modeling the effects of dnr forest management … the effects of dnr forest management alternatives...
TRANSCRIPT
-
Modeling the Effects of DNR Forest Management Alternatives on Marbled Murrelets in Washington:
A Population Viability Analysis Approach
Zach Peery and Gavin JonesDepartment of Forest and Wildlife Ecology
University of Wisconsin-Madisoncc: Glenn Bartley
cc: Nicholas Hatch
-
There is no standard definition, most scientists favor a definition that explicitly requires a mathematical model.
We define PVA as an analysis that uses data in an analytical or simulation model to calculate the risk
of extinction or a closely related measure of population viability (Ralls et al. 2002)
What is a PVA?
Population Viability Analysis (PVA)
Why use a PVA? To help make management decisions about species of conservation concern.
-
Assessing extinction risk for a population or species Determine MVP needed to achieve the desired level of protection Informing population recovery - is the species still imperiled? Identifying key life stages or threats that should be managed (sensitivity
analyses) Evaluating risks or benefits associated with different management options
Common PVA Applications
cc: Bill and Dot Bell cc: Sheila Whitmore
-
Challenges to Strong Inference from PVAs Insufficient data to parameterize model Little information on species-environment relationships Little ability to test model-based predictions of risk
PVAs: Panacea or Wishful Thinking?
Panacea?Population Viability Analyses represent the flagship technology of the field of Conservation Biology.
Wishful Thinking?
Reality
Michael Soule (2002)
All models are wrong, but some are usefulBox and Draper (1987)
-
For example, Akakaya and Atwood (1997. Cons. Bio) ranked management options for California Gnatcatchers based on relative risk of extinction
Absolute projections of extinction risk are generally considered unreliable Assessment of risk should instead be made on a relative basis (e.g. among
management alternatives)
Management Option
Extin
ctio
n Ri
sk
Absolute vs Relative Risk
cc: PJ Rambler
-
Biology of Small Populations
Small populations are more likely to go extinct than large populations
Image Credit: http://www.randykinnick.com/wp-content/uploads/2011/09/Lost-Springs-Population.jpg
Richard Primack (2010) Essentials of Conservation Biology, 5th edition
Perc
enta
ge o
f Po
pula
tions
Per
sistin
g
-
Small Populations affected by both Deterministic and Stochastic Factors
Deterministic factors: factors that change population size in a relatively predictable manner such as habitat loss.
Stochastic factors: factors that result in less predictable changes in population size (e.g., genetics, random demography, weather, food supplies)
cc: Tom MacKenziehttp://www.fws.gov/caribbean/ES/Parrot-Gallery.html
Beissinger et al. 2009. Ecological Monographs.
Num
ber o
f Par
rots
Year
-
Deterministic and Stochastic Factors can Interact
0
2000
4000
6000
8000
10000
12000
0 10 20 30 40 50 60
Area of HabitatPopulation Abundance
PVAs can measure risk during the bottleneck (small population) phase
Area
or A
bund
ance
Years into Future
-
Expert Opinion Conservation Principles Rules of Thumb
Degree of ThreatLow High
High
Low
Unc
erta
inty
Uncertainty, Threat, and Assessment Method
Habitat Mapping
PVAMarbled Murrelets
cc: Glenn Bartley
-
Simplest PVAs are based only on numbers of individuals in a population
Need estimates of historic population growth rates, current population size, and effects of stochasticity
Use mathematical model (equations) to project simulated populations in forward in time
0
10
20
30
-10 -9 -8 -7 -6 -5 -4 -3 -2 -1
Nt
Year, t
N observed in pastPredicted future N
Time
Present
Population Viability AnalysisPV-what?Po
pula
tion
Size
(N)
Extinction probability:proportion of simulated populations where N becomes zero
Expected change in population size: average difference in beginning and ending N
Risk Metrics
Chart2
2022
18.4625.4865.486
152.4262.426
172.7332.733
1433
1044
1533
1155
1622
1355
Year, t
Nt
data
Survey TypeLand TypeYearCountSE
RadarMMCAs-1020.0002.000
RadarMMCAs-918.4625.486
RadarMMCAs-815.0002.426
RadarMMCAs-717.0002.733
-614.0003.000
-510.0004.000
-415.0003.000
-311.0005.000
-216.0002.000
-113.0005.000
0113
110.452
29.92753
39.4311253
48.959568752
58.51159031251
68.08601079692
77.6817102573
87.29762474422
96.9327435073
106.58610633162
data
22
5.4865.486
2.4262.426
2.7332.733
33
44
33
55
22
55
Year, t
Nt
MMCAs
Reserves
AV Count (SE)
1
1
1
1
1
1
1
-
Incorporating Environmental Effects with Model Parameters
How Do We Estimate Model Parameters?
Preferably with data!Change in fox abundance = x + x
Effect of rain on foxes
Coyote abundance
Effect of coyotes on foxes
Amount of rainfall
0100200300400500600700800
50 75 100 125 150 175 200
Coyote Abundance
Fox
Abun
danc
e
Cc: Peterson Moose
Modified from: Dennis et al. (2000) Journal of Wildlife Management
Wikipedia Commons
-
Applying PVA methods to Marbled Murreletsand Forest Management in Washington
Developed a demographic, metapopulation PVA model that projects murreletpopulations forward in time
Estimates risk under various forest management alternatives.
