpin the scientific advice given to fishery managers. However, discard estimates are often imprecise as a result of limited sampling coverage. Estimating discard rates from length frequency distributions (LFDs) in commercial catches may complement information from observer trips. The accuracy of estimates depends greatly on careful investigation of the discard variability. Here, the impact of three essential factors was quantified for beam-trawl fisheries in the southern North Sea: (i) market prices, (ii) landings per trip (LPT) limitations, and (iii) selectivity of the commercial fishing gear. Observed discard rates for cod, plaice, sole, and whiting were compared with estimates based on length frequency data, taking account of the variability attributable to LPT limitations and market price. Observed discard estimates of cod and whiting differed significantly from LFD-derived estimates because of highgrading. The results indicate that LFD-derived discard estimates are only reliable if the crucial driving factors are quantified. LFDs can be collected from research vessels or by fishers in partnership with scientists. Based upon many of these LFDs and the discard-variability factors identified in observer programmes, discard rates can be estimated better.Depestele, J., Vandemaele, S., Vanhee, W., Polet, H., Torreele, E., Leirs, H., and Vincx, M. 2011. Quantifying causes of discard variability: an indispensable assistance to discard estimation and a paramount need for policy measures. – ICES Journal of Marine Science, 68: 1719–1725.
Introduction
North Sea fisheries are responsible for the highest level of discards in the world (Kelleher, 2005), although discard rates have dropped in recent years (Aarts and Poos, 2009; Enever et al., 2009). Fishery management is expected to reduce fisheries discards (Enever et al., 2009), especially discards of the beam-trawl flatfish fishery in the North Sea (Catchpole et al., 2008). This fishery is among the greatest generators of discards in the North Sea (Catchpole et al., 2005).
Reliable estimates of discard rates and the causes of their variability are of paramount importance both for stock assessment (Dickey-Collas et al., 2007) and effective fishery management (Rochet and Trenkel, 2005). Currently, estimates of discards are highly imprecise, owing to the limited coverage of the fleet for discard sampling (Dickey-Collas et al., 2007).
Estimating discard rates from gear selectivity characteristics, in addition to observer data, may provide valuable information for commercial fish species. Several exercises illustrate this theory, e.g. for discards of non-commercial fish species (Piet et al., 2009), for historical discards (Aarts and Poos, 2009), and for studying spatio-temporal variation in discarding (Welch et al., 2008). Discard rates can be inferred from the assumption that commercial fish below minimum landing size (MLS) are discarded, whereas those above MLS are landed. The discarding of marketable fish, defined as highgrading (Gillis et al., 1995a), can reduce the accuracy of such discard estimates (Piet et al., 2009). Length frequency distributions (LFDs) associated with each type of commercial fishing gear can be obtained from gear-technology trials or directly from fishers. Additionally, they can be deduced from selectivity parameters and estimates of absolute abundance. The latter are obtained by raising fish-density data while taking account of the catchability of survey gear (Piet et al., 2009). Adding this method of discard estimation to current monitoring programmes could expand the number of records and has the potential to extend the spatio-temporal distribution of discard data at low cost (Polet et al., 2010).
Discard estimates from modelling exercises depend on fish abundance, fishing effort, MLS, and gear selectivity (Piet et al., 2009). Market incentives and policy measures have been identified as crucial drivers of discarding, but there seems to be no quantitative assessment of their contribution to the discarding problem (Rochet and Trenkel, 2005).
Here we elucidate the market- and policy-driven causes of discarding. To illustrate it, four commercial fish species were selected from the Belgian flatfish beam-trawl fishery in the southern North Sea, Dover sole (Solea solea), plaice (Pleuronectes platessa), cod (Gadus morhua), and whiting (Merlangius merlangus). Understanding the reasons for discarding may have implications for policy measures such as quota restrictions and gear regulations. It also clarifies what parameters other than gear selectivity and MLS need to be taken into account when estimating discards of commercial fish species.
Material and methods
The Belgian discard-sampling programme
The study focused exclusively on the Belgian flatfish-directed beam-trawl fishery in the southern North Sea (ICES Division IVc) during 2006, 2007, and 2008. Fishery-dependent data are collected for the Belgian beam-trawl fleet in accordance with the EC Data Collection Regulation (EC, 2000). These include landing and discard data from commercial vessels, obtained since 2004 through on-board observation. Selection of the vessels for this observer programme is random but conditional on the cooperation of fishers. The data were collected during fishing activities in one or more ICES Divisions, e.g. IVb and IVc. Fishing trips were defined as the primary sampling units, and hauls within a fishing trip as secondary sampling units. The latter are correlated within trips, implying that hauls are not independent and can confound statistical analyses (see below). The sampled gear was a beam trawl, 4 or 12 m wide, equipped with a chain mat and a codend of 80 mm mesh.
