Progress Report

CICEET Progress Report for the period 9/30/04 Through 3/15/05

Project Title: Development of an Automated Mapping Technique for Monitoring and Managing Shellfish Distributions
Principal Investigator(s): Steven R. Schill, Dwayne E. Porter, Loren D. Coen, Dave Bushek

Figures
Figure 1

Figure 1
Figure 2

Figure 2
Figure 3

Figure 3
Figure 4

Figure 4
Figure 5

Figure 5
Figure 6

Figure 6
Figure 7

Figure 7
Figure 8

Figure 8
Figure 9

Figure 9
Figure 10

Figure 10
Figure 11

Figure 11
Figure 12

Figure 12
Figure 13

Figure 13
Accomplishments
Scheduled Tasks
  • LiDar Data classification of intensity and elevation data is nearly complete.
  • Draft manuscript of J.Vincent’s dissertation on the use of Mixture Tuned Matched Filtering with HyMAP and AISA imagery to map shellfish is nearly complete. Defense is expected before the end of the current spring semester.
  • Automation of feature extraction using Feature Analyst is on-going.

Progress on Tasks

  • Summary of the research and documentation to create a dissertation has occupied a majority of the time spent on this project in the last several months. Progress has been made in the results and conclusions sections of the dissertation. The dissertation proposal was initially based upon using linear spectral unmixing. Since then it was found that using Mixture Tuned Matched Filtering is a more efficient method of mapping. Linear spectral unmixing assumes all endmembers and endmember variability is accounted for in the image. Due to this change in the research methodology, the literature review was also changed to reflect the change in methodology. Work is still progressing in the quantitative analysis of spectral endmembers. ANOVA statistical measures were first explored but found to be unsatisfactory when comparing spectral curves. Currently spectral binary encoding methods have been employed but methods of quantifying differences using this method is being explored.

  • Analysis of LiDAR data is past the exploratory stage and a scheme for subset and analysis of the data is now ongoing. As many areas of interest that can be subset and classified, (shellfish patch reefs) will be used to derive training files for the Feature Analyst portion of the research. Currently, four areas of interest have been subset and analyzed. We anticipate another five or six areas of interest will be used to complete the LiDAR training set.

  • We plan on using the output from Mixture Tuned Matched Filtering in ArcGIS coverage or shapefile format as training files within the Feature Analyst extension. In addition, training files from the classification of the LiDAR intensity data will be used along with the MTMF training files. Currently only three LiDAR sample sites have been classified and results using three sample sites have not had a satisfactory result, more classified sample sites are needed to further the analysis.

Difficulties Encountered

  • Dissertation;
    A portion of this research is being utilized for a doctoral dissertation. Specifically, the derivation of endmembers and mapping of using Mixture Tuned Matched Filtering mapping methods with HyMAP and AISA imagery is within the framework of the doctoral research. Difficulties encountered include finding an adequate method to compare field and imagery derived spectral endmembers. Normal statistical measures fall short of adequately comparing spectral curves but methods of comparison include spectral binary encoding, autoregressive spectral analysis, and Chi-Square Goodness of Fit analysis. Currently, the binary encoding method is being explored as a viable option.

    Field Surveys for accuracy assessment of classified shellfish habitat was delayed last fall due to weather. It was decided to wait until the LiDAR data was classified and then an accuracy assessment of the HyMAP, AISA and LiDAR classified imagery will be carried out simultaneously.

  • LiDAR;
    Classification of LiDAR data has taken longer than expected due to the unique nature of the research. Statistical data exploration revealed a high degree of skewnes in the data, which was corrected by a logarithmic transformation of the data. Initial focus was on the elevation data and while the data yielded exceptional visual interpretations of the study area elevations, the elevations by themselves were insufficient for discriminating finer classes of shellfish. A small problem encountered was the level of accuracy in deriving areas of interest without the inclusion of water or vegetation. Elevation data has been shown to accurately discriminate between water and vegetation to subset the LiDAR intensity points into areas of interest. The intensity return values were used classify shellfish. The intensity return values are not calibrated to any external reference standard but differences in the intensity of the energy returning to the sensors are sufficient enough to discriminate between water, mud, and shellfish. The LiDAR data sets are two flight lines of 0.25ft spacings. One flight line is comprised of ten tiles with each tile have approximately 1.5 million data points. Initially, the large amount of data points and file sizes hampered analysis but creating subsets of the data has eased the computation overhead. Another encountered problem is the noticeable lack of either geographic information systems or remote sensing fields to develop algorithms that can handle the large numbers of points. Traditional point algorithms cannot accommodate the number of points in a LiDAR data set.

  • Spectral Library;
    Getting student help to provide data entry of the spectral library has been lacking. While not a high priority, the input of the shellfish data library is important in the dissemination of the data and completion of a large portion of this research.

Anticipated Success in Meeting Project Objectives
We anticipate being able to finish the dissertation and accuracy assessment in a timely manner. Entering the spectral library and metadata in to the CDMO database is a bit more nebulous due to finding a graduate research assistant to enter the data. Once an assistant has been found we will have a better idea of a completion date for this task.

We do not anticipate any delays in the final analysis utilizing the output from the different classifications.

Preliminary Data
See Figures.

Tasks and activities for next reporting period

Tasks for the next reporting period

  • Defend dissertation before the end of the 2005 spring semester and complete field accuracy assessments.
  • Spectral library entered into Central Data Management Office database.
  • Three articles for publication
  • Final report for CICEET research compiled

Work plan to accomplish tasks

  • Finalize draft dissertation within the next three weeks and complete field accuracy assessment, (two days in field). Defend by end of April, 2005.
  • Utilize graduate research assistant to input spectral library into database. Accomplish this task by mid-summer.
  • Finalize drafts of three manuscripts: 1- Using MTMF to classify Intertidal Shellfish Resources, 2- Integrating Automated Feature Extraction And Spectral Mixture Analysis, and 3- Shellfish Classifications Using LiDAR Intensity. Have manuscripts accepted for publication by mid-summer.

Concerns or difficulties
Finding a graduate research assistant may have to coincide with the summer term when the current semester expires. The timeline for finalizing draft manuscripts and getting them accepted to a peer-reviewed journals is subject to delays due to the sending and receiving of manuscripts and delays with the publisher.

Expenditures
All expenditures were in the range anticipated for the work accomplished to date. As of March 15, 2005, we have billed for approximately 75% of the total budget and the remaining funds have been rolled over into year 3 extension. We anticipate the need and expenditure of the total remaining budget in year 3 work based on our no-cost extension that has been approved through August 2005.

SCDNR:
Total Budget: $45,745.00
Billed to Date: $28,195.03
Remaining: $17,549.97 (38%)

USC:
Total Budget: $99,266.00
Billed to Date: $57,919.59
Remaining: $41,346.41 (42%)