1. GIS and Remote Sensing Projects Portfolio
by
Kristen Hestir
Maps
Image Processing
Charts
Tables
Graphs
Geospatial analysis of invasive species
Posters
2. California Organic Crops (2003)
versus Pesticide Use (2007)
Geographic
Analysis Pounds of pesticide per crop acre
0.00 - 1.00
1.01 - 2.50
2.51 - 5.00
5.01 - 10.00
Compares 10.01 - 12.60
Pesticides types include:
insecticides
acreage of hebicides
microbiocides
fungicides
rodenticides
organic
crops ®
to pesticide
usage in One dot represents 50
acres of organic crops
50
California. 250
1000
Representative densities: number
of acres per 100 square kilometers
Crops include:
field crops
fruit and nutsl
livestock and apiary
vegetables
nursery and floriculture Projection: California Teale Albers
Kilometers
0 50 100 200 300 400 500
Kristen Hestir
5/01/200
Source: University of California at Davis, Statistical Review of California's Organic Agriculture, 1998-2003 and Pesticide Action Network GIS and Cartography
3. Cartography Cartogram of Banana Exports
to the United States, 2002
Banana Mexico
exports from
South America Honduras
Jamaica
Guatemala
to the USA. Nicaragua
Costa Rica
Venezuela
Cartograms Panama
Colombia
use distorted Banana Exports in 1000 Kilogram Units
1 - 200,000
map geometry 200,001 - 400,000
400,001 - 600,000
Ecuador
in order to 600,001 - 800,000
Bolivia Brazil
800,001 - 1,022,347 Peru
convey Includes all bananas as food either fresh or dried
thematic
information in Exports in 1000 Kilogram Units
a visually
stimulating
225 450 900
±
way.
Krsiten Hestir, 4/24/2009, Cartography & GIS
Source: Tariff and trade data from the U.S. Department of Commerce, the U.S. Treasury, and the U.S. International Trade Commission.
4. Spatial Analysis
Viewshed
AM/FM Radio
Coverage
of
Dona Ana
County
Viewshed
illustrates an area
of land that is
“visible” from a
fixed vantage
point.
5. Top Ranked 100 Countries by
Gross Domestic Product and Quality of Life, 2005
Top 5 countries: Ireland, Switzerland, Norway, Luxemborg, Sweden
Rank by GDP per capita Rank (best to least) 9 Criteria for Quality of Life
quality of life
1-5 Material wellbeing Climate and geography
6 - 10 1-5 Life expectancy Job security
Political stability and security Political freedom
³
11 - 50 6 - 10 Low divorce rate Gender equality
51 - 100 11 - 50 Community life
Not in top 100 51 - 100
0 1,750 3,500 5,250 7,000
Source: The Economist Intelligence Unit Quality of Life Index Kilometers
Kristen Hestir, 4/15/2009
6. Autism Prevalence (2006), Superfund Sites (2007) and Arsenic Groundwater Contamination (2001)
WA
MT MN
ND ME
OR VT
WI NY NH
SD
ID MA
WY
MI
RI
CT
IA PA
NV NE OH NJ
IN
IL
DE
UT
CO MO MD
VA
KS WV
KY
CA
NC
AZ OK TN
NM AR
SC
GA
MS AL
TX
TX LA
Wells with Unsafe Arsenic Levels Representative Densities: FL
Percent of Children
per 1,000 Square Kilometers Number of Superfund Sites 0.16 to 0.20
per 125 Square Kilometers
No unsafe wells
0.21 to 0.40
±
