Vous êtes sur la page 1sur 44

Myth meets Machine:

Using Individual-based Models to Test Common


Natural Resource Management Beliefs

Steve Railsback
Lang Railsback & Associates
and
Department of Mathematics,
Humboldt State University
Arcata, California, USA

Acknowledgements
Humboldt State University Department of
Mathematics, Environmental Modeling Graduate
Program

Software Developers
Steve Jackson

Colin Sheppard

Fish ecologists:
Bret Harvey & Jason White
US Forest Service

Today I will:
Introduce inSTREAM, an individual-based
simulation model of stream trout
Discuss what inSTREAM tells us about
4 common beliefs about fish and wildlife
managementare they myths?*
Draw conclusions about the value of complex
models in management ecology
*A fiction or half-truth, especially one that forms part of an ideology.
5
American Heritage Dictionary

The motivation:
Flow and temperature management of
hydropower dams
How do changes in
- Flow
- Temperature
- Channel shape
etc

affect river trout


populations?
6

The approach:
Individual-based modeling
Create a virtual river
Populate it with digital fish that
select good foraging habitat while competing
for food
grow
die
spawn
Population dynamics emerge from survival,
growth, and reproduction of individuals
7

inSTREAM (since 1999)

Belief 1:
Animals select habitat to
maximize growth

N. F. Hughes 1992
9

How We Model Growth


Food intake per fish:
food concentration velocity capture area.

Depth

Capture area:
Reactive
distance

10

10

30

Reactive distance:
Detection distance (cm)

Increases with fish size


Decreases with velocity

25
20
15
10
Observations
Model

5
0

10
15
Fish length (cm)

20

Reactive
distance

Capture success

Depth

0.8
0.6
0.4

Hill and Grossman (1993)


observations

0.2

Model

0
0.0

0.5

1.0

11
Ratio of cell velocity to fish maximum sustainable
11
swim speed

1.5

Testing the model: Can inSTREAM reproduce


patterns of habitat selection observed in real
trout, when individuals maximize growth?
Observed pattern

Response to high
flow
Hierarchical feeding
Competitorinduced shift
Predator-induced
shift
Higher velocity at
higher
temperatures
Response to
reduced food

Maximize
growth

12 of 60

What about avoiding risk?

Very few fish die of old age!!


13

Predation risk in inSTREAM


Large fish are eaten by birds, mammals
Risk is lower where fish are difficult to see
(deep, fast habitat)
Small fish are eaten by big fish
Risk is lower in shallow habitat
Risk decreases as fish grows
14

How to include risk avoidance in


modeling habitat selection?
State-based dynamic modeling literature:
select habitat that maximizes
future fitness:
Survival of predation
Survival of starvation

over a future time period


15

How can we include risk avoidance in


modeling habitat selection?
Fish assume that todays conditions persist over
90-day time horizon
Select the cell that maximizes expected
future fitness over the time horizon:
Survival of predation = f(fish size, depth, hiding cover)
Survival of starvation = f(growth rate, current condition)

16

Pattern-oriented test of theory for


habitat selection in trout
Observed pattern

Maximize
growth

Maximize survival

Maximize expected
future fitness

Response to high flow

Hierarchical feeding

Competitor-induced
shift

Predator-induced
shift

Higher velocity at
higher temperatures

Response to reduced
food

Railsback and Harvey. 2002. Ecology 83:1817-1830.

17

Belief 1: Do animals select habitat to


maximize growth?
To model adaptive behaviors such as habitat
selection, we must often assume individuals act
to maximize future fitness, which depends on:
Growth (or: not starving)
Current condition
Life stage (juvenile, adult, spawner)
Predation and other risks
Railsback and Harvey. 2013. Trait-mediated trophic interactions: is foraging theory keeping up?
Trends in Ecology & Evolution 28:119-125.

18

Belief 2: Trout populations are inherently


nocturnal or diurnal
Are trout species or populations
nocturnal or diurnal by nature or do
they adaptively switch between daytime
and night feeding?
Can we model
switching as an
adaptive behavior?
19

Modeling diel selection of habitat and activity


Evaluate each combination of day and night
activity for each habitat cell
Day

+ Night

Select the combination that provides highest


expected fitness over the time horizon
20

Additional Complexities
Growth and mortality risk vary with:
Day vs. night
Lower feeding success at night
Lower predation risk at night
Activity
Negative growth when hiding
Much lower predation risk when hiding
21

Does our model of


diurnal vs. nocturnal feeding work?
Many patterns of diel activity selection in
trout have been observed

Are these patterns reproduced in the IBM?


