The coast-wide collapse in marine survival of west coast Chinook and steelhead: slow-moving catastrophe or deeper failure?

Accelerating decreases in survival are evident for northern Hemisphere salmon populations. We collated smolt survival and smolt-to-adult (marine) survival data for all regions of the Pacific coast of North America excluding California to examine the forces shaping salmon returns. A total of 3,055 years of annual survival estimates were available for Chinook (Oncorhynchus tshawytscha) and steelhead (O. mykiss). This dataset provides a fundamentally different perspective on west coast salmon conservation problems from the previously accepted view. We found that marine survival collapsed over the past half century by a factor of at least 4-5 fold to similar low levels (~1%) for most regions of the west coast. The size of the decline is too large to be compensated by freshwater habitat remediation or cessation of harvest, and too large-scale to be attributable to specific anthropogenic impacts such as dams in the Columbia River or salmon farming in British Columbia. Within the Columbia River, both smolt survivals during downstream migration in freshwater and adult return rates (SARs) of Snake River populations, often singled out as exemplars of poor survival, appear unexceptional and are in fact higher than estimates reported from other regions of the west coast lacking dams. Formal Columbia River rebuilding targets of 2-6% SARs may therefore be unachievable if regions with nearly pristine freshwater conditions also fail to achieve these targets. Finally, we present case studies demonstrating that the historical response to evidence that the salmon problems are primarily ocean-related was to re-emphasize freshwater actions and to stop work on ocean issues. With ocean temperatures forecast to increase far further, the failure of management to identify the drivers of salmon collapse and respond appropriately suggest that the future of most west coast salmon populations is bleak.


Introduction
The total abundance of salmon in the North Pacific has now reached record  Table S1for reported estimates). We used these data to make regional comparisons of 182 smolt survival and survival scaled by distance travelled during the downstream 183 freshwater migration to the sea. The red horizontal line shows the median value for all Snake River data in a given panel (red coloured bars). Data are shown as a box and whisker plot with associated sample size listed above the appropriate boxes. Abbreviations: LRE, Lower Columbia River and estuary (i.e., the river reach from just below the lowest (Bonneville) dam to the river mouth); Release to BON measures Snake River survival from hatchery release through the Snake River above Lower Granite Dam and down through the 8-dam hydrosystem to the last dam (Bonneville). Full river measures survival from release to the mouth of the Columbia River. Data sources and annual survival estimates are reported in Supplementary Table S1. Coast-wide trends in adult survival (SARs) 230 Adult survival data for Chinook salmon are available for a varying range of years.

231
The most extensive data sets are for the upper Columbia (both subyearling and yearling 232 Chinook) and Snake rivers (yearlings), which extend back to the 1960s (Table S1). In essentially all regions where time series extend back to the 1970s or earlier, 236 survival to adult return has substantially decreased with time (Fig 3). A large drop in 237 SARs for yearling Snake River Chinook is evident from the 1960s to approximately the over time but at a persistently higher mean level relative to the Snake. 252 253 Regional survival differences 254 When compared by region (Fig 4), median Snake River yearling (Spring) Chinook 255 SAR (1%) is higher than the regional median SARs for Puget Sound (0.55%) and the 256 Strait of Georgia (0.53%), and is virtually identical to median survival for the Upper 257 (0.96%) and Lower (1.08%) Columbia River populations. Regional Fig 4).

266
As the time series plot (Fig 3)  survival rates of Columbia basin stocks (Fig 2). The regional-scale aggregation of SAR data provides a useful overview of 295 survival between regions. However, important population-specific differences are 296 potentially obscured because small numerical differences may in fact reflect large 297 differential impacts on survival when SARs are low. For example, when regional 298 SARs are only 1%, a population-specific SAR of 0.5% actually represents a population 299 whose survival rate is only half that of the other populations. In addition, regional Chinook, and survival rates per 100 km are much better than those of all steelhead 421 populations located outside the Snake River (Fig 2).

523
As we will discuss, many other "single factor" reasons for poor salmon survival along 524 the west coast also suffer from the same logical flaw that survival now seems to be poor 525 everywhere.

