Issue #1: Data Concerns

An important component of the PATH analysis is a comparison of upriver/downriver stocks. At the time NMFS prepared its status reviews on Columbia Basin stocks, NMFS was told that population data was unavailable for many tributaries. Now that information has been produced and used in the PATH analyses. The available documentation does not identify actual vs. "reconstructed data".

Can PATH disclose precise details on the methodologies used to gather the actual base redd/adult data over this 30-something year period. Were they the same in both areas? Is accounting for redds easier or harder in the lower rivers than the Snake system; i.e., might there be some bias between the two areas? Were methodologies the same every year, or did they change? Would changing methodologies change the potential biases? Can one really count redds and adults from a trail bike in some Salmon River streams? Was a trail bike also used in some parts of the lower river streams? Why didn’t PATH include a detailed section on basic field methodologies?

Is it true that some sort of assumptions/expansions were used to fill in probably 75% or so of the data used in the analyses? Please disclose the raw redd counts and explain and justify all assumptions used to expand data.

Isn't it true that 2-3 more years of additional stock abundance data are now available than those data used in the Report's analysis? Wouldn't this new data, including recent upturns in some upriver stocks, change the results? How so?

The Report suggests that downstream spring chinook stocks "have not experienced an equivalent increase in 'extra mortality'" to upstream stocks (p. 198). The Report does not contain an adequate explanation of the data upon which it reaches this conclusion; please summarize that data.

Is there any possibility that productivity differences between lower and upper areas (if they aren’t simply an artifact of data gathering differences or faulty assumptions) is related to something other than the hydrosystem, like possible differences in oceanic distributions or productivity differences of the natal streams themselves? Speaking strictly scientifically, can salmon from one ESU be used as a control for salmon in another? (Of course, without assuming there are no other differences between them other than the hydrosystem that might affect productivity?) If yes, explain the empirically-derived data that demonstrates this.

Two of the stocks apparently identified by PATH as having better survival are stocks in the John Day and Deschutes Rivers. What conclusions, if any, has PATH reached concerning the absence of significant hatchery influence on the John Day stocks on their survival, as compared with Snake River stocks? Didn't Snake River hatchery releases increase dramatically after 1970 (p. 217; Figure 6.2-1)? Did any of the downriver comparison stocks suffer similar increases?

What conclusions, if any, has PATH reached concerning the effect of the Deschutes River dams on Deschutes River stocks (cf. Lost Creek summer flow/temperature effects)? Aren't such effects absent for most Snake River tributaries, with the exception of the Clearwater (owing to Dworshak)—and does the Clearwater stock (not listed) show less of the "extra mortality" than other upriver stocks? Has PATH considered differences in temperature increases since 1970 in the areas inhabited by upriver and downriver stocks?

How does PATH respond to the criticism that you can "pick a stock and pick an answer"? Do trends with respect to the John Day and Deschutes stocks hold up for other downriver stocks? What fraction of the downriver stocks has PATH considered?

Is there empirical evidence that the productivity of salmon increases exponentially as populations grow smaller? Isn't it true that the classic Ricker curve was derived based on data sets from more abundant populations, and that the model does not fit empirical data from depressed populations? Is it correct that PATH did not take account of the spatial scales involved in studying productivity, and instead used generalized models (e.g., p. 134)? Did PATH assume the same measurement errors in all stocks? Doesn't this bias the estimation of parameters for the Ricker curve? How so?

We understand that PATH included a depensation term "p" in the spawner-recruit model (e.g., p. 134-5) to see if productivity decreases with decreasing stocks and concluded there is no depensation (p. 145). What empirical evidence supports the form of the equations was realistic and that as the stocks go down the productivity goes up exponentially as you conclude in the PATH model? Did PATH explore other forms of spawner-recruitment functions and their implications?

Isn't it contrary to sound statistical techniques to use a form of the equation with both the D and STEP variables when the existing data cannot estimate both parameters (p. 136)?

Isn't it true that the use of the classic Ricker model in the lifecycle analysis is predicting larger-than-actual smolt production? Is it true that the "disappearance" of these "extra" smolts is attributed to hydrosystem mortality?