Basic model structure developed previously for a corvid-based murrelet PVA in California
-
A Meta-population Model
A metapopulation consists of a group of spatially separated subpopulations linked by the dispersal of individuals.
Dispersal of individuals influences local population dynamics
The murrelet PVA model assumes two simplified populations (DNR and non-DNR)
DNRsubpopulation
Individual dispersal
non-DNRsubpopulation
-
A Demographic PVA Model The model projects the WA murrelet population forward in time based on
demographic rates - reproductive and survival rates - within subpopulations
100Juveniles
100Adults
Adult survival rate = 0.90(90 adults survive and remain adults)
Year 2
Year 1
140Adults
Reproductive rate = 0.5 (50 juvs produced by
100 adults)
50Juveniles
140Adults
Year 1 Population = 200
Year 2 Population = 210
Quiz: how do we incorporate stochasticity?
-
Estimating Survival Rates
Non-juvenile survival rates estimated based on a mark-recapture study of murreletsin California (Peery et al. 2006)
Estimates ranged from ca. 0.87 to 0.90
Juvenile survival rates assumed to be 70% of non-juvenile survival rates (Beissinger 2005)
Half Moon Bay
Santa CruzAno
Nuevo Baycc: Zach Peery
cc: Zach Peery
cc: Zach Peery
-
Reproductive Rule Sets:Linking the Analytical and PVA Models
Habitat quality (6 Pstages) affects max nesting density - Platform density, canopy layers, stand origin, forest type
Increasing Max Nesting Density
Pstage: 0 0.25 0.36 0.47 0.62 and 0.89 1
For example, max nesting density in pstage 1 is 4x greater than in pstage 0.25
-
Max Nest Density
Soft: 0.60 x DHard: 0.25 x D
Edge conditions affect max nesting density
Reproductive Rule Sets:Linking the Analytical and PVA Models
OE
IE
IF
Soft: 0.80 x DHard: 0.625 x D
D
Interior Forest
-
Area of Nesting Habitat
Nes
ting
Carr
ying
Cap
acity
Slope = 1
Nesting carrying capacity = nesting habitat area x max nesting density (Pstage and edge)
Area = AArea = A
Reproductive Rule Sets:Linking the Analytical and PVA Models
Pstage = 0.47
Area = A
Carrying capacity = K
Nesting habitat area affects nesting carrying capacity in a 1 to 1 manner
Carrying capacity = 2 x K
Pstage = 0.47
Area = 2 x A
A
K
2A
2K
-
Relationship between MurreletsNumbers and Nesting Habitat
Raphael (2006). Conservation Biology.
Mea
n Co
unt
Amount of Habitat
Raphael et al. (2002). Condor.
Nesting Habitat (km2)
At se
a Po
pula
tion
Size
-
Nest Success
0.380
Edge conditions also affect nest success
Reproductive Rule Sets:Linking the Analytical and PVA Models
OE
IE
IF
0.465
0.550
Interior Forest
-
Habitat Conditions and the Landscape Scale Current habitat conditions were aggregated across each
landownership to determine:- Nesting carrying capacity - Nest success
Habitat conditions on DNR lands projected forward in time (50 years) using the Forest Vegetation Simulator
Assumed no change in habitat on non-DNR lands
cc: M. Hobson
-
ModelInputs Outputs
Initial Abundance(by stage class)
Carrying Capacity (number of nests)
DNR Lands
Non-DNR Lands
dispersal
Demography(survival, breeding,
dispersal)
RelativeExtinction Risk
RelativeChange in
Population Size
Nest Success
Habitat Quantity and Quality
(Pstage)
Edge Conditions
Habitat Population
Environmental Variability
A Conceptual Representation of the PVA Model
Habitat Quantity (acreage)
-
Demographic Submodel
-
New breeders do not preferentially select high-quality habitat
Breeders stay in the same landownership unless they are displaced by habitat loss
Displaced breeders become nonbreeders for at least one year
Displaced breeders become breeders again if nesting habitat becomes available
Some Additional Model Rules and Assumptions
cc: Tom Johnson
-
Matching the Model to RealityThe Reality
At-sea monitoring indicates ~5% annual declines in WA from 2001 to 2014
0
3000
6000
9000
12000
15000
2000 2005 2010 2015 2020 2025 2030
Mur
rele
tPop
ulat
ion
Size
in W
ashi
ngto
n
The Problem
Using values for survival and reproductive rates that yielded 5% declines resulted in little ability for recruits to fill into potential new nesting habitat
Future HabitatModeled Future Population
The Solution
Conduct parallel Population Risk and Enhancement analyses with different capacities for murrelets to fill into new nesting habitat
Falxa et al. (2014), Lance and Pearson (2015)
-
Risk vs Enhancement AnalysesEnhancement analysis: How do the alternatives differ in ability to
enhance regional (WA) and local (DNR) murrelet populations?
