Transcript Powerpoint

Extent and Mask
•
•
•
•
Extent of original data
Extent of analysis area
Mask – areas of interest
Remember all rasters are rectangles.
Examining the Pattern of Land
Surface Characteristics
Describing and Quantifying
Landscape Patterns
Why?
Examine cause and effect
Detect changes
Develop relationships
Landscape Metrics
• Landscape Composition
• Landscape Configuration
Quantifying Landscape Pattern
• Important Issues to Consider
– 1st need a clear a priori statement of the
objectives and/or hypothesized pattern
changes.
– Classification scheme is critical.
– Scale (resolution and extent) must be defined.
– How should you identify a “patch”.
– Most metrics are correlated.
– What constitutes a significant change?
How should you identify a “patch”?
• Four-neighbor vs. eight-neighbor rule
• 4-way (no diagonals): 13 patches
• 8-way (with diagonals): 11 patches
• A Patch is also a Region in ESRI Grid.
RegionGroup
• Region Group Function will assign a unique value to all regions.
• You can use 4 or 8 neighbor cells to define a region. Default = 4.
• You can define connection “Within” the current zones or
“Cross” zones if you specify values to exclude.
• I always “Link” the original zone values to the new regions.
Landscape Metrics
• Composition
– Fraction or proportion (pi) occupied by type i.
– Richness (s): number of types present.
– Relative Richness (R): R = (s / smax) * 100.
• Smax = Potential number of types, hard to determine.
– Diversity (H):
s
H  [ pi ln( pi )] / ln( s )
i 1
– Dominance (D):
s
D  [ln( s )   pi ln( pi )] / ln( s )
i 1
Landscape Metrics
• Configuration: Patch-based Metrics
– Patch-based measurements: number, size,
perimeter (i.e. edge), shape, and density
– Perimeter/Area Ratio – indication of shape
– Connectivity (RSi)
• RSi = LCi / (pi * TotalArea)
• where LCi = size of the largest patch in type i
• Fragmentation index for type i = 1 - RSi
– Fractals
• Used as a measure of patch size.
Landscape Metrics
• Configuration: Contagion
– Probabilities of adjacency (qi,j):
qi , j  ni , j / ni
• where ni = # of cells in type i
• ni,j = # of cells when type i is adjacent to type j
– Contagion (C):
s
s
C  [1   Pij ln( Pij )] / 2 ln( s)
i 1 j 1
• where Pij = probability that two randomly chosen adjacent cells
belong to types i and j.
Diversity
• The concept of diversity has two components:
– Variety or Richness (# of species or cover types)
– Relative Abundance
• s (Richness) addresses the first
• H (Diversity) and D (Dominance) indices addresses the
second
– H is also called an Evenness Index
• Example: 2 sites, both with 10 types
– Site A: Type 1 = 90%, Types 2-10 = 10% combined
– Site B: All types are equal 10%
– Which is more diverse?
Kepner et al., EPA-NERL/ESD http://www.epa.gov/nerl1/land-sci/san-pedro.htm
Change in Land Cover Extent
Hectares
350000
300000
250000
200000
Grassland
Desertscrub
Mesquite
150000
100000
50000
0
Urban
1973
1986
1992
1997
Landscape Change Statistics
Grassland
1973
Area (ha)
1997
Desertscrub
% Rel.
Change
1973
1997
% Rel.
Change
Mesquite Woodland
1973
1997
Urban
% Rel.
1973
Change
312,850 263,432 -15.80 296,330 229,953 -22.40 20,821 101,602 +387.98 3,205
34.84
-15.80
39.19
30.41
-22.40
2.75
1997
% Rel.
Change
16,494 +414.63
% Cover
41.37
13.44 +387.98
0.42
2.18
+414.63
# of Patches
50,715 58,142 +14.64 26,260 39,991 +52.29 15,558 53,310 +242.65
418
3,010 +620.10
Largest
Patch (ha)
126,258 53,173 -57.89 201,165 37,361 -81.43 461.52
3,574 +674.34
982
4,938 +402.82
Ave Patch
Size
6.18
4.54
-26.54
11.3
5.76
-49.03
1.34
1.91
+42.54
7.86
5.55
-29.39
Connectivity
0.62
0.56
-9.68
0.66
0.55
-16.67
0.31
0.37
+19.35
0.74
0.69
-6.76
Courtesy Bill Kepner, US-EPA
Other Software
• If you are interested in landscape analysis there are
other software produces available.