Sampling coverage ranged between 2.4 and 4.5% of landings, comparable with other discard-sampling programmes (e.g. Stratoudakis et al., 1998; Rochet et al., 2002). The spatial distribution of sampling is a good reflection of commercial fishing activity in the southern North Sea (Figure 1), which was primarily in the west of the area. In all, 28 trips (374 tows) were sampled on board nine beam trawlers. Every alternate haul was sampled by the observers, unless that haul was rejected (e.g. as a result of severe net damage, <3%), in which case the next haul was sampled. Sampling took place around the clock to reflect typical working conditions. Fishers sorted the marketable fish from the conveyor, and the observers also collected all commercial fish species discarded, both undersized and highgraded. The observers did not interfere with fishing operations. However, an observer effect resulting from changes in fishing practice and location could bias the observed discard rates (Liggins et al., 1997; Benoît and Allard, 2009).
Discard rates and highgrading
Calculating discard rates based on LFDs and MLS (Destim) partly reflects the selectivity of the beam trawl and does not account for the sorting behaviour of fishers. Fish below MLS were considered discards (d), whereas those above MLS were assumed to be harvested (k). Consequently, highgrading was not considered in Destim. Equation (1) gives Destim in numbers, and a length–weight conversion is needed to obtain the corresponding weight. Relevant data were collected in the southern North Sea during August and September of 2007 and 2008, during research surveys with the RV “Belgica”. The total length of the fish (L, cm) was measured to the nearest 0.5 cm below, and ungutted wet weights (wt, g) were also recorded at sea. Numbers were converted to length using a weight–length relationship (Table 1).
Species . | a . | b . | Pearson r² . | n . | Minimum (cm) . | Maximum (cm) . |
---|---|---|---|---|---|---|
Sole | 0.007568 | 3.0617 | 0.9862 | 2 724 | 5.5 | 40.5 |
Plaice | 0.009641 | 3.0319 | 0.9935 | 2 897 | 8.5 | 57 |
Cod | 0.010137 | 2.9912 | 0.9904 | 83 | 15.5 | 62 |
Whiting | 0.009030 | 2.9508 | 0.9898 | 1 021 | 5.5 | 39 |
Species . | a . | b . | Pearson r² . | n . | Minimum (cm) . | Maximum (cm) . |
---|---|---|---|---|---|---|
Sole | 0.007568 | 3.0617 | 0.9862 | 2 724 | 5.5 | 40.5 |
Plaice | 0.009641 | 3.0319 | 0.9935 | 2 897 | 8.5 | 57 |
Cod | 0.010137 | 2.9912 | 0.9904 | 83 | 15.5 | 62 |
Whiting | 0.009030 | 2.9508 | 0.9898 | 1 021 | 5.5 | 39 |
Species . | a . | b . | Pearson r² . | n . | Minimum (cm) . | Maximum (cm) . |
---|---|---|---|---|---|---|
Sole | 0.007568 | 3.0617 | 0.9862 | 2 724 | 5.5 | 40.5 |
Plaice | 0.009641 | 3.0319 | 0.9935 | 2 897 | 8.5 | 57 |
Cod | 0.010137 | 2.9912 | 0.9904 | 83 | 15.5 | 62 |
Whiting | 0.009030 | 2.9508 | 0.9898 | 1 021 | 5.5 | 39 |
Species . | a . | b . | Pearson r² . | n . | Minimum (cm) . | Maximum (cm) . |
---|---|---|---|---|---|---|
Sole | 0.007568 | 3.0617 | 0.9862 | 2 724 | 5.5 | 40.5 |
Plaice | 0.009641 | 3.0319 | 0.9935 | 2 897 | 8.5 | 57 |
Cod | 0.010137 | 2.9912 | 0.9904 | 83 | 15.5 | 62 |
Whiting | 0.009030 | 2.9508 | 0.9898 | 1 021 | 5.5 | 39 |
Factors affecting the highgrading of cod and whiting
Discard rates estimated directly from LFDs were compared with on-board observations using t-tests. Any discrepancies between the observed and estimated discard rates may be related to market price, landings per trip (LPT) limitations, and/or catch composition. These factors were evaluated as explanatory factors for discarding fish above MLS. Mean fish prices were retrieved from the fishing ports of Ostend, Zeebrügge, and Nieuwpoort. The LPT constraints for each sampled vessel were calculated from national regulations, which specify the maximum allowable LPT-day based on engine power. The maximum allowable landings varied over time, according to national management decisions, ensuring that the total landings of the Belgian fleet matched national quota restrictions. The maximum allowable landings of