Railsback, Harvey, et al. 2005. Tests of theory for diel variation in salmonid feeding activity
and habitat use. Ecology 86:947-959.
22

Observed Pattern (1)


More night feeding when temperature is low
Night feeding

100%
80%
60%
40%
20%
0%
0

10

15

Temperature, C

Reduced metabolism allows fish to meet energy


needs by feeding at night
23

Observed Pattern (2)


More daytime feeding when food is scarce

Daytime feeding

100%
80%
60%
40%
20%
0%
0%

100%

200%

300%

Food availability, percent of baseline

24

Belief 2: Are trout inherently nocturnal


or diurnal?
Field and laboratory studies indicate that
individual trout adaptively switch between
day and night feeding

We can model this switching as a fitnessseeking behavior that depends on habitat

25

Belief 3: Food limitation

Population

Is this conceptual model useful?

Food availability
Railsback, Harvey. 2011. Importance of fish behaviour in modelling conservation problems:
26
food limitation as an example. Journal of Fish Biology 79:16481662.

The simulation experiment

Green River, Utah


below
Flaming Gorge Dam
1.0 km reach
27

Population

The simulation experiment

Food availability
Increase food availability over wide range
7-year simulations with spawning
28

Model results: Food always limits!


(more food more trout)
0.24 trout / m2

29

Why this result??

Is it due to behavior?? Which behavior?


30

Activity selection is how model trout


convert food to survival

31

No! (in our model)


Instead:

Population

Is this conceptual model of


food limitation useful?

Food availability

Because of tradeoff behaviors,


any factor that affects growth or survival likely
has some effect on abundance

(Food is always important!)


32

Belief 4:
Selected habitat is good habitat
Fundamental assumption of
habitat selection modeling:
Habitat with high densities of animals is
high-quality habitat
This assumption is controversial
Railsback, Stauffer, Harvey. 2003. What can habitat preference models tell us? Tests using a
virtual trout population. Ecological Applications 13:1580-1594.
33

inSTREAM as a virtual laboratory


In the model,
we know exactly what habitat quality is
(the fitness measure that simulated trout
use to select cells)
we can observe habitat and trout
densities without error

34 of 60

Definition of a cells habitat quality


Fitness potential =
expected survival of starvation and
predation for next 90 days,
assuming no competition for cells
resources (food, feeding habitat)
35

Fitness potential

Do simulated densities predict


habitat quality? Age 2+ trout

Density

36

Fitness potential

Selected habitat is generally


high quality habitat, but...

Wide variation
in density at
highest habitat
quality

Density

37

Fitness potential

Many high-quality
cells with no fish

Density

38

Fitness potential

Do simulated densities predict


habitat quality? Age 1 trout

Occupied cells
with low quality

Density

39

Belief 4: Why is observed animal density


a poor predictor of fitness potential?
Competition for local resources
Unused habitat
Individual variation (good habitat for one adult
trout may not be so good for another)

Relations between resource availability and


individual fitness can be nonlinear
More food may have little fitness benefit to
each fish, but could support more fish
40

Conclusions about modeling


inSTREAM is a large, complex, uncertain model
But none of the analyses I presented required
precise, site-specific results

Often the process of building and applying a model


is illuminating by itself...
by forcing us to document, explore, and test our
assumptions and beliefs, and see the
inconsistencies

41

Conclusions for management ecology


These beliefs / myths are really are simple
(simplistic?) conceptual models
Individual-based simulation, with adaptive
behavior, is a good way to test such
conceptual models
Our experience suggests a different way
to think about animal populations...
42 of 60

Thinking about how changes in habitat


affect animal populations:
What do the animals eat
(where, when, how)?
What eats the animals (where, how)?

What behaviors do the animals use to


trade off growth and risk?
43 of 60

More information:
www.humboldt.edu/ecomodel

44

Vous aimerez peut-être aussi