I have approximately 14-15 years of experience doing research on Columbia River salmon. The first half of that period was devoted primarily to socio-economic aspects of salmon enhancement and recovery actions. The last 6-7 years of work has focused on quantitative analysis of the biological effects of measures to aid Endangered Species Act (ESA) listed stocks. These efforts have been funded largely by the Bonneville Power Administration (BPA), and to a lesser degree by the Corps of Engineers. However, the views expressed here are strictly my own, and should not be interpreted as representing those of BPA or any other organization.
My remarks focus on three broad areas:
1. The Cumulative Risk Initiative (CRI) life-cycle analysis that forms the basis for the population status analysis in the NMFS September 2000 Draft Biological Opinion (Bi-Op);
2. The Plan for Analyzing and Testing Hypotheses (PATH) results, how they compare to recent data, and how they are used in Bi-Op;
3. Uncertainty in the biological effects of the Bi-Op "Reasonable and Prudent Alternatives" (RPA's) and in the possible effects of drawdown.
Obviously, this covers a lot of territory, and my comments will only touch on the highlights of each topic.
CRI LIFE-CYCLE APPROACH
At the risk of over-simplifying a complex modeling exercise, one can characterize the CRI approach as assuming that the future will be a straight-line projection of the past, with a great deal of random noise around that projection. It uses estimates of salmon population abundance over time -- generally derived from dam counts or redd (nest) counts of fish returning to spawn -- and extrapolates trends in past estimates into the future. Unlike PATH efforts, each population is assumed to be independent of the others, except in sensitivity analyses (the latter are not used in the Bi-Op itself). Because salmon abundance varies greatly from year to year, the CRI approach, in common with other attempts to predict future fish numbers, is not very precise. While the "average" trend for a population may be upward or downward, one cannot place very much confidence in any particular value of the trend estimate. However, it is clear that populations of most ESA-listed stocks have shown marked declines over time, regardless of the model used to project those trends into the future. The downward trend, after all, is the rationale behind listing the stocks under the ESA.
As others have noted in previous hearings, the CRI model development and application was not nearly so strongly collaborative as PATH. The approach by NMFS was to develop models, have public workshops to review them, and modify the models in response to the comments received. In contrast, PATH held far more meetings and workshops, with participation in both being restricted largely to the scientists who were actively working on the analyses.
However, given NMFS' decision to analyze 12 Evolutionarily Significant Units (ESUs) in just over 12 months, the level of collaboration in PATH would have been impossible within that timeframe. PATH required almost 5 years to do a thorough job on two ESU's -- Snake River spring/summer chinook and Snake River fall chinook -- and a more perfunctory analysis of Snake River steelhead. In addition (see section on PATH, below), in light of recent information, the PATH conclusion regarding drawdown of the lower Snake projects might in the end have been similar to the Bi-Op: do what's feasible to improve anadromous fish passage at the existing hydrosystem, make improvements in other phases of the life cycle, and defer a decision on drawdown.
That having been said in defense of NMFS, the CRI method for population projections clearly has some problems. For anyone trying to follow their analysis from the outside, the many changes over the past year make it very difficult to be certain what version of the model is being used in any given version of the Bi-Op. For example, substantial changes in extinction estimates occurred between the July release of the Bi-Op and the current (September) version. In addition, the so-called "lambda criteria" in the draft Bi-Op (section 9.2.2.1) -- that the population growth rate must be at least 10% per year, or consultation will be re-initiated -- may lead one to believe that growth rates can be estimated very accurately. In fact, as noted above, the growth rates are very imprecise and noisy, because population abundance varies widely from year to year. This makes decision criteria based on growth rates extremely problematic. In fact, for some stocks it appears that even if populations reach recovery levels -- several hundred to several thousand spawners -- within a decade, the 10% growth rate may still not be met.
Finally, for wild stocks that have relatively large numbers of hatchery-origin spawners, the CRI population projections are very sensitive to "hatchery effectiveness." Again, to risk over-simplification, hatchery effectiveness concerns whether or not hatchery-born fish spawning in the wild are as effective as their wild-born cousins at producing viable offspring -- fish that will eventually return in the future and spawn. The Bi-Op rightly points to the need to obtain empirical estimates of this for different stocks -- at present, the numbers used as a sensitivity are based on professional judgment -- but is silent on whether or not deliberate supplementation with hatchery fish will continue or not. This question needs to be resolved soon, since many stocks are supplemented heavily at present, and may in fact be sustained largely by hatchery fish.