whiting were 250 kg d−1 for Belgian beam trawlers in 2006, but no limitations were set in 2007 and 2008. On 27 November 2008, the fishery for whiting was closed. Constraints on catch composition are set by EU regulation. The catch must consist of at least 70% of listed species at all times before return to port (EC, 1998). The list includes a wide range of species, including, for instance, sole, plaice, dab (Limanda limanda), brill (Scophthalmus rhombus), and turbot (Psetta maxima). Cod is not in the list, but whiting is. To investigate the influence of catch composition, whiting was considered at risk of being discarded if the composition of trip landings had >75% of listed species. Similarly, cod could not exceed 20% of the landings. This factor was included in the analysis by considering cod discards to be likely if a trip landing of cod exceeded 15%.
Statistical analysis
Results
Discard rates and highgrading
To estimate the discard rates by weight, numbers-at-length were converted to weight-at-length, using length–weight relationships. Parameters a and b of the length–weight relationships are given in Table 1, together with the regression coefficient (Pearson r2), the number of individuals measured (n), and the sizes of the smallest and largest fish measured.
Discard rates are presented as numbers and weights for sole, plaice, cod, and whiting (Table 2). The observed and the estimated discard rates differed significantly (p < 0.01 in a paired t-test) for all four species. Examining LFDs shows that the discard rates were apparently dictated by MLS for sole and plaice, but less so for whiting and cod, of which there was substantial discarding above MLS (Figure 2). This is equally demonstrated by the higher standard deviations of the differences between the observed and estimated discard rates for cod and to some degree for whiting. The LFD curves for cod and whiting imply that factors other than MLS determine discarding. The mean proportion of discarded fish above MLS was calculated for each quarter in each sampling year (Table 3). Few cod were apparently highgraded in 2006, but in the final quarter of 2007 and the first and second quarters of 2008, highgrading was recorded. Whiting was highgraded in each quarter in each year, although highgraded proportions were less in the second quarter of 2006 and the first of 2007. Highgraded proportions of whiting were higher in the fourth quarter of 2006 and the second quarter of 2008; no data were available for the first quarter of 2007. The findings suggest that discards of fish above MLS did not change predictably between years. Factors driving highgrading of cod and whiting were then investigated in more detail.
Species . | Observed discard rate mean (s.d.) . | Estimated discard rate mean (s.d.) . | Difference observed and estimated discard rate mean (s.d.) . |
---|---|---|---|
Sole | 0.29 (0.18) | 0.25 (0.16) | 0.04 (0.04) |
0.13 (0.11) | 0.11 (0.10) | 0.02 (0.03) | |
Plaice | 0.42 (0.24) | 0.39 (0.22) | 0.02 (0.07) |
0.27 (0.21) | 0.25 (0.18) | 0.03 (0.10) | |
Cod | 0.65 (0.29) | 0.55 (0.31) | 0.10 (0.27) |
0.47 (0.31) | 0.36 (0.31) | 0.12 (0.24) | |
Whiting | 0.70 (0.33) | 0.55 (0.29) | 0.15 (0.51) |
0.61 (0.33) | 0.46 (0.27) | 0.15 (0.16) |
Species . | Observed discard rate mean (s.d.) . | Estimated discard rate mean (s.d.) . | Difference observed and estimated discard rate mean (s.d.) . |
---|---|---|---|
Sole | 0.29 (0.18) | 0.25 (0.16) | 0.04 (0.04) |
0.13 (0.11) | 0.11 (0.10) | 0.02 (0.03) | |
Plaice | 0.42 (0.24) | 0.39 (0.22) | 0.02 (0.07) |
0.27 (0.21) | 0.25 (0.18) | 0.03 (0.10) | |
Cod | 0.65 (0.29) | 0.55 (0.31) | 0.10 (0.27) |
0.47 (0.31) | 0.36 (0.31) | 0.12 (0.24) | |
Whiting | 0.70 (0.33) | 0.55 (0.29) | 0.15 (0.51) |
0.61 (0.33) | 0.46 (0.27) | 0.15 (0.16) |
Upper rows, number-based discard rates; lower rows, weight-based rates. The observed rates have been determined at sea, whereas the estimates are obtained from LFDs of landings and MLSs. Their differences indicate the validity of using LFDs and MLSs for discard estimation.