Has PATH evaluated the production predictions of its models against indices of smolt abundance at Lower Granite? Why not?

We understand that only some participants within PATH are permitted to run the FLUSH model, and understand that the State and Tribal participants in the PATH process have refused for more than a year to provide details of the model to other PATH participants, including the FLUSH > model executable code, data files, source code and in particular the > values of the reservoir mortality coefficients. Unless and until these materials are disclosed, we can have no confidence in the output of the PATH process. Many of our questions concerning the model could be clarified and sharpened if we had a better understanding of how the model works.

Is it true that the time steps in the FLUSH model are in one month increments? Is it true that the FLUSH model does not predict survival over the same river stretch that PIT-tag studies have been conducted? If so, how can PATH evaluate the validity of the FLUSH model?

The Report discloses "upper bound" in-river survival estimates for CRiSP and FLUSH that are substantially below measured survival in 1998 (p. 50). Shouldn't "upper bound" survival estimates, based on "optimistic assumptions" be more than measured survivals?

The Report discloses A3/B1 in-river survival estimates for CRiSP and FLUSH (p. 51) that are substantially higher, on a per-mile basis, than measured survivals of migrants in the free-flowing Snake River. On what basis does PATH conclude that survival through the Lower Snake and Lower Columbia will be higher per mile than survival per mile above Lower Granite? Aren't predator densities known to be higher lower in the River?

The Report notes that 1995-97 survival estimates for fall chinook "are possibly the only data available for use in either the calibration or validation of the passage model", but that "[a]ppropriate survival values for this process, to-date, have not been agreed upon by the workgroup" (p. 83). Is PATH saying that some within the process are trying to use survival estimates other than those measured for model calibration and validation"? On what basis? If not, what does this discussion mean.

What is the significance of the differences in "emigrant distribution" between CRiSP and FLUSH models, and why does FLUSH use less than all the available data for assessing emigrant distribution (p. 142)?

On what basis can scientists reasonably conclude to give greater weight to a computer model that does not fit the available reach survival data (FLUSH), and lesser weight to a computer model (CRiSP) that fits such data better (p. 36)? What evidence was before the WOE scientists concerning the "fit" of these models to the data?

Why is the "higher mortality before 1980" hypothesis linked to the demonstrably untrue hypothesis that "turbine and bypass mortality are due to descaling alone" (p. 24)? If this contra-factual assumption about the nature of mortality is removed from TURB4, would the weightings change, or did the WOE scientists assume, in connection with TURB4 and TURB5, that they could not identify improvement in passage mortality since the 1970s (see p. 36)? On what basis did they ignore measured improvements?

On what basis can scientists reasonably conclude that ocean conditions for salmon have not deteriorated since the 1970s (see p. 36)? Please identify all studies and data used to support this conclusion.

On what basis can scientists reasonably conclude that BKD will dissipate if transportation of smolts ceases (see p. 36)? Aren't there other vectors for the disease? What is known about BKD prevalence in tributaries below McNary? Is the prevalence below McNary in untransported stocks markedly lower?

Why are the BKD, Hydro and Regime Shift hypotheses weighted as if they were mutually exclusive? (Aren't they treated as mutually exclusive in the life cycle modeling as well (p. 137)?)

Given that flow augmentation relies upon within-year effects of reservoir releases upon migrating salmon populations, isn't within-year flow-survival data the best available scientific data to predict the effects of flow augmentation measures? Has PATH considered such data?

What effort has PATH made to resolve the uncertainties created by colinearity between flow and other independent variables such as temperature, Julian date, and smolt condition? Has PATH considered and evaluated the best available scientific evidence indicating that in multiple-regression analyses, these variables have greater predictive power on the survival of migrating juveniles than flow?

What are the theoretical and empirical bases for assuming that the rate of mortality increases exponentially per day while migrating juveniles pass through the hydrosystem?

