Broadhurst et al.: Modified sorting technique to mitigate the collateral mortality of Metapenaeus macleayi 
291 
River (approximately 14 minutes) was 
greater than that in the Hunter River 
(approximately 6 minutes; Table 1). 
The first mixed-effects model that 
was applied to all of the data for the 
trawled-and-discarded school prawns 
in both experiments included the fixed 
factors of sorting method (conventional 
vs. water tray), delay in sorting (2 min- 
utes vs. 15 minutes), and their inter- 
action, sample time, and the weight of 
total catch, and salinity at the bottom 
of the river during fishing. The sec- 
ond and third models were applied to 
the water-tray and conventional-tray 
deployments. Both models included 
sorting method (conventional vs. wa- 
ter tray), delay in sorting (2 minutes 
vs. 15 minutes), and their interaction, 
sampling time, and the weight of to- 
tal catch. Water tray salinity and dis- 
solved oxygen were also included in 
the second model, and air temperature 
and cloud cover were used in the third 
model. The fourth model was applied 
to the control data with sampling time 
as the only the fixed effect. The stage 
of molt was not included in any of 
the models for the Hunter River data 
because it was not possible to quan- 
tify the condition of all deceased in- 
dividuals. In any case, only 13.0% of 
all school prawns sampled at T 0 had 
soft exoskeletons. Similarly, it was not 
possible to measure the CL of all dead 
school prawns at T 24 , T 72 , and T 120 , 
and therefore size could not be consid- 
ered in the analyses. 
Clarence River experiment 
A total of 4800 school prawns were 
caged during the experiment, of which 
240 individuals were in each of the con- 
trol and four treatment groups at each of the four sam- 
pling times. Seventy school prawns escaped (as a result 
of faulty lids on their cages) before being sampled; they 
escaped from one cage for the 15-min conventional-tray 
treatment at each of T 24 and T 72 , two cages for the 15- 
min conventional-tray treatment at T 120 , one cage for 
the 2-min water-tray treatment at T 24 , and two cages of 
controls at T 72 . Up to 12.0% of the control school prawns 
were dead at each sampling time (attributed to handling 
during transfer, being caged, or natural causes, Fig. 2A), 
providing adjusted temporal mortalities of up to 41.4% 
and 18.2% for the 15-min conventional and water-tray 
treatments, respectively (Table 2). 
Irrespective of the four mixed-effects models fitted to 
the various data sets, or the treatment of school prawns, 
there was a strong significant impact of sampling time 
on mortality that was largely attributable to the death 
of most individuals during their first 24 hours of cag- 
ing, after which the rate of attrition stabilized (Tables 
2 and 3 , Fig. 2A , PcO.Ol). All three analyses of the 
trawled-and-discarded school prawns also showed a sig- 
nificant impact of the delay in the start of sorting, with 
an overall predicted (and unadjusted) mean ±standard 
error (SE) of 31.6 ±2.4% after 15 min, compared to only 
21.2 ±1.9% after 2 min (Table 2, P<0.05). 
In addition to the main effects of delay in sorting 
and sampling time, sorting method had a significant 
impact across all deployments, and there were consis- 
tently greater mortalities to those individuals discarded 
after sorting in the conventional tray (predicted unad- 
justed mean of 33.5 ±2.4%) than in the water tray (19.3 
±1.8%), and especially after a 15-min delay (predicted 
