Reducing the short-term mortality of juvenile school prawns ( Metapenaeus macleayi ) discarded during trawling

Abstract

A field experiment was carried out in southeastern Australia to assess the short-term mortality and stress incurred by juvenile school prawns ( Metapenaeus macleayi ) discarded from an estuarine trawler. Some 35% of the prawns died up to 72 h after being caught in a trawl, exposed to air during sorting and separation from the retained catch (as per normal commercial procedures), then discarded into replicate cages. Total mortality was partitioned into that caused by trawling (about 16% of mortalities), and by subsequent sorting and grading (about 19%). Assuming that the majority of the non-penaeid bycatch is excluded from trawls (by the use of bycatch reduction devices), the latter mortalities could be almost eliminated by sorting and separating unwanted school prawns in water-filled compartments. Emersion stress was measured as concentrations of l -lactate in the haemolymph, which were elevated for at least 40 min following capture, but similar among all trawled treatments. l -lactate levels decreased within the first 24 h post-capture, then remained constant over at least the next 48 h, and were greater than baseline levels. The potential benefits associated with subtle changes to handling practices onboard estuarine trawlers are discussed.

Introduction

The mortality of bycatch from penaeid-trawl fisheries throughout the world has attracted considerable attention during the past two decades and has led to the development of various strategies designed to alleviate the negative impacts that such wastage may have on marine populations. Excluding temporal and spatial closures, bycatch management options can be divided into two general categories of input controls that include: (i) specific physical modifications to gear to improve species and/or size selectivity ( Broadhurst, 2000 ); and (ii) changes to operational and/or post-capture handling procedures that reduce the fishing mortality of discards ( Wassenberg and Hill, 1989 ; Gamito and Cabral, 2003 ; Purves et al ., 2003 ).

The first option has been the most commonly adopted. Depending on the species or sizes to be excluded from trawls, improvements to selection have been achieved via simple alterations to the sizes and configurations of mesh used (e.g. Broadhurst et al ., 2004a ), and/or the development of physical bycatch reduction devices (BRDs; see Broadhurst, 2000 , for a review). In New South Wales (NSW), Australia, inherent variation among the characteristics of different estuarine penaeid-trawl fisheries has resulted in a range of modifications to conventional gears, including BRDs such as the Nordmøre-grid ( Broadhurst and Kennelly, 1996 ) and Hawkesbury River square-mesh panel ( Broadhurst and Kennelly, 1995 ), both of which are designed to reduce the catches of key non-target species. More recently, in response to concerns over poor size selectivity and mortality of large quantities of juvenile penaeids (<15 mm carapace length, CL), and especially school prawns ( Metapenaeus macleayi ), new codends made entirely of square mesh (between 25-mm and 29-mm mesh hung on the bar) have been tested and adopted in some fisheries ( Broadhurst et al ., 2004a ; Macbeth et al ., 2004 ).

Research in NSW's estuarine trawl fisheries has demonstrated that square-mesh codends can allow up to some 60% of small, unwanted school prawns to escape during fishing ( Broadhurst et al ., 2004a ; Macbeth et al ., 2004 ). Further, as part of simulated trawling experiments in the laboratory, Broadhurst et al . (2002) observed that escaping school prawns incurred minimal short-term mortality (<11%). While these results have obvious benefits for stocks of school prawns, some discarding of unwanted juveniles persists, because the size selectivity of the new square-mesh codends is not 100% efficient, and catches often include various size cohorts. The vulnerability of these discarded prawns warrants quantification of their post-release fate. The use of simple ancillary management options was also investigated as a means to improve survival.

The short-term mortality of organisms discarded or released after capture is often a consequence of several cumulative sublethal disturbances caused by a range of inter-related factors, which include, but are not limited to, the amount of time spent in the fishing gear, the quantity or type of bycatch ( Kaiser and Spencer, 1995 ), duration of exposure to air ( Hill and Wassenberg, 1990 ; Lancaster and Frid, 2002 ; Gamito and Cabral, 2003 ), and onboard handling techniques ( Lancaster and Frid, 2002 ; Gamito and Cabral, 2003 ). In NSW's estuarine trawl fisheries, school prawns can spend up to 1 h in the codend before being landed on the vessel, then exposed to air for 10 min during separation from non-target species, and then an additional 5–10 min during separation into retained and discarded categories (using a device called a “riddler”; Figure 1a ).