Species . | Observed discard rate mean (s.d.) . | Estimated discard rate mean (s.d.) . | Difference observed and estimated discard rate mean (s.d.) . |
---|---|---|---|
Sole | 0.29 (0.18) | 0.25 (0.16) | 0.04 (0.04) |
0.13 (0.11) | 0.11 (0.10) | 0.02 (0.03) | |
Plaice | 0.42 (0.24) | 0.39 (0.22) | 0.02 (0.07) |
0.27 (0.21) | 0.25 (0.18) | 0.03 (0.10) | |
Cod | 0.65 (0.29) | 0.55 (0.31) | 0.10 (0.27) |
0.47 (0.31) | 0.36 (0.31) | 0.12 (0.24) | |
Whiting | 0.70 (0.33) | 0.55 (0.29) | 0.15 (0.51) |
0.61 (0.33) | 0.46 (0.27) | 0.15 (0.16) |
Species . | Observed discard rate mean (s.d.) . | Estimated discard rate mean (s.d.) . | Difference observed and estimated discard rate mean (s.d.) . |
---|---|---|---|
Sole | 0.29 (0.18) | 0.25 (0.16) | 0.04 (0.04) |
0.13 (0.11) | 0.11 (0.10) | 0.02 (0.03) | |
Plaice | 0.42 (0.24) | 0.39 (0.22) | 0.02 (0.07) |
0.27 (0.21) | 0.25 (0.18) | 0.03 (0.10) | |
Cod | 0.65 (0.29) | 0.55 (0.31) | 0.10 (0.27) |
0.47 (0.31) | 0.36 (0.31) | 0.12 (0.24) | |
Whiting | 0.70 (0.33) | 0.55 (0.29) | 0.15 (0.51) |
0.61 (0.33) | 0.46 (0.27) | 0.15 (0.16) |
Upper rows, number-based discard rates; lower rows, weight-based rates. The observed rates have been determined at sea, whereas the estimates are obtained from LFDs of landings and MLSs. Their differences indicate the validity of using LFDs and MLSs for discard estimation.
. | Highgraded proportion of discards per quarter . | |||
---|---|---|---|---|
Year . | Quarter 1 . | Quarter 2 . | Quarter 3 . | Quarter 4 . |
Cod | ||||
2006 | 0.03 (0.09) | 0 (0.01) | 0 | 0.04 (0.08) |
0.04 (0.09) | 0.01 (0.02) | 0 | 0.04 (0.09) | |
2007 | No data | 0.01 (0.05) | 0.05 (0.11) | 0.47 (0.35) |
0.01 (0.04) | 0.06 (0.11) | 0.37 (0.32) | ||
2008 | 0.62 (0.25) | 0.28 (0.26) | 0.02 (0.06) | 0.03 (0.13) |
0.49 (0.22) | 0.31 (0.24) | 0.04 (0.09) | 0.03 (0.14) | |
Whiting | ||||
2006 | 0.17 (0.24) | 0.07 (0.07) | 0.18 (0.13) | 0.23 (0.18) |
0.18 (0.23) | 0.08 (0.04) | 0.22 (0.08) | 0.22 (0.14) | |
2007 | No data | 0.12 (0.13) | 0.15 (0.09) | 0.16 (0.11) |
0.12 (0.11) | 0.23 (0.06) | 0.20 (0.05) | ||
2008 | 0.10 (0.12) | 0.32 (0.24) | 0.18 (0.08) | 0.17 (0.13) |
0.15 (0.11) | 0.37 (0.22) | 0.26 (0.09) | 0.23 (0.16) |
. | Highgraded proportion of discards per quarter . | |||
---|---|---|---|---|
Year . | Quarter 1 . | Quarter 2 . | Quarter 3 . | Quarter 4 . |
Cod | ||||
2006 | 0.03 (0.09) | 0 (0.01) | 0 | 0.04 (0.08) |
0.04 (0.09) | 0.01 (0.02) | 0 | 0.04 (0.09) | |
2007 | No data | 0.01 (0.05) | 0.05 (0.11) | 0.47 (0.35) |
0.01 (0.04) | 0.06 (0.11) | 0.37 (0.32) | ||
2008 | 0.62 (0.25) | 0.28 (0.26) | 0.02 (0.06) | 0.03 (0.13) |
0.49 (0.22) | 0.31 (0.24) | 0.04 (0.09) | 0.03 (0.14) | |
Whiting | ||||
2006 | 0.17 (0.24) | 0.07 (0.07) | 0.18 (0.13) | 0.23 (0.18) |
0.18 (0.23) | 0.08 (0.04) | 0.22 (0.08) | 0.22 (0.14) | |
2007 | No data | 0.12 (0.13) | 0.15 (0.09) | 0.16 (0.11) |
0.12 (0.11) | 0.23 (0.06) | 0.20 (0.05) | ||
2008 | 0.10 (0.12) | 0.32 (0.24) | 0.18 (0.08) | 0.17 (0.13) |
0.15 (0.11) | 0.37 (0.22) | 0.26 (0.09) | 0.23 (0.16) |
Upper rows, number-based highgraded proportion; lower rows, weight-based proportion.