0.01 to 0.04 5 Superfund sites 0.41 to 0.60
0.05 to 0.25
30 Superfund sites 0.61 to 0.80
0.26 to 0.50
0.51 to 1.00 60 Superfund sites 0.81 to 0.95
1.01 to 2.13 One dot represents 5 superfund sites Kilometers
Dot placement is randomized at the state level 0 250 500 1,000 1,500
7. Wyoming Nebraska
Mesilla Valley,
New Mexico Nevada Utah
Colorado
California Kansas
Oklahoma
Arizona
New Mexico
Study Ri
o Gr
an
Texas
de
Area
Ri v
MEXICO
er
Maps Projection: Lambert Conformal Conic
Yuma Valley,
Projection: UTM, WGS 84, Zone 13S
River Arizona
o
rad
lo
Mesilla Valley Study Area
Co
Yuma Valley Study Area
Las Cruces Metro
Yuma 1990 Metro
Yuma 2007 Metro
0 10 20
Kilometers ¯ Projection: UTM, WGS 84, Zone 11S
8. Leaf-On
(Min: 300°K; Max: 329°K)
Image
Derivative
Temperature
Land Surface (Degrees Kelvin)
High: 329
Temperature
Maps
Low : 279
Yuma
Valley, AZ Leaf-Off
(Min: 279°K; Max: 311°K)
0 5 10 15
Kilometers ¯
10. Leaf-On
Band 1 Min: 46 Max: 16811
Band 2 Min: -729 Max: 5877
Band 3 Min: -7101Max: 3270
Image
Derivative
Tasseled Cap TCT
Transformation RGB
Red: Band 1
Green: Band 2
Yuma Valley, Blue: Band 3
AZ
Leaf-Off
Band 1 Min: 469 Max: 14905
Band 2 Min: -913 Max: 4457
Band 3 Min: -6160 Max: 3622
0 5 10 15
Kilometers ¯
11. Landcover Assessment from Landsat TM5 Image,Mesilla Valley 2009
Digitized
Land
Cover
Change
Maps Land Cover Classes
Residential Cropland and Pasture Streams and Canals Strip Mines, Quarries, Gravel Pits
Industrial and Commercial Orchards, Etc. Reservoirs Transitional Areas
Transportation Confined Feeding Operations Forested Wetland Mixed Barren Land
Mixed Urban or Built-Up Land Mixed Rangeland Sandy Areas other than Beaches
0 7.5 15 22.5 30
Kilometers
/
12. Land Agricultural Land
Cover Barren Land
Rangeland
Change Urban
Water
Maps Wetland
and Pie 2% 1%
Charts 9%
13%
9%
1985 66%
Overall Accuracy = 76%
13. Land Agricultural Land
Cover Barren Land
Rangeland
Change Urban
Water
Maps Wetland
and Pie 3% 2%
Charts 5%
16% 10%
2009 64%
Overall Accuracy = 83.7%
14. Process Flow Chart
Stage 1: Leaf-on Tasseled Cap Principal Land Surface Normalized
Leaf-on Component Analysis Temperature Leaf-on Difference
Leaf-off Leaf-on Impervious Surface
Tasseled Cap Land Surface Leaf-on
Leaf-on, Leaf- Leaf-off Principal Temperature Leaf-off
off Component Analysis Normalized
Tasseled Cap Leaf-off Land Surface Difference
Leaf-on, Leaf- Temperature Leaf-on, Impervious Surface
off Principal Leaf-off Leaf-off
Component Analysis
Leaf-on, Leaf-off Normalized
Difference
Classify: Maximum Likelihood Evaluate: Confusion Matrices and McNemar tests. Impervious Surface
Leaf-on, Leaf-off
Stage 2: Select top performers and apply:
5 textures: entropy, angular second moment, homogeneity, correlation, contrast
3 x 3, 5 x 5 and 7 x 7 windows,
Classify: Maximum Likelihood
Evaluate: Confusion Matrices and McNemar tests.
Stage 3: Select top performers and apply:
Combined feature stacks: textures, derivatives etc.