Figure 1

Diagrammatic representation of (a) school prawns being riddled from a sorting compartment, and (b) the design of the cylindrical cages used to hold school prawns.

Depending on their severity, the treatments listed above could contribute towards immediate mortality of discarded prawns or, alternatively, increase their susceptibility to factors such as disease and/or reduced ability to evade predation ( Ryer, 2002 ). While quantifying the short-term fate of discarded bycatch is fairly straightforward and has received considerable attention ( Wassenberg and Hill, 1989 ; Hill and Wassenberg, 1990 , 2000 ; Evans et al ., 1994 ; Lancaster and Frid, 2002 ; Castro et al ., 2003 ; Gamito and Cabral, 2003 ), obtaining information on sublethal disruptions that might negatively influence longer-term survival is more difficult and requires examination of physiological responses to stress. For crustaceans, an appropriate method of determining the severity of anaerobic stress is to measure the changes in the levels of lactic acid in their haemolymph, then to compare these with base or resting levels ( Taylor and Spicer, 1987 ; Spicer et al ., 1990 ). Appropriate temporal monitoring of l -lactate can provide an accurate indication of the physiological reaction and recovery of prawns, and therefore of the severity of anaerobic stress. Our aims in this study were to quantify the mortality and stress of juvenile school prawns discarded from estuarine trawlers using (i) current conventional practices, and (ii) simple modified post-capture handling procedures.

Material and methods

Equipment used and collection of control prawns

The experiment was carried out during October 2004 in Lake Woolooweyah (29°26′S 153°22′E), a shallow, coastal lagoon (<4 m deep) in which prawns are trawled commercially (see Liggins and Kennelly, 1996 , for details). The equipment used included a commercial prawn trawler rigged with conventional, paired trawls (see Broadhurst et al ., 2004a , for details), two 3000-l fibreglass tanks, 15 plastic 30-l transport tanks, and 72 cylindrical cages (0.27 m in diameter × 1.2 m in length) made from black plastic 6-mm polyvinyl chloride (PVC) mesh and solid PVC bases, filled with sediment ( Figure 1b ). The 3000-l tanks were located on the bank of the lake, and were supplied with continuously flowing, aerated water at a rate of 45 l min ⁻¹ . The 15 plastic transport tanks were used on the trawler, while each of the 72 cylindrical cages was secured in a tributary leading into the lake, approximately 2 m apart, at depths between 1 and 1.5 m.

Two weeks before the experiment started, some 1000 school prawns were caught in the lake using the trawls rigged with codends made from 10-mm knotless mesh and towed for <10 min. At the end of each tow, live school prawns were emptied from the codends directly into holding tanks onboard the trawler and transported to the two 3000-l fibreglass tanks, where they were left to acclimatize and be monitored daily for mortality. These school prawns were used as controls in the experiment.

Experimental protocol

The experiment examined five treatment groups associated with school prawns being trawled and released, and one control group ( Figure 2a, b , respectively; Table 1 ). On the first day of the experiment the trawler completed four consecutive 1-h tows between 07:15 and 13:30 local time, according to normal commercial operations. At the end of the first tow, the codends were lifted onboard and their entire contents equally distributed onto a tray partitioned into four separate compartments (each 52 × 32 × 10 cm; Figure 1a ). Two compartments were dry (as per normal commercial practice), but the remaining two compartments were filled with water to a depth of about 5 cm. Immediately after the catch was distributed among the compartments, 30 small (<15 mm CL) school prawns were randomly selected from the dry compartments and placed into three of the 30-l plastic tanks (i.e. 10 prawns per tank). This first treatment (termed “unsorted”; Table 1 ) was designed to provide information on the fate of school prawns after being trawled, landed on the vessel, then released immediately. The bycatch of non-target species was then concurrently removed and discarded (within 5 min), before all school prawns in each compartment were separated into retained and discarded categories using a riddler (which took 5–10 min; Figure 1a ). For the second and third treatments (termed “dry–dry” and “dry–wet”, respectively; Table 1 ), school prawns from the dry compartments were passed through the riddler and into containers (55 × 35 × 35 cm deep; Figure 1a ) that were empty (dry–dry treatment) or filled with water (dry–wet treatment; Table 1 ). In the fourth and fifth treatments (termed “wet–dry” and “wet–wet”, respectively), the school prawns from the two water-filled compartments were riddled into empty and water-filled containers as above ( Table 1 ). The order in which each of the treatment compartments was riddled was determined randomly.