. | Highgraded proportion of discards per quarter . | |||
---|---|---|---|---|
Year . | Quarter 1 . | Quarter 2 . | Quarter 3 . | Quarter 4 . |
Cod | ||||
2006 | 0.03 (0.09) | 0 (0.01) | 0 | 0.04 (0.08) |
0.04 (0.09) | 0.01 (0.02) | 0 | 0.04 (0.09) | |
2007 | No data | 0.01 (0.05) | 0.05 (0.11) | 0.47 (0.35) |
0.01 (0.04) | 0.06 (0.11) | 0.37 (0.32) | ||
2008 | 0.62 (0.25) | 0.28 (0.26) | 0.02 (0.06) | 0.03 (0.13) |
0.49 (0.22) | 0.31 (0.24) | 0.04 (0.09) | 0.03 (0.14) | |
Whiting | ||||
2006 | 0.17 (0.24) | 0.07 (0.07) | 0.18 (0.13) | 0.23 (0.18) |
0.18 (0.23) | 0.08 (0.04) | 0.22 (0.08) | 0.22 (0.14) | |
2007 | No data | 0.12 (0.13) | 0.15 (0.09) | 0.16 (0.11) |
0.12 (0.11) | 0.23 (0.06) | 0.20 (0.05) | ||
2008 | 0.10 (0.12) | 0.32 (0.24) | 0.18 (0.08) | 0.17 (0.13) |
0.15 (0.11) | 0.37 (0.22) | 0.26 (0.09) | 0.23 (0.16) |
. | Highgraded proportion of discards per quarter . | |||
---|---|---|---|---|
Year . | Quarter 1 . | Quarter 2 . | Quarter 3 . | Quarter 4 . |
Cod | ||||
2006 | 0.03 (0.09) | 0 (0.01) | 0 | 0.04 (0.08) |
0.04 (0.09) | 0.01 (0.02) | 0 | 0.04 (0.09) | |
2007 | No data | 0.01 (0.05) | 0.05 (0.11) | 0.47 (0.35) |
0.01 (0.04) | 0.06 (0.11) | 0.37 (0.32) | ||
2008 | 0.62 (0.25) | 0.28 (0.26) | 0.02 (0.06) | 0.03 (0.13) |
0.49 (0.22) | 0.31 (0.24) | 0.04 (0.09) | 0.03 (0.14) | |
Whiting | ||||
2006 | 0.17 (0.24) | 0.07 (0.07) | 0.18 (0.13) | 0.23 (0.18) |
0.18 (0.23) | 0.08 (0.04) | 0.22 (0.08) | 0.22 (0.14) | |
2007 | No data | 0.12 (0.13) | 0.15 (0.09) | 0.16 (0.11) |
0.12 (0.11) | 0.23 (0.06) | 0.20 (0.05) | ||
2008 | 0.10 (0.12) | 0.32 (0.24) | 0.18 (0.08) | 0.17 (0.13) |
0.15 (0.11) | 0.37 (0.22) | 0.26 (0.09) | 0.23 (0.16) |
Upper rows, number-based highgraded proportion; lower rows, weight-based proportion.
