BEHAVIORAL CHARACTERISTICS BEHAVIORAL CHARACTERISTICS OF LANDSCAPE STRUCTURE METRICS OF LANDSCAPE STRUCTURE METRICS

IN NEUTRAL LANDSCAPESIN NEUTRAL LANDSCAPES

FRAGSTATS WorkshopFRAGSTATS Workshop18, July 200318, July 2003IALE World Congress IALE World Congress Darwin, AustraliaDarwin, Australia

Step 1: Generate binary neutral landscapes using the Step 1: Generate binary neutral landscapes using the computer program RULE (computer program RULE (Gardner 1999Gardner 1999).).

256 x 256 cell grids256 x 256 cell grids

5%5%

Factorial Design Factorial Design H (n = 21) x P (n = 19)H (n = 21) x P (n = 19)

100 replicates of each of 399 100 replicates of each of 399 H x P combinationsH x P combinations

11Increasing aggregation (H)Increasing aggregation (H)00

IncreasingIncreasingarea (P)area (P)

95%95%

FRAGSTATS Specifications:FRAGSTATS Specifications:• 30 m cell size30 m cell size

• 90 m edge depth90 m edge depth

• 500 m search radius500 m search radius

• 8 cell neighbor rule8 cell neighbor rule

• No borderNo border

• No backgroundNo background

• Boundary not included as edgeBoundary not included as edge

Step 2: Calculate 55 applicable class-level metrics Step 2: Calculate 55 applicable class-level metrics on all 39,900 neutral landscapes using on all 39,900 neutral landscapes using FRAGSTATS.FRAGSTATS.

Conceptual Metric ClassificationConceptual Metric ClassificationArea/Edge/Area/Edge/

DensityDensityClass AreaClass AreaPercent of LandscapePercent of LandscapePatch DensityPatch DensityEdge DensityEdge DensityLandscape Shape IndexLandscape Shape IndexLargest Patch IndexLargest Patch IndexNormalized Shape IndexNormalized Shape IndexPatch Area*Patch Area*Radius of Gyration*Radius of Gyration*

ShapeShapePerimeter Area Perimeter Area

FractalFractalPerimeter Area Perimeter Area

Ratio*Ratio*Shape Index*Shape Index*Fractal Dimension Fractal Dimension

Index*Index*

Core AreaCore AreaTotal Core AreaTotal Core AreaCore Percent of Core Percent of

LandscapeLandscape# Disjunct Core Areas# Disjunct Core AreasDisjunct Core Area Disjunct Core Area

DensityDensityCore AreaCore AreaDisjunct Core Area*Disjunct Core Area*Core Area Index*Core Area Index*

Isolation/Isolation/ProximityProximity

Proximity Index*Proximity Index*Similarity Index*Similarity Index*Euclidean Nearest Euclidean Nearest

Neighbor*Neighbor*

Contagion/Contagion/InterspersionInterspersion

Percent Like AdjacenciesPercent Like AdjacenciesClumpiness IndexClumpiness IndexAggregation IndexAggregation IndexIntersperson and Intersperson and

Juxtapostion IndexJuxtapostion IndexLandscape DivisionLandscape DivisionSplitting IndexSplitting IndexEffective Mesh SizeEffective Mesh Size

ConnectivityConnectivityPatch Cohesion IndexPatch Cohesion Index

ContrastContrastContrast Weighted Edge Contrast Weighted Edge

DensityDensityTotal Edge Contrast IndexTotal Edge Contrast IndexEdge Contrast*Edge Contrast*

Conceptual Metric ClassificationConceptual Metric ClassificationArea/Edge/Area/Edge/DensityDensityCACAPLANDPLANDPDPDEDEDLSILSILPILPInLSInLSIAREA*AREA*GYRATE*GYRATE*

ShapeShapePAFRACPAFRACPARA*PARA*SHAPE*SHAPE*FRAC*FRAC*

Core AreaCore AreaTCATCACPLANDCPLANDNDCANDCADCADDCADCORE*CORE*DCORE*DCORE*CAI*CAI*

Isolation/Isolation/ProximityProximityPROX*PROX*SIMI*SIMI*ENN*ENN*

Contagion/Contagion/InterspersionInterspersionPLADJPLADJCLUMPYCLUMPYAIAIIJIIJIDIVISIONDIVISIONSPLITSPLITMESHMESH

