WALTERS: ECOLOGY OF HAWAIIAN SERGESTID SHRIMPS 



intermediate tow of 400-600 m; and a deep tow of 

 400-1,200 m, eacli night of the cruise. Daytime 

 trawling investigated possible moon-related 

 changes in the daytime distribution of sergestids 

 and included a 400- to 800-m tow and a 600- to 

 1,000-m tow each day. The actual depths sampled 

 by the trawl deviated somewhat from the protocol, 

 as we used no telemetry on the trawl. The last 

 daytime tow was an all-day affair sampling from 

 1,100 to 1,900 m. The sergestids from this cruise 

 were identified to species and counted, but not 

 sexed or measured. 



Feeding Study: DSB III 



An important problem in any study of feeding 

 in mid-water animals is the effect of the sampling 

 gear on feeding behavior. A mid-water trawl 

 concentrates animals in the cod end to unnaturally 

 high densities. Often the trawl lumps together 

 animals from different depth zones. A predator 

 feeding on the contents of the cod end is likely to 

 eat prey it would not normally take in the natural 

 state, either because predator and prey do not 

 occur at the same depth or because the prey can 

 normally escape the predator. Examination of 

 sergestid stomach contents from the Teuthis 

 series suggested that many shrimp had been 

 feeding in the trawl. A modification of the trawl 

 became necessary to get reliable feeding data. 



The DSB III cruise of 2-3 February 1973 was 

 designed to investigate the feeding behavior of 

 mid-water animals. The MT was modified by 

 tieing off the cod end ahead of the plankton net, 

 allowing zooplankton to escape through the 

 meshes. The trawl mouth was tied open. Daytime 

 and nighttime oblique and horizontal tows were 

 taken, the main objective being to obtain as large 

 and varied a collection of mid-water animals as 

 possible without much concern for their depth of 

 capture (Table 1). The samples were preserved in 

 5% Formalin seawater and returned to the labo- 

 ratory, where the sergestids were sorted out and 

 their stomach contents identified. 



Using the MT in this fashion produced one 

 unexpected bonus. In addition to flushing out 

 prey-sized zooplankton, the water current forced 

 the catch and the inner lining of the net through 

 the coarse outer net in pockets. Within each pocket 

 the animals were firmly held by the force of the 

 water, preventing movement and feeding. Future 

 feeding studies might profit from deliberately 

 designing this effect into the sampling gear. 



Analysis of Vertical Distribution Data: 

 The Contamination Problem 



Most previous studies of vertical distribution 

 (e.g., Foxton 1970, T. A. Clarke 1973, Donaldson 

 1975) have assumed that all the animals captured 

 in a horizontal tow were taken at a single depth. 

 While such an assumption simplifies the presen- 

 tation and interpretation of the data, it can 

 produce a misleading picture of the vertical struc- 

 ture of the mid-water community if the tows 

 actually fish over a substantial depth range. Open 

 trawls like the IKMT are the most susceptible to 

 contamination of the catch by animals from other 

 depths, since they fish during setting and re- 

 trieval. In this case, contamination usually takes 

 the form of shallow-living animals appearing to 

 have been captured below their normal depth. 

 Rapid setting and retrieval can minimize but not 

 eliminate the problem (T.A. Clarke 1973). Foxton 

 (1970) and Donaldson (1975) have shown that 

 animals from other depths can contaminate IKMT 

 samples even when the trawl is fitted with open- 

 ing-closing cod end buckets. Some animals become 

 temporarily entangled in the net early in the tow. 

 When they break free later on, the trawl may be 

 fishing at a different depth, resulting in a sample 

 that mixes shallow and deep animals in an un- 

 known proportion. 



Even an opening-closing trawl like the MT can 

 give misleading results if it is allowed to wander 

 vertically while open. In such a case, assigning the 

 entire catch to the modal depth broadens out the 

 apparent vertical range in both directions. Our 

 experience has shown that towing the MT deeper 

 than 200 m results in substantial vertical wander- 

 ing unless its depth is constantly monitored and 

 adjusted. Since a working telemeter was available 

 only during the latter part of our program, most of 

 our "horizontal" tows actually have a vertical 

 range of 50-100 m. The problem increases with 

 depth; tows below 800 m commonly wander 200 m 

 or more. Assigning the catch to a modal depth 

 would produce a misleading vertical distribution 

 pattern. 



The vertical distribution diagrams presented in 

 this paper allow for vertical wandering of the 

 trawl and for unequal sampling time with depth. 

 Only horizontal tows are considered. The water 

 column is divided into 25-m zones, and the amount 

 of time each tow spent in each zone is determined 

 from the various depth zones in proportion to the 

 time towed in each zone. Let q be the number of 



803 



