Habitat Effectiveness Monitoring Subgroup Report

1
Habitat Effectiveness Monitoring Subgroup
Report CSMEP Chris Beasley Nick Bouwes Tim
Copeland Steve Katz Ian Parnell Charlie
Paulsen Keith Wolf Bonneville Workshop July
20-21, 2005 The task Apply the DQO to
effectiveness monitoring Steps 1-5 generate a
question Steps 6 7 strategy to identify an
optimal design
2

• Steps 1-5 were intended to generate the design
  parameters for monitoring
• Identified questions
• To what degree have restoration and/or protection
  actions affected key habitat attributes?
• To what degree have these actions affected the
  subject population?
• Are particular classes of projects more effective
  than others (e.g., scale of effect v. cost and/or
  time)
• What are the mechanistic connections between
  recovery actions, key habitat attributes and fish
  population responses?
• Unfortunately key requirements of the DQO process
  were not available for incorporation
• Management Input
• Eg step 2
• Identify principal study questions (?)
• Define alternative actions
• Combine the principal study question and
  alternative actions into a decision statement and
  state each decision in terms of whether to take
  action.
• Organize multiple decisions into an order of
  priority

Princilpal Study Question Alternative Action
Are there significant levels of lead in floor dust at a childrens residence? Remove the children from the residence
Are there significant levels of lead in floor dust at a childrens residence? Initiate a clean-up removal of lead-based paint.
Are there significant levels of lead in floor dust at a childrens residence? Take no action.
3

• The absence of management input made it unclear
  how to proceed to steps 6 7
• Applied a new process to help compose
• generic utility
• Clarification of the question
• What are all the species, down to life-history
  type and gender, of interest?
• What is the spatial boundary of the population
  for which inferences will be made?
• What is the population response variable you want
  to evaluate to determine whether a change has
  occurred?

   
4

• Picked an example Lemhi
• The Lemhi Habitat Conservation Plan (HCP)
• Heavily impacted with agriculture diversions and
  consequent passage issues
• Plan for a series of phased in reconnection
  projects in the upper main stem
• 10-17 tributary reconnection projects
• 4 year round reconnection, seasonal for the rest
• Phased in over the next 35 years
• 4 completed by 2010
• 10 completed by 2025

5

• Clarified questions for the Lemhi
• Have the actions implemented under the Lemhi HCP
  expanded the distribution of rearing juvenile
  salmonids within the basin and increased the
  density of rearing juvenile salmonids relative to
  average mainstem densities by X over 30 years
  (with some precision) when the number of
  spawners, natural disturbances, climate
  indicators, and habitat conditions not-impacted
  by the actions have been accounted for?
• Have the actions implemented under the Lemhi HCP
  produced at least a 100 increase in the number
  juvenile spring Chinook salmon leaving the Lemhi
  River in 30 years (/- X) when the number of
  spawners, natural disturbances, climate
  indicators, and habitat conditions not-impacted
  by the actions have been accounted for?
• Have the relative magnitudes of the seasonal
  migration pulses and size distribution of
  migrating Chinook juveniles leaving the Lemhi
  River changed over the life of the Lemhi HCP?
• Have the actions implemented under the Lemhi HCP
  increased the abundance of bull trout in
  reconnected tributaries relative to unconnected
  tributaries by X over 30 years (with some
  precision)?
• Have the actions implemented under the Lemhi HCP
  increased parr-smolt survival (X /-specified
  precision) of juvenile spring Chinook salmon
  leaving the Lemhi River in 30 years when the
  number of spawners, natural disturbances, climate
  indicators, and habitat conditions not-impacted
  by the actions have been accounted for?
• Have the returns of adults Chinook salmon to the
  Lemhi basin increased X (/-specified precision,
  see VSP criteria developed by ICTRT) over the
  period of the Lemhi HCP?