ConnectivityConnectivityCOHESIONCOHESION

ContrastContrastCWEDCWEDTECITECIECON*ECON*

Metric BehaviorMetric Behavior

P

H

• Calculate metrics in landscapes from three geographically distinct regions in Calculate metrics in landscapes from three geographically distinct regions in the United States:the United States:– Idaho (221 landscapes, 5 classes)Idaho (221 landscapes, 5 classes)

– Western Massachusetts (155 landscapes, 7 classes)Western Massachusetts (155 landscapes, 7 classes)

– Colorado (152 landscapes, 4 classes)Colorado (152 landscapes, 4 classes)

• Superimpose values from real landscapes onto values from neutral landscapes.Superimpose values from real landscapes onto values from neutral landscapes.

Step 4: Plot the range of the H x P space that real Step 4: Plot the range of the H x P space that real landscapes occupylandscapes occupy

Metric BehaviorMetric Behavior

P

H

• Use cluster analysis to classify metrics based on Use cluster analysis to classify metrics based on behavior along P and H gradients.behavior along P and H gradients.

• Graphically compare behavior of metrics.Graphically compare behavior of metrics.

Step 5: Evaluate patterns of class-level metric Step 5: Evaluate patterns of class-level metric behavior in using mean metric values for 48 behavior in using mean metric values for 48 metrics.metrics.

Primarily a Function of PPrimarily a Function of P

AREA_AMAREA_AMAREA_SDAREA_SDGYRATE_AMGYRATE_AMGYRATE_SDGYRATE_SDCORE_AMCORE_AMCORE_SDCORE_SDTCATCADCORE_AMDCORE_AMDCORE_SDDCORE_SDPROX_MNPROX_MNPROX_CVPROX_CVPROX_SDPROX_SDDIVISIONDIVISIONMESHMESH

PH

LPILPI

Primarily a Function of H:Primarily a Function of H:Strongly Related to HStrongly Related to H

nLSInLSIPARA_SDPARA_SDFRAC_CVFRAC_CVFRAC_SDFRAC_SDCAI_SDCAI_SDCLUMPYCLUMPY

PAFRACPAFRAC

CLUMPYCLUMPY

Related to Interaction of P and H Related to Interaction of P and H Parabolic Parabolic Response Along PResponse Along P

LSILSIPDPDGYRATE_CVGYRATE_CVFRAC_AMFRAC_AMSHAPE_AMSHAPE_AMSHAPE_CVSHAPE_CVSHAPE_SDSHAPE_SD PROX_AMPROX_AMDCORE_CVDCORE_CVDCADDCAD

EDED

DCADDCAD

SHAPE_SDSHAPE_SD

FRAC_AMFRAC_AM

Related to Interaction of P and HRelated to Interaction of P and H

GYRATE_MNGYRATE_MNPARA_AMPARA_AMCORE_MNCORE_MNDCORE_MNDCORE_MNCAI_AMCAI_AMCAI_MNCAI_MNSPLITSPLITPLADJPLADJAIAICOHESIONCOHESIONENN_AMENN_AMENN_MNENN_MNENN_CVENN_CVENN_SDENN_SD

AREA_MNAREA_MN

COHESIONCOHESION

ENN_AMENN_AMGYRATE_MNGYRATE_MN

PARA_AMPARA_AM

Differential Metric SensitivityDifferential Metric Sensitivity

AREA_MNAREA_MN

AREA_AMAREA_AM

P = 5%P = 50%P = 95%

H = 0

P = 5%P = 50%P = 95%

H = 1

**Limitations****Limitations**

• Results are based on Results are based on

– binary neutral landscapes.binary neutral landscapes.

– at one scale.at one scale.

– only one configuration gradient (H). Varying shape, inter-patch only one configuration gradient (H). Varying shape, inter-patch distance, etc. would yield different behavior.distance, etc. would yield different behavior.

• Identified 7 behavioral groups with varying relationships Identified 7 behavioral groups with varying relationships with P and H.with P and H.

• Conceptual similarity Conceptual similarity ≠ ≠ behavioral similarity.behavioral similarity.• Many metrics have non-linear behavior and lack of Many metrics have non-linear behavior and lack of

sensitivity in at least part of the H x P space.sensitivity in at least part of the H x P space.– Problematic conditions do not always exist in real landscapes.Problematic conditions do not always exist in real landscapes.

• Very few metrics measure configuration independent of area Very few metrics measure configuration independent of area – most confound P & H.– most confound P & H.

Main PointsMain Points