Title: Land Cover Mapping for the Southwest Regional Gap Analysis Project
1Land Cover Mapping for the Southwest Regional Gap
Analysis Project
UGIC Conference, 2003 Sherwood Hills Resort, Utah
- John Lowry
- College of Natural Resources
- RSGIS Laboratory
- Utah State University
- Collaborators Doug Ramsey, Lisa Langs,
- Gerald Manis, Jessie Kirby, Wendy Rieth, Marie
Ducharme
2Presentation Overview
- I. Project Description Participants
- II. Mapping Methodology
- III. Comparison with 1995 GAP Vegetation Map
- IV. Timeline Summary
3I. Project Description Participants
What is GAP?
4State-Based GAP Projects
- State-based vegetation classification systems
(cover type legends) - State-based mapping methods
- State-based mapping area
5Salt Desert Shrub in the 4 Corners
6Objectives for SWReGAP
- Regional vegetation classification system (land
cover legend) - Regionally standardized data and mapping methods
- State mapping responsibilities defined by
eco-regional areas - Improvements in vegetation land cover map
7II. Mapping Methodology
- A) Nature of the Classification Problem
- Spatial Resolution Extent
- Thematic Resolution
- B) Predictor Data
- Variables Used to Predict Land Cover
- C) Training Samples
- Sample size and adequacy
- D) Properties of the Classifier
- Basics of Decision Tree Classifiers
- Tools developed
8Classification Problem Spatial Resolution
- 85 Landsat 7 scenes
- 30 meter resolution
- Three seasons spring, summer fall
9Classification Problem Spatial Extent
- 40 Mapping zones
- Spectrally consistent
- Eco-regionally distinct
10Classification Problem Thematic Resolution
NatureServe Ecological Systems
NVC Formation
NVC Alliance
NVC Association
NVC Class/Subclass
1,800 units
10 units
5,000 units
700 units
300 units
MRLC 2000 Proposal
Gap Analysis Program
National Park Mapping
(Natural/Semi-natural types)
(Slide Courtesy Pat Comer, Nature Serve)
11Ecological Systems
- Groups of plant communities and sparsely
vegetated habitats unified by similar ecological
processes, substrates, and/or environmental
gradients. - Manifest in a landscape as a spatial aggregation
at an intermediate scale (10 ha 100,000 ha),
persisting for 100 or more years.
(Slide Courtesy of Pat Comer, NatureServe)
12Inter-Mountain Mixed Salt Desert Shrub System
Shadscale Shrubland Alliance
Four-wing Saltbush Alliance
13Predictor Data Spectral Characteristics
14Predictor Data Elevation
15Predictor Data Aspect and Landform
16Training Data Sources
3000 air photo interpretation sites US Forest
Service
17Training Data Sources
3000 air photo interpretation sites US Forest
Service
5200 Sites from other organizations (USGS
Landfire BLM)
18Training Data Sources
3000 air photo interpretation sites US Forest
Service
5200 sites from other organizations (USGS
Landfire BLM)
7800 field work RSGIS Lab in collaboration
with BLM UDWR
16000 total sample sites
19Properties of the Classifier Decision Trees
- Data-mining software for decision-making and
exploratory data analysis, also called CART
(Classification and Regression Trees (Breimann
1984)) - Identify complex relationships between multiple
independent variables to predict a single
categorical occurrence - Predictor variables may be categorical or
continuous - Recursively splits the predictor data cloud
to create prediction rules or a decision tree
with branches nodes - Several software packages SPLUS, See5/C5, CART,
R - Gaining popularity remote sensing land cover
classification (Brown de Colstoun et. al, 2003
Larwrence Wright 2001 Hansen et al. 2000
Friedl Brodley 1997 Hansen, Bubayah Defries
1996))
20Mining the Predictor Layers
21(No Transcript)
22Example Splits on 2 variables
