by: geotrek. hunter krenek: remote sensing analyst & gis analyst joe dowling: assistant project...
TRANSCRIPT
By: GeoTrek
Implementation of a Geodatabase for Freeman
Center
Hunter Krenek: Remote Sensing analyst & GIS analyst
Joe Dowling: Assistant Project Manager & GIS analyst
Peter Vogt: Website Designer & GIS analystAlfredo Perez: Project Manager & GIS
specialist
Team Members
Purpose: Build a functioning geodatabaseCreate a brochure with map atlasLand cover classification of huisacheThis project is designed to enable future
research and assist in the maintenance required to keep up the ranch.
The map atlas and brochure will also be used to help navigate the ranch to visitors, research partners, and stakeholders
Project Description
Freeman CenterFreeman Center is
4,200 acres of land in the Texas hill country
That is owned by Texas State University
Goals are to provide effective stewardship of the center’s ecosystem and infrastructure
Task 1: Building a Geodatabase
Advantages of ArcGIS
Digital copy of filesPerform spatial
analysisOrganizationAttribute tablesFile type Display with multiple
projections
Task 1: Building a Geodatabase
Cultural Features
Roads Boundary Fences Creeks Buildings Pastures Wells Drinkers
Cultural Features
Data Texas Parks and Wildlife Department (TPWD)Texas Natural Resources Information System
(TNRIS) Capital Area Council of Governments
(CAPCOG)Digital Ortographic Quarter Quad (DOQQ) .kmz file from Google EarthGPS data collection
Data & Methodology
Our First step was to convert the .kmz files to shape files that could be used in ArcMap.
Data & Methodology
Ranch Bondries.kmz Ranch Bondries_1.lyr
The newly created shapefiles were then used to populate the geodatabase.
Data & Methodology
Creating TopologyA topology feature class was created to set
rules for feature relationships.
Data & Methodology
Creating TopologyRules
Must not have gapsContains pointsMust be insideMust not have danglesMust not self-intersectMust not overlapMust be properly inside boundary
Data & Methodology
Creating TopologyRun error reports based on rules created in the
topology feature classFix or omit errorsTopology edit tools used:
ExtendTrim
Data & Methodology
Creating Feature Classes A polygon feature class was created to
encompass the pastures and buildings present on the ranch.The pasture feature class was created based of the
fences layer digitized from the original .kmz.The buildings feature class was created by
performing a trace while using the 6’’ satellite imagery as a georeference.
Data & Methodology
Edit Attribute TablesFor the roads, fences, pastures, and creeks
layers the calculate geometry tool was used to convert distance from decimal degrees to meters and acres.
For the fence layer the field calculator was also used to categorize the fences into high fence or low fence.
Data & Methodology
Data & Methodology
The boundary layer was used to extract by mask the DOQQ to limit the extent to our study area.
Data & Methodology
Data & Methodology
6’’ satellite imagery was used to georeference the locations of features such as: Buildings
DrinkersWellsFencesRoads
Website Compilation Data & Methodology
Data & Methodology
Manifold was used to generate a map document that was then exported to a website file.
The Freeman Center is experiencing a huisache encroachment on its property.
The Freeman Center has requested a land cover classification for huisache be created for inclusion in the geodatabase provided by GeoTrek.
Task 2 Abstract
2008 and 2012 1m resolution NAIP Imagery (TNRIS)
Freeman Center Boundary Line Shapefile SSURGO Soil Survey ShapefileHuisache tree GPS points
Data
Obtain the NAIP imagery from 2008 and 2012Create mosaic images using Erdas Imagine
Extract AOI from mosaic imagesThe extracted false-color composite images of the
Freeman Center were then used to perform a landcover classification with the following classes:WaterDevelopedBarrenForest/ShrublandHerbaceous (grassland)
Methodology
Training dataMade using the Grow Region and Polygon Tools
with the Signature Editor in Erdas Imagine Google Earth used to make a visual confirmation
of the landcover type. Due to poor spectral separability in the
training data the Forest and Shrubland classes were merged to yield one class with greatly improved separability.
Methodology
Accuracy Assessment:Accuracy Assessments were performed with an
expected accuracy of 80% and an acceptable error of 10%.
A sample size of 64 reference points was derived using Binomial Probability Theory.N=[22 (80)(20)]/102
Methodology
After completion of the initial landcover classification, the Forest/Shrubland class was extracted to narrow the AOI for the huisache classification.
The GPS point data gathered on site was then used in creating training data. Identified huisache trees in the false-color
composite images
Methodology
Spectral Separability of Huisache vs Mesquite:Positively identifying the preferred soil types of
huisache and mesquiteGPS Data Point Sampling Design:
Training Data and Accuracy AssessmentsDetermining the Best Imagery to Use:
NAIP imagery:Good:
1 meter spatial resolution Publicly available
Bad: First included the NIR band in 2008 2008 images comprised of 3 bands, eschewing the blue band in
exchange for NIR.
Constraints and Limitations
Task 1:Implement functioning geodatabaseAttribute tables with correct measuring units
for each feature I e: pastures = acresMaps
FEMA Q3 Navigation SoilsEcological Systemswebsite with interactive map
Results
Functional Geodatabase
Results
Correct Unit Measurement in Attribute Tables
Results
Correct Unit Measurement in Attribute Tables
Results
Database Composite
Fences
Roads
Drinkers
Wells
Pastures
Buildings
Texas Ecological Systems Map
Remote Sensing analysis produced the following:Classified 2012 Map Image
Training Data with polygons2012 NDVI Map Image2012 Tasseled Cap Transformation Map ImageMethodology Constraints and Advice for Future
Researchers
Results Task 2
2012 Classified Map Image
Accuracy = 85.94%
Kappa statistic = .8187
2012 Tasseled Cap Transformation
Map Image
2012 NDVI Map Image
Interactive web map
Task 1-Before our team started working on FreemanCenter.gdb., The Freeman Center only possessed .kmz files created using GoogleEarth, and had zero spectral data. The attribute fields oflength and area were calculated from degrees into meters andacres respectively. The Freeman Center can now have access toa geodatabase that can be used to make maps and for surface modeling. Task 2-A viable Freeman Center huisache tree classification will be a topic for ongoing research at Texas State University. Although this initial investigation met with setbacks, the project report will contribute to future progress in identifying the spec-tral signature of this species.
Conclusion