As with many other uncertainties, one may well wonder why, if it is so important, it has not been the subject of more research. The answer, I think, lies in the ESA focus on wild stocks. NMFS has interpreted this to mean that only wild-born fish "count" when it comes to achieving recovery goals -- that supplementation with hatchery fish to sustain a run cannot be part of a long-term management strategy. Therefore, hatchery supplementation can only be used as a safety net to sustain runs over the short term, though the exact definition of "short term" is an open question.
This in turn means that the question of how effective hatchery fish are is very important for future management actions. If effectiveness is very low, then hatchery fish are producing very few progeny, and supplementation is likely a waste of resources. If it is high, and the hatchery-origin fish are helping sustain the runs. The question then becomes whether or not this situation is desirable from a scientific and/or policy perspective.
PATH AND ITS USE IN THE Bi-Op
I was closely involved in PATH. While I disagree with many of its conclusions, I feel that I am qualified to make some statements as to how those conclusions stand up to recently available information.
PATH participants analyzed a host of uncertainties, ranging from minor ones having little effect on the models' output to some that turned out to be very important. Among the latter, three stand out as being very influential:
1. Survival rates for juveniles migrating in-river;
2. "D", i.e., survival of transported juvenile fish after release below Bonneville Dam relative to survival of in-river migrants;
3. "Extra" mortality, i.e., mortality not explained by simple life-cycle models or by the effect of downriver passage of juveniles through the hydrosystem.
While the PATH process is, of course, now defunct, these are still relevant, because they are used in the Bi-Op to estimate the effects of drawdown on Snake and Upper Columbia River stocks. In this section, my remarks apply primarily to fish that migrate as juveniles in the spring -- spring chinook, summer chinook and steelhead -- not to fall chinook. In particular, I am concerned with how PATH conclusions stack up when compared to recently available information.
First, NMFS' recent estimates of in-river survival rates for Snake River spring/summer chinook are substantially higher than those predicted by the passage models (CRISP and FLUSH) used in PATH. In particular, they are 2-10 times higher than those predicted by FLUSH, and somewhat higher than CRISP predictions. Therefore, both passage models to some degree overstated the direct effects of the hydrosystem on Snake stocks, FLUSH by a very substantial margin.
Second, NMFS estimates of "D", while very noisy and imprecise, are higher than those used by either passage model in PATH -- somewhat higher than CRISP, and much higher than those used by FLUSH. The implication is that transported fish survive at a higher rate overall -- from Lower Granite dam as juveniles back to Lower Granite as adults -- than do fish that migrate inriver.
Clearly, in these two areas, PATH predictions are at odds with recent observations. When combined with very high returns of spring chinook this year, they may also be at odds with various PATH hypotheses about extra mortality. In particular, high numbers of returning adults and jacks (immature fish that spend only 1 year in the ocean), combined with other indicators, suggests that an ocean regime shift may have occurred. I say "may" because such shifts have, in the past, lasted for 20 years or more, and a few years of high returns do not a 20-year shift make. If such a shift has occurred, however, it casts serious doubt over the other PATH extra mortality hypotheses -- that it is caused by the existence of the Lower Snake dams, or that it is "here to stay" due to diseases transmitted by hatchery fish or some other unknown cause.
All of this casts considerable doubt on PATH's conclusion that dam breaching is clearly the best alternative to recover Snake River stocks. In addition, it has some important implications for the way PATH results are used in the current Bi-Op. Basically, the Bi-Op uses two values for extra mortality when analyzing the effects of drawdown.One version assumes that extra mortality is zero, the other uses an average of the values derived in PATH, and applies this to both Snake and mid/upper Columbia stocks. If the argument outline] above is correct, then the Bi-Op use of PATH results is clearly wrong. This matters because the projected effects of drawdown manifest themselves primarily via a reduction in extra mortality. Therefore, the value used for this is crucial to predicting the effects of dam removal on fish survival.
