Landscape Disturbance Succession Modeling: Linking comprehensive ecosystem simulations for managemen

1
Landscape Disturbance Succession Modeling
Linking comprehensive ecosystem simulations for
management applications

• Bob Keane,
• USDA Forest Service,
• Rocky Mountain Research Station,
• Missoula Fire Sciences Laboratory,
• Missoula, MT

2
Guiding philosophy

• All models are WRONG,
• but some are useful
• George Box

Relative Comparison vs Absolute Answers
3
Landscape Disturbance Succession Modeling

• Simulation of the interaction of disturbances
  with vegetation development in a spatial domain

Spatially Explicit vs Spatial
4
Why build landscape disturbance succession models
(LDSMs)?

• Research
• Understand and explore ecosystem dynamics
• Management
• Compare alternatives
• Compute effects
• Plan activities
• Allocate resources

5
History of landscape disturbance succession models

• Started in 1980s
• Needed other technological advances
• GIS
• Computers
• Remote sensing

6
Reality of landscape disturbance succession
models

• Not prognostic or predictive
• Not well suited for individual events
• Simulate regimes
• Data intensive
• Computer intensive

7
Model Approaches

• Stochastic
• Empirical
• Mechanistic

8
LDSM simulation approaches
stand watershed region
year decades centuries

• Mechanistic
• Empirical
• Stochastic

Approaches often overlap and the best models
probably contain elements of all these approaches.
9
Uses of landscape disturbance succession models.
stand watershed region
Evaluate risk
year
Schedule treatments
decades
Provide targets
centuries
10
Uses of landscape disturbance succession models.
stand watershed region
Prescribe
Locate
Prioritize/ Allocate resources
year
decades
centuries
11
Stand
Watershed
Region
FOFEM
FARSITE
NWS Models
Consume Burnup
Year
Flammap
GCMs
SIMPPLLE FETM
FVS
Decades
CRBSUM
RMLANDS
Gap Models
Centuries
MAPPS
Fire-BGC
FOREST-BGC
DGVMs
LANDSUM
12
LDSM ConundrumDelicate balance between realism
and reality

• Detailed, Complex
• Hard to interpret
• Computationally demanding
• Difficult to parameterize
• Difficult to initialize

• Simple, Efficient
• Inadequate ecological representation
• Limited applications
• Not robust
• Inaccurate
• Insensitive

13
Landscape Processes
Seed abundance
Climate
and dispersal
Weather
Fire
Land use
Landscape
Patch
Level
Dynamics
Insect/
Disease
Hydrology
Pollution/
Fire effects
Deposition
Decomposition
Soil and
Stand and
fuel moisture
Plant Level
Carbon/nutrient
cycling
Photosynthesis
respiration
Stand
Understory
Evapotranspiration
development
dynamics
14
Landscape-Level Model ComponentsClimate and
Weather

• Basic driving variables common to all other model
  components
• Long term weather stream
• Daily measurements
• Mountain weather generator
• Major weather variables
• Temp (max, min), humidity, radiation,
  precipitation, wind

15
Landscape-Level Model ComponentsDisturbance
Wildland Fire

• Disturbance simulation broken into three
  processes
• Initiation
• Fire starts on landscape
• Spread
• Fire spread, rate, intensity, termination
• Effects
• Results of fire on ecosystem

Wildfire in Selway-Bitterroot
16
Landscape-Level Disturbance ComponentsFire
Ignition

• Difficult process to mechanistically model
  because detailed information needed at multiple
  scales
• Fire start dynamics
• Lightning, human
• Fuel bed receptivity
• Size, moisture content
• Ambient weather
• Temperature, wind

Lightning Storm
17
Landscape-Level Disturbance ComponentsFire
Behavior (spread)

• Most studied and modeled fire component
• Landscape simulation
• Independent of polygons
• Link to weather stream
• Wind, moisture content
• Provide appropriate fuels
• Describe fuel types and condition
• Some models available
• FARSITE, cell automata, etc.