23Example Tree Output for 2 Variables
24Example Rules Output
See5 Release 1.17 Wed Apr 23 134202
2003 Options Rule-based
classifiers Class specified by attribute
dep' Read 7097 cases (10 attributes) from
t3.data Rules Rule 1 (17, lift 45.4)
band01 1 band03 gt 115 band03
lt 122 band05 lt 81 band06 lt
1419 -gt class 1 0.947 Rule 2 (9,
lift 43.6) band01 1 band02 lt
102 band03 gt 115 band03 lt 118
band04 lt 117 band06 lt 1419
-gt class 1 0.909 Rule 3 (6, lift 42.0)
band01 13 band03 lt 110
band05 lt 73 band07 4
Generated with cubistinput by EarthSat
Training samples 10260 Validation samples
2565 Minimum samples 0 Sample method
Random Output format See5 dep. h/mgz
n_5/trainingdata/mrgpts1.img(Layer_1) Xcoord i
gnore. Ycoord ignore. band01 1,2,-30
h/mgzn_5/img_files/sum30cl.img(Layer_1) band02
continuous. h/mgzn_5/img_files/subrt.img(Layer
_1) band03 continuous. h/mgzn_5/img_files/sundv
i.img(Layer_1) band04 continuous. h/mgzn_5/img
_files/fandvi.img(Layer_1) band05 continuous. h
/mgzn_5/img_files/fabrt.img(Layer_1) band06 con
tinuous. h/mgzn_5/img_files/elev.img(Layer_1) b
and07 0,1,2,3,4,5,6,7,8,9,10. h/mgzn_5/img_file
s/landf.img(Layer_1) dep 1,2,3,4,5,6,7,8,9,10,
11,12,13,14,15,16,17,18,19,20. h/mgzn_5/training
data/mrgpts1
25Tools for Spatial Applying Decision Trees
- Rulemaker (SPLUS Imagine)
- Vinod Chowdary (USU, MS Computer Sci.)
- http//www.gis.usu.edu/docs/projects/swgap/rulemak
er.html - STATMOD (SPLUS Arcview)
- Christine Garrard (USU, MS Biology)
- http//bioweb.usu.edu/gistools/statmod/
- Imagine CART Module (See5 Imagine)
- Eros Data Center (Earth Satellite Corp)
- http//www.gis.usu.edu/7Eregap/download/C5Module/
26II. Comparison with 1995 Utah GAP Vegetation Map
- Utahs Great Basin Eco-Region ( 17.5 M acres,
300 x 120 Miles) - Approximately 5 mosaicked Landsat 7 scenes
- 3000 sample sites (1700 USU/1300 other sources)
27Comparison for Great Basin Eco-Region (Partial
List)
281995 GAP Vegetation Map
2003 GAP Veg. Map (preliminary)
29Park Valley Example
301995 GAP 30 M
2003 GAP 30 M
1995 GAP Pub.1KM
31Tooele Example
Tooele
32Tooele
Tooele
Tooele
1995 GAP 30 M
2003 GAP 30 M
1995 GAP Pub.1KM
33Clarkston Example
Clarkston
341995 GAP 30 M
2003 GAP 30 M
1995 GAP Pub.1KM
Clarkston
Clarkston
Clarkston
35Validation, Accuracy and Appropriate Uses
- Accuracy Assessment when map is completed
- Internal Validation concurrently with mapping
effort - Large landscape monitoring and planningscales of
1 100 k 1 250k
36IV. Timeline Summary
- 2000 Work began
- 2001 Field data collection imagery acquisition
- 2002 Field data collection, imagery acquisition,
land cover mapping began - 2003 Field data collection land cover mapping
continue - 2004 Land Cover map complete. Final products
complete
37Comparison with 1995 Utah GAP Vegetation Map
Product
- 1995 GAP Veg. Map
- Utah state boundary
- 36 land cover types
- Re-sampled to 1 km resolution
- Distributed on CDROM
- 2003 GAP Veg. Map
- Independent of state boundaries
- Anticipated 50 cover types
- 30 meter resolution with 0.5 ha MMU
- Distributed via WWW and CDROM
- Improved Accuracy
- Nationally consistent vegetation classification
(NatureServe NVCS)
38Acknowledgements
- Doug RamseyProject Principle Investigator
- Lisa LangsGraduate Student
- Wendy RiethGraduate Student
- Jessie DenhamField Botanist
- Marie DucharmeField Botanist
- Gerald ManisPlant Ecologist/Mapper
- Chris GarrardProgrammer
- Rob Johnson--Cartographer
- Eric SantGraduate Student
- Chris McGintyGraduate Student
- Jarom GilbertUndergraduate Student
- Sheryl BoyackUndergraduate Student
- Wendy HurdUndergraduate Student
- Meg PoulsonField Technician
- Todd SajwajGraduate Student