GEOG 8450 Geospatial Tools in Landscape Analysis 8/12/2008 Madden Fall 2006 1 Kim, Xu and Madden UGA-CRMS GEOBIA 2008, Univ. of Calgary Object-based Vegetation Type Mapping using Ikonos Imagery: Spectral, Spatial and Topographic Information www.crms.uga.edu CRMS Minho Kim Bo Xu Marguerite Madden Center for Remote Sensing and Mapping Science (CRMS) Department of Geography University of Georgia, USA . Kim, Xu and Madden UGA-CRMS GEOBIA 2008, Univ. of Calgary Presentation Objectives: Background on UGA CRMS-NPS vegetation database development of Great Smoky Mountains National Park by manual interpretation. Assess the use of geospatial object-based image analysis (GEOBIA) to capture expert knowledge and semi/automate forest cover mapping. Present results from Kim and Xu on classification accuracies with spectral, texture, topographic and contextual information – and vegetation aggregation. Kim, Xu and Madden UGA-CRMS GEOBIA 2008, Univ. of Calgary Presentation Objectives: Background on UGA CRMS-NPS vegetation database development of Great Smoky Mountains National Park by manual interpretation. Assess the use of geospatial object-based image analysis (GEOBIA) to capture expert knowledge and semi/automate forest cover mapping. Present results from Kim and Xu on classification accuracies with spectral, texture, topographic and contextual information – and vegetation aggregation. Kim, Xu and Madden UGA-CRMS GEOBIA 2008, Univ. of Calgary Presentation Objectives: Background on UGA CRMS-NPS vegetation database development of Great Smoky Mountains National Park by manual interpretation. Assess the use of geospatial object-based image analysis (GEOBIA) to capture expert knowledge and semi/automate forest cover mapping. Present results from Minho Kim and Bo Xu on classification accuracies with spectral, texture, topographic and contextual information – and vegetation aggregation. Kim, Xu and Madden UGA-CRMS GEOBIA 2008, Univ. of Calgary Kim, Xu and Madden UGA-CRMS Digital Vegetation Databases/Maps for National Park Units of the Southeast Everglades National Park Big Cypress National Preserve Biscayne National Park Great Smoky Mountains National Park Mammoth Cave National Park Little River Natl. Canyon National Preserve Big South Fork Natl. River & Recreation Area Cumberland Gap National Historical Park Blue Ridge Parkway Obed Wild & Scenic River Guilford Courthouse Natl. Military Park Ninety Six National Historic Park USGS/NPS National Vegetation Mapping Program Carl Sandburg Home National Historic Site Abraham Lincoln National Historic Site Fort Donaldson National Battlefield Stones River National Battlefield Cowpens National Battlefield Russell Cave National Monument Kings Mountain National Military Park Shiloh National Military Park Chickamauga National Military Park Chattanooga National Military Park
8
Embed
GEOG 8450 Geospatial Tools in 8/12/2008 Landscape ...people.ucalgary.ca/~gjhay/geobia/linkedpresentations...GEOG 8450 Geospatial Tools in Landscape Analysis 8/12/2008 Madden Fall 2006
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
GEOG 8450 Geospatial Tools in Landscape Analysis
8/12/2008
Madden Fall 2006 1
Kim, Xu and Madden UGA-CRMS GEOBIA 2008, Univ. of Calgary
•Object-based Vegetation Type Mapping using
•Ikonos Imagery: Spectral, Spatial and
Topographic Information
www.crms.uga.eduCRMS
Minho KimBo Xu
Marguerite Madden
Center for Remote Sensing and Mapping Science (CRMS)Department of GeographyUniversity of Georgia, USA
. Kim, Xu and Madden UGA-CRMS GEOBIA 2008, Univ. of Calgary
Presentation Objectives:
Background on UGA CRMS-NPS vegetation database development of Great Smoky Mountains National Park by manual interpretation.
Assess the use of geospatial object-based image analysis (GEOBIA) to capture expert knowledge and semi/automate forest cover mapping.
Present results from Kim and Xu on classification accuracies with spectral, texture, topographic and contextual information – and vegetation aggregation.
Kim, Xu and Madden UGA-CRMS GEOBIA 2008, Univ. of Calgary
Presentation Objectives:
Background on UGA CRMS-NPS vegetation database development of Great Smoky Mountains National Park by manual interpretation.
Assess the use of geospatial object-based image analysis (GEOBIA) to capture expert knowledge and semi/automate forest cover mapping.
Present results from Kim and Xu on classification accuracies with spectral, texture, topographic and contextual information – and vegetation aggregation.
Kim, Xu and Madden UGA-CRMS GEOBIA 2008, Univ. of Calgary
Presentation Objectives:
Background on UGA CRMS-NPS vegetation database development of Great Smoky Mountains National Park by manual interpretation.