Assumes nesting habitat loss is the primary factor cause of murrelet population declines
Uses a relatively optimistic values for adult survival (0.90)
Greater capacity for new recruits to fill into new nesting habitat
Assumes number of breeders > nesting carrying capacity
Risk analysis:
How do the alternatives differ in their effects on risk to regional (WA) and local (DNR) murrelet populations?
Assumes both nesting habitat loss andchronic environmental stressors caused murrelet population declines
Uses a relatively pessimistic values for adult survival (0.87)
Less capacity for new recruits to fill into new nesting habitat
Assumes number of breeders > nesting carrying capacity
-
Annu
al p
opul
atio
n gr
owth
rate
0.9
0.91
0.92
0.93
0.94
0.95
0.96
0.97
0.98
0.99
1
1.01
1.02
0 10 20 30 40 50
Risk Analysis
Stable
Enhancement Analysis
0.9
0.91
0.92
0.93
0.94
0.95
0.96
0.97
0.98
0.99
1
1.01
1.02
0 10 20 30 40 50
Years into future
Stable
Deterministic Expectations under the Risk and Enhancement Analyses
Years into futureAn
nual
pop
ulat
ion
grow
th ra
te
-
Incorporating Stochasticity
0
2000
4000
6000
8000
10000
12000
14000
16000
2000 2002 2004 2006 2008 2010 2012 2014
Estimated amount of annual variation in population size from at-sea monitoring
Used this variation to determine how much survival and reproductive rates should vary from year to year
Initial population size = avg from
2010-2014
Mur
rele
tPop
ulat
ion
Size
in W
ashi
ngto
n
Falxa et al. (2014), Lance and Pearson (2015)
-
or = 0(Regional) (DNR)
Parameter Risk EnhancementAnnual non-juvenile survival rate 0.87 0.90Annual juvenile survival 0.70 x non-juvenileAnnual dispersal rate
DNR non-DNR = 0.91non-DNR DNR = 0.09
WA:DNR non-DNR = 0.91non-DNR DNR = 0.09
DNR: 0Initial female population size DNR: 311
non-DNR: 3,129
Initial nesting carrying capacity 40% > Initial number of females of breeding age
Variance in reproductive rates 0.012Variance in survival rates 0.003
Parameters Used (non-reproductive)
-
0.500
0.505
0.510
0.515
0.520
0.525
0.530
0 5 10 15 20 25 30 35 40 45 50 -
50
100
150
200
250
300
350
400
0 5 10 15 20 25 30 35 40 45 50
Management Alternatives Considered(DNR-managed Lands)
Num
ber o
f Nes
ts
Nesting Carrying Capacity
Nes
t Suc
cess
Rat
e
Nest Success
Years into the Future
-
Risk Analysis State of Washington
-
Risk Analysis DNR Lands Only
-
Enhancement Analysis State of Washington
-
Enhancement Analysis DNR Lands Only (no dispersal)
-
Proposed Sensitivity Analyses
What are the most important habitat conditions?
- Habitat amount- Habitat quality- Edge-interior configuration
How sensitive are result to model uncertainties?
What are the consequences of less nesting habitat development than predicted?
(for illustration purpose only, analyses in progress)
Incr
ease
in R
isk
-
Preliminary Thoughts on PVA Results
Differences in state-level risk were small among the four alternatives considered
While differences were small, Alternative E reduced state-level risk the most
The ability to contribute to state-level enhancement was very similar among alternatives
Alternative E led to somewhat more murreletson DNR lands than other alternatives under the enhancement analyses
Greater risk was estimated for the no change scenario than for other alternatives under some assumptions and at some scales
-
Next Steps
Model Alternatives C and D Formalize and conduct
sensitivity analyses Write report and
manuscript for publication Peer review both
documents
Modeling the Effects of DNR Forest Management Alternatives on Marbled Murrelets in Washington: A Population Viability Analysis ApproachSlide Number 2Slide Number 3Slide Number 4Slide Number 5Biology of Small PopulationsSlide Number 7Deterministic and Stochastic Factors can InteractSlide Number 9Population Viability AnalysisIncorporating Environmental Effects with Model ParametersApplying PVA methods to Marbled Murrelets and Forest Management in WashingtonA Meta-population ModelA Demographic PVA ModelEstimating Survival RatesReproductive Rule Sets:Linking the Analytical and PVA ModelsReproductive Rule Sets:Linking the Analytical and PVA ModelsReproductive Rule Sets:Linking the Analytical and PVA ModelsSlide Number 19Reproductive Rule Sets:Linking the Analytical and PVA ModelsHabitat Conditions and the Landscape ScaleSlide Number 22Slide Number 23Some Additional Model Rules and AssumptionsMatching the Model to RealityRisk vs Enhancement AnalysesSlide Number 27Incorporating StochasticityParameters Used (non-reproductive)Management Alternatives Considered(DNR-managed Lands)Slide Number 31Slide Number 32Slide Number 33Slide Number 34Proposed Sensitivity AnalysesPreliminary Thoughts on PVA ResultsNext Steps