6

• Clarified questions for the Lemhi
• Have the actions implemented under the Lemhi HCP
  expanded the distribution of rearing juvenile
  salmonids within the basin and increased the
  density of rearing juvenile salmonids relative to
  average mainstem densities by X over 30 years
  (with some precision) when the number of
  spawners, natural disturbances, climate
  indicators, and habitat conditions not-impacted
  by the actions have been accounted for?
• Have the actions implemented under the Lemhi HCP
  produced at least a 100 increase in the number
  juvenile spring Chinook salmon leaving the Lemhi
  River in 30 years (/- X) when the number of
  spawners, natural disturbances, climate
  indicators, and habitat conditions not-impacted
  by the actions have been accounted for?

• Information Needs
• Distribution of juvenile salmonids
• Density of juvenile salmonids
• Abundance of juvenile Chinook
• Size distribution of juvenile Chinook
• Parr-smolt survival of juvenile Chinook
• Abundance of Bull trout
• Trend in Adult Chinook
• Number of Spawners
• Natural Disturbance Record
• Climate Indicators
• Habitat Conditions

7


8


9
Low Design - Effort

some targeted habitat tagging _at_traps
seines
10
Mid Design - Effort
tagging _at_traps seines
11
High Design - Effort
tagging _at_traps seines
12
Data Products
Data Information Low Mid High
Redd Counts Adult Abundance ? ? ?
Juvenile Counts _at_ traps Juvenile Emigrant Abundance ? ? ?
Parr Smolt Tag detections Parr-Smolt Survival ? ? ?
Snorkel counts (targeted) Juvenile Distribution ? ? ?
Habitat survey (presence) Effect of Actions on Habitat ? ? ?
Snorkel counts (extensive) Parr Density Distribution ? ?
Habitat survey Covariates for fish population response ? ?
Adult detection ( tagging) _at_ Weirs Adult Returns Distribution ?
PIT detections at treatment/control sites Fish distribution Juvenile movement ?
Carcass surveys Prespawn mortality Adult distribution ?
13
Question Low Mid High
1) ...distribution of rearing juvenile salmonids density of rearing juvenile salmonids? ? - ? ? ? ?
2) ... increase in the number spring Chinook... Leaving? ? ? ?
3) migration pulses and size distribution of migrating Chinook increased? ? ? ?
4) increased the abundance of bull trout - ? ?
5) increased parr-smolt survival of juvenile ? ? ? ?
6) increased returns of Adult Chinook? ? ? ?
altered the habitat in the manner anticipated? - ? ?
Addressed with greater confidence
Addressed with less confidence
14
Cost
Design Top-Down Bottom-Up
Low 323,000.00 354,000.00
Mid 377,000.00 493,400.00
High 580,000.00 643,600.00
Top-Down based on per project costs and
contracting history for previous
projects. Bottom-up based on cost per unit
time per person multiplied by the sample sizes
identified in the plans.
15

• What are the opportunities to economize these
  designs?
• Is one of these plans good enough?
• Does this plan get me what I need?
• Why should I not just pick the cheapest design?
• If I have a fixed amount of money, what parts of
  each design can I compromise on?
• If so what performance do I lose for the dollar
  saved?

16

• What inference is good enough?
• Comparisons
• Less replication
• Low A B relative to C - Before and After
• Greater Replication
• Mid Reconnected vs unconnected trib. Areas -
  Before and After (narrow data)
• High Reconnected vs unconnected trib. areas -
  Before and After (broad data)
• The Client has to decide if this is good enough
  Not monitoring technicians.

17

• What are the opportunities to economize these
  designs
• Is one of these plans good enough?
• Does this plan get me what I need?
• Why should I not just pick the cheapest design?
• If I have a fixed amount of money, what parts of
  each design can I compromise on?
• If so what performance do I lose for the dollar
  saved?
• Usually this means statistical performance
  quality of inference, power, precision, accuracy
  per dollar.