Figure 2

Diagrammatic summary of the experimental design depicting (a) the five treatment groups, and (b) the one control group, for which a total of 120 prawns were collected in each case; (c) 12 replicate cylindrical cages (in total) into which 10 prawns were placed for each of the treatment groups; (d) each of the four sampling times to which three of the 12 cages were allocated; and (e) two live prawns taken from each cage for l -lactate measurements (i.e. one measurement taken for each prawn) upon sampling. Note: l -lactate measurements were only made for treatment groups and sampling times marked *.

Table 1

Summary of the five treatment groups applied to school prawns.

	Treatment
Description	Unsorted	Dry–dry	Dry–wet	Wet–dry	Wet–wet
Trawled for 1 h	Yes	Yes	Yes	Yes	Yes
Discarded
Immediately	Yes	No	No	No	No
After riddling	No	Yes	Yes	Yes	Yes
Sorted in
Dry compartment	No	Yes	Yes	No	No
Water-filled compartment	No	No	No	Yes	Yes
Riddled into
Dry container	No	Yes	No	Yes	No
Water-filled container	No	No	Yes	No	Yes

Immediately after the various discarded prawns were sorted according to their treatments, 30 were randomly selected (from each treatment) and placed into three of the 30-l plastic tanks (i.e. 10 school prawns per tank). After the onboard sorting procedure was completed, the 15 30-l tanks were immediately transported and emptied into 15 of the cylindrical cages (i.e. three cages for each treatment). The time required to transport the prawns was about 40 min. The above process was repeated for an additional three tows, providing a total of 60 cylindrical cages, each containing a total of 10 school prawns (i.e. 12 replicate cages for each of the five treatment groups; Figure 2c ).

Immediately after the last of the treatment prawns were placed into their cages, four prawns were randomly selected from the 3000-l fibreglass holding tank, secured in aluminium satchels, and kept frozen in liquid nitrogen for lactate analysis ( Broadhurst et al ., 2002 ; see below). The water level in one of the two 3000-l holding tanks was then lowered and ice added to reduce the temperature to 15°C, to anaesthetize the school prawns. A total of 120 prawns was randomly selected and placed in groups of 10 into 12 of the 30-l plastic tanks ( Figure 2b ). After 40 min (the same time between the catch and release into the cages for treatment prawns), the 30-l plastic tanks were transported to the cylindrical cages and emptied into the remaining 12 cylindrical cages ( Figure 2c ). These cages were used as controls ( Table 1 ). A Horiba U10 water quality meter was used to record the temperature (°C) and dissolved oxygen (mg l ⁻¹ ) levels in the lake, in each container housing school prawns, and in the 3000-l holding tank prior to removing the control prawns. One measurement was taken at each of these stages during the first tow only.

Three of the 12 cylindrical cages containing school prawns in the control, and each of the five treatment groups, were destructively sampled at each of four times: immediately after school prawns were placed in the cages ( T₀ ), and at exactly 24 ( T₂₄ ), 48 ( T₄₈ ), and 72 ( T₇₂ ) h post- T₀ ( Figure 2d ). Sampling to a maximum of 72 h was chosen because Wassenberg and Hill (1993) reported negligible mortality after three days during a laboratory experiment addressing similar issues relating to non-penaeid trawl bycatch in Moreton Bay, Australia. At each sampling time, the relevant cylindrical cages were removed from the lake. For each cage, the numbers of dead school prawns were recorded, and two of the live prawns were selected randomly and secured in aluminium satchels prior to freezing in liquid nitrogen ( Figure 2e ). Owing to the costs involved, analyses of l -lactate (μmol g ⁻¹ ) were done for only the control and three of the treatment groups (unsorted, dry–dry, wet–wet) at T₀ , T₂₄ , and T₇₂ ( Figure 2e ).

Statistical analyses

A χ² goodness-of-fit test was used to test for no bias in the sexes of surviving school prawns used in the experiment. Two-sample Kolmogorov–Smirnov tests were used to compare the size frequencies of prawns between treatments.

