assessing the impact of land cover spatial resolution on forest fragmentation modeling
DESCRIPTION
Assessing the Impact of Land Cover Spatial Resolution on Forest Fragmentation Modeling. James D. Hurd and Daniel L. Civco C enter for L and use E ducation A nd R esearch (CLEAR) Department of Natural Resources Management & Engineering The University of Connecticut - PowerPoint PPT PresentationTRANSCRIPT
Assessing the Impact of Land Assessing the Impact of Land Cover Spatial Resolution on Cover Spatial Resolution on
Forest Fragmentation Forest Fragmentation ModelingModeling
James D. Hurd and Daniel L. CivcoJames D. Hurd and Daniel L. CivcoCCenter for enter for LLand use and use EEducation ducation AAnd nd RResearch esearch (CLEAR)(CLEAR)
Department of Natural Resources Management & Department of Natural Resources Management & EngineeringEngineering
The University of ConnecticutThe University of ConnecticutU-4087, Room 308, 1376 Storrs RoadU-4087, Room 308, 1376 Storrs Road
Storrs, CT 06269-4087Storrs, CT 06269-4087
OutlineOutline
• The Forest Fragmentation Model
• Application to land cover of different spatial resolutions
• Concluding remarks
Forest FragmentationForest FragmentationModelingModeling
Developed by Riitters Developed by Riitters et al.et al. (2000) to assess (2000) to assess global forest fragmentation from 1 km land cover global forest fragmentation from 1 km land cover data.data.
Adapted for use on Landsat derived Adapted for use on Landsat derived land coverland cover information information (30-meter spatial resolution).(30-meter spatial resolution).
Categorizes Categorizes forestforest pixels into 6 types: pixels into 6 types:• Core forest• Perforated forest• Edge forest• Transition forest• Patch forest• Undetermined forest
Background
Forest Fragmentation Forest Fragmentation ModelModel
9x9 Analysis Window
Forest PixelNon-Forest Pixel
Based on Image Convolution.Based on Image Convolution.
Uses a roving analysis window of fixed size to Uses a roving analysis window of fixed size to determine vales of Pf (amount) and Pff determine vales of Pf (amount) and Pff (adjacency).(adjacency).
Forest Fragmentation Forest Fragmentation ModelModelHow It Works
Pf =57 forest pixels81 total pixels = 0.70
9x9 Analysis Window
Forest PixelNon-Forest Pixel
Pf = proportion of forest pixels in analysis window.
Forest Fragmentation Forest Fragmentation ModelModelHow It Works
Forest PixelNon-Forest Pixel
Pf =57 forest pixels81 total pixels = 0.70
Pff =86 pixel pairs, both forest119 pixel pairs, at least one forest
= 0.72
144 pixel pairs in a 9x9 window(72 vertical, 72 horizontal)
9x9 Analysis Window
Pff = how connected are those forest pixels.
Forest Fragmentation Forest Fragmentation ModelModelHow It Works
• Core forest, Pf = 1.0• Perforated forest, Pf > 0.6 and Pf – Pff > 0• Edge forest, Pf > 0.6 and Pf – Pff < 0• Transition forest, Pf < 0.6 and >0.4• Patch forest, Pf < 0.4• Undetermined forest, Pf > 0.6
and Pf = Pff
Forest Fragmentation Forest Fragmentation ModelModelHow It Works
(Riitters et al., 2000)
Core ForestCore Forest - all surrounding grid cells are - all surrounding grid cells are forest.forest.Perforated ForestPerforated Forest - the interior edge of a - the interior edge of a forest tract such as would occur around a small forest tract such as would occur around a small clearing or house lot.clearing or house lot.Edge ForestEdge Forest - the exterior edge of a forest tract - the exterior edge of a forest tract such as would occur along a large agricultural such as would occur along a large agricultural field or urban area.field or urban area. Transitional ForestTransitional Forest - about half of the - about half of the surrounding grid cells are forest.surrounding grid cells are forest.Patch ForestPatch Forest - less than 40% of surrounding - less than 40% of surrounding grid cells are forest.grid cells are forest.
Definitions
Forest Fragmentation Forest Fragmentation ModelModel
Pf = 0.70
Forest PixelNon-Forest Pixel
Forest Fragmentation Forest Fragmentation ModelModelHow It Works
Pf = 0.70
Pf - Pff = 0.70-0.72 = -0.02How It Works
Forest PixelNon-Forest Pixel
EDGE FOREST
Forest Fragmentation Forest Fragmentation ModelModel
9x9 analysis window27x27 analysis window81x81 analysis window
Analysis windows ofdifferent sizes can beapplied.
As analysis window size increases, the amount of core forest decreases.
