Transcript Assessment and monitoring

MONITORING and
ASSESSMENT:
Fish 7380. Dr. e. irwin
(many slides provided by Dr. Jim Nichols)
Rivers are inherently difficult to assess
 Diverse fauna (hard to enumerate)
Populations change through time
Abundance estimates (measures?)
 Habitat specialists (or not)
 Unidirectional flow
Pseudoreplication
Upstream affects downstream
 Standardization
 Efficiency
 Detectability
Why Monitor?
Science
Understand ecological systems
Learn stuff
Management/Conservation
Apply decision-theoretic approaches
Make smart decisions
Key Component of Science: Confront
Predictions with Data
Deduce predictions from hypotheses
Observe system dynamics via monitoring
Confrontation: Predictions vs.
Observations
Ask whether observations correspond to
predictions (single-hypothesis)
Use correspondence between observations and
predictions to help discriminate among
hypotheses (multiple-hypothesis)
Use of Monitoring in Science
Strength of inference:
Manipulative experiment > Impact study >
Observational study
Strength of inference for observational
studies:
Prospective (a priori hypotheses) >
Retrospective (a posteriori stories)
Claim: monitoring is most useful to science
when coupled with manipulations of
system
“Monitoring of populations is
politically attractive but
ecologically banal unless it is
coupled with experimental work to
understand the mechanisms behind
system changes.” (Krebs 1991)
Management/Conservation
Key Elements
Objective(s): what do you want to achieve
Management alternatives: stuff you can do
Model(s) of system response to
management actions (for prediction)
Measures of model credibility
Monitoring program to estimate system
state and other relevant variables
Role of Monitoring in Management
Determine system state for statedependent decisions
Determine system state to assess degree
to which management objectives are
achieved
Determine system state for comparison
with model-based predictions to learn
about system dynamics (i.e., do science)
How to Monitor?
Basic Sampling Issues
 Detectability
Counts represent some unknown fraction of animals in
sampled area
Proper inference requires information on detection
probability
 Geographic variation
Frequently counts/observations cannot be conducted
over entire area of interest
Proper inference requires a spatial sampling design that
permits inference about entire area, based on a sample
Detectability: Monitoring Based on
Some Sort of Count
 Ungulates seen while walking a line transect
 Tigers detected with camera-traps
 Birds heard at point count
 Small mammals captured on trapping grid
 Bobwhite quail harvested during hunting season
 Kangaroos observed while flying aerial transect
Detectability: Conceptual Basis
N = abundance
C = count statistic
p = detection probability; P(member of N
appears in C)
E (C )  pN
Detectability: Inference
Inferences about N require inferences
about p
C
ˆ
N
pˆ
Indices Assume Equal Detectability
Ni = abundance for time/place i
pi = detection probability for i
Ci = count statistic for i
ˆij  C j / Ci
ij  N j / N j
E (ˆij )  E (
Cj
Ci
)
pjN j
pi N i
How Do We Generate System
Dynamics? Study Designs
 Use design that imposes, or takes advantage of,
a manipulation of some sort
Manipulative experimentation (randomization,
replication, controls)
Impact study (lacks randomization and perhaps
replication, but includes time-space controls)
 No manipulation - observational study
Prospective (confrontation with predictions from a priori
hypotheses)
Retrospective (a posteriori story-telling)
Spatial Sampling Designs
Simple random sampling
Stratified random sampling
Systematic sampling
Cluster sampling
Double sampling
Adaptive sampling
Dual-frame sampling
Measurement Error
Recognize
Account for it
Scale of study
Match to critter
Detectability
Efficiency
P of capture
Patchy organisms (and/or habitat)
Nested designs
Quantify spatial patchiness
Identify scale
Spatial and Temporal Variation
BACI design
Before-After/Control-Impact
Disturbances
Biological response to disturbances
Anthropomorphic
Pulse
Press
Catastrophes
Time scale of recovery
Rapid Techniques
Categorical and regression trees
Other Multimetric techniques
How much to sample
Logistics
Time
$$$$
What State Variable to Monitor:
3 Levels of Inference
 Community – multiple species
State variable: Species richness
Vital rates: rates of extinction and colonization
 Patch – single species
State variable: Proportion patches occupied
Vital rates: P(patch extinction/colonization)
 Population – single species
State variable: abundance
Vital rates: P(survival, reproduction, movement)
What State Variable to Monitor?
Choice Depends On:
Monitoring objectives
Science: what hypotheses are to be
addressed?
Management/conservation: what are the
objectives?
Geographic and temporal scale
Effort available for monitoring
Required effort: species richness, patch
occupancy < abundance
Indices: Dealing with Variation in
Detectability
Standardization (variation sources that we
can identify and control)
Covariates (variation sources that we can
identify and measure)
Prayer (variation sources that we cannot
identify, control or measure)
CONCLUSION:
ESTIMATE DETECTABILITY!
Patch Occupancy Estimation and
Modeling: Applications
 Amphibian monitoring
Wetlands: anurans, aquatic salamanders
Terrestrial plots: salamanders
 Spotted owl monitoring and patch-dynamic
modeling
 Waterbird colony dynamics
 Tiger distribution surveys
 Landbird monitoring
 Fish monitoring
Animal Abundance:
Estimation and Modeling
 Traditional monitoring foci:
Variation over time: trend
Variation over space or species: relative abundance
 Many estimation methods (e.g., Seber 1982,
Williams et al. 2002)
 Each estimation method is simply a way to
estimate detection probability for the specific
count statistic of interest
 Final step is always:
C
ˆ
N
pˆ
Observation-based Count Statistics:
Detectability
Distance sampling
Double sampling
Marked subsets
Multiple observers (dependent,
independent)
Sighting probability modeling
Temporal removal modeling
Capture-based Count Statistics:
Detectability
Closed-population capture-recapture
models
Open-population capture-recapture
models
Removal models (constant and variable
effort)
Trapping webs with distance sampling
Change-in-ratio models
Rate Parameters Relevant to
Changes in Abundance
 Population growth rate
 Survival rate, harvest rate
 Reproductive rate (young per breeding adult)
 Breeding probability
 Movement rate
 Process variance
 Slope parameters for functional relationships
Recommendations:
Why Monitor?
 Monitoring is most useful when integrated into
efforts to do science or management
 Role of monitoring in science
Comparison of data with model predictions is used to
discriminate among competing models
 Role of monitoring in management - determine
system state for:
State-specific decisions
Assessing success of management relative to objectives
Discrimination among competing models
Recommendations:
What to Monitor?
Decision should be based on objectives
Decision should consider required scale
and effort
Reasonable state variables
Species richness
Patch occupancy
Abundance
Recommendations:
How to Monitor?
 Detectability
Counts represent some unknown fraction of animals in
sampled area
Proper inference requires information on detection
probability
 Geographic variation
Frequently counts/observations cannot be conducted
over entire area of interest
Proper inference requires a spatial sampling design that
permits inference about entire area, based on a sample
Adaptive Management
Seeks to optimize management decisions in
the face of uncertainty,
using learning at one stage to influence
decisions at subsequent stages,
while considering the anticipated learning in
the optimization.
Final considerations
Other disciplines are kicking our butts
Standardization can be harmful at times
Remember scale…
New (?) techniques for analysis are
emerging
Bayesian methods
Heirarchical method (account for spatial
dependancy)