We understand that Dr. James Anderson has attempted to "reverse engineer" some of this data from the limited disclosures by the PATH modelers, and thereby deduce the survival assumptions implicit in FLUSH. For 1998, is it true that that FLUSH would predict LGN-BON survival of 20%, CRISP would predict 52%, and NMFS estimated actual survival at 58%? For 1997, is it true that FLUSH would predict 41% survival, CRiSP would predict 46%, and NMFS estimated actual survival at 46%? Is it true that higher flows in 1997 made the modeled travel time around 16 days, while travel time in 1998 was modeled at around 22 days. Isn't the longer travel time and higher survival in 1998 entirely inconsistent with the fundamental assumptions of the FLUSH model?

The Report relies on "Earl Weber's analysis" (p. 47 n.12) to assess the effects of adding 1 MAF of Snake River flow augmentation, suggesting that such an increase in flow augmentation would increase reservoir survival by 5%. Please provide copies of this analysis, and explain why such ad hoc analysis was used instead of a comparison of effects through both CRiSP and FLUSH, as elsewhere in the Report.

The Report discloses that the CRiSP fall chinook modeled survivals in the reach below Lower Granite did not fit observed survivals (p. 122; Figure 3.2.1-2). Why did PATH present model results contrary to the available data on flow and survival for fall chinook when PATH knew, because it cited the work (p. 122), that the analysis of available data does not detect a flow-survival relationship for fall chinook within the hydrosystem? Why does the Report omit this sort of "ground truthing" information entirely when discussing the FLUSH fall chinook results? Isn't it true that the modeled survivals for FLUSH are an even poorer fit than CRiSP?

Why does the FLUSH model cap fish travel time at the flow-travel time 90^th percentile when the dams are in place, and remove this limitation when dams are removed (p. 142)?

If the flow-survival relationship is removed entirely from the assumptions underlying the PATH results, how would those results change?

On p. 82 of the Annual Report, PATH assumes 10% direct turbine mortality (see also p. 72). Isn't the best scientific data concerning direct turbine mortality derived from balloon-tag studies? Wouldn't the best estimate of direct turbine mortality, based on such studies, be on the order of 5%, half of the mortality assumed by PATH? Can one reasonably conclude that assuming excessive direct turbine mortality is one factor responsible for the fact that both CRiSP and FLUSH tend to predict lower-than-measured passage mortality?

The discussion of fall chinook suggests that passage models apply a special, higher predation rate in the boat-restricted-zone (BRZ) immediately below the dams (p. 72). The same discussion also refers to mark-recapture studies of turbine mortality that would appear to include some or all of the predation in the BRZ. Isn't PATH double-counting turbine mortality by using modeling parameters based on such studies and counting predation mortality in the BRZ in addition to turbine mortality?

If a 5% figure for direct turbine mortality is used in the modeling underlying the PATH results, how would those results change?

On p. 82 of the Annual Report, PATH also assumes 2% spillway mortality. We are aware that substantially higher spillway mortality has recently measured at The Dalles Dam, and are aware of other dam-specific measurements that exceed the 2% figure; on what basis do some within PATH challenge how "applicable these studies are" (p. 72). The "historical" discussion of spill survivals on p. 72 does not include the full range of measured spillway survival, and omits estimates of high mortality obtained at Lower Monumental Dam on steelhead. On what basis does PATH assume a single common spillway mortality figure rather than use dam-specific numbers? Is this some sort of limitation of the models? Why doesn't it make more sense to use models with dam-specific across-the-concrete mortality parameters?

The report also assumes spillway effectiveness of 1.0, citing "several studies" (p. 82). We are unaware of any such studies and request that PATH identify them. Every study specifically listed in the Report shows a value greater than one (p. 73; Table 3.1.1-3); the list omits pre-1997 radio-tag studies also demonstrating evidencing higher spillway effectiveness.

Please explain why the many studies showing spillway effectiveness greater than 1.0 were discounted and not incorporated into the spring/summer analysis. Please explain why, given the availability of many spill effectiveness estimates for projects other than The Dalles, PATH chose a single estimate of spill effectiveness of 1.0 at all projects other than The Dalles.

