152 



steelhead), tiaan would be flio case without spill, given otherwise comparable conditions, 

 bi my risk analysis in Question s* 3, I estimate that spill is, on balance, hurting survival. 



The benefits of spill will be a difficult problem to research and resolve, mainly 

 because the calculated incremental survival benefit of spill (assuming uo mortality from 

 gas) is a small percentage of the total run (ca. 1-3 %). The basic experimental design to 

 test this has been worked out over man>' years and generally requires about 10 years to 

 complete from beginning to end. Perhaps for this reason. NMFS has not de\'eloped a 

 similar experimental design to evaluate spill, despite millions of smolts available at federal 

 hatcheries in the Snake and Columbia Rivers. As a general rule, these evaluations require 

 smolt releases from at least four consecutive brood years, each having marked 

 experimental and control groups, each group with two or more replicates, and each 

 replicate with about 100,000 smolts. This design is necessary to estimate variation within 

 and between years, and thus form the confidence intervals about tiie mean for the 

 experimental and coiitrol group. Since variation expands the confidence interval, and 

 since variation is usually high, differences of less than 5 % survival are not likely to be 

 statistically significant. This alone may preclude evaluation of spill, 



.^s an altemati^'e to the nece8.saril>' long and difficult experiment, some entities 

 have addressed the spiil/gns evaluation problem using laboratory data and computer 

 modeling of various schemata. Many of these analyses are filled witli good intentions, but 

 bad assumptions. For example, the Fishen' Agencies make the unwarranted assumption 

 ih^X all fish will detect and avoid gas supersaturatioii. This assumption must be rejected 

 because several reports document that supersaturation has killed large numbers of wild 

 lish in areas where the fish could tiave sounded to e.sc : ipe the excess gas. If 

 supersaturation avoidance exists in Nature, it is cenaini\' unreliable and untrustwortln', 



Unfortunately, the resulting and often bitter debate-? over the validity of models, 

 assumptions, appi-opriateness of data, and accuracy of results simply demonstrates that no 

 consensus exists in the scientific community on the potential^j^enefits of spill, except as 

 dictated by agency policy. \ 



The NMFS policy on spill may represent a well intended preference, but it is not 

 science. Tlie NMFS spill policy puts the entire river at risk, and itb monitoring and 

 evaluation is inadequate to lest whether spill will benefit or depress imolt survival, 

 NMFS apparently hasnt developed a research hypothesis for testing, criteria for judging 

 the resulting data, or a coherent step down plan with protocols for acquiring tlie critical 

 researcli data. The result is a hodge-podge of e.\panded ongoing efforts cy various 

 agencies who avoid peer review and work mostly in camera (secret). The physical gas 

 monitoring is pioviding a glut of expensive but unreliable information which does not 

 appear to be connected to administrati\'e mechanisms for triggering lower gas levels. 

 NMFS's biological monitoring of gas bubble disease is adequate to detect a serious fish 

 kill, but it is inadequate to reveal a low or indirect moitalitN', and tlie investigative 

 approaches and methodology skew dovmward the apparent incidence of GBD. 