Factors affecting the highgrading of cod and whiting
The LPT limitation for cod varied over the year, without any obvious, repeated pattern (Figure 3). Fish prices were also variable: €2.9–4.5 per kg for cod and €1.0–2.4 per kg for whiting (Figure 3). Five fishing trips (120 hauls) yielded landings of at least 15% cod, and on three fishing trips (58 hauls), the landings included <75% of listed species.
None of the examined explanatory variables in the GLMM could explain the variation in the response variable [Dobs/(Dobs + Destim)] for whiting, demonstrating that discarding of whiting above MLS was not the result of a variation in fish price, LPT, or the catch composition. Backward selection of the fixed-effect variables for cod resulted in a model with LPT as the only significant variable (F = 6.806, d.f. = 1.486, p < 0.001). The smoother had 1.486 effective degrees of freedom, indicating a nearly linear decreasing trend of the response as a function of LPT (Figure 4). However, the adjusted r² is 0.28, indicating that the explanatory power of the model is limited.
Discussion
Here we investigated the discarding behaviour of the Belgian beam-trawl fishery in the southern North Sea. Sole and plaice were highgraded, but the low quantities demonstrate that highgrading is unimportant. Cod above MLS are possibly discarded as a consequence of LPT restrictions, whereas discards of whiting above MLS cannot be attributed to seasonal changes in fish price, LPT limitations, or catch composition. Estimating the discard rates of sole and plaice from LFDs and MLS is therefore proposed for the Belgian beam-trawl fishery in the southern North Sea, but additional factors need to be considered if discard rates are to be estimated for cod and whiting.
Highgrading and discard estimates of target species
The observed and estimated discard rates for all species were significantly different, although only a limited quantity of sole and plaice above MLS was discarded and just few undersized sole and plaice were landed. These differences are more likely attributable to the visual sorting process. For example, fishers may unintentionally discard plaice 27 cm long and retain undersized fish. The limited landings of undersized fish and discards of fish above MLS contributed to the differences between the observed and estimated discard rates. Therefore, we conclude that highgrading of sole and plaice is negligible in the Belgian beam-trawl fishery in the southern North Sea.
Kell and Bromley (2004) hypothesized that differences in highgrading can be explained largely by the interplay of targeting behaviour and LPT limitations. Targeting behaviour depends on species abundance and fish price (Gillis et al., 2008), so is indicated by the total landed value. The proportion of landed value was 49–53% for sole, 11–13% for plaice, 4–5% for cod, and <1% for whiting for the Belgian fishing fleet in the period 2006–2008. Assuming sole and plaice to be real target species for the Belgian beam trawlers in the southern North Sea, this study might indicate that fishers can eliminate highgrading of target species. For example, the choice of fishing grounds may preclude highgrading, with relative abundances of target species matched with availability of LPT (Gillis et al., 2008; Quirijns et al., 2008). If an excessive quantity of marketable plaice is caught, fishers may redirect fishing effort towards patches of lesser abundance. Bromley (2000) similarly suggested that if marketable plaice are of low value (as is typical in early spring), fishing effort is reduced.
In contrast to these findings, highgrading of plaice has been suggested to take place when the largest size classes are in poor condition. Highgrading low-value fish, typically done in early spring, is meant to save LPT for high-value fish at the end of the year (Poos et al., 2010). The monthly plaice landings of the Belgian beam-trawl fishery ranged between 2 and 7% of the annual total from February to August, and between 10 and 18% from September to December. Combining these results with Figure 2 seems to confirm that Belgian fishers do not highgrade plaice in the southern North Sea. Interestingly, this contrasts with the results from the Dutch beam-trawl fishery. A possible explanation lies in the overlap between the spatial distribution of plaice and fishing effort. Landings of the Belgian beam-trawl fishery in the southern North Sea primarily originate in fishing grounds with a rocky seabed along the western part of ICES Division IVc (Figure 1). Dutch beam trawlers mostly operate in the southeastern and central North Sea (Poos and Rijnsdorp, 2007), where plaice abundance is greater (Bogaards et al., 2009). However, avoiding LPT limitations on plaice could also be realized by misreporting plaice catches to a bordering ICES Division with unfulfilled LPT (B. Deputter, pers. comm.). Finally, the differences in discarding behaviour could be due to an observer effect, which has been demonstrated in the Gulf of St Lawrence (Benoît and Allard, 2009). Perhaps Belgian fishers chose not to highgrade sole and plaice because of the observers’ presence. A post hoc analysis of this effect is difficult to address quantitatively within the current design of the discard-sampling programme (Benoît and Allard, 2009), so was not tested in this study. However, the potential observer effect implies the need for caution when drawing conclusions on the apparent lack of highgrading of sole and plaice in the beam-trawl fishery. Understanding the preconditions of the fishery is advisable if sole and plaice discards are estimated from LFD and MLS.
