Integrating plant and animalprocesses in grazing system models

Peter J. WeisbergUniversity of Nevada, Reno

How many ungulates should there be?

Is this a difficult question?

• Variable forage base

• Variable intake rate

Stocking Rate = (Production * Allowable Use) / Animal Intake

Sources of Variation

• Weather

• Soils

• Topography

• Hydrology

• Grazing Behavior

• Population Management

• Grazing Management

Savanna Model, M. Coughenour, NREL

1000

1200

1400

1600

1800

2000

2000

4000

6000

8000

10000

12000

Elk Population Level

Elk

Inta

ke

(kg/

elk/

year

)

Dry Year

Wet Year

DeterministicModel

1000

1200

1400

1600

1800

2000

2000

4000

6000

8000

10000

12000

Elk Population Level

Elk

Inta

ke

(kg/

elk/

year

)

StochasticWeather

Why Model Grazing Systems?

• Aid understanding – complex problems

fuzzy notions -> Data + Algorithms

• Evaluate logical outcomes of our beliefs

Mule Deer Decline

Coyote Predation Winter Habitat Loss

Shrub SenescenceCompetition from Elk

? ?

? ?

1935 1945 1955 1965 1975 1985 1995 2005

Coyote H0

Habitat Loss H0

Coyote * Habitat Loss

OBSERVED

1935 1945 1955 1965 1975 1985 1995 2005

1935 1945 1955 1965 1975 1985 1995 2005

1935 1945 1955 1965 1975 1985 1995 2005

HYPOTHETICAL SIMULATIONS

Why Model Grazing Systems?

• Aid understanding – complex problemsfuzzy notions -> Data + Algorithms

• Evaluate logical outcomes of our beliefs

• Scenario modeling for assessingmanagement alternatives (forecasting)

• Knowledge transfer– SCIENTIST MANAGER

• Why model grazing systems?

• Types of grazing system models

• Spatial heterogeneity and scale

• Management models

• Where do we go from here?

Energetics

Population ForageIntake

Distribution

AvailableForage

AvailableEnergy

HabitatMosaic

Energetics

Population ForageIntake

Distribution

AvailableForage

AvailableEnergy

Energetics

Population ForageIntake

Distribution

AvailableForage

AvailableEnergy

HabitatMosaicHabitatMosaic

Animal Model

Plant Model

Production

PopulationAllocation

PropaguleDispersal

Solar Energy

BiomassRemoval Production

PopulationAllocation

PropaguleDispersal

Solar Energy

BiomassRemoval Production

PopulationAllocation

PropaguleDispersal

Solar Energy

BiomassRemoval

Integrated Model

Energetics

Population

Distribution

Productionand

Allocation

Population

Dispersal

HERBIVORY

Solar Energy

Habitat

Energetics

Population

Distribution

Productionand

Allocation

Population

Dispersal

HERBIVORY

Solar EnergySolar Energy

Habitat

Integrated Grazing Models andSpatial Heterogeneity

• Animal distribution, seeddispersal

• Dynamic climate variablesas drivers

• Climate interpolated overa spatially heterogeneouslandscape

• Linkages to underlyingGIS (soils, vegetation,topography)

Mean Elk Density,Feb. 1994 - 99

Aspen RegenerationProbability,

Landscape ScaleAll Winter Jan.-Feb.

Modeling HerbivoreDistribution Patterns-Difficult to validate-Time-dependent-Dependent on factors otherthan forage

(Weisberg and Coughenour, 2003

Larsen, G. 1988.The Far Side

When grazing system models areintegrative and spatially-explicit:

• We improve our ability to make predictionsabout the real world, IF– Good data– Understanding of how processes and data

translate across spatial and temporal scales

Herbivore Vegetation:-one bite at a time-effects may amplify overlarge areas and long timeperiods

Vegetation Herbivore:-influences over multiple scales-coarse scale effects constrainfiner scale herbivore processes

ShRSRI

MAX

MAX

+=

I=f(B,D,CI, Snow, …)

SaplingModel

GapModel

New Tree Establishment

Light Availability,Overstory Composition

HerbivoreModel

Browsing

Forage Availability

Habitat Selection:Intermediate time stepsCoarse spatial resolution

Forage Intake andPlant ResponseShort time stepsFine spatial resolution

Forest Gap DynamicsLong time stepsIntermediate resolution

Linking models across scales

SaplingPhysiology and

Growth

SaplingMortality

ForestForestRegenerationRegenerationBrowsing

Temperature

Light

Small-SeedlingEstablishment

Winter Browsing of Current Annual Growth (distributed over 6-months)

Light

Browsing

Browsing HH HH

HH

HH

Years to Exceed 150 cm

Picea abies

SaplingModel

GapModel

New Tree Establishment

Light Availability,Overstory Composition

HerbivoreModel

Browsing

Forage Availability

Linking models across scales

• We are working in the realm of limited theory,limited observations (often at the wrong scales)

When grazing system models areintegrative and spatially-explicit:

DATA

UNDERSTANDINGAfter Holling 1978, fromStarfield and Bleloch 1986

1

4

3

2

• Why model grazing systems?

• Types of grazing system models

• Spatial heterogeneity and scale

• Management models

• Where do we go from here?

• Management models are especiallydifficult…– Focus is integration not abstraction

– Management decisions often made overheterogeneous landscape or regional units

– Objective is accurate forecasts

Modeller’schallenge

COMPLEX

SPATIALLYEXPLICIT PREDICTIVE

Management Model Requirements

• Accessibility and transparency

• Strong connections to real-world data

010203040506070

0 10 20 30 40 50

Predicted Live Herbaceous Biomass (g/sq m)

Obs

erve

d

Highly Productive

Moderately Productive

Poorly Productive

Pasture Quality Obs NDVI Sim NDVI

Model-Data Linkages

• Data collected overextensive areas

• Historical data tovalidate models forcurrent conditions

• Experimental,long-term controlledgrazing studies

We need more:

Management Model Requirements

• Accessibility and transparency

• Strong connections to real-world data

• Stakeholder and End-User interactionduring model development

Manager-Scientist Interaction in theModeling Process

Que

stio

nsD

ecid

e on

Out

puts

Mat

hem

atic

al M

odel

Dec

ide

onD

rivin

g Va

riabl

esC

once

ptua

lizat

ion

Link

ages

toEm

piric

al D

ata

Impl

emen

tatio

nM

odel

Tes

ting

Inte

rpre

t Res

ults

New

Site

s an

d Sp

ecie

sN

ew Q

uest

ions

Institutionalized Model-Making• Emphasizes process over

product

• Facilitates and structuresmanager-scientist-stakeholder interactions

• Integrates and organizesknowledge

• Guides research direction

• Placeholder for continuingdevelopment

Sage, Patten & Salmon. 2003. J. Nat. Cons. 10:280-294.

• “Process over Product”– Decision-maker involvement a must

– Requires transparency, “gaming”

OR• Accurate Predictions

– Elusive target

– Only useful product of “black box” models

How can Models SupportDecision Making?

ABSTRACTION– Isolate system

components ofparticular interest

– Assume homogeneity,equilibrium

– Incremental approach,limited scope

– Easier to do goodscience

INTEGRATION– Inclusive of relevant

system components;Focus on interactionsand linkages

– Incorporate & explaincontingency

– “reverse reductionism”– Fuzzy “answers” to

big-picture questions

For Management Models:Need cross-fertilization of the two approaches

Happy Holidays!!!