Classify with: Maximum Likelihood, Support Vector Machine, Artificial Neural Network
Evaluate: Confusion Matrices
15. Matrices - Error Assessment and Statistical Test of
Significance
Ground Reference Data (Pixels)
Map Data Agriculture Barren Rangeland Urban Water Wetland Total
Confusion Agriculture 45 0 0 1 1 0 47
Barren 5 28 14 7 1 1 56
Matrix: Rangeland 61 5 308 7 1 4 386
Urban 9 21 33 155 1 5 224
Water 27 0 5 21 51 0 104
Wetland 64 3 55 11 7 43 183
Total 211 57 415 202 62 53 1000
Accuracy
Overall accuracy, Kappa coefficient
Measures
Map 1 ₂₁
wrong
₁₂ correct
McNemar Map 2 wrong sum both wrong M total wrong Map 2
Matrix: correct M sum both right total right Map 2
total wrong Map 1 total right Map 1
16. Comparative Analysis – Bar Chart
Confusion Matrices Results
Mesilla Valley Yuma Valley
83
83
78
Overall Accuracy (%)
78
73 L-On 73
68 68
L-Off
63 63
12B
58 58
No LST NDISI PCA TCT No LST NDISI TCT PCA
Derivative Derivatives
Feature Stacks Feature Stack
18. Comparative Analysis – Line Chart
Confusion Matrices Results
74
72
Overall Accuracy (%)
Mesilla Valley
70 overall
accuracy
68
66
64 Yuma Valley
overall
62
accuracy
60
Feature Stacks from Stage 1
19. Comparative Analysis – Line Chart
Confusion Matrices Results
86
Mesilla Valley
Stage 1
76
Mesilla Valley
Overall Accuracy (%)
Stage 2
66
Mesilla Valley
56 Stage 3
Yuma Valley
46
Stage 1
36 Yuma Valley
Stage 2
26 Yuma Valley
1 3 5 7 9 11 13 15
Stage 3
High ------------ Overall Accuracy Rank ------------ Low
20. An
Assessment
Using
Remote
Sensing and
GIS
Salt Cedar
Dynamics in
Northern
Doña Ana
County, N
M
21. New Mexico
Site 1
Site 2
Salt Cedar
Dynamics in
Las Cruces
Site 3
Study Areas
Site 4
0 5 10 20
Kilometers
Projection: UTM Zone 13N, NAD 83
M. Smith, T. Jones, V. Prileson, and K. Hestir, 2010/04/11 ¯
22. Site 1 Site 2
1936 Site 3
Site 4
Land
Cover
Built-up
Barren
Land Cover Type
Salt cedar high
Salt cedar medium
Row crops
Pecans 0 250
¯
500
Meters
1,000
Projection: UTM, Zone 13N, NAD83
Water Salt cedar low Other vegetation
Data source: NAIP 2009 Natural Color Aerial Photography Authors: K. Hestir, T. Jones, V. Prileson, and M. Smith (04/05/2010)
23. Site 1 Site 2
1955 Site 3
Site 4
Land
Cover
Built-up
Barren
Land Cover Type
Salt cedar high
Salt cedar medium
Row crops
Pecans 0 250
¯
500
Meters
1,000
Projection: UTM, Zone 13N, NAD83
Water Salt cedar low Other vegetation
Data source: NAIP 2009 Natural Color Aerial Photography Authors: K. Hestir, T. Jones, V. Prileson, and M. Smith (04/05/2010)
24. Site 1 Site 2
1983 Site 3
Site 4
Land
Cover
Built-up
Barren
Land Cover Type
Salt cedar high
Salt cedar medium
Row crops
Pecans 0 250
¯
500
Meters
1,000
Projection: UTM, Zone 13N, NAD83
Water Salt cedar low Other vegetation
Data source: NAIP 2009 Natural Color Aerial Photography Authors: K. Hestir, T. Jones, V. Prileson, and M. Smith (04/05/2010)
25. Site 1 Site 2
2009 Site 3
Site 4
Land
Cover
Built-up
Barren
Land Cover Type
Salt cedar high
Salt cedar medium
Row crops
Pecans 0 250
¯
500
Meters
1,000
Projection: UTM, Zone 13N, NAD83
Water Salt cedar low Other vegetation
Data source: NAIP 2009 Natural Color Aerial Photography Authors: K. Hestir, T. Jones, V. Prileson, and M. Smith (04/05/2010)
26. Site 1 Site 2
Land
Site 3 Site 4
Cover
Dynamics
1936-
1955
¯
Salt Cedar Dynamics
Salt cedar increase Water persistent
Salt cedar persistent Other land covers persistent Meters
0 250 500 1,000
Salt cedar decrease Other land cover changes Projection: UTM, Zone 13N, NAD83
Authors: K. Hestir, T. Jones, V. Prileson, and M. Smith (04/05/2010)
