A Bayesian Approach to Combine Multiple Sources of Escapement

1
A Bayesian Approach to Combine Multiple Sources
of Escapement Data to Estimate Wind River
Steelhead Abundance

• Dan Rawding and Charlie Cochran

2
Steelhead Abundance Estimate

• Goal - provide an unbiased estimate with an
  estimate of uncertainty (confidence interval) for
  Wind River summer steelhead
• Motivation - Hemlock dam was removed during the
  summer 2009, so we want to determine if
  mark-resight (snorkeling) estimates are a robust
  alternative to mark-recapture (trap) estimates
  from the recoveries at Hemlock dam.
• Test Petersen assumptions and correct to obtain
  an unbiased estimate

3

• Wind River is located 55 miles east of Portland
• Drainage Area is 226 sq miles
• Only summer steelhead can jump Shipherd Falls
• Nov-Apr 100 ladder
• May-Oct 25 ladder
• Hatchery steelhead excluded since 1999
• Hatchery steelhead are less than 2 of steelhead
  escapement

4
(No Transcript)
5
Shipherd Falls Trapping

• Trap operation (Jun 1 May 1) all fish tagged
• Fluorescent Floy tags (Jun 1Nov 15) rotate tag
  colors so repeats jumping falls can be identified
  
• White tag for summer steelhead fish (Nov 15 - May
  1)
• Blue tag for winter steelhead (Nov 15 to May 1)
• Double Floy tags
• Ensure resight during snorkels
• To estimate tag loss

6
Recapture Resight Methods

• Petersen Estimate is N mc/r, where m marks,
  c captures, and r recaptures
• Trout Creek
• Two trap design based on all marks released at
  Shipherd Falls and trapping at Trout Creek
• Winter Snorkel
• Mark-resight abundance estimate for fluorescent
  tags plus a census of all fish passed from Nov
  15- early May.

7
Jumper Method

• Aug Sep snorkel estimate (200 300)
• Sep Aug snorkel estimate (100)
• Jumper Rate (Sep - Aug estimate) / trapped fish
  (100/20) 5
• of trapped fish until end of jumper period
  jumper rate (305) 150
• of trapped fish after jumper period 100
• Population estimate is the sum of the
  estimates 300 150 100 550

8
Unbiased Estimates

• Petersen assumptions must be met for an unbiased
  estimate
• Population is closed
• No tag effects
• No tag loss
• Tagging does not effect catchability/survival
• All tags are identified and reported
• Equal Catchability
• all fish have the same probability of being
  marked
• or all fish have the same probability of being
  captured in the second sample
• or marked and unmarked fish mix completely

9
Closure

• Tag recoveries
• adjacent basins,
• PIT tag detections from mainstem Columbia River
  dams
• tributary instream interrogators, and
• anglers
• Probably a slight under reporting of tags leaving
  basin, since all out of basin recoveries are not
  complete
• 99.3 to 99.7 of tagged fish stayed in the Wind
  River

10
Tag Loss

• Double Floy tag all fish at Shiphed Falls
• Inspect fish at Trout Creek trap to estimate
  single tag loss, and then estimate the
  probability of losing both tags, and apply to
  Trout Cr. Estimate.
• Annual probability of retaining at least one tag
  is 97
• What about snorkeling?
• Obtain tag recovery information from Trout Cr,
  trap, Shipherd Falls re-ascension repeat
  spawners, and angler recoveries
• Use logistic regression to estimate tag loss
  based on days after tagging and probability of
  losing two tags
• Use median day between tagging and snorkel to
  estimate tag loss, and the probability of
  retaining at least one tags

11
(No Transcript)
12
Tag Induced Mortality

• Skamania Hatchery test
• 500 controls hatchery summer steelhead swim into
  hatchery raceway untouched
• 500 treatment held in raceway then netted,
  anesthetized, and double Floy tagged
• 21 day survival was over 99 for controls and
  treatment.
• Test hypothesis that handling reduced survival
  estimate survival of treatment relative to
  control
• Bayesian P-value 0.20, so cant reject the
  hypothesis that survival of handled steelhead is
  reduced
• Survival 99.5 w/95CI (98.4 -100)

13
Equal Mixing Aug Sep Snorkel Surveys

• Petersen estimate unbiased if systematic survey
  complete mixing
• Summer steelhead hold in the mainstem Wind River
  before ascending into tributaries during freshets
  Oct May Freshets
• Best estimate is 93-100 of summer steelhead are
  in the mainstem Wind River during Aug Sep
  snorkel surveys
• Systematic survey of mainstem Wind River
• Complete mixing test of the upper lower snorkel
  survey section showed complete mixing of tagged
  an untagged fish in 29 of 32 surveys (?0.05)

14
Snorkel Efficiency Proportion Marked

• Is snorkel efficiency (ratio of recaptures to
  marks) the same? 14 of 18 consecutive day surveys
  no difference (?0.05)
• Interpretation there is some evidence that some
  crews see tagged fish at higher rate than other
  crews
• Is the proportion of marked fish (ratio of
  recaptures to observed fish) the same? 18 of 18
  consecutive day surveys no difference (?0.05)
• Interpretation even though there is some
  evidence of varying snorkel efficiency, there is
  a lack of evidence that tagged and untagged fish
  are being observed at different rates.
  Population estimates are consistent between
  consecutive day surveys.
• Pool snorkel surveys from consecutive days to
  improve abundance estimate

15
 
Bayesian Mark-Recapture Model
                   
Beta prior (ab1), Uniform prior (minT,
max2000), psnorkel efficiency, Nt Pop.
Est, Ttags, Rrecaptures,Ccaptures
16
Jumper Model
17
DAG for tag loss, handling, closure incorporated
into Petersen Estimate
18
2 of 27 estimates Significantly Different
(?0.05)
19
(No Transcript)
20
Summary

• Precision (CV) of individual Trout Cr, Jumper,
  and Winter Snorkel survey estimates ranged from
  11 to 22 from 2001 onward.
• Jumper estimate was the most precise in 8 of 10
  years with CV from 11 to 17.
• When all method were combined the precision (CV)
  improved to 7 to 10.
• Multiplying the number of tags released by 96 to
  99 is needed to obtain an unbiased
  mark-recapture estimate to adjust for closure
  (lt1), handling mortality(lt1), and tags loss
  (probability of retaining at least one tags lt3
  from tagging to prespawning) violations

21
Summary

• There was no difference in the proportion of
  marked fish between 18 consecutive snorkel
  surveys but there was some evidence of different
  snorkel efficiency on consecutive surveys (4 of
  18)
• Mark-recapture (trap) and mark-resight (snorkel)
  estimates were similar in 25 of 27 comparisons
• Mark-resight estimate is a consistent alternative
  to the mark-recapture estimates in the Wind
  River.
• Summer steelhead mark-resight (snorkeling) may be
  a more cost-effective alternative to
  mark-recapture (trapping) in larger system where
  holding a trap is expensive and problematic.