HOME      

     RESEARCH 

         LINKS     

       RESUME   

For my Masters thesis project, I examined the use of high-resolution satellite imagery and ancillary data for creating detailed land-cover and land-use maps. I primarily worked with eCognition software for all classifications and also used ArcGIS and Erdas Imagine for some analytical, image pre-processing, and cartographic tasks.

Brief project description:
I analyzed Ikonos imagery for a part of Mississauga, Ontario,  to create a land-cover map and then derived land use from land cover.

Ancillary data (i.e. building and road data) were included in the classification and assessed in terms of their accuracy improvements for the land-cover classification. An object-oriented approach was used throughout the classification process. The object-oriented approach allows for the grouping of pixels into homogenous image objects on several scales. These image objects were classified in terms of their characteristics (i.e. shape, texture, size, spectral values, etc.) to generate three land-cover maps with a total of 17 classes and varying inclusion of ancillary data.

The most accurate land-cover map built the base for the land-use map. Through examination of how the different land-cover classes relate to one another spatially, 10 land-use classes were distinguished. For example, large buildings surrounded by parking lots and little vegetation along a major street indicated a industrial or commercial land use. In similar fashion, all other land-use classes were extracted.

Thesis excerpts (links open pdf files)

• Abstract, Acknowledgements, Table of Contents
   

• Satellite imagery pre-processing
   

• Flowchart, outlining general project steps
   

• Flowchart, outlining land-cover classification process
   

• Final land-cover map (using road ancillary data)
   

• Final land-use map (10 classes)

Presentation of preliminary results at the AAG Annual Conference (Chicago, March 2006)
Illustrated Paper Presentation

(Click on image for bigger pdf version)

Undergraduate Research
1) "Urban Forest Canopy Cover: Portland, Oregon, 1972-2002"
                                       2) "2002 Urban Forest Canopy & Land Use in Portland's Hollywood District"

1)
During my undergraduate years, I worked on a research project with Prof. Joe Poracsky, analyzing urban forest canopy cover in Portland, OR, over a three-decade period. This project gave me a wonderful opportunity to explore various aspects of remote sensing and GIS while, at the same time, helped me to financially support my studies.

Brief project description:
Landsat TM and MSS data for three different years were classified in terms of forest cover. Then canopy extents were compared and analyzed on a neighborhood level. It was observed that some neighborhoods declined in canopy extent, while others (particularly older and well-established neighborhoods) showed an increase in their canopy extent.

Canopy cover was also related to different land-use classes in order to derive recommendations for canopy-cover targets for the respective land-use classes.

More information about this project can be found on the website
Urban Forest Canopy Cover: Portland, Oregon, 1972-2002, which was the result of a course I took about presenting geographic data online. The website also includes a link to the final project report.

The project also led to several presentations, including a poster presentation (The poster was awarded "Best Student Entry" among all student posters.) at the "GIS in Action" conference.
                  (Portland, OR, April 2003)

(Click on image for bigger pdf version)
 

2)
Receiving additional funding through an "Undergraduate Research and Creative Activity Award," I explored Portland's urban forest in more detail.

Brief project description:
I  first classified tree-canopy cover in Portland's Hollywood district, using multi-spectral one-meter resolution aerial photography. In a second step, I compared canopy-extents of three different land-use classes. Finally, I suggested possible tree-planting sites, based on calculations of average-sized columnar and spreading trees.

The final report "2002 Urban Forest Canopy & Land Use in Portland's Hollywood District" includes more detail about this project.

Results were also presented in poster format at the
"GIS in Action" conference.
             (Portland, OR, May 2004)

(Click on image for bigger pdf version)

Term project for GGR 1907 (Advanced GIS - UofT)

The project explored determining random sample points for different clusters of an unsupervised classification and then connect them through a cost surface in the most time-efficient way possible.

The modeling approach described in the project report can be used to efficiently plan sample-site visits during field research.

(Click on image for pdf version of final report)
 

Mapping historic US travel routes (UofT)

As part of some contract work, I used the ArcScan extension in ArcMap to semi-automatically digitize line features from scanned paper maps that also had to be georeferenced to modern-day US boundaries.

The ArcMap screenshot shows small zoomed-in portions of the final maps, which are part of ongoing research and are not yet published.

(Click on image for bigger pdf version)

In this project, several landscape metrics were evaluated by exploring ovenbird habitat in Southern Ontario. Correlation of ovenbird abundance and 19 different landscape metrics and variables revealed that certain metrics are better indicators for ovenbird breading habitat than others.

(Click on image for pdf version of final report)
 

As part of my capstone course in the final year of my undergraduate studies, I was responsible for creating a map of the Johnson Creek watershed in Portland, OR. The map shows point locations of streets crossing Johnson Creek and major arterials crossing the creek, among other features.

(Click on image for bigger pdf version)