Lethal fire in Douglas-fir
18
Landscape-Level Disturbance ComponentsFire
Behavior (spread)

• Several methods for simulating disturbance spread
• Cell automata/percolation
• Spread to surrounding pixels
• Mechanistic pixel
• Spread across pixels based on physical processes
• Mechanistic vector
• Vector based spread simulations (Hygens )

19
Stand Model ComponentsDisturbance Effects

• Important fire effects to include in simulation
• Fuel consumption
• Mechanistic/Empirical approach
• Plant mortality
• Empirical/stochastic approach
• Soil heating
• Heat transfer approach
• Smoke
• Linked to consumption
• Emission characteristics

Fire-scarring on lodgepole pine
20
Landscape-Level Model ComponentsSeed Dispersal
/ Species Migration

• Provide mechanism for species to move across
  landscapes
• Seed dispersal
• Include topography, wind, seed characteristics,
  animal, water effects
• Propagule survival
• Include vital attributes
• Seed crop frequency and amount

Whitebark pine cones
21
Landscape-Level Model ComponentsHydrology

• Provide for water routing within landscapes
• Many models available
• e.g. WEPP, TOPMODEL
• Important for specialized habitats
• Seeps, riparian areas, valley bottoms
• Important for aquatic processes
• Stream flow, water quality

McDonald Lake
22
Landscape-Level Model ComponentsInsects and
Diseases

• Provide for multiple disturbance applications
• Link to fire simulations
• Disturbance effects
• Important for long-term simulations
• Exotic diseases, insect epidemics
• Difficult to implement
• Limited process models available

Mountain pine beetle
23
Landscape-Level Model ComponentsOther Important
Components

• Inclusion based on simulation objective
• Land use
• Harvesting, settlement
• Pollution
• Deposition, ozone, etc.
• Exotic invasions
• Blister rust, weeds

24
Seed abundance
Climate
and dispersal
Weather
Fire
Land use
Landscape
Patch
Level
Dynamics
Insect/
Disease
Hydrology
Pollution/
Fire effects
Deposition
Decomposition
Soil and
Stand and
fuel moisture
Plant Level
Carbon/nutrient
cycling
Photosynthesis
respiration
Stand
Understory
Evapotranspiration
development
dynamics
Stand Processes
25
Stand-Level Model ComponentsSuccession Driver
Examples

• Simple, single variable models
• FIRESCAPE, SEM-LAND
• State and transition models
• SIMMPPLE, LANDSUM, RMLANDS
• Empirical stand models
• FVS, DFSIM
• Cohort models
• LANDIS, QLAND
• Traditional Gap-phase models
• JABOWA, SILVA, FIRESUM, FORET, LINKAGES, ZELIG
• Mechanistic Gap-phase
• FORSKA, FORSUM, FORCLIM, FORCYTE
• Mechanistic ecosystem models
• HYBRID, Fire-BGC, ECOPHYS, TREEDYN, TREE-BGC

Glacier NP 1967 fire
26
Succession DriverSimple, single variable models

• Use a single variable to represent all of
  vegetation development processes
• Most used variables
• Stand age
• Fuel accumulation
• Canopy cover
• Example models
• FIRESCAPE
• SEM-LAND

27
Succession DriverState and transition models

• Use linked vegetation classes (communities)
  along pathways of successional change driven by
  simulation time steps
• Other names
• Succession pathway diagrams
• Box models
• Example models
• SIMMPPLE
• LANDSUM
• RMLANDS

28
Succession DriverEmpirical stand growth models

• Use empirical data to drive growth in tree and
  stand attributes
• Other names
• Growth and Yield
• Yield tables
• Statistical growth models
• Individual tree growth models
• Example models
• FVS
• DF-SIM

29
Succession DriverCohort models

• Represent stand using fixed number of categories
  and simulate transition between these categories
• Common cohorts
• Diameter
• Species
• Height
• Cover
• Example models
• LANDIS
• QLANDS

30
Succession DriverTraditional gap-phase models

• Simulate individual tree diameter and height
  growth using general climate and soils drivers
• Other names
• Gap models
• Other simulations
• Mortality
• Regeneration
• Example models
• JABOWA, SILVA, FIRESUM, FORET, LINKAGES, ZELIG
• ZELIG, LAMOS,