18

• Why collect so much other data?
• (more data more , no?)
• Use diverse data to explain/subtract other
  sources of variability
• Filter out noise to leave treatment effects

Mainstem
parr to smolt survival
Hayden
treatment - control survival
Year
Difference between sites (treatment, spawners,
temp residual) Difference between sites
(treatment, spawners residual) Difference
between sites (treatment residual)
19

• How to conceptually link the design data to the
  test?
• There are numerous potential analytical designs
• w/ different relationships between data tests
• BACI design. Low design Develop a linear
  predictor model and evaluate relative importance
  of treatment Paulsen and Fisher (2004)
• Asymmetric BACI design - Underwood (1994)
• Randomized Intervention analysis - Carpenter et
  al. (1989)
• Picking the right design remains a research
  effort

20
Optimization Outputs Statistical
performance Precision Bias Power Cost Inputs Co
ntrasts Replication Duration Experimental
design BACI/Staircase Pulse/Press Response
design Complicated question O(outputs) lt
O(inputs) Therefore, no single solution. A
template is not appropriate for all effectiveness
monitoring
C
B
P
R
C
D
21
The analytical design defines the relationship
between all the variables in the monitoring data
and how differences constitute a test. Different
relationships can be traded to economize based on
priorities established in steps 1-5 of DQO. This
is a continuing research project. There is still
a need for policy/ management to step in and
inform where appropriate.
22
How to conceptually link the design data to the
test? Example 1 BACI design Underwood (1994)
Source of Variartion df a a b b d d
Source of Variartion df MS F MS F MS F
Before vs. After B 1 526.2 901.7 4212.3
Among Locations L 3
Impact vs. Controls I 1 2108.2 980.6 1881.2
Among Controls C 2 35790.2 35790.2 35790.2
T(B) 10 635.0 908.1 552.5
B x L 3
B x I 1 1125.6 2340.2 12625.6 26.55
B x C 2 335.5 335.5 335.5 0.75
T(B) x L 30
T(Bef) x L 15
T(Bef) x I 5 515.6 515.6 515.6
T(Bef) x C 10 435.9 435.9 435.9
T(Aft) x L 15
T(Aft) x I 5 1041.4 3.47 266.0 8.89 435.3 1.45
T(Aft) x C 10 453.6 1.51 453.6 1.51 453.6 1.51
Residual 192 300.0 300.0 300.0
Pulse impacts
Press impacts
23
Is any of this good enough?
Design objective Evaluation Criteria for design objective L H
High inferential ability Ability to answer each question at appropriate scale. ? ?
High inferential ability Ability to meet clients needs (supply adequate information for decisions). ? ?
High inferential ability Spatially representative of larger unit of interest. Ability to legitimately aggregate data to multiple scales. ? ?
High inferential ability Continuity of time series so as to estimate trends
High inferential ability Reliability of data and assumptions required to move from data to decisions can be completely documented ? ?
Strong Statistical Performance Precision (relative to required precision for management decisions). ? ?
Strong Statistical Performance Statistical power (e.g., graphs showing power to detect various effect sizes of management importance over a 10-year period). ? ?
Strong Statistical Performance Coverage how often does true value fall within 95 CI. ? ?
Strong Statistical Performance Accuracy or Bias as estimated by direct methods comparisons relative to truth ? ?
Reasonable Cost Cost/year at scale of interest.
Reasonable Cost Hybrids Precision / cost, coverage/cost, accuracy/cost ? ?
Reasonable Cost Ability to leverage other funding sources. Use overlapping domains of interest from different agencies. ? ?
Reasonable Cost Opportunities for collaboration.
Practical Safety
Practical Feasibility of implementation.
Practical Flexibility to accommodate different types of indicators (e.g. physical habitat, juvenile fish) or different monitoring protocols ? ?
Practical Widely applicable to different regions, environmental conditions. ? ?
Practical Integration with existing programs.
Environment Environmental impact of monitoring protocol.
24

• Consequences of altered response design
• Create a time series with 8 components
• Fit with increasing number of monitored
  variables to a selected sub-segment

3000
Spawners -60.26time1294.7
Linear model
2500
r25.3
2000
1500
1000
500
0
-500
3000
3 Component model
2500
r228.9
2000
Spawners
1500
1000
500
0
-500
3000
6 Component model
2500
r277.7
2000
1500
1000
500
0
-500
0
2
4
6
8
10
12
Years
More variables is a better description/discriminat
ion But how complicated a question are you ready
to ask? The Client has to answer not the tech