• FRAGSTAT is on open source software package that
is heavily used in landscape analysis. It is designed to
compute a wide variety of landscape metrics for
categorical map patterns.
– http://www.umass.edu/landeco/research/fragstats/fragstats.
html
• USGS Landscape Analysis Tool Portal
– http://rmgsc.cr.usgs.gov/latp/index.shtml
The Resolution must be Defined – Small features will “drop-out”
as resolution is decreased which occurs when:
Vector data – increase in minimum mapping unit (MMU) size
Raster data – increase in cell size
Factors That Must be Considered
Classification Scheme is Critical – Resolution of attributes
The more “detailed the classification the more complicated
(rich/diverse) the landscape.
The Extent of the analysis area affects the measurement of landscap
features. The larger the extent the greater the probability that rare
types will be present. A smaller extent will give more weight to
dominate features. Bigger is usually “better”, BUT it is Important
that your analysis area extent matches the portion of the landscape
you want to study.
Factors to consider when determining the optimal extent and
resolution:
1. O’Neill’s rules:
a. Resolution should be 2 to 5 times smaller than the spatial
features being analyzed.
b. Extent should be 2 to 5 times larger than the largest patches.
2. Estimate the “zone of influence” of the phenomenon being
studied.
3. Important to stay within portion of the landscape being studied.
Potential stratifications: climatic, elevation, land ownership,
land use
4. Minimum resolution may be control by measurement technique
5. When in doubt conduct a study on the affect of scale and extent.
Vector to Raster Conversion
• Points and Lines are relatively easy compared to polygons.
– With points and lines you always change dimensions
• The basic problem is how to assign values to a uniform set of cells from a nonuniform set of polygons.
Centroid Method
• Using this method, each cell is assigned the value of the
feature that passes through the center of the cell. This method
can be used for any feature type, but is particularly useful in
coding continuous data, such as elevation, temperature or
density. Assign a cell values based on samples or predictions
based on the cell center location. Ex: Used in interpolation.
Predominant Type Method
(majority weighting)
• The value of the feature that fills the majority of the cell is
assigned to the location. This method is good for discrete or
noncontinuous data such as land cover, vegetation, or soils,
where the boundaries of the objects can be defined and their
associated value assigned to the cell when it occupies the
majority of the cell.
Most Important Type Method
(Unequal Weighting)
• Each cell is assigned the value associated with the features that
have been specified as more important to the study.
• For example, you may want to retain the identity of locations
or zones that contain endangered plant species, even if the
endangered plants don’t fill the majority of the grid cell.
Percentage Breakdown Method
• In this method, a cell is assigned several
values, one per feature, according to the
percent each feature occupies within the cell.
This is a difficult and costly method to
implement at data entry. However, it can be
especially useful for statistical data.
• Typically used for “large” cells – i.e. polygons
Polygon to Raster Tool
• Cell_Center
• Maximum_Area
– Assignment will be based on the largest single feature.
• Maximum_Combine (Majority)
– Assignment will be based on type with the largest combined
area.
• Priority Field
– Defined the field that will used to determine preference.
Organ Pipe NM
Landscape Metrics
•
•
Affect of Resolution (grain size) on Landscape Representation
Based on Organ Pipe Cactus National Monument Vegetation
Resolution
1x1m
5x5m
10 x 10 m
20 x 20 m
50 x 50 m
100 x 100 m
250 x 250 m
500 x 500 m
1000 x 1000
2000 x 2000 m
3000 x 3000 m
# of Type
19
19
19
19
19
19
17
15
11
9
4
# of Patches
158
158
158
156
172
247
104
60
28
13
5
Average Patch Size (ha)
106.6
106.6
106.6
108.0
97.0
68.4
162.8
274.2
625.0
1,230.8
3,060.0