CICEET Progress Report for the period 3/15/05 Through 9/15/05
Project Title: Discriminating and Mapping Invasive Plant Species Using Advanced Remote Sensing Technology
Principal Investigator(s): Stephen D. Smith
Additional Investigator(s): Eugenia M. Barnaba, Milo E. Richmond, Magdeline Laba, Susan Hoskins
Project Start Date:1 September 2003
Project Objectives for This Reporting Period
Objectives
- Develop vegetation community maps for each HRNERR
- Complete accuracy assessment
- Undertake comparative analysis of satellite image and aerial photographic vegetation community maps
- Finalize technology transfer component and implement
- Prepare final report
Tasks to meet objectives
1. Develop vegetation community maps for each HRNERR
- Continue processing of satellite imagery
2. Complete accuracy assessment
- Develop accuracy assessment methodology
- Assemble field crew and equipment needed
- Analyze field data
3. Undertake comparative analysis of satellite image-based and aerial photographic-base vegetation community maps.
- Finalize development of comparative boundary analysis methodology Implement methodology
4. Finalize technology transfer component and implement
- Finalize agenda and logistics
- Prepare workshop materials
- Present workshop
5. Prepare final report
Progress on Tasks
Develop vegetation community maps for each HRNERR site.
Utilizing the training data acquired from the 2004 field season, the Quickbird images, collected during the same season as the training data, were processed to produced a maximum likelihood classification product (as outlined earlier). These products utilize the same classification scheme as the 1997 vegetation mapping developed from aerial photography. Satellite image based vegetation community maps were produced for the Stockport Flats, Tivoli Bays, Iona Island, and Piermont Marsh HRNERR sites.
Complete accuracy assessment
a. An accuracy assessment is required to verify the integrity of the image-base map product. Issues and methodologies were discussed during the April 6, 2005 Advisory Committee meeting. Input from that meeting and subsequent meetings with individual committee members assisted in the development of the accuracy assessment methodology. Collection of accuracy assessment data was undertaken during the period of June 23rd September 15th, 2005.
b. Accuracy assessment is being performed on vegetation community polygons derived from the maximum likelihood classification product. Target polygons were selected using a stratified random sampling method. Not every vegetation type is present in every HRNERR site. An initial sampling method was devised with a goal of 50 polygons per vegetation community type across all HRNERR sites. If all community types were equally present then 1000 polygons would need to be visited. Given time and cost constraints, the number of polygons had to be reduced. This reduction was accomplished by removing community types for which accuracy assessment was deemed unnecessary due to their obvious signatures (e.g. upland), temporal consistency (e.g. railroad), or impossible delineation due to tide levels at the time of image acquisition (e.g. un-vegetated flats).
c. A two-person field crew was assembled, outfitted with a vehicle, canoe, GPS unit, cell phone and other equipment. We were fortunate to obtain Roger Downs for the Crew Chief position again this year. Having led the training data acquisition crew in 2004, Roger is very familiar with all four sites which aided in the successful completion of the 2005 field season. The Data Manager for this season was Dane Klinger (graduate student in Environmental Studies, Bard College, Annandale, NY). The project orientation covered an overview of the project, accuracy data collection requirements, safety, and management procedures to be followed. The orientation was given by Stephen Smith (IRIS) and Chuck Nieder (HRNERR Research Coordinator). The HRNERR facilities at the Bard College Field Station provided a base of operations and Chuck Nieder served as primary contact for daily operations.
d. After the initial week of field collection, data were reviewed and procedures verified by Magdeline Laba during a visit to the Field Station. To insure data quality and coordinate data flow to and from the field crew, visits were scheduled approximately every two weeks throughout the field season. Many of these visits were accomplished in conjunction with other projects in the area.
e. The field crew was able to visit virtually all of the 700+ polygons they were assigned. Only a few sites were not visited due to tide and time constraints.
3. Undertake comparative analysis of satellite image and aerial photographic vegetation community maps
An investigation of the comparative boundary analysis issue is ongoing. Current plans include a methodology incorporating a systematic point sampling applied to each map to identify polygons for which various characteristics will be assessed and compared. Specific characteristics and assessment methods are being developed from shape metrics such as: area, perimeter, area-perimeter ratio, and comparison to standard.
4. Finalize technology transfer component and implement
Coordinating with HRNERR staff to identify best possible workshop time and location. A tentative agenda has been developed. CICEET project team is developing presentations designed to concisely explain the project goals, objectives, and results. Specific methodologies will need to be imparted through additional hands-on exercises.
5. Prepare final report
Preparation of the final report will begin as soon as the results from the accuracy assessment and map comparison are available and workshop evaluations have been compiled.
Accomplishments
1. Over 700 field data observations have been collected this summer to be used as part of the accuracy assessment process. Satellite image-based vegetation community maps have been produced for all four HRNERR sites.
2. Development of the comparative boundary analysis methodology is progressing and will be ready to implement early this fall.
3. Workshop planning has begun.
4. A CICEET Advisory Committee meeting was held at Cornell on April 6, 2005. Focus of the meeting was a review of a tentative accuracy assessment methodology and discussion of foreseen difficulties.
Difficulties
At this time we are experiencing no unforeseen difficulties. We were able to assemble a field crew to collect accuracy assessment data this summer. This will again result in a re-alignment of some budget expenditures.
Project Objectives for Next Reporting Period
Objectives
- Complete accuracy assessment evaluation
- Undertake comparative analysis of satellite image and aerial photographic vegetation community maps.
- Implement regional workshop
- Prepare final report
Tasks to Meet Objectives
1. Complete accuracy assessment evaluation
- Analyze field data and develop final accuracy assessment tabulations
- Finalize development of comparative boundary analysis methodology
- Implement methodology
- Finalize agenda, logistics and participant list
- Complete preparation of workshop materials
- Host workshop
Work Plan for Next Reporting Period
Work plan to follow the above tasks and proposal timeline.
Anticipated Success in Meeting Project Objectives
No anticipated problems in meeting project objectives as outlined above.
Overall Project Timeline Update
Project is currently running under a no-cost extension. The new project termination date is January 31, 2006. This extension was requested to: 1) permit the collection and processing of accuracy assessment points during the 2005 field season, 2) enable the planned workshop to incorporate all the results of the project, and 3) permit the scheduling of the workshop at an optimum time for NERR staff participation.
Preliminary Data
Acquire satellite imagery
High resolution (2.6m multi-spectral and 0.6m panchromatic) images of all HRNERR sites.
Piermont Marsh 8/04/04
Stockport Flats 8/22/04
Tivoli Bays 8/22/04
Iona Island 9/04/04
2. Collect training data for image analysis - 2004
Vegetation community type identification data for 398 individual field locations and an additional 74 delineated polygons was collected. All field data referenced with a GPS units to nominal 3-5 meter accuracy.
3. Development of vegetation community maps
Vegetation community maps have been developed for each of the four HRNERR sites based on 2004 high-resolution satellite imagery. These maps use the same 20 class scheme developed for the 1997 aerial photographic mapping.
4. Collected accuracy assessment data - 2005
Vegetation community type identification data for 700+ polygons was collected. All field data referenced with a GPS units to nominal 3-5 meter accuracy.
Dissemination
Dissemination of data or results is premature at this time.
Expenditures
Overall expenditures are within the budget. The need to fund field crews for 2004 and 2005 has required a substantial re-alignment of specific expenditures.