Asymmetrical analysis of variance (ANOVA) was used to test the hypothesis of no differences in the rates of mortality (expressed as a proportion for each cylindrical cage) or levels of l -lactate (μmol g ⁻¹ ) between the various treatments and the control. Data were transformed (sin ⁻¹ (√ x ) and ln( x + 1), respectively), tested for heteroscedasticity, then analysed using the appropriate two- and three-factor ANOVAs, respectively. Both models included the handling of prawns and sampling time as orthogonal and fixed factors ( n = 3 cages for the former analysis). Handling of prawns was further partitioned into treatment vs. control, and among treatments terms. The third factor used in the model to analyse l -lactate levels was termed cages, and was nested in the interaction between handling of prawns and sampling time ( n = 2 prawns). Where the interaction terms were non-significant at p > 0.25, pooling with the residual was done to increase the power of the test for the main effects and the partitioned terms ( Winer et al ., 1991 ).

Where required, significant f -ratios were investigated using Student–Newman–Keuls (SNK) multiple comparisons ( Underwood, 1981 ). Standard logistic regression was used as a substitute for SNK tests to examine mean rates of mortality, where the latter failed to elucidate any definitive order among means for which a significant difference was detected by ANOVA. This technique required fitting logistic regression models to a data set in which each caged prawn represented one data point in the analyses. One of two outcomes was possible for each prawn (binary response variable, live or dead), whereas handling of prawns was assigned as the explanatory variable (see Results; SAS Institute, 2003 ). Each treatment in turn was assigned as the comparison (i.e. −1) treatment to ultimately produce a matrix of odds-ratio estimates (ORE), modelling the probability of a prawn dying ( SAS Institute, 2003 ). The closer the ORE to a value of 1, the less different the odds of a prawn dying are between the two treatments. Statistical significance of differences between treatments was inferred from 95% Wald confidence limits ( SAS Institute, 2003 ).

Results

Experimental school prawns and environmental data

Approximately 1000 school prawns were caught and released into the two 3000-l fibreglass tanks. Of these, about 30 died in each tank during the first 24 h, and between one and five died each day over the next three days. During the nine days immediately prior to the start of the experiment, about 35 prawns died in one of the tanks, but none died in the other tank. Consequently, only prawns from the latter tank were used in the experiment.

During the first day of the experiment, 10 school prawns were collected and released into the majority (89%) of replicate cylindrical cages. Because some school prawns escaped during their transfer and/or there was researcher error during their collection, either 8, 9, or 11 prawns were released into the remaining cages, except for one replicate for the wet–dry treatment (sampled at T₇₂ ), into which no prawns were released.

The surviving prawns used in the experiment ranged in size between 7.8 mm and 17.1 mm CL, and their sex ratio was not significantly biased towards either sex (male:female ratio of 1:1.12; n = 429, χ² = 1.54, p > 0.05). Kolmogorov–Smirnov tests failed to detect significant differences in the size frequency distributions of surviving school prawns among the control and various treatment groups ( p > 0.05). Mean CLs ranged between 12.96 (±0.15) mm for the control group and 13.52 (±0.15) mm for the unsorted treatment group.

The dissolved oxygen recorded during the first tow was 7.65 mg l ⁻¹ in the lake; 4.90 mg l ⁻¹ in the wet sorting compartments immediately before riddling; 5.25 mg l ⁻¹ in the containers immediately after school prawns were riddled; 5.30 mg l ⁻¹ in the 30-l transport tanks immediately prior to school prawns being released into the cylindrical cages; and 7.80 mg l ⁻¹ in the two 3000-l fibreglass tanks. The water temperature in the lake was 24.4°C. Similar temperatures were recorded in all of the various containers used to house school prawns during the work (24.1–24.4°C).

Mortality and stress of school prawns

ANOVA detected significant f -ratios for the main effect of handling of prawns and the asymmetrical, partitioned terms between treatment and control groups, and among treatment groups ( Figure 3 ; Table 2 ). There was no significant difference in mortality among sampling times because most mortalities caused by being trawled and/or discarded occurred between prawns being caught and first sampled at T₀ (i.e. approximately 40 min; Figure 4 ). Only three school prawns died in the control group during the experiment, providing a mean (±s.e.) percentage mortality rate of 2.6 (±1.4)% per cage, significantly lower than the 28.1 (±2.3)% from all treatments combined ( Figure 3 ; Table 2 ). SNK tests could not detect a definitive order for the proportions of school prawns that died among the five treatment groups (ranging from 17.3 ± 2.8% to 38.0 ± 4.1% for the wet–wet and dry–dry treatments, respectively; p > 0.05; Figure 3 ). OREs from logistic regression analyses demonstrated that the probability of mortality was not significantly different between the unsorted and wet–wet treatments, or between the wet–dry, dry–wet, and dry–dry treatments ( Figure 3 ; Table 2 ). School prawns in the last three treatments were, however, between 2.18 and 2.90 times (and significantly) more likely to die than those in the other treatments ( Figure 3 ; Table 2 ).