Analysis Window Sizes
9x927x2781x81
Forest Fragmentation Forest Fragmentation ModelModel
Three Related Items to Consider
1. Spatial resolution of input land cover.
2. Width of desired edge (i.e. how far from a non-forested feature do you need to be before you are in core forest?).3. What analysis window size will you use?
Forest Fragmentation Forest Fragmentation ModelModel
Core Forest to Non-forestCore Forest to Non-forest
n x n pixel analysis window, where n is oddminimum pixel distance = 1 + [(n – 1) / 2]
Minimum distance = minimum pixel distance * pixel resolution
A core forest pixel will be at least the minimum distance from a non-forest pixel (pixel center to pixel center).
Core Forest to Non-forestCore Forest to Non-forest9x9 analysis window = 1 + [(9 - 1) / 2] = 5 pixelsFor 30 m pixels = 5 x 30 = 150 m (pixel centers)
(5 x 30) – 30 = 120m (pixel edges)
5 pixels(150 m)(120 m)
No longer core forest
Three Related Items to Consider
1. Spatial resolution of input land cover.
2. Width of desired edge (i.e. how far from a non-forested feature do you need to be before you are in core forest?).3. What analysis window size will you use?
Forest Fragmentation Forest Fragmentation ModelModel
Application to Land Cover Application to Land Cover of Different Spatial of Different Spatial
ResolutionsResolutions
Using Analysis Windows of Same Pixel Dimensions
1 - meter 5 - meter 10 - meter 30 - meter
Larger spatial resolution created by degrading the 1-meter spatial resolution land cover
Land CoverLand Cover
Forest Fragmentation ResultForest Fragmentation Result
1 - meter 5 - meter 10 - meter 30 - meter
5 x 5 Analysis Window
Forest Fragmentation ResultForest Fragmentation Result
1 - meter 5 - meter 10 - meter 30 - meter
15 x 15 Analysis Window
Forest Fragmentation ResultForest Fragmentation Result
1 - meter 5 - meter 10 - meter 30 - meter
27 x 27 Analysis Window
Forest Fragmentation ResultForest Fragmentation Result
1 - meter 5 - meter 10 - meter 30 - meter
81 x 81 Analysis Window
Application to Land Cover Application to Land Cover of Different Spatial of Different Spatial
ResolutionsResolutions
Using Analysis Windows of Same Areal Extent
Land CoverLand Cover
Larger spatial resolution created by degrading the 1-meter spatial resolution land cover
1 - meter 5 - meter 10 - meter 30 - meter
271x271 55x55 27x27 9x9
Analysis Window Area = approx. 7.3 hectares
Forest Fragmentation ResultForest Fragmentation Result
1 - meter 5 - meter 10 - meter 30 - meter
Developed, 21.46
Non-fores t, 2.54
Perforated, 36.89
Patch, 0.76Trans ition,
9.46
Edge, 12.38
Core, 14.03
Water, 2.48
30 – meter, 9x9
water, 0.94
interior, 10.91
perforated, 9.51
edge, 46.43
patch, 0.39 developed, 20.50
non-forest, 2.64
transition, 8.67
water, 0.91
interior, 11.80
perforated, 1.82
edge, 53.50
patch, 0.46 developed, 20.34
non-fores t, 2.60
trans ition, 8.57
water, 0.93
interior, 11.67
perforated, 0.34edge, 55.09
patch, 0.46 developed, 20.35
non-forest, 2.57
transition, 8.60
1 – meter, 271x271 5 – meter, 55x55
10 – meter, 27x27
Pf = 0.765Pff = 0.744
Pf - Pff = 0.021
9x9 Analysis Window
30 – meter Land Cover30 – meter Land Cover9x9 Analysis Window9x9 Analysis Window
PERFORATED FOREST
Pf = 0.844Pff = 0.916
Pf - Pff = -0.072
27x27 Analysis Window
10 – meter Land Cover10 – meter Land Cover27x27 Analysis Window27x27 Analysis Window
EDGE FOREST
Pf = 0.848Pff = 0.966
Pf - Pff = -0.118
55x55 Analysis Window
5 – meter Land Cover5 – meter Land Cover55x55 Analysis Window55x55 Analysis Window
EDGE FOREST
Pf = 0.847Pff = 0.983
Pf - Pff = -0.136
271x271 Analysis Window
1 – meter Land Cover1 – meter Land Cover271x271 Analysis Window271x271 Analysis Window
EDGE FOREST