27. Site 1 Site 2
Land
Site 3 Site 4
Cover
Dynamics
1955-
1983
¯
Salt Cedar Dynamics
Salt cedar increase Water persistent
Salt cedar persistent Other land covers persistent Meters
0 250 500 1,000
Salt cedar decrease Other land cover changes Projection: UTM, Zone 13N, NAD83
Authors: K. Hestir, T. Jones, V. Prileson, and M. Smith (04/05/2010)
28. Site 1 Site 2
Land
Site 3 Site 4
Cover
Dynamics
1983-
2009
¯
Salt Cedar Dynamics
Salt cedar increase Water persistent
Salt cedar persistent Other land covers persistent Meters
0 250 500 1,000
Salt cedar decrease Other land cover changes Projection: UTM, Zone 13N, NAD83
Authors: K. Hestir, T. Jones, V. Prileson, and M. Smith (04/05/2010)
29. Site 1 Site 2
Land
Site 3 Site 4
Cover
Dynamics
1983-
2009
¯
Salt Cedar Dynamics
Salt cedar increase Water persistent
Salt cedar persistent Other land covers persistent Meters
0 250 500 1,000
Salt cedar decrease Other land cover changes Projection: UTM, Zone 13N, NAD83
Authors: K. Hestir, T. Jones, V. Prileson, and M. Smith (04/05/2010)
30. WILL THE JAGUAR (Panthera onca) PERSIST IN NEW MEXICO AND
ARIZONA?
Kristen Hestir Department of Geography, New Mexico State University
Figure 1. JungleWalk.com (*). Figure 2. JungleWalk.com (*).
Introduction Methods Conservation Efforts and Threats
Jaguars (Panthera onca), the largest felids in the Americas, once The methods in this study are Conservation Efforts:
were common in the southwestern United States. based on a literature review: In 1997 the jaguar was placed on the endangered species list by the
United States Department of the Interior, Fish and Wildlife Service.
Jaguars have been sighted in Arizona and New Mexico but with
decreasing frequency in the past 100 years (McCain and Childs • General description of the In 2009, the U. S. Fish and Wildlife Service declared designation of
2008). Only four males sighted in last 20 years. species. critical habitat is necessary and is developing proposed sites.
Why try to conserve the Arizona and New Mexico part of their • Range (historical and current) Disagreement within the jaguar conservation community. Use time
range? Populations that reside on the periphery of ranges can habitat requirements. and money to save peripheral populations, essential to survival of
be critical for the long-term survival of the species. species
• Conservation efforts and OR
threats to survival in Arizona concentrate time and money on the more densely populated
Research Question and New Mexico. Figure 7. http://www.destination360.com/south-
america/brazil/images/st/amazon
ranges.
-animals-jaguar.jpg.
Will jaguars persist in the New Threats:
Mexico and Arizona part of U.S.-Mexico border fence (from 2007), partitions northern
their range given the current range, reduces natural prey, limits water supplies, reduces mating
status of the species and Results potential, shifts migrant traffic and law enforcement activities into
ongoing conservation mountain habitats (further degrading habitats and increasing
efforts? Species Description:: encounters with humans).
Northern jaguars are smaller than their South American relatives.
Figure 3 JungleWalk.com (*). Jaguars have fur with small dots, large irregular spots and rosette Illegal killing continues due to cattle depredations, pelts
markings (Figures 1-3, 5-7). No two are alike, distinctive patterns (Figure 9) and incidental takes from traps and snares.
are used to identify individuals. Loss of habitat due to urban expansion, mineral
Study Site mining, increased cattle grazing, water mining.