Figure 3

Differences in the mean (±s.e.) percentages of school prawns per cage that had died by the final sampling time ( T₇₂ ) for the five treatments (unsorted, wet–wet, wet–dry, dry–wet, and dry–dry) and the control during the experiment (i.e. n = 12; one wet–dry replicate was excluded from the data; see text). Relevant results from the asymmetrical ANOVA (treatment vs. control term) and logistic regression analyses (among treatments) are shown (< or =).

Figure 4

Total proportions of prawns that had died from the control (Co), unsorted and wet–wet treatments combined (Wa), and the treatments involving prolonged exposure to air (i.e. wet–dry, dry–wet, and dry–dry) combined (Ex), immediately after being placed into the cages ( T₀ ), then after 24 ( T₂₄ ), 48 ( T₄₈ ), and 72 ( T₇₂ ) h. Data are combined across three replicate cages. The total numbers of prawns at T₀ for each data point is shown in italics.

Table 2

Summary of mean-square (MS) estimates and f -ratios from asymmetrical ANOVA comparing the proportions of school prawns that survived for five treatment groups (unsorted, wet–wet, wet–dry, dry–wet, and dry–dry) and the control group, sampled at four times ( T₀ , T₂₄ , T₄₈ , and T₇₂ ). n = 3, but one wet–dry/ T₇₂ replicate was excluded from the analysis (see text). All data were sin ⁻¹ (√ x ) transformed. Pld means that the f -ratio for the interaction term was not significant at p = 0.25, so the sums of squares and d.f. were pooled in the residual. ** p < 0.01; * p < 0.05 (top panel). Odds-ratio estimates generated from logistic regression analyses of the five trawled treatments, using each treatment in turn as the comparison (i.e. −1) treatment. * p < 0.05 inferred from 95% Wald confidence limits. na means not applicable (bottom panel).

Source of variation	d.f.	MS estimate	f -Ratio
Handling of prawns ( H )	5	0.522	13.448**
Treatment vs. control	1	1.994	12.932*
Among treatments	4	0.154	3.971**
Sampling time ( S )	3	0.053	1.366
H × S –Pld
Residual (pooled)	62	0.039

	Odds-ratio estimates for treatments
Designated comparison (−1)	Unsorted	Wet–wet	Wet–dry	Dry–wet	Dry–dry
Unsorted	na	0.926	2.331*	2.177*	2.681*
Wet–wet	1.080	na	2.517*	2.351*	2.896*
Wet–dry	0.429*	0.397*	na	0.934	1.150
Dry–wet	0.459*	0.425*	1.071	na	1.232
Dry–dry	0.373*	0.345*	0.869	0.812	na

The four school prawns sampled from the 3000-l plastic tank at the start of the experiment had a mean (±s.e.) l -lactate level of 0.98 (±0.45) μmol g ⁻¹ . ANOVA testing the hypothesis of no differences in the levels of l -lactate among the control, unsorted, wet–wet, and dry–dry groups at T₀ , T₂₄ , and T₇₂ revealed significant main effects for the handling of school prawns and sampling time, and the partitioned term, between treatment and control groups ( Table 3 ). SNK tests of these means revealed that, irrespective of the particular handling method after being trawled or the sampling time, treatment school prawns had a significantly greater mean level of l -lactate (combined mean ± s.e. of 7.75 ± 0.60 μmol g ⁻¹ ) than control school prawns (1.96 ± 0.16 μmol g ⁻¹ ; Figure 5 ). Further, irrespective of the handling of prawns, all individuals sampled at T₀ had a significantly greater mean concentration of l -lactate (9.26 ± 1.17 μmol g ⁻¹ ) than those sampled at T₂₄ and T₇₂ (4.70 ± 0.61 and 4.94 ± 0.61 μmol g ⁻¹ , respectively). No other significant effects were detected ( Table 3 ).