UNCERTAINTY IN THE EFFECTS OF Bi-Op ACTIONS
The alert reader will have noticed the prevalence of terms such as "imprecise," "noisy," and "uncertain" in much of the text above. In some cases this is simply the nature of the beast. Reasonably reliable salmon abundance data for the Columbia dates back to the closing of Bonneville Dam in 1938. It shows very high variability from year to year, with runs often increasing or decreasing by a factor of 10 over the course of a few years. Reliable life-stage survival rates (at least for Snake River fish) extend back almost a decade, when PIT tag technology was first used, and these too show a high degree of variability. No statistical model or experimental design will make this variability vanish. The best one can hope for is that carefully designed studies will account for it properly. Under some circumstances, clever designs can make some of the variation "cancel out" by controlling for it, while assessing whether management actions work as planned.
The high variability matters when one is trying to assess the effects of management actions. In the next few paragraphs, I discuss two RPA's mandated in the draft Bi-Op, and an uncertainty that is critical to the projected success of the Bi-Op in avoiding jeopardy. I look first at the actions, flow augmentation and off-site mitigation, and then at the potential effects of drawdown.
Flow Augmentation
Recent research by NMFS has found almost no relationship between spring flow and reservoir survival for spring/summer chinook and steelhead juveniles in the Snake. Although high flows are associated with faster downstream migration through the Snake and Columbia reservoirs, this apparently does not lead to increased survival. Given the lack of evidence that higher spring flow leads to higher survivals within the hydropower system, one can fall back on the hypothesis that increased flow may lead to higher survival in the Columbia estuary or in the Columbia River "plume" just off-shore, where fresh water from the river mixes with salt water. Unfortunately, all one can do at present is make guesses about this, because no direct estimates of estuary/early ocean survival are available. It may be possible to obtain estimates of this using so-called sonic tags, where "fingers" are placed in migrating juveniles, and the signals are picked up by an array of floating microphones placed at intervals in the estuary and just off-shore. Personally, I suspect that it will turn out that estuary survival does not vary much with flow for these stocks, but the only way to find out is to do the studies.
Fall chinook which migrate downstream as juveniles in the summer, present a very different picture. NMFS research shows a fairly strong, positive relationship between flow and survival. However, this is confounded with temperature and turbidity. In addition, it is confounded with the date that study fish were released into the river: fish leaving early do better than fish leaving later in the season, and flow decreases over the migration season. Therefore, separating the effects of flow, time of release, etc. using data for years past is impossible. Deliberate manipulation of flows, accompanied by intensive field studies of tagged fish, is the only way I know of to resolve this uncertainty.
Off-site Mitigation
For many stocks, off-site mitigation is a very important part of the Bi-Op's reasonable and prudent alternatives (RPA's). This is because relatively modest changes in survival rates are expected from additional hydrosystem actions. In combination with large increases needed to avoid jeopardy for some stocks, the result is that much of the increase must come from freshwater spawning/rearing improvements. The Bi-Op calls for many actions -- screening diversions, reducing subbasin irrigation withdrawals, and the like. Actually measuring the effects of these off-site actions on survival rates will be a complex undertaking.
Previous work I have conducted shows that there are large variations in survival rates from parr (immature fish tagged in their natal streams) to smelts arriving at Lower Granite Dam. More specifically, survival rates vary both across years and across rearing areas with different types of land use. The variation across time and space seems sensible: parr survive at higher rates in cool, moist years than in dry ones, and fish in wilderness areas have higher survival than fish rearing in agricultural areas. However, the variability over time and space is high. Even if the effects of habitat modification are substantial, detecting survival increases will require carefully designed studies that need to start very soon to produce results within the Bi-Op's 5- to 8-year time frames.
Potential Effects of Drawdown
For Snake River stocks, breaching of the four Lower Snake dams, and perhaps McNary and John Day, is used as the "gold standard" in the Bi-Op: if breaching is undertaken, the action agencies would then have done all they could to reduce the effects of the hydropower system on Snake River stocks. However, a couple of important uncertainties underlie this conclusion. First, recent work by NMFS researchers concludes that for Snake spring chinook and steelhead, survival rates for juveniles migrating through the Snake and Columbia (from McNary to Bonneville Dam) is about the same now as it was before the Lower Snake projects were built. Therefore, it follows that any benefit the fish derive from breaching would come about not because of acute, short-term effects but because of the elusive "extra mortality" noted above. As previously noted, it is my opinion that whatever extra mortality there may have been since the 1970's, it may well vanish if an ocean regime shift results in higher ocean survival. This leads me to be skeptical of the projected biological benefits that may results from breaching. Most scientists, including me, believe that the fish would be better off without dams, but recent information suggests that the benefits are probably quite modest.