Assess the use of geospatial object-based image analysis (GEOBIA) to capture expert knowledge and semi/automate forest cover mapping.
Present results from Minho Kim and Bo Xu on classification accuracies with spectral, texture, topographic and contextual information – and vegetation aggregation.
Kim, Xu and Madden UGA-CRMS GEOBIA 2008, Univ. of Calgary Kim, Xu and Madden UGA-CRMS GEOBIA 2008, Univ. of Calgary
Digital Vegetation Databases/Maps for National Park Units of the Southeast
Everglades National ParkBig Cypress National PreserveBiscayne National ParkGreat Smoky Mountains National ParkMammoth Cave National Park Little River Natl. Canyon National PreserveBig South Fork Natl. River & Recreation AreaCumberland Gap National Historical Park Blue Ridge ParkwayObed Wild & Scenic RiverGuilford Courthouse Natl. Military ParkNinety Six National Historic Park
USGS/NPS National Vegetation Mapping Program
Carl Sandburg Home National Historic Site Abraham Lincoln National Historic SiteFort Donaldson National BattlefieldStones River National Battlefield Cowpens National BattlefieldRussell Cave National MonumentKings Mountain National Military ParkShiloh National Military ParkChickamauga National Military ParkChattanooga National Military Park
GEOG 8450 Geospatial Tools in Landscape Analysis
8/12/2008
Madden Fall 2006 2
Kim, Xu and Madden UGA-CRMS GEOBIA 2008, Univ. of Calgary
Mapping and GIS Analysis of Vegetation in Great Smoky Mountains National Park, NPS
Cherokee
Gatlinburg
2025 m
10 km
250 m
Elevation range from 250 to 2025 m AMSLNearly continuous forest cover= Photogrammetric challenge Kim, Xu and Madden UGA-CRMS GEOBIA 2008, Univ. of Calgary
Manual interpretation of forest communities from color infrared (CIR) photos acquired in the fall
Kim, Xu and Madden UGA-CRMS GEOBIA 2008, Univ. of Calgary
1400 Photos at 1:12,000 Scale
Kim, Xu and Madden UGA-CRMS GEOBIA 2008, Univ. of Calgary
Spatial Accuracy ~ +/- 5 – 10 m RMSE
Kim, Xu and Madden UGA-CRMS GEOBIA 2008, Univ. of Calgary
Wear Cove Ortho-corrected Raw Vectors
Wear Cove Edited Vectors
Wear Cove Final Vegetation Map
Kim, Xu and Madden UGA-CRMS GEOBIA 2008, Univ. of Calgary
Fieldwork is conducted by CRMS, NPS and NatureServe to verify interpretation, establish vegetation classes and refine the rule sets for fire fuel classes.
Kodak Digital Field Imaging System (FIS) =
DC265 Digital Camera +Garmin III Plus GPS
GEOG 8450 Geospatial Tools in Landscape Analysis
8/12/2008
Madden Fall 2006 3
Kim, Xu and Madden UGA-CRMS GEOBIA 2008, Univ. of Calgary
Great Smoky Mountains National ParkVegetation Classification System
I. FORESTA. Sub Alpine Forest
1. Fraser Fir Fa. Formerly Fraser Fir (F)
2. Red Spruce – Fraser Fir S-F, S/F, F/Sa. Red Spruce – Fraser Fir/Rhododendron S-F/Rb. Red Spruce – Fraser Fir/Low Shrub-Herb S-F/Sh
3. Red Spruce Sa. Red Spruce/Rhododendron S/Rb. Red Spruce/Birch S/NHx:Bc. Red Spruce/Hemlock S/T, T/Sd. Red Spruce and/or Hemlock (uncertain) S.T
Kim, Xu and Madden UGA-CRMS GEOBIA 2008, Univ. of Calgary
Analysis and visualization of vegetationdistributions with respect to environmental factors
Kim, Xu and Madden UGA-CRMS GEOBIA 2008, Univ. of Calgary
Thunderhead Mountain
Overstory Vegetation
Thunderhead Mountain
Digital Elevation Model
Kim, Xu and Madden UGA-CRMS GEOBIA 2008, Univ. of Calgary
Elevation
Slope
Aspect
GEOG 8450 Geospatial Tools in Landscape Analysis
8/12/2008
Madden Fall 2006 4
Kim, Xu and Madden UGA-CRMS GEOBIA 2008, Univ. of Calgary0
100
200
300
400
500
600
700
800
900
1000
N NE E SE SE SW W NW
Aspect
Are
a (Ha)
Cove Hardwoods (CHx)
+
Thunderhead Mountain Vegetation and Elevation Range
0
200
400
600
800
1000
1200
1400
1600
1800
PIRD W
OzH
OzHf
HxCHx
Om
H TO
cHM
AL HINHx
MO
MO/H
th KR-K R P
General Vegetation Classes
Ele
vati
on
(m
)
Min
MaxMean
+
Thunderhead Mountain Vegetation and Slope
0
10
20
30
40
50
60
PIRD W
OzHOzH
fHx
CHxOm
H TOcH
MAL HINHx
MO
MO/Hth K
R-K R P
General Vegetation Classes
Slo
pe
(deg
rees
)
MinMaxMean
55
+
P(CHx) = .29 (North aspects 0 to 23o or 338 to 360o) + .19 (Northeast aspects 23 to 68o) + .31 (Northwest aspects 293 to 338o).
