Comparison of sensitivity of landscape-fire-succession models to variation in terrain, fuel pattern and climate

GCTE Task 2.2.2

Relationships between Global Change and Fire Effects at Landscape Scales

ContributorsGeoff Cary Australian National UniversityRobert Keane USDA Forest ServiceRobert Gardner Appalachian Laboratory, U. Maryland Sandra Lavorel Université Joseph Fourier, FranceMike Flannigan Canadian Forest ServiceIan Davies Australian National UniversityChao Li Canadian Forest ServiceJim Lenihan USDA Forest ServiceScott Rupp University of AlaskaFlorent Mouillot Carnegie Inst. of Washington, Stanford

Context

• Interactions between fire, climate and vegetation strongly influence landscape dynamics

• Simulation models are a critical tool for understanding these dynamics

• There has never been a uniform comparison that objectively evaluates model behaviour

Context

• Inherent difficulties in comparing models that were developed for a variety of purposes:

• Difficulty in obtaining validation data

• Difficulty in removing bias from the context in which the model was originally developed

Approach

• Unique approach to resolving these difficulties

• Standardise the variation in terrain, fuel pattern and climate

• Compare the sensitivity of model output to variation in terrain, fuel pattern and climate (i.e. not validation)

Models and native landscapes

EMBYR Yellowstone NP

FIRESCAPE South-east Australia

LAMOS-DS Generic (Corsica)

LANDSUM North-west US

SEM-LAND West-central Alberta

Models and native landscapes

• Lattice models

• Link fire ignition, fire spread, succession and fire effects

• Large landscapes, long duration

• Represent a variation in LFSM formulation

Experimental design

Treatment Levels Replicates

Terrain Flat 1 map/levelRollingMountainous

Fuel clumping Fine 10 maps/levelCoarse

Climate Observed 10 single yearsWarmer & wetter /levelWarmer & Drier

Terrain

Valley

Mid slope

Peak

50 km

Elevation range

Flat 1250 m

Rolling 625 – 1875 m

Mts 0 – 2500 m

Landscape position

Fuel clumping

Low

Moderate

High

50 km

Fine25 ha clumps

Fuel loads

Coarse625 ha clumps

Weather – Glacier National Park

1

1.5

2

2.5

3

9 10 11 12 13 14 15

Average daily maximum T (oC)

Ave

rage

dai

ly p

pt (

mm

)

weather repsall years

Climate scenarios

Temperature Precipitation

Warmer / Wetter + 3.6 oC x 1.2

Warmer/ + 3.6 oC x 0.8 Drier

Replication and simulations

Terrain (3 levels) 1 map / levelFuel clumping (2 levels) 10 maps / levelClimate (3 levels) 10 single years / level

3 x 2 x 10 x 3 x 10 = 1800 single year simulations / model

• Fires affected fuel load/age• Vegetation succession “removed”

Analysis

• Sensitivity of ln (area burned) to

TerrainFuel patternClimate changeInter-annual variation in climate

… and their interactions

• Variance in area burned explained (r2) by factors and interactions amongst them

• determined from fully factorial GLM (SAS)

Results

• Variation in climate explained considerable total variance for some models

Model Weather replicate r2

EMBYR 0.33*

FIRESCAPE 0.09*

LAMOS-DS 0.04*

LANDSUM 0.33*

SEM-LAND 0.54*

Climate

• Climate change explained considerable total variance in area burned for most models

Model Climate r2 Effect

EMBYR 0.03* (0.04) WW < WD = OB

FIRESCAPE 0.42* (0.46) OB < WW = WD

LAMOS-DS 0.28* (0.29) OB < WW = WD

LANDSUM 0.18* (0.27) OB < WW < WD

SEM-LAND 0.37* (0.81) OB < WW < WD

Fuel

• Fuel explained considerable total variance in area burned for EMBYR

Model Fuel r2 Effect

EMBYR 0.21* (0.32) Fine < Coarse

FIRESCAPE 0.02* Fine < Coarse

LAMOS-DS 0.00

LANDSUM 0.00* Fine < Coarse

SEM-LAND 0.01

Terrain

• Terrain explained considerable total variance in area burned for FIRESCAPE

Model Terrain r2 Effect

EMBYR 0.00FIRESCAPE 0.29* (0.32) Flat < Rolling < Mountainous

LAMOS-DS 0.00

LANDSUM 0.00

SEM-LAND 0.00

SummarySource No. models where: r2 > 5% Pr > F (0.001)

Terrain 1 1Fuel 1 4Terrain * fuelWeather replicate 4 5Terrain * weather replicateFuel * weather replicateTerrain * fuel * weather replicateClimate 4 5Terrain * climateFuel * climateTerrain * fuel * climateClimate * weather replicate 3 5Terrain * climate * weather replicateFuel * climate * weather replicateTerrain * fuel * climate * weather replicate

Conclusion

• Models largely sensitive to annual variability in weather

• Important to understand changes in climate variability

• Importance might diminish with particular changes in climate

Conclusion

• Models generally more sensitive to climate than fuel pattern or terrain

• Warmer/wetter and warmer/drier climates result in significant increases in area burned

• More important to understand climate trends and annual variability in weather, than fuel pattern or terrain, in explaining variation in area burned at the landscape scale

Conclusion

• Individual models sensitive to fuel pattern and weather because key processes represented in them.