Study site located in southern portions of Arizona and New Size ranges from 1.7 to 2.4 meters (nose to tail tip) in
Mexico (Figure 4), bordering Mexico. Based on historical ranges length, weighing between 45 to 113 kilograms. Climate change: models predict widespread Figure 9.
http://www.flickr.com/photos
and recent remote camera sightings. Prey: cattle (57% of biomass consumed), white-tailed deer, wild ecosystem disruptions in Mexico. /barcdog/2409633979/.
pig, rabbits,jackrabbits, coatis (raccoon family),
skunk, coyote, and reptiles
(Rosas-Rosas 2006). Conclusions
Range: Persistence in Arizona and New Mexico depends largely upon the
Variety of habitats from rain forest critical habitat proposal by the U.S. Fish and Wildlife Service and
to arid scrub. In the Sonoran the fate of the U.S.-Mexico border fence. Jaguars have a grim
Figure 5. JungleWalk.com (*). desert they use scrub, mesquite, prognosis for survival in the study area.
grassland, woodlands.
Range size varies widely, Acknowledgements
33 km2 to 1300 km2 per individual I would like to thank Dr. Carol Campbell for the interesting topic.
(Figure 8).
Density 1 to 10 individuals per
100 km2. depending on resource (*) http://www.junglewalk.com/photos/jaguar-pictures-I6147.htm
References
Brown, D. E. 1983. On the status of the jaguar in the southwest. The Southwestern Naturalist 28 (4):459-460.
Conde, D. A., F. Colchero, H. Zarza, N. L. Christenssen, J. O. Sexton, C. Manterola, C. Chávez, A. Rivera, D. Azuara, and G. Ceballos. Sex matters: modeling male and female habitat differences for jaguar conservation. Biological Conservation 143:1980-1988.
availability and habitat fragmentation. Federal Register, January 13 75 (8): 1741-1744.
Foster, R. J., B. J. Harmsen, and C. P. Doncaster. 2010. Habitat use by sympatric jaguars and pumas across a gradient of human disturbance in Belize. Biotropica 42 (6):724-731.
Grigione, M. M., K. Menke, C. López-González, R. List, A. Banda, J. Carrera, R. Carrera, A. J. Gordano, J. Morrison, M. Sternberg, R. Thomas, and B. Van Pelt. 2009. Identifying potential conservation areas for felids in the USA and Mexico: integrating reliable knowledge
across an international border. Fauna and Flora International, Oryx 43 (1):78-86.
Haag, T., A. S. Santos, D. A. Sana, R. G. Morato, L. Cullen. P. G. Crawshaw, C. De Angelo, M. S. Di Bitetti, F. M. Salzano, and E. Eizirik. 2010. The effect of habitat fragmentation on the genetic structure of a top predator: loss of diversity and high differentiation among
Figure 8. Estimated historical range of remnant populations of Atlantic Forest jaguars (Panthera onca). Molecular Ecology. 19:4906–4921.
Figure 4. Study area and locations of jaguars reported Hamilton, S. D. 2010. Investigative Report Macho B. U.S. Fish and Wildlife Service.
Figure 6. JungleWalk.com (*). jaguars based on expert opinion (Grigione et Hatten, J. R.., A. Averill-Murray, and W. E. Van Pelt. 2005. A spatial model of potential jaguar habitat in Arizona. Journal of Wildlife Management 69 (3):1024-2005.
killed in Arizona and New Mexico 1900-1980 (adapted McCain, E. B., and J. L. Childs. 2008. Evidence of resident jaguars (Panthera onca) in the southwestern United States and the implications for conservation. Journal of Mammalogy 89 (1):1-10.
Navarro-Sermentc, C., C. A. López-González, J. P. Gallo-Reynoso. 2005. Occurrence of jaguar (Panthera onca) in Sinaloa, Mexico. The Southwestern Naturalist 50 (1):102-106.
from Brown 1983). al. 2009). Rabinowitz, A., and K. A. Zeller, 2010. A range-wide model of landscape connectivity and conservation for the jaguar, Panther onca. Biological Conservation 143 (4):939-945.
