DQO Steps 67 Template Status and Trends Subgroup

1
DQO Steps 6-7 TemplateStatus and Trends Subgroup

• IDFG ODFW USFWS CRITFC
• Claire McGrath Tom Rien Paul Wilson Earl Weber
• Sam Sharr Eric Tinus
• Charlie Petrosky
• NPT NOAAF
• CBFWA WDFW Paul Kucera Chris Jordan
• Frank Young Annette Hoffman Jay Hesse
• Pete Hahn Chris Beasley

2
Primary management question

• The Problem Delisting of Snake River Sp/Sum
  Chinook ESU
• The Decision SRSS Chinook ESU is no longer at
  risk of extinction (5 in 100 yrs)
• Inputs to the Decision must define
• Performance measures
• Uncertainty in data
• Natural variability spatial and temporal
• Sampling measurement
• Evaluate sensitivity of decision to inputs
• Test scenarios ( monitoring designs)

3
Objectives

• The Problem Delisting of Snake River Sp/Sum
  Chinook ESU
• Evaluate sensitivity of decision to inputs
• Test scenarios ( monitoring designs)

Specific example
General approach
4
Decision Rules - A/P

• A/P viability curve Risk lt 5 of decreasing to
  below critical number of spawners/year for a
  generation in a 100 years

5
Decision Rules SS/D

• SS/D Categorical, weight of evidence approach
• Uncertainty ? increased risk ratings

6
Decision Rules A/P x SS/D Viability Matrix
7
Acceptable limits on decision errors

• TRT is still working on values for a and b
• Formal consideration of uncertainty rare
• Acceptable is a policy decision
• TRT will make the decision
• We will provide probability level of decision
  error from alternate designs

8
Analytical methods to evaluate PMs
9
Analytical methods for redd surveys
10
Sources of variation - Natural

• Spatial variation Site to site at point in time
• ? Ensure that sampling sites are spatially
  representative
• Temporal variation Interannual
• ? Ensure that sampling is of sufficient duration
• Space x Time interaction

11
Sources of variation - Measurement

• Residual sampling / measurement uncertainty
• Inaccurate or biased methods
• ? Reduce by improving methods of data
  collection, e.g., random sampling based methods
• Precision arises from sampling based estimators
• ? Reduce uncertainty by increasing sample size
  or effort.

12
Optimizing design to minimize error

• Simulation model
• Tradeoffs among potential designs investigated
• Cost / year
• Benefit reduce decision error
• Use information on variance
• Natural variability
• Measurement uncertainty

13
Acknowledged variation

• Natural variation
• Inter-annual at a fixed location
• Intra-annual across locations
• Interaction between above
• Measurement uncertainty
• Accuracy (bias)
• Precision (variance)

14
Monitoring Design Simulation

• Generate data simulated time series by
  population. Reflect reality for a) high, b) low,
  and c) moderate risk. Use realistic spatial and
  temporal variance structure.
• Take input data and generate monitoring data
  using alternate monitoring programs.
• Take monitoring data, put into decision rules.
  Re-sample iteratively.
• Conduct sensitivity analysis, to investigate
  influence of model components.

15
Monitoring Design Simulation

• Generate data simulated time series by
  population. Ex Pop A at moderate risk.
• Temporal variance structure

16
Spatial variation
Snake River
Snake River
Input dataset Mean spatial variance estimate
Salmon River
Lake Cr. Weir
Secesh River
Johnson Cr. Weir
South Fork Weir
Middle Fork Salmon River
South Fork Salmon River
Johnson Creek
17
1, cont. Assign level of uncertainty for natural
spatial variation

• Example, for 2 populations
• A. Small spatial variance
• B.Large spatial variance

18
Spatial and temporal variation

• Example for Population A with known status
• decreasing trend, variance is realistic
• Mean truth
• Error bars denote spatial variance

19
1. Generate data simulated time series (x
years) by population (n31).
20
1, cont. Assign level of uncertainty associated
with data given a specific measurement method

• Example, for 3 methods
• A. Precise and unbiased
• B. Imprecise and unbiased
• C. Imprecise and biased

21
Alternate design templates

• Relatively high accuracy and precision
• Moderate accuracy and precision
• Relatively low level of accuracy and precision

22
Filter input data with monitoring design Do
for all metrics necessary for the decision

• Example
• Method with high precision
• 5 of 31 populations ?
• Method with mod. precision
• 18 of 31 populations ?
• Method with low precision and positive bias
• 8 of 31 populations ?

23
3. Take monitoring data and put into decision
rules

• Data ? unfiltered ? result ? decision correct
  decision?
• If not, problem with model input data or with
  rules for turning data into decision
• Data ? design 1 ? result ? decision ? run
  multiple times, how often is decision correct?
• Data ? design 2 ? result ? decision ?
• Data ? design 3 ? result ? decision ?

24
4. Sensitivity analyses

• Sensitivity analysis, things to vary
• Input data sets
• Monitoring design templates
• Levels of measurement uncertainty
• Decision rules

25
Evaluating alternate designs additional work
required to improve designs

• Analyze existing datasets to improve estimates
  for
• Spatial variation
• Temporal variation
• Residual variation
• Conduct calibration studies to improve estimates
  of
• Variance associated with particular methods
• Symmetric - precision
• Assymetric - bias

26
Example of uncertainty estimationerror
associated with redd surveys
27
July Deliverables

• Immediate need work through all pieces of input
  
• Input simulation data (estimates of natural
  variability)
• Turning monitoring programs in to quantitative
  changes in input data
• Build the model (code, programming) start in
  July
• Products By July workshop, wont yet have a
  working model, but will have a framework and
  explanation of the model to present