On p. 82 of the Annual Report, PATH also assumes 12% bypass mortality. Was this value assumed for all bypass systems? PATH purports to justify this result using a single study at Little Goose Dam (pp. 76, 82, 118), without reviewing extant data from other bypass systems. NMFS recently presented 1994-1995 estimates of bypass survival at Lower Granite Dam, Little Goose Dam, and Lower Monumental Dam. In all cases, bypass survival was higher than turbine survival and ranged from 96-100%. Please provide further details on the single, contrary value used by PATH ("the mean of five replicates with substantial variability" (p. 118), and explain why PATH chose to disregard all data other than the unpublished 1997 results.

We note that PATH's assumption suggests that the Region's massive investment in bypass systems, over many years, increased across-the-concrete mortality by 2% for every fish that successfully avoided the turbines. This assumption seems to contradict generally increased in-river survivals in the wake of turbine screening. It is also hard to believe that adaptive management failed so profoundly. If PATH were correct, we could significantly improve in-river survival by shutting down all bypass systems, yet we do not see any fishery management agency advocating this course of action.

The Report suggests that PATH considered alternative hypotheses under which the FGE of extended length submersible bar screens was greater than the FGE of standard length submersible travelling screens, and a hypothesis that the FGEs were equal in both cases (p. 24; see also pp. 76-78). Weren't the many millions of dollars invested in extended-length screens supported by evidence that such screens would increase FGE? On what factual basis does PATH continue to give life to the hypothesis that all this money was wasted?

The above issues regarding the choice of parameters also raise the question of how these numbers were chosen. Was this a consensus decision of PATH or a subgroup thereof?

The Report suggests that PATH considered two alternative hypotheses for survival through the Lower Snake Reach after dam removal: 85% and 96% (p. 25). Do these assumptions depend upon historical measurements of pre-dam survival, e.g., those conducted by Raymond. Has PATH accounted for the numerous errors in Raymond's analysis found by Williams and others?

Don't both of these assumptions provide for substantially higher survival than that measured for both hatchery and juvenile migrants through the free-flowing reaches of the Snake River above Lower Granite Dam? On what basis does PATH discard survival per mile data from the Clearwater River? On what basis does PATH discard data concerning survival per mile for hatchery fish? What fraction of the overall data available on survival per mile did PATH utilize in coming up with survival per mile estimates? What were the estimates?

Is there any reason to believe that dam removal would affect the incidence of fish diseases and parasites? Isn't it true that these diseases and parasites are more prevalent today than prior to construction of the four Lower Snake Dams? How has PATH accounted for the prevalence of such diseases and parasites in assessing post-removal natural mortality?

Is there any reason to believe that dam removal would affect the density of predators on salmon smolts, such as walleye and Northern pikeminnow? Has PATH compared the density of predators in undammed reaches, such as the Clearwater River, and in dammed reaches?

What assumptions has PATH made about steady-state predator populations after a transition to dam removal? On what basis has PATH rejected the hypothesis that predator densities will increase, thereby increasing "natural" mortality?

What assumptions has PATH made about river temperatures after dam removal, particularly peak temperatures inimical to salmon survival, which are reported to have been higher before the dams were constructed?

The discussion of the FLUSH model states that "pre-dam survivals [are] derived from model relationships and second set of runs with fixed survival per mile" (p. 142). Please describe this "second set of runs" and the survival per mile number that was used.

Has PATH considered the extent to which manmade predator habitat (e.g., dam tailraces), will be replaced by restored predator habitat (stillwater below riffles). Does PATH have a rational basis for concluding whether the amount of predator habitat will be increased or decreased?

How do the PATH results change depending upon assumptions concerning natural mortality? How would they change by adding an alternative hypothesis that juvenile survival per mile through the Lower Snake reach would equal measured survival per mile above Lower Granite Dam?

The Executive Summary to the 1998 Annual Report suggests (p. 9; Table 1.3.2-1) that under the "SRP Weights", the probability of recovery falls from 0.27 to 0.25 as transportation is increased. What increase in transportation was assumed? By what causal mechanism does transportation reduce the probability of recovery? Is there any empirical evidence that indicates that such a mechanism exists?

Does PATH ascribe negative effects on survival to stress during transportation? What is the empirical basis for such an assumption? Isn't it true that observed negative effects of handling stress (e.g., ability to evade predators) disappear within a day? Is there any evidence whatsoever that stress resulting from juvenile passage persists beyond a period of days?