Rosas-Rosas, O. C. 2006. Ecological status and conservation of jaguars (Panthera onca) in northeastern Sonora, Mexico. Dissertation, New Mexico State University, Las Cruces, New Mexico, USA.
1. Spangle, S. L. 2007. Biological opinion 22410-2007-F-0416: pedestrian fence projects at Sasabe, Nogales and Naco-Douglas, Arizona. United States Fish and Wildlife Service, Phoenix, Arizona..
31. LAND COVER CLASSIFICATION IN AN ARID REGION: AN
EVALUATION OF REMOTE SENSING APPROACHES
Kristen Hestir1 and Dr. Michaela Buenemann1
1Department of Geography, New Mexico State University
PROBLEM STATEMENT CHALLENGES OF CLASSIFYING LAND COVER IN ARID REGIONS
• Human induced land cover change is occurring at unprecedented rates worldwide and is affecting an estimated 39 to 50% of Earth’s land
surface. • Spectral responses of bright desert 4500
4000
5000
4500
• Drylands are of particular concern, they cover 41% of Earth’s land surface, are home to 35% of world population and are experiencing soils are often confused with the spectral 3500 4000
Reflectance x100
Reflectance x 100
3500
rapid population growth. response of impervious (urban) surfaces 3000
3000
2500
• Land cover change information can provide a basis for understanding what dryland areas are at risk, what this means for desert (Figure 3). 2000
2500
Impervious Surface Barren Land
ecosystems. • Soils dominate the spectral response 1500
Rangeland
2000
1500 Rangeland
• Landsat Thematic Mapper satellite imagery can provide spatially explicit and continuous information on land cover change. By using the weaker signal of sparse vegetation 1000 1000
500 500
various classification algorithms and feature stacks, land cover types can be differentiated in the imagery based on their unique spectral can be lost. 0 0
and spatial characteristics. • Physiological qualities of dryland 1 2 3 4 5 6 1 2 3 4 5 6
Bands
• There are, however, some characteristics of drylands which make land cover classification challenging. vegetation decreases the strong red Bands
edge and reduces absorption in the Figure 3. Comparison rangeland spectra (pink) and Figure 4. Comparison of rangeland spectra (white)
visible bands compared to typical impervious (urban) surfaces . and barren land (yellow).
OBJECTIVES non-dryland vegetation.
• Dryland vegetation is highly sensitive to resources, so the same species at different locations can have variable spectral responses
• Classify land cover of the Mesilla Valley (Figures 1 & 2) using two classification algorithms and various combinations of Landsat TM- (Figure 4).
derived spectral and textural information • Soils dominate spectral responses; however, they can have heterogeneous mineral content, causing variable spectral responses (Figure4).
• Compare the land cover maps in terms of their overall accuracies.
METHODS AND ACCURACY ASSESSMENT RESULTS AND DISCUSSION
• A leaf-on image of July 29, 2009 was georectified to a 2009 National Aerial Imagery Program Digital Ortho-Quarter Quad (DOQQ) and 95.00%
A land cover map (Figure 6) was produced for
radiometrically corrected using ENVI FLAASH atmospheric correction module. A leaf-off image of March 23, 2009 was georectified to 90.00%
each classification algorithm and various
the leaf-on image and radiometrically corrected to the leaf-on image using empirical line calibration. 85.00%
combinations of Landsat TM-derived spectral
• 1000 GPS and DOQQ points representing 5 land covers (agriculture, barren, rangeland, water, built-up) and shadow were used to train
O verall Accuracy
80.00%
and textural information. 75.00%
the two classifiers, Maximum Likelihood (MLC) and Support Vector Machine (SVM). Leaf-on
70.00% Leaf-off
• Image stacks (Figure 5) included combinations of 6 bands leaf-on, 6 bands leaf-off, Principal Components Analysis (PCA), Tasseled Cap Stage 1: Initial classifications show stacking leaf-on Leaf-on Leaf-off
65.00%
(TC), Land Surface Temperature (LST), and Normalize Difference Impervious Surface Index (NDISI). and leaf-off imagery gives equal or improved accuracy 60.00%
• Map accuracies were assessed using error (confusion) matrices based on 1000 randomly generated reference points.