526
Overfishing alone can't explain the decline 527 Wasser et al [48] cite this blanket statement: "Anadromous salmonids (ocean-type) Chinook harvest levels of 50%-70% that were formerly sustainable are no 535 longer sustainable because marine survival dropped 4-5 fold over the past few decades.

536
The drop in marine survival is too large (75-80%) to be compensated by even the 537 complete cessation of harvest. The magnitude of the gap is widely unappreciated, and 538 the relatively small percentage difference between the harvest rate (50-70%) and SARs 539 (75-80%) is misleading. (1 ) (1 ) The maximum compensation management can make for declining marine 561 survival occurs when all fisheries are curtailed completely (h 2 =0). In this case, ceasing 562 or reducing harvest can only fully compensate if the initial rate of sustainable harvest is    The data analyzed in this paper demonstrate both a long term coast-wide decline 668 in survival for Chinook  Island. Managing salmon research 848 We are troubled that the increase in monitoring evident as survival has dwindled anthropogenically altered freshwater ecosystems will help redress the problems (e.g.,

867
[91]).   there is now some developing evidence for differential salmon survival in the sea; e.g.,

918
Assuming that the region of poor survival progressively expanded northward  unlikely that recovery can be achieved without an improvement in ocean survival.

1019
Unfortunately, current scientific knowledge is simply insufficient to understand how to 1020 promote this.    Table S1. In the other regions, smolt survival was estimated from hatcheries or traps, to 1170 acoustic receiver arrays near the river mouths. In some cases, fish were transported 1171 either upstream or downstream of their tagging location, e.g., Chilko Chinook were 1172 reared at a lower Fraser River hatchery but released ~500 km upstream in the Chilko

1173
River. Migration distances to the sea after release were typically much shorter than in 1174 the Columbia or Fraser Rivers (see Table S1).  River basin and for a few stocks located in the Columbia River basin ( In contrast, systematic survival data based on PIT tags first came into 1206 widespread use in the Columbia River Basin in 1997 (Table S2). PIT tags are long-1207 lived but extremely short distance radio frequency tags that can successfully transmit  Table S2.       , where μ and σ are respectively the mean and standard 1489 deviation of log e (S), and Z t is the standard normal variable Z~N(μ, σ). This is 1490 important because the log-normal distribution is skewed, exhibiting occasional rare 1491 high survivals which increases the expected value above the mean. As a result, the 1492 expected value of a log-normally distributed SAR time series is neither the simple mean   In this numeric example, where survival falls from 6% at the start of the record to 1% at 1538 the end, the uncertainty relative to the point estimate increases almost 2.5-fold (√6).

1539
(Taking into account that both the number of outgoing smolts and the number of 1540 returning adults is not known without error, as is implicitly assumed in using the 1541 binomial probability distribution, the actual uncertainty will be even larger when these 1542 uncertainties are taken into account).

1544
It is interesting to note that should survival fall from the current ca. 1% level to 0.1%--1545 a ten-fold further decline-it would in fact be difficult to recognize this massive decline 1546 (a fall as large as the decline from 100% to 10% or 10% to 1% survival) because of the   shows the median value for all Snake River data in a given panel (red coloured bars). Data are shown as a box 15 and whisker plot with associated sample size listed above the appropriate boxes. Abbreviations: LRE, Lower 16 Columbia River and estuary (i.e., the river reach from just below the lowest (Bonneville) dam to the river 17 mouth); Release to BON measures Snake River survival from hatchery release through the Snake River above 18 Lower Granite Dam and down through the 8-dam hydrosystem to the last dam (Bonneville   Supplementary Table S1. The black horizontal line within each bar is the median of the SAR data 52 available for that population. Median survival across all available data for each geographic region is shown as 53 a blue line; median Snake River survival for all populations combined is shown as a red line and overplotted 54 on all panels for comparison. The number of years of data is shown to the right. 55 and each year) by the median SAR calculated across all available Snake River SARs for that year. The 58 median Snake River SAR is overplotted in red. Note the logarithmic scale on the x-axis. 59 above each group (green font) and the ratio of median SARs relative to the Snake River is shown immediately 64 above the upper whiskers (black font). 65