Figure 5

Differences in the mean (±s.e.) l -lactate levels of prawns (μmol g ⁻¹ ) for three treatments (unsorted, wet–wet, and dry–dry) and the control group, sampled immediately after being placed into the cages ( T₀ ), then at 24 ( T₂₄ ) and 72 ( T₇₂ ) h. Data are combined across three cages, each providing two measurements (i.e. n = 6).

Table 3

Asymmetrical ANOVA assessing the effect of three treatments (unsorted, wet–wet, and dry–dry) and the control group on the l -lactate levels of school prawns sampled at three times ( T₀ , T₂₄ , and T₇₂ ). Data are for measurements from two prawns in each of three cages for each Handling × Sampling time (i.e. H × S ) combination ( n = 2). All data were ln( x + 1) transformed. ** p < 0.01; * p < 0.05.

Source of variation	d.f.	MS estimate	f -Ratio
Handling of prawns ( H )	3	4.356	20.004**
Treatment vs. control	1	13.014	484.123**
Among treatments	2	0.027	0.123
Sampling time ( S )	2	18.01	8.270**
H × S	6	0.400	1.839
Cage [ H × S ]	24	0.218	1.784
Residual	36	0.122

Discussion

The experiment revealed mortality of up to 35% (adjusted to account for the mortality of control prawns) for juvenile school prawns discarded from estuarine prawn trawlers according to current conventional practices, most of which was in the first 40 min post-landing. This estimate is more than three times greater than that for prawns escaping through codend meshes during trawling (i.e. <11%; Broadhurst et al ., 2002 ). It is also apparent, however, that simple alterations to the onboard handling of discarded school prawns could considerably reduce mortality. Specifically, the direct mortality resulting from the current practice of sorting and grading into dry compartments and containers could be almost completely eliminated by the use of containers filled with water.

It is clear that a certain proportion of discarded school prawns died as a direct consequence of being caught in the trawl; estimated here at approximately 16% (adjusted for mortality to control prawns). While the exact causes of the deaths are unknown, there are several likely contributing factors. In particular, previous studies have demonstrated that the time spent in the codend and the amount or type of bycatch appears to be negatively correlated with survival. For example, in a Portuguese beam trawl fishery, the mortality of brown shrimp ( Crangon crangon ) immediately discarded without being sorted was virtually halved by reducing tow duration from 30 to 20 min ( Gamito and Cabral, 2003 ). Moreover, Macbeth et al . (2005) noted a relatively poorer condition of penaeids and other bycatch when jellyfish were present in the catches of an estuarine seine, probably because of their stinging nematocysts. The potential for interaction between non-penaeid bycatch and school prawns was probably reduced in the present work because the Nordmøre-grid prevented all organisms >20 mm wide from entering the codend. However, some school prawns may still have made contact with other species (including jellyfish) in the trawl body or as they were separated by the grid, which might have contributed to the mortality.

Other studies have demonstrated that once catches are landed on a vessel, the main factors influencing the survival of discards are exposure to air and/or different temperatures ( Taylor and Spicer, 1987 ; Lancaster and Frid, 2002 ; Gamito and Cabral, 2003 ). Positive correlations between air exposure and/or temperature and the mortality of discarded bycatch have been demonstrated for other trawl fisheries in Australia ( Wassenberg and Hill, 1989 ) and around the world ( Lancaster and Frid, 2002 ; Gamito and Cabral, 2003 ). The results from the present experiment suggest that exposure to air for as little as 5–10 min probably contributed towards an additional 19% mortality to the total discarded prawns (i.e. in addition to the 16% mortality directly caused by their capture). Clearly, a simple preventative measure to avoid such mortality is to keep the catch immersed as much as possible during onboard sorting and grading.