What empirical evidence suggests that transported and non-transported fish have differing rates of survival below Bonneville? How, if at all, has PATH accounted for effects arising from differing release times of sample groups in estimating "D"?

Please provide a better explanation of the transportation assumptions discussed at p. 25 of the Report. If the assumption of delayed mortality is removed from the models, such that the efficacy of transportation is modelled based on measured transportation/in-river ratios, how would the PATH results change?

Is it true that PATH assumed that transported fish survived approximately 17% less to the spawning grounds than in-river migrating fish above Lower Granite Dam? What empirical data support such an assumption? If one analyses, for example, the available data concerning the return of smolts to the spawning grounds (e.g., Tables in the Transportation Review Group Ad Hoc Review), isn't it true that positive TIRs appear (albeit without statistical significance). What is the posited mechanism by which the survival of transported fish is affected three or four years later after transportation? Do any available studies of transportation, using fish that have migrated hundreds of miles down the Snake River, document any impaired homing to the spawning grounds.

The Report states that "for non-drawdown actions (A1, A2, and A2') the vast majority of fish are transported". Isn't there a significant difference between A1 and A2? Figure 2.2.6-3 suggests an error in the FLUSH model, in that there is no way that A1 operations collect 94-100% of Snake River fish. How do the FLUSH predictions square with other estimates of the proportion of fish transported in recent years, which are used in the TMT and IT? The CRiSP and FLUSH models also have dramatic differences in the modeled percentages of fish transported in earlier years (retrospective analysis) (p. 144; Figure 3.3.1-2)—aren't there historical estimates of transport percentages that can be used to identify which assumptions are correct?

Are the PATH participants suggesting that positive TBRs for fall chinook transported from McNary will not be experienced by Snake River stocks, and if so, why? Is it true that PATH is assuming that transported Snake River fall chinook salmon will suffer 80-99% higher mortality below Bonneville than untransported fish?

The Report indicates three principal sources of "extra" mortality: the rise of BKD, some sort of "Hydro extra mortality hypothesis", and climate changes (p. 23 & n.2). Please provide a discussion of the empirical evidence supporting each of these hypotheses identified by PATH, and how dam removal would affect such "extra mortality".

To what extent do the Report's estimates of "D" depend upon low flow/low survival data points from 1977 and 1973? Has PATH considered operational features of the FCRPS rendering those years not comparable to other years (very high spill before turbine installation; massive trash buildup at Lower Granite) as described by Williams and others?

Please provide a fuller explanation of the PROSPA-ALPHA and PROSPD-DELTA assumptions used in life cycle modeling (p. 25).

Is there a posited mechanism by which removal of dams would affect survival in the ocean? What empirical evidence exists concerning the posited mechanism?

On what basis has PATH assumed that four-dam removal would increase adult survival by roughly 50% through the reach presently occupied by the four Lower Snake dams and reservoirs (p. 15)? Isn't it true that adult travel time through this reach is now lower than before the dams were built, based upon radio tag data comparing travel time through the reach and above it? Is PATH assuming, in effect, that adult survival through an undammed reach of the Lower Snake is 100%?

The general presumption in fisheries science seems to be that higher velocity and more turbulent flows impose substantial energetic costs on migrating salmon. In light of these considerations, how has PATH analyzed the incremental effect of dams and reservoirs, as opposed to natural rapids, on adult passage?

The Report includes escapement forecasts for Johnson Creek (at p. 13; Table 1.3.6-3). Can PATH provide estimates of increases in escapement for the entire listed ESUs across the various alternatives? We note that the change in mean escapements from alternative A1 to A3 are relatively small, suggesting that the total number of adults realized from dam removal is relatively small; is it reasonable to infer that the other stocks will display behavior similar to the Johnson Creek stock?