• Core forest, Pf = 1.0• Perforated forest, Pf > 0.6 and Pf – Pff > 0• Edge forest, Pf > 0.6 and Pf – Pff < 0• Transition forest, Pf < 0.6 and >0.4• Patch forest, Pf < 0.4• Undetermined forest, Pf > 0.6
and Pf = Pff
Edge/Perforated DefinitionEdge/Perforated Definition
Land Cover 9x9 Analysis Window
9x9 Analysis Window applied to 30-meterspatial resolution land cover
Altering Definition of Altering Definition of Edge/PerforatedEdge/Perforated
Results of Forest Frag. Model using a 27x27 Analysis Window on 10m Land Cover
Riitters Original DefinitionsPerforated Pf – Pff > 0
Edge Pf – Pff < 0
Altered DefinitionsPerforated Pf – Pff > -0.05
Edge Pf – Pff < -0.05
Altered DefinitionsPerforated Pf – Pff > -0.1
Edge Pf – Pff < -0.1
NOT SO GOOD BETTER MORE LIKE 30m9x9 FF RESULTS
Altering Definition of Altering Definition of Edge/PerforatedEdge/Perforated
Riitters Original DefinitionsPerforated Pf – Pff > 0
Edge Pf – Pff < 0
Altered DefinitionsPerforated Pf – Pff > -0.05
Edge Pf – Pff < -0.05
Altered DefinitionsPerforated Pf – Pff > -0.11
Edge Pf – Pff < -0.11
NOT SO GOOD BETTER MORE LIKE 30m9x9 FF RESULTS
Results of Forest Frag. Model using a 55x55 Analysis Window on 5m Land Cover
Altering Definition of Altering Definition of Edge/PerforatedEdge/Perforated
Riitters Original DefinitionsPerforated Pf – Pff > 0
Edge Pf – Pff < 0
Altered DefinitionsPerforated Pf – Pff > -0.1
Edge Pf – Pff < -0.1
Altered DefinitionsPerforated Pf – Pff > -0.15
Edge Pf – Pff < -0.15
NOT SO GOOD BETTER MORE LIKE 30m9x9 FF RESULTS
Results of Forest Frag. Model using a 271x271 Analysis Window on 1m Land Cover
Altering Definition of Altering Definition of Edge/PerforatedEdge/Perforated
Water, 4.19
Core, 32.19
Edge, 6.79
Transition, 1.71 Patch, 1.90
Perforated, 35.56
Non-forest, 0.00
Developed, 17.65
Water, 4.18
Core, 25.39
Edge, 6.20
Transition, 1.45 Patch, 2.11
Perforated, 42.89
Non-forest, 0.00
Developed, 17.78
Water, 4.38
Core, 26.53
Edge, 6.29
Transition, 1.36 Patch, 2.07
Perforated, 41.87
Non-forest, 0.00
Developed, 17.50
Water, 4.33
Core, 26.16
Edge, 5.43
Transition, 1.34 Patch, 2.06
Perforated, 42.99
Non-forest, 0.00
Developed, 17.68
30- meter 9x9 (original definitions)10- meter 31x31 (Pf-Pff > OR < -0.1) 30 – meter, Pf – Pff threshold = 0.0
5 – meter, Pf – Pff threshold = -0.11
10 – meter, Pf – Pff threshold = -0.05
1 – meter, Pf – Pff threshold = -0.16
Class Edge Bias = 0.0488 * ln(r) - 0.1601
Class Edge Bias CurveClass Edge Bias Curve
where r = land cover spatial resolution
CLEAR WebsiteCLEAR Website
clear.uconn.edu
http://clear.uconn.edu/projects/landscape/forestfrag/ff_tool.htm
CLEAR WebsiteCLEAR Website
Downloadable ToolDownloadable Toolfor ESRI ArcGIS 9.2for ESRI ArcGIS 9.2
Concluding RemarksConcluding Remarks
Concluding RemarksConcluding RemarksWhen applying Forest Fragmentation model, three
items to consider:- land cover spatial resolution- desired width of edge/perforation- analysis window size
The spatial resolution (and landscape pattern) of the land cover impacts the result of the model:- it may be desirable to alter the definition of edge and perforated.
Ultimately the results are only an approximation, but:- provide a powerful visual impact- allow for the quantification of forest fragmentation.
Assessing the Impact of Land Assessing the Impact of Land Cover Spatial Resolution on Cover Spatial Resolution on
Forest Fragmentation Forest Fragmentation ModelingModeling
CCenter for enter for LLand use and use EEducation ducation AAnd nd RResearch esearch (CLEAR)(CLEAR)Department of Natural Resources Management & Department of Natural Resources Management &
EngineeringEngineeringThe University of ConnecticutThe University of Connecticut
U-4087, Room 308, 1376 Storrs RoadU-4087, Room 308, 1376 Storrs RoadStorrs, CT 06269-4087Storrs, CT 06269-4087