Methods over single date stacks (Figure 7). 55.00%
50.00%
6 bands PCA 4 TC LST NDISI
PROCESS FLOW
Land Covers
STUDY AREA Built-Up
Agriculture
Water
Barren
Figure 7: Classification accuracies for Stage 1.
Rangeland
94.00%
Utah Colorado Stage 1: Leaf-on Tasseled Cap Principal Component Land Surface Normalized Difference 92.00%
Leaf-on Analysis Leaf-on Temperature Leaf-on Impervious Surface
90.00%
§
¦
¨
I-25 Leaf-off Leaf-on
O verall Accuracy
Tasseled Cap Principal Component Land Surface 88.00%
Leaf-on, Leaf-off Leaf-off Analysis Leaf-off Temperature Leaf-off 86.00%
Normalized Difference Initial Accuracies
Figure 6: Example classified map.
Tasseled Cap Principal Component Land Surface Impervious Surface 84.00% Entropy
Arizona New Mexico Leaf-off Stage 2: The texture filters entropy and Homogeneity
Leaf-on, Leaf-off Analysis Leaf-on, Temperature Leaf-on, 82.00%
Leaf-off Leaf-off homogeneity, with 7 by 7 80.00%
New Mexico Normalized Difference
Impervious Surface window, improved stage 1 initial 78.00%
Texas Leaf-on, Leaf-off accuracy by 2.5 %, 8.9% , 8.3%, 5.0 % 76.00%
6 Bands PCA 4 TC LST NDISI
Select top 5 and apply textures: and 2.1% for 6
§
¦
¨I-10 Stage 2: bands, PCA4, TC, LST, and NDISI Figure 8: Classification accuracies for Stage 2 with top
Mexico 3 x 3, 5 x 5 and 7 x 7 windows stacks respectively (Figure 8). two textures.
5 textures
94
Stage 3: Multiple image derivatives improved
Texas 93.5
classification accuracy even further (1.2%, 1.5%
Overall Accuracy
U.S. Bureau of the Census, Map of United States 93
0 125 250 500 and 1.8% improvement over stage 2 for the 3
Kilometers Select top 3 and apply combined feature stacks: 92.5
mlc
combinations. MLC and SVM classification
92 svm
Boundaries and Roads textures, derivatives etc. 91.5
algorithms performed equally well. Differences
Stage 3: Add classification algorithm: Leaf-on Leaf-off + Leaf-on Leaf-off + PCA 4 + homo + TC
homo + pca 4 homo + TC homo
in overall accuracy ranged from ( 0.2 % to 1.6 %)
Las Cruces Study Interstate Maximum Likelihood Image Stacks and Multiple Derivatives between the two classifiers (Figure 9).
Metro Area Highway
Projection: UTM Zone 13N, Datum: WGS 84
Support Vector Machine Figure 9: Classification accuracies for Stage 3.
Ü 0 2.5 5 10 15 20
Kilometers
ACKNOWLEDGMENTS
Figure 1: Location of the study area. Figure 2: Imagery from: USGS Global Visualization Figure 5: Flowchart of Image Processing. This work was supported by NSF Grant DEB-0618210, as a contribution to the Jornada Long-Term Ecological
Viewer. Research (LTER) program, by the United States Department of Agriculture, Agricultural Research Service
Editor's Notes
Cover 41% of Earth’s land surface, home to 35% of world population, experiencing rapid population growth, a driver of land cover change
Add image M12 = # of pixels misclassified in Map1 and not in Map 2M21 = # of pixels misclassified in Map 2 and not in Map 1If M12 + M21 > 19 then: Χ2 = (|M12 – M21| - 1)2 / (M12 + M21)At 1 degree of freedom, 0.05% confidence interval, If Χ2 > 3.84 then differences are statistically significant