WHERE TO GO FROM HERE
While the model wars -- a central feature of PATH -- have abated to some degree, they still continue on other fronts. For example, the Bi-Op has recently been criticized for not using decision analysis techniques, as was done in PATH. Given the uncertainties outlined above, and many more that I've not gone into here, there are two things of which I am absolutely certain:
1. Computer models and projections, however derived, are no substitute for empirical data.
2. Surprises, both pleasant and otherwise, will be a prominent feature of salmon management and biology for a long time to come.
Given these features of the problem, what should be done in future? First, management agencies in the Columbia have a long history of taking actions largely on faith and best professional judgment. Because they were convinced that the actions would work, they did not monitor the effects to assess their success or failure. This will have to change in the future if we are to retain any credibility with the public and with elected officials. I have identified two actions in the Bi-Op -- flow augmentation and off-site mitigation -- that I think are especially problematic in this regard. Previous estimates of costs for flow augmentation and water spilled for fish averaged $180 million per year. These assumed electricity prices well below the current market. The Bi-Op's requirements for off-site mitigation are still too general to enable meaningful cost estimates, but actions to reduce withdrawals, screen diversions, and improve riparian habitat will not be cheap. We owe it to ratepayers and taxpayers to monitor the biological consequences of actions closely, to see that these expenditures are having the intended effects.
Second, we need to be humble in the face of our ignorance. Our ability to accurately forecast the future -- with or without the effects of new management actions -- is very limited. The plethora of models that do such forecasting should not distract one from this. Therefore, one should not place too much confidence in anyone who says "I know what will happen to the fish if we do. . . The best we can hope for is to continue learning as we go along, and not take predictions -- optimistic or pessimistic -- too seriously.
Finally, the monitoring called for in the Bi-Op is an enormous undertaking in its own right. Given the scale of the effort involved, I am concerned that there will be a movement to try to monitor everything that swims, creeps, or crawls, since almost anything might be related to problems for ESA-listed stocks. In the absence of some guiding principles -- e.g., that monitoring be directed at assessing the effects of Bi-Op actions -- I worry that efforts will be too broad, general, and diffuse. If that happens, in 5 to 10 years scientists and policy makers will be rehashing the same arguments about flow augmentation, transportation, hatchery effectiveness, and the like that vex us today. Instead, I would recommend a set of closely monitored management experiments to see what works and what doesn't. Unless this is the foundation for future research efforts, money spent on monitoring will be money down the drain.
NMFS, March 2000. Salmonid travel time and survival related to flow in the Columbia River. Concludes that there is almost no relationship between flow and survival for spring migrants, and a highly confounded relationship for fall chinook.
NMFS, April 2000. Passage of juvenile and adult salmonids past Columbia and Snake River dams. Concludes that current juvenile passage survival is similar to that in the 1960's, before the Snake River dams were constructed.
NMFS, April 2000. Summary of research related to transportation of juvenile anadromous salmonids around Snake and Columbia River dams. Contains "D" estimates for Snake River fish.
Paulsen, C. and T. Fisher. Statistical relationship between Snake River spring/summer chinook salmon parr-to-smolt survival and indices of land use. In press, Transactions of the American Fisheries Society. Demonstrates that juvenile survival from Parr to smolt varies widely across years and tagging locations.
Oosterhout, G. Analysis of a decision: A critique of the National Marine Fisheries Service's draft Biological Opinion on the Operation of the Federal Columbia River Hydropower System, from the perspective of the sciences of decision analysis and risk assessment. Criticizes the draft Bi-Op for the failure to use decision analysis techniques.
Paulsen, C. and R. Hinrichsen. Experimental Management for Snake River Spring/Summer Chinook: Trade-offs Between Conservation and Learning for a Threatened Species. In review, Canadian Journal of Fisheries and Aquatic Sciences. Examines how many years would be required to detect the effects of a variety of management actions.