31
Succession DriverMechanistic gap-phase models

• Gap models that mechanistically simulate
  ecological processes using physical relationships
• Other simulations
• Hydrology (water balance)
• Photosynthesis
• Respiration
• Decomposition
• Example models
• FORSKA, FORSUM, FORCLIM, FORCYTE
• HYBRID, FireBGCv2, ECOPHYS, TREEDYN, TREE-BGC

32
Succession DriverMechanistic ecosystem
simulation models

• Stand level models that mechanistically simulate
  ecological processes using physical relationships
• These models DO NOT simulate at tree level
• Other names
• Big leaf models
• Ecosystem dynamics models
• Other simulations
• Hydrology (water balance)
• Photosynthesis
• Respiration
• Decomposition
• Example models
• BIOME-BGC, FOREST-BGC, MAPPS, LPG

33
Landscape Disturbance Succession ModelsThree
Examples

• LANDSUM (Keane et al. 2001)
• LANDIS (Mladenoff 1998)
• FireBGCv2 (Keane et al. 2008)

34
LANDSUMThumbnail description

• Stand-level model no plants
• State and transition succession driver
• Stochastic disturbance initiation simulation
• Independent fire growth simulation
• Stochastic simulation of effects
• Parsimonious

35
LANDSUM Fire Simulation

• No fuels or weather inputs
• Rothermel (1991) spread
• Uses only slope and wind
• Fuels specified by fire probabilities
• Pixel to pixel spread
• Simulates fire independent of polygon boundaries

Fires
36
LANDSUM Applications

• Describe HRV landscape patch dynamics
• Compare management alternatives
• Explore advantages and disadvantages of
  simulation approach

37
LANDISThumbnail description

• Cross between a gap and vital attributes model
• Tracks presence and absence of species age
  cohorts at decade time steps
• Contains many disturbances
• Contains seed dispersal
• No dynamic fuel simulation
• Stochastic ignition simulation
• Independent fire growth simulation using
  percolation model
• Generalized simulation of effects
• Management oriented

38
LANDIS Fire Simulation

• Ignition is stochastic from a fire cycle
  probability distribution
• Fine fuel keyed to vegetation type
• Coarse fuels from tree mortality
• Spread to surrounding pixels based on fuel
  rating, slope, wind vectors
• Pixel to pixel spread
• Classification of fire effects

39
LANDIS Applications

• Describe HRV landscape patch dynamics
• Compare management alternatives
• Prioritize large regions
• Explore advantages and disadvantages of
  simulation approach

40
FireBGCv2Thumbnail description

• Highly complex, mechanistic process model
• Tree-level simulation
• Multiscale implementation
• Integration of many ecosystem submodels
• Ecosystem simulation
• Merger of BGC and FIRESUM models

Glacier NP landscape Net Primary Productivity
41
FireBGCv2 Stand-Level Modeling Diagram
42
FireBGCv2 Applications

• Explore effects of climate change on fire,
  landscape, and vegetation dynamics
• Understand ecosystem response to climate and
  disturbance regimes

Fire-maintained shrubfield
43
LANDSUM vs FireBGCv2 Comparison
LANDSUM
Fire-BGCv2

• Slow
• High memory use
• More disk space
• Difficult to run
• Greater output detail
• Robust application
• Greater understanding
• Highly deterministic

• Fast
• Low memory use
• Less disk space
• Easy to run
• Limited output
• Limited applications
• Little cause-effect
• Highly stochastic

44
LDSM Simulation Problems

• Changing map resolution
• Fire, climate, vegetation linkage
• Fire regime simulation
• Stochastic effects
• Initialization influences
• Landscape shape effects
• Parameterization problems
• Landscape size dilemma
• Output interval quandary
• Validation concerns

45
SummaryLandscape Disturbance Succession Modeling

• Many landscape disturbance succession models
  available and many more will be developed
• Development and selection of best model will
  always depend on simulation objective
• The perfect landscape disturbance succession
  model has not and probably will never be built
• Model selection and development is nearly always
  based on compromise between development cost,
  detail, expertise, input requirements, and time