Empirically-Based Rule Set
Elevation Range
Slope
Aspect
Kim, Xu and Madden UGA-CRMS GEOBIA 2008, Univ. of Calgary
Spatial correlations used to define rule sets for vegetation types
F
MO-Hth MO CHx
Montane Oak-Heath (MO-Hth)
Montane Oak(MO)
Cove Hardwood (CHx)
Kim, Xu and Madden UGA-CRMS GEOBIA 2008, Univ. of Calgary
Manually interpreted vegetation database as reference data set for accuracy assessment
Kim, Xu and Madden UGA-CRMS GEOBIA 2008, Univ. of Calgary
Graminoid PasturePPPGrass
Heath Bald/Northern HardwoodsHeath Bald /Shrub (Sb)
Hth/NHxHth/Sb
HthHthShrub
Cove Hardwoods (CHx)Rich Type (CHxR) with Hemlock (T)
CHxRCHxR/T
CHxRCHxMixed
Northern HardwoodsYellow Birch Type (NHxB)Beech Gap (NHxBe)Acid Type (NHxA)Rich Type (NHxR)Montane Northern Red Oak (MOr) withRhododendron-Kalmia (R-K), Heath Bald (Hth) or graminoidMixed Hardwoods Acid Type (HxA) with Eastern Hemlock (T)
NHxBeNHxA
NHxRMOr/R-KMOr/Hth
MOr/GHxAHxA/T
NHxNHxA
NHxRMOrHx
NHxMOr
Hx
Deciduous
Eastern Hemlock (T) and Mixed Northern Hardwood-Acid Type (NHxA)or Successional Mixed, Acidic (HxA)
T/NHxAT/HxA
TTConiferous
Association Description15-Class9-Class7-Class5-Class
Segments over Vegetation Manual Classified Segments
GEOG 8450 Geospatial Tools in Landscape Analysis
8/12/2008
Madden Fall 2006 8
Kim, Xu and Madden UGA-CRMS GEOBIA 2008, Univ. of Calgary
Comparison of Classification Performance
+ 5.6- 2.2+ 4.9- 10.0Grass
- 3.2- 6.6- 8.3- 3.9Shrub
+14.2+5.6+1.5- 7.0Mixed
+15.5-1.7+7.7-3.4Evergreen
+2.1+14.3+0.5+10.0Deciduous
User’s Accuracy %
Producer’s Accuracy %
User’s Accuracy %
Producer’s Accuracy %
VegetationType
Spectral Mean - Topography Spectral Mean – Topo/Streams
Kim, Xu and Madden UGA-CRMS GEOBIA 2008, Univ. of Calgary
Deciduous ForestMaximum gain of 14.3 % in producer’s accuracy by adding
topographic variables and buffer distance of stream channels
Evergreen forestIncrease of producer’s accuracy with slight decrease of
user’s accuracy by adding topography and stream buffers. Maximum gain of 15.5 % in user’s accuracy
Mixed forestIncreased individual accuracies by adding topographic
variables and buffer distance of stream channelsMaximum gain of 14.2 % in user’s accuracy
Summary
Kim, Xu and Madden UGA-CRMS GEOBIA 2008, Univ. of Calgary
• Shrub- Decrease of accuracies when adding ancillary
information. Maximum loss of 8.3 % in user’s accuracy by adding topographic variables
• Grass- Decrease in producer’s accuracy after adding ancillary
information. Maximum loss of 10 %
- Increase in user’s accuracy with the addition of ancillary information. Maximum gain of 5.6 %
Summary
Kim, Xu and Madden UGA-CRMS GEOBIA 2008, Univ. of Calgary
Manual Spectral
SpectralTopo
SpectralTopoStreams
Kim, Xu and Madden UGA-CRMS GEOBIA 2008, Univ. of Calgary
Spectral information alone did not produce appropriate segmentation results particularly for continuous features such as forest types and narrow shaped vegetation communities.
Texture, topography and context (i.e., proximity to stream channels) improved segmentation quality and classification, especially for forest types in mountainous areas.
Future research will continue to explore GEOBIA methods for expert knowledge preservation and semi-automation of vegetation mapping.
Conclusions
Kim, Xu and Madden UGA-CRMS GEOBIA 2008, Univ. of Calgary
Thank you
Center for Remote Sensing and Mapping ScienceDepartment of Geography, The University of Georgiahttp://www.crms.uga.edu