In addition to the observed mortality, the results indicated that surviving school prawns sustained at least some stress as a result of the various treatments. The concentrations of l -lactate in prawns in the 3000-l tanks immediately before the experiment (0.98 ± 0.45 μmol g ⁻¹ ), and in most of the control cages (0.7–3.3 μmol g ⁻¹ ), compared well with the baseline estimates (between 0.5 ± 0.21 and 3.72 ± 0.38 μmol g ⁻¹ ) of Broadhurst et al . (2002 , 2004b , respectively). The elevated level of l -lactate in all landed prawns suggests that the extent of stress and exercise of trawled individuals were similar to those observed by Broadhurst et al . (2002) . Other studies examining the responses of Crangon crangon ( Onnen and Zebe, 1983 ) and Palaemon spp. ( Taylor and Spicer, 1987 ) to anoxia or exercise suggest that further effects attributable to onboard handling practices might have been expected. However, we observed that the l -lactate levels of school prawns were already elevated at landing, and the relatively brief time of onboard handling was clearly not conducive to a significant removal or increase in l -lactate for prawns from the wet–wet and dry–dry treatments, respectively. Given the observed benefit of the wet–wet treatment on the survival of discarded school prawns, the accumulation of l -lactate per se was clearly not the main factor influencing mortality, so other factors should be considered in trying to explain mortality during onboard handling. An obvious difference between prawns in the dry–dry and wet–wet treatments is likely to be the resumption of acid–base regulation in the latter ( Taylor and Spicer, 1988 ), which may have helped to contain further physiological damage.

Elevated l -lactate levels across the treatments do not necessarily mean that the prawns were completely exhausted. A further increase or delay in l -lactate removal may actually reflect a recovery process that preferentially restores arginine phosphate levels and energy charge ( Onnen and Zebe, 1983 ; Gäde, 1984 ). Other processes might also have caused the l -lactate levels to plateau. The levels (in whole body terms) were relatively high compared with those reported in the other crustacean exercise studies mentioned above, possibly the consequence of some sort of feedback mechanism that prevented further accumulation of l -lactate. For example, the trawled school prawns may have been relatively inactive on the tray, or if they were active, then l -lactate production was not directly involved. Some level of aerial respiration by newly exposed 1–2 g juveniles in the dry–dry treatment might also have occurred ( Taylor and Spicer, 1988 ).

Recovery from an exercise peak in crustaceans is generally completed in a matter of hours ( Onnen and Zebe, 1983 ; Head and Baldwin, 1986 ), so the higher l -lactate levels of treatment school prawns compared with their controls after 24 and 72 h is of concern; it suggests a potentially chronic effect on discards. Perhaps this, and the similar departures in baseline l -lactate levels seen in previous studies (e.g. Broadhurst et al ., 2002 ), was related to the “residual anaerobic stress” observed in captive lobsters ( Jasus edwardsii ; Speed et al ., 2001 ). In any case, such an apparent protracted physiological recovery may translate to increased susceptibility to disease or even predation ( Ryer, 2002 ). However, the potential for the latter may be limited to the period between when the prawns are discarded and when they reach the seabed, because we observed that at T₀ and each subsequent sampling time the vast majority of the prawns were buried in the sediment at the base of the cages. Despite being stressed, the probability of predation would presumably be similar to other (non-discarded) buried prawns. Future research would benefit from examination of the potential for such deleterious longer-term effects.

While at least 65% of school prawns currently survive the short-term effects of being trawled, sorted, and discarded from trawlers in Lake Woolooweyah according to conventional practices, simple modifications that involve reducing air exposure by sorting in water could significantly improve survival by up to a further 20%. In the absence of information on the natural mortality and the potential for other unaccounted fishing mortality (attributable to subsequent predation or disease), it is difficult to determine the longer-term benefits associated with adopting such changes to operational methods. Nevertheless, the minimal additional effort and financial expense associated with the recommended changes, combined with the potential for a reduction in mortality, warrant their adoption by commercial fishers. Given the potential benefits to prawn stocks to be gained by sorting in water, it is likely that the fishers would voluntarily affect such operational changes. In combination with the impending introduction of square-mesh codends throughout NSW's estuarine prawn-trawl fisheries as a management strategy to reduce the capture of small prawns ( Broadhurst et al ., 2004a ; Macbeth et al ., 2004 ), these simple operational changes would probably further reduce negative impacts on stocks. Further work examining other factors influencing the survival of discarded school prawns, such as the effects of different trawl durations, or the longer-term influence of elevated levels of stress, would be most useful in developing strategies that could potentially minimize the negative impacts.

Funding for the work was provided by NSW Department of Primary Industries and the Fisheries Research and Development Corporation (Grants 2001/031 and 2005/056). Thanks are extended to Damian Young, Ben Black, Dennis Reid, Paul Exley, John Nagle and Don, Barry, Geoff and Deborah Johnson.