The Report assumes a carrying capacity of 18,000 fall chinook, based on the highest estimated number of spawners in 1968 (p. 141). But the Hells Canyon Complex eliminated fish passage in 1967, so that 1968 returns would have included adults generated by a much larger above-Hells-Canyon carrying capacity. After 1971, when few or no pre-dam adults would still have been returning, the highest number of estimated adults (p. 104; Table 3.1.2-7) was 2,208. How do the PATH results change if the Ricker parameters are adjusted for a maximum carrying capacity of only 4,000 fish, or some other, substantially lower number for carrying capacity? Wouldn't such an adjustment tend to reduce the extraordinarily high recovery probability (1.0) for A3? Don't the estimated returns suggest that a recovery target of 2500 fish (larger than ever measured since 1971) is unrealistic?

Why has PATH rejected the use of salmon metapopulation models in assessing the risk to ESA-listed stocks? Isn't it true that metapopulation models indicate a virtually certain persistence of Snake River spring/summer chinook salmon?

The Report suggests (at pp. 14, 56), that some habitat improvements can reduce abundance by triggering mixed-stock harvest increases. Why doesn't PATH run a habitat sensitivity analysis holding harvest constant?

What is the basis for PATH's assumption that in-river harvest rates on spring/summer chinook have ranged from 3% to 8% since 1975? What assumptions did PATH make about commercial, sport and tribal harvest? What is the nature and quality of the underlying data?

Why does PATH assume that spring/summer chinook are not taken in ocean fisheries? What data is available on the ocean distribution and feeding habits of spring/summer chinook that would suggest that such fish are not at risk from ocean fisheries? What studies have tested the hypothesis that such fish may be taken "incidentally" in ocean fisheries?

The harvest results seem to suggest that a 15% change in adult survival translates into a 2.5-3% change in recovery probabilities (p. 15). It appears that dramatically larger changes in recovery probabilities are associated with changes from A1 to A3. Can PATH provide a harvest sensitivity analysis that involves changes in harvest on the order of 50% or more, to provide a consistency check with the alternatives analysis? If this is not appropriate, please explain why not.

The Report states that the tributary harvest rate is assumed to be 5% at all run sizes for both spring and summer chinook (p. 60). Table 2.3.2.-1, however, suggests that tributary harvest is only accounted for at certain run sizes above 100% of MSP. Which approach was used in modeling?

How would the jeopardy assessments change under a sensitivity analysis that assumed zero harvest rates at all run sizes? Is it true that the PATH models yield nonsensical (chaotic) results if both zero harvest and dam removal is assumed? Wouldn't such an analysis be prudent to test the specifications of the models?

Some have suggested that up to 1.5 million spring/summer chinook once spawned in Snake River tributaries. We assume the available historical PATH data, aggregated across all Snake River stocks, does not begin to reach this level of abundance. We also assume (see Issue #12) that forecasted escapement, even under A3, is nowhere near this level. From this perspective, how does PATH believe that A3 could result in overescapement (p. 14; see also p. 61)?

Why does PATH state that "predation on salmon smolts by Caspian terns and other bird predators in the estuary is hypothesized to have increased dramatically in the late 1980s and early 1990s" (p. 15; emphasis added)? What is hypothetical about these colonies of birds?

Why hasn't PATH considered evidence of increased marine mammal mortality?

Why hasn't PATH considered evidence of competition from exotic species, such as shad, as limiting factors in salmon production?

Why doesn't PATH address the relationship between recent increases in avian and marine mammal mortality and the WOE panel's assumptions concerning changes in system survival over time?

In the Report, PATH analyzes five general actions for spring/summer chinook (p. 22; Table 2.2-1). A sixth alternative, formerly known as "A4", reduced flow augmentation, is not covered in the analysis. Was it PATH or someone else who removed this alternative from the scope of biological analysis to be undertaken by PATH? Who? Inclusion of this analysis would not only provide an evaluation of a rational recovery strategy—rational in the sense that resources presently wasted on spring flow augmentation could be more effectively employed elsewhere—but also provide some indication of the sensitivity of the PATH results to assumptions about the efficacy of flow augmentation.

The Report recommends that the "A6/A6' [surface bypass alternative] be given a low priority at this time for further spring/summer chinook passage/lifecycle modeling" (p. 49). Please explain how the assumptions used to reach this conclusion are consistent with the assumptions used in modeling A2.