Highgrading and discard estimates of bycatch species
Discard rates of bycatch species must account for market- and management-induced variability. Species with low abundance and value lack profitability, so are more likely to be highgraded. This statement was formally tested for whiting, which was highgraded in nearly all quarters (Table 3), independent of LPT restrictions. Stratoudakis et al. (1998) obtained similar findings for demersal trawlers and seiners catching low-value gadoids. For bycatch species with low profitability, LPT was not the main limitation. Vessel-storage capacity and/or catch composition should reveal the incentives for highgrading, especially when low-value species are discarded to make space for more profitable fish (Gillis et al., 1995b), when catches of target species are below expectations (Redant and Polet, 1994), or when legislation requires the bulk of the landings to be of the target species (Catchpole et al., 2008; Benoît and Allard, 2009). The analysis of highgraded whiting in this study did not indicate catch composition at the trip level as an important driver. However, examining the variability of catch composition on a haul-by-haul basis might uncover a relationship between catch composition and highgrading of bycatch species. Our analysis shows substantial cod discarding above MLS during periods with stringent LPT limitations. Catch-composition requirements might also have influenced highgrading, but a haul-by-haul approach for detecting this effect is again preferable. Owing to the lack of landings and discards data by haul, however, that hypothesis could not be tested.
Management implications
This study has indirect implications for fishery management. First, it indicates that estimation of discard rates may complement the data available from discard observer programmes. This could lead to improved stock assessment and advice on total allowable catches. A protocol for such estimates should require that commercial species be regulated by MLSs, and that LFDs of the catch using a particular fishing gear can be estimated for the species under investigation. One conceivable way of collecting those LFDs would be to involve fishers along with electronic monitoring and automated observations of fish species and lengths using a digital camera (White et al., 2006; Benoît and Allard, 2009). Another way to obtain LFDs of fish catches using commercial gear would be through combining abundance estimates and selectivity parameters (Piet et al., 2009). As LFDs for each type of fishing gear and fish species can be collected more easily than actual discard data, their spatial and temporal range could be expanded considerably. However, estimates would need to be verified against observed discard rates from observer programmes, because variability in discarding can also result from market- and management-induced factors (Rochet and Trenkel, 2005). Such a protocol, when applied to the Belgian beam-trawl fishery in the southern North Sea, would result in justifiable estimates for the target species (sole and plaice). For highly profitable bycatch species (e.g. cod), LPT incentives need to be accounted for when making an estimate. For bycatch species with low profitability (e.g. whiting), LPT and seasonal variations in fish price cannot explain the variation in discarding. For low-value species, a more-detailed investigation of discard-variability factors, such as fishers’ behaviour in response to catch composition, is needed.
Second, specifying the drivers of discard variability expedites the identification of discard-mitigation measures and the evaluation of introduced actions (Borges et al., 2006; Enever et al., 2009). Whiting is a clear example where gear selectivity needs to be changed to avoid the conditions leading to highgrading. Highgrading takes place in nearly all quarters, although there are feasible techniques that reduce catches of whiting without major losses of sole and plaice (van Marlen, 2003). In addition to establishing factors of discard variability and mitigating the causes, these results need to be followed up. Only in this way will fishery management have any hope of succeeding (Graham et al., 2007; Enever et al., 2009).
Acknowledgements
We thank the shipowners, crew, and on-board observers for their assistance in the ILVO discard sampling programme. The research was facilitated by information from the Administrative Centre of the Department for Agriculture and Fisheries of the Flemish Government (Eddy Tessens). We are also indebted to Stefan Van Dongen for helpful discussions and valuable suggestions on the statistical analyses, and to Miriam Levenson for grammatical polishing. We also acknowledge the inputs of the Journal guest and in-house editor and two anonymous referees, for advice that considerably improved the quality of this paper. The study was partly carried out within the WAKO-II project, funded by Contract SD/NS/08A of the Belgian Science Policy. Further support for the work was provided by the Flemish Government and the European Commission (through the Data Collection Framework).