 |
|
|
|
|
|
|
|
|
|
|
|
|
 |
|
|||||||||||
 |
 |
 |
|
|||||||||
 |
|
|||||||||||
 |
 |
 |
|
|||||||||
 |
|
|||||||||||
 |
 |
 |
|
|||||||||
 |
|
|||||||||||
 |
 |
|
||||||||||
 |
|
|||||||||||
 |
 |
 |
|
|||||||||
 |
|
|||||||||||
 |
 |
 |
|
|||||||||
 |
|
|||||||||||
 |
 |
 |
|
|||||||||
 |
|
|||||||||||
 |
|
|||||||||||
 |
|
|||||||||||
Change Detection for Coastal North Carolina Using Landsat Thematic Mapper Data
Funded by National Oceanic and Atmospheric Administration (NOAA) Coastal Change Analysis Program (C-CAP), through the North Carolina Sea Grant Program, and the North Carolina Division of Coastal Management (DCM)
Principal Investigator: Siamak Khorram
Project Scientists: Heather Cheshire, Xiaolong Dai, and Jeff Morisette
OBJECTIVE
The goal of this project was to develop a regional land cover change database for coastal North Carolina. Primary objectives were: to develop a wall-to-wall land cover product for the Coastal Plain of N.C. utilizing current TM data. to analyze changes in land cover between Tb (1993/1995) and Tb-1 (1987/1989), and to assess and document accuracy of the current land cover product and the change database.
APPROACH
Multitemporal Landsat TM scenes covering the costal plain were acquired, processed, and classified into land cover and changes classes. Gepcoded, terrain-corerected TM data were acquired for leaf-off (winter) and leaf-on (summer or spring) conditions. The geo-referencing accuracy of each scene was checked and the corresponding scenes at two times were then co-registered. The binary change mask was developed using the technique of Ellipsoidal Change Detection based on the Mahalanobis distance of multispectral difference images. The five TM scenes within each seasonal coverage were normalized and mosaicked. The coastal plain was then stratified into urban and non-urban areas; non-urban areas were also stratified by physiographic regions. Urban and non-urban areas were classified separately into nine classes usiing supervised and unsupervised approaches. National Wetlands Inventory digital data were incorporated to identify estuarine and palustrine wetlands. In cooperation with the DCM, CEO developed field data requirements and field tally sheets for collectoin of ground data used for image classification and accuracy assessment. Classificatioin accuracies were determined based on data from over 740 sample sites.
Landsat 5 Thematic Mapper FCC of mosaicked winter images (1993-1995) covering the Coastal Plain of North Carolina Land cover classification of Tb data for NC Coastal Plains, 16 classes of C-CAP protocol
The Spatial Modeler Language in Erdas Imagine has extensively been used in module development in the project. Click on the thumnail for the model for Change Mask Module. Areas identified as having spectrally changed from time Tb-1 to time Tb in the entire study area Areas identified as having actual changes in the ground from time Tb-1 to time Tb in the entire study area. Thirty from-to categories are represented in this image. RESULTS
- Complete land cover classification of the coastal plain from current TM data
- Accuracy assessment of the current land cover classes
- Identification of areas which have changed between the previous tiime and the current time. the accuracy assessments of the classified areas for the previous time period and the from to change matrix for predominant change categories. The resultant land cover product adheres to C-CAP classification protocols. Satisfied classification accuracy for Tb and excellent change detection accuracy were achieved
- Two Ph.D. dissertations related to this project were successfully completed
- Eight presentations have been delivered at ASPRS and IEEE/IGARSS international conferences
- More than ten articles have been published in prestigious refereed journals and national and international conference proceedings.
Contact the Webmaster Contents Copyright © 2000 NC State University
