Comparison of the New Jersey Landscape Alison Burnett Mark June-Wells December 13, 2006.
-
Upload
wilfred-summers -
Category
Documents
-
view
214 -
download
1
Transcript of Comparison of the New Jersey Landscape Alison Burnett Mark June-Wells December 13, 2006.
Comparison of the Comparison of the New Jersey Landscape New Jersey Landscape
Alison BurnettAlison BurnettMark June-WellsMark June-Wells
December 13, 2006December 13, 2006
QuestionsQuestions
How do multi-fractal, random, and a NJ How do multi-fractal, random, and a NJ map relate at the landscape scale?map relate at the landscape scale?
In what landscape characteristics does In what landscape characteristics does the NJ map differ from the two random the NJ map differ from the two random maps?maps?
What can class-level indices tell us What can class-level indices tell us about the spatial distribution of NJ state about the spatial distribution of NJ state patches vs. random patch distribution.patches vs. random patch distribution.
HypothesisHypothesis NJ State map will differ significantly NJ State map will differ significantly
from both random and multi-fractal from both random and multi-fractal random maps for the measured random maps for the measured indices at the class and landscape indices at the class and landscape levels.levels.
ConsiderationsConsiderations
Equal probability inputs for random Equal probability inputs for random map classes…Nature is not random.map classes…Nature is not random.
Shape of NJ State vs. rectangle.Shape of NJ State vs. rectangle.
M&M ProgramsM&M ProgramsRULERULE
RULE:RULE: Generate a series of random maps.Generate a series of random maps. Rows = 274; Columns = 148Rows = 274; Columns = 148 34 classes – probability for each = .02934 classes – probability for each = .029 Nearest Neighbor Rule 1 chosenNearest Neighbor Rule 1 chosen No analysis…output as ASCIINo analysis…output as ASCII Multifractal Map was created using 8 Multifractal Map was created using 8
levels and an h=.5levels and an h=.5
M&M ProgramsM&M ProgramsArcArc
Converted all ascii files or NJ, Converted all ascii files or NJ, Random, and Multifractal maps to Random, and Multifractal maps to grid files for analysis.grid files for analysis. I love ARC!!!!I love ARC!!!!
M&M ProgramsM&M ProgramsFragstatsFragstats
Fragstats:Fragstats: Conducted basic analysis at the class Conducted basic analysis at the class
and landscape levels with a broad and landscape levels with a broad assemblage of indices.assemblage of indices.
Neighbor Rule of 4Neighbor Rule of 4 Parameters were chosen based on Parameters were chosen based on
simplicity and efficacy of design.simplicity and efficacy of design.
Fragstats ParametersFragstats ParametersLandscapeLandscape
Edge DensityEdge Density FRAC meanFRAC mean FRAC standard FRAC standard
deviationdeviation CONTIG meanCONTIG mean CONTIG std. dev.CONTIG std. dev.
PROX meanPROX mean PROX std. dev.PROX std. dev. COHESIONCOHESION
Fragstats Parameters Fragstats Parameters ClassClass
NPNP PDPD TETE PAFRACPAFRAC
PROX weighted meanPROX weighted mean PROX standard PROX standard
deviationdeviation CLUMPYCLUMPY COHESIONCOHESION
M&M ProgramsM&M ProgramsArcViewArcView
ArcView3: Made maps for visual purposes
Results- Landscape MetricsResults- Landscape Metrics
Map Type Edge Density
NJ 4.9
Random 19538.6
Multi-fractal 17821.8
NJ map significantly less than created NJ map significantly less than created mapsmaps
Expected result for NJExpected result for NJ Multi-fractal slightly less than randomMulti-fractal slightly less than random
Results- Landscape MetricsResults- Landscape Metrics
Map TypeFractal Dimension
mean std. dev.
NJ 1.0193 0.0316
Random 1.0626 0.2232
Multi-fractal 1.0745 0.1912
Subsets of geometrical space within which Subsets of geometrical space within which they residethey reside
Random/multi-fractal maps more complexRandom/multi-fractal maps more complex
Results- Landscape MetricsResults- Landscape Metrics
Map TypeContig
mean std. dev.
NJ 0.1221 0.1659
Random 0.0143 0.0445
Multi-fractal 0.0403 0.0778
High=clumped; low=disperseHigh=clumped; low=disperse Expected that NJ by more clumped Expected that NJ by more clumped Random least clumpedRandom least clumped
Results- Landscape MetricsResults- Landscape Metrics
Map TypeProximity
mean std. dev.
NJ 23.835 76.293
Random 0.695 0.209
Multi-fractal 0.991 0.474
Lower= more isolatedLower= more isolated Higher=more connectedHigher=more connected Also expected from NJ mapAlso expected from NJ map
Results- Landscape MetricsResults- Landscape Metrics
Map Type Cohesion
NJ 95.1672
Random 8.1095
Multi-fractal 19.9875
NJ’s classes are very connectedNJ’s classes are very connected Multi-fractals classes still more connected Multi-fractals classes still more connected
than randomthan random
Results- Class MetricsResults- Class MetricsRandom and Multi-fractalRandom and Multi-fractal
GeneralGeneral Random and multi-fractal maps show similar Random and multi-fractal maps show similar
values for each index across all patchesvalues for each index across all patches Random maps, 34 patch types, all the same Random maps, 34 patch types, all the same
probability probability As expected multi-fractal maps show higher As expected multi-fractal maps show higher
clumpiness and cohesion than random mapsclumpiness and cohesion than random maps Random map “total edge” values more Random map “total edge” values more
contiguous throughout class typescontiguous throughout class types ““Patch density” same for both maps, slightly Patch density” same for both maps, slightly
more evenly distributed for randommore evenly distributed for random ““Number of patches” for random map almost Number of patches” for random map almost
all the same; more variation in multi-fractalall the same; more variation in multi-fractal
Results- Class Metrics Results- Class Metrics NJ MapNJ Map
Patch Density significantly lower than Patch Density significantly lower than created mapscreated maps
Greater variation in total edge as Greater variation in total edge as compared to created mapcompared to created map
Proximity mean overall greater, showing Proximity mean overall greater, showing more connectedness of classesmore connectedness of classes
Fractal dimension Fractal dimension NOTNOT significantly significantly different across mapsdifferent across maps
Patch cohesion variable “could be” highly Patch cohesion variable “could be” highly correlated to patch typecorrelated to patch type Moderately developed- 95.426Moderately developed- 95.426 Estuarine marsh- 92.4115Estuarine marsh- 92.4115 Cultivated- 94.5218Cultivated- 94.5218 Golf courses- 7.933Golf courses- 7.933
ConclusionsConclusions
NJ map is significantly different than NJ map is significantly different than both random and multi-fractal maps both random and multi-fractal maps in class and landscape analysisin class and landscape analysis
Further, more rigorous investigation, Further, more rigorous investigation, would require producing a wider would require producing a wider variety of map types and imputing variety of map types and imputing real-life patch probabilities.real-life patch probabilities.
AcknowledgementsAcknowledgements
GIS lab people who let us in on GIS lab people who let us in on SaturdaySaturday
Software companiesSoftware companies FragstatsFragstats RuleRule ArcArc ArcView3ArcView3
Questions?Questions?