Program: FRAGILE ENVIRONMENTS
Project Title: Land Characteristics from Remote Sensing
Location of Study Area: South Florida/Everglades
Project Start Date: March 1996
Project End Date: October 1999
Project Number:
Project Chief: Greg Desmond
Region/Division/Team/Section: Eastern Region/National Mapping Division/Mapping Applications Center/Science and Applications Branch
E-mail:gdesmond@usgs.gov
Phone: 703/648-4728
Fax: 703/648-4165
Mailing Address: M.S. 521 National Center, Reston, VA 20192
Program Element(s)/Task(s)
The task number assigned to this work is 1.10 Land Characteristics from Remote Sensing.
However, we treat this task as three components. The element/task to which each component contributes and the amount of effort allotted each component follows:
1) Vegetation mapping for hydrologic modeling (task 1.5), 45%.
2) Evapotranspiration (ET) regionalization for hydrologic modeling (task 1.4), 45%.
3) Periphyton mapping for geochemical research and environmental monitoring (element 3), 10%
Panel:
Collaborators, Clients:
Collaborators:
Water Resources Division researchers (e.g., Edward German) will provide technical/science expertise and will evaluate the utility of generated data and information products. Personnel costs are expected to be covered under Task 1.4).

Geologic Division and Water Resource Division researchers (e.g., Dr. Nancy Simon) will provide expertise regarding periphyton characteristics, pigment analysis, etc.. Personnel costs are expected to be covered under element 3. Specific WRD collaborators are currently being solicited from the Mercury programs and on-going research activities in South Florida.

Researchers at the University of Maryland (e.g., Dr. Ralph O. Dubayah) will provide technical and scientific expertise regarding satellite remote sensing, climatology, and hydrology. No personnel costs will be incurred.

NASA scientists will provide technical expertise in hyperspectral data analysis and vegetation mapping as part of current DOLNASA collaborative efforts. We are currently in the process of identifying specific collaborators through on-going discussions with NASA headquarters. No personnel costs will be incurred.

Corps of Engineers researchers (e.g., Dr. John E. Anderson) will provide assistance in DMSV data collection and processing. Personnel costs are covered by data acquisition funding.

Clients:
USGS hydrologic modelers
Everglades National Park
South Florida Water Management District
Academic Process Modeling Community
Other State and local researchers
General public

BACKGROUND NARRATIVES

Project Summary: Remote sensing, image processing, geographic information system technologies and current research into hydrological, climatological, and ecological applications of these technologies will be employed to derive scientific information used in describing, understanding, and modeling vegetation patterns and hydrologic processes for the purposes of natural resource decision making and restoration monitoring in South Florida.

Project Justification: The extrapolation of processes typically measured or modeled at point locations or at micro scales to macro scales is an extremely difficult undertaking. However, this capability is necessary in order to identify the important components of the natural system, quantify the impacts of human activity on the system, forecast system behavior, and monitor the effects of restoration actions. This work will develop the techniques and produce the data sets necessary to conduct hydrologic modeling (surface water flow and water budget) at the regional scale. Additionally, success in developing periphyton mapping techniques would produce a critical new tool for biogeochemical and water quality research being conducted in tasks 3.3, 3.5, and other programs within the USGS (for example the WRD Mercury Initiative).

Project Objectives: In-situ and remotely sensed data from numerous platforms and sensors, each possessing different spatial, temporal, and spectral resolution, will be processed, analyzed, and combined to produce fields of information about biophysical variables (e.g., vegetation species composition, vegetation density, vegetation structure, and components of the surface energy balance). These data will be evaluated for their utility in specific modeling and monitoring contexts. Developed techniques will be transferred to other scientists and management agencies via existing program and other professional outlets. The data sets developed will be documented using established metadata standards and made available through collaborative research and the results of task 1. 11 (Ecosystem Database Development).

Overall Strategy, Study Design, and Planned Major Products: The overall task has been divided into three primary components with specific objectives and general classes of clients. For each component, we will systematically develop, evaluate, and apply the techniques which yield the most appropriate, spatially distributed information possible on the vegetation, climate, and hydrologic variables of interest to South Florida project scientists. The study designs for each component include the combination of field campaigns (to provide information for data calibration and accuracy assessment) and extensive techniques development using state-of-the-art hardware, software, and physical process modeling. Collected data will be processed (e.g., to mitigate atmospheric effects and conduct data fusion to infer biophysical fields) to maximize their information content. Collaborative use of the resulting information products will allow us to evaluate their utility for process modeling and environmental monitoring while facilitating outreach and technology transfer.

WORK PLAN

Overall: This overall project consists of three components. The work plan for each component is described individually below.

Component 1: Vegetation Mapping Study:
Vegetation density information over large areas is needed for large scale hydrodynamic models to help quantify the resistance vegetation offers to water flow. The approach is to use multisensor data fusion and other classification/segmentation techniques to improve vegetation classification performance obtainable from moderate resolution, single sensor data. High resolution (0.5m), aircraft collected, digital multispectral video (DMSV) data for small study areas and on site data are being used for ground truth. Major effort is directed towards the sensor data integration and classification of derived texture information from synthetic aperture radar (SAR) and multispectral imagery. Products include proof-of-concept examples and reports on advanced techniques for generating land cover/vegetation density maps, texture based segmented SAR imagery for interpretation, and enhanced/merged high resolution (6m) SAR -and- multispectral imagery.

Component 2: Evapotranspiration (ET) Regionalization
Methods currently being developed using satellite systems for ET modeling and mapping span a broad range of complexity. Empirically-derived statistical relationships, physically- based analytical approaches, numerical models, and the exploitation of observed relationships between radiant surface temperatures (Tr) and spectral vegetation indexes like the Normalized Difference Vegetation Index (NDVI) are all being investigated. In pursuit of the objectives of task 1.4, several recently developed methods will be implemented in the South Florida environment to generate spatially and temporally distributed fields of relative and absolute evapotranspiration. These fields will in turn be compared against ET modeled from the data collected under task 1.4. Standard statistical techniques will be used to assess the accuracy and precision of the estimates generated using the remote sensing inputs. Finally, through our own modeling efforts and those of our collaborators, we will evaluate whether these data are adequate for anticipated modeling requirements.

Component 3: Periphyton Mapping Feasibility Study
Expanding existing data collection and analysis in a "new direction", we will determine the efficacy of mapping periphyton composition and extent using prototype hyperspectral data sets and advanced processing techniques. The overall goal of this component is to develop and evaluate hyperspectral analysis tools for mapping characteristics of periphyton (e.g., its composition, distribution, and its community associations) which are useful in monitoring environmental quality and modeling ecological processes. The 1998 research constitutes a feasibility study to determine the potential for the hyperspectral capability. FY 1999 research would develop means of extrapolating the techniques through space and time.

Timeline:
Vegetation Mapping Timeline:
FY 1998
1) Collection of remotely sensed satellite imagery (Lemeshewsky)
2) Develop and apply image preprocessing tools (Lemeshewsky)
3) Distribute preliminary vegetation density classification from multispectral data and SAR
data for review (Lemeshewsky).
4) Produce improved classification using multidate data (Lemeshewsky)

FY 1999
1) Apply developed techniques to a broader region (Lemeshewsky)

Evapotranspiration (ET) Regionalization Timeline:
FY 1998
1) Data collection (Jones, German).
2) Data calibration to radiance, reflectance, and radiant surface temperature (Jones).
3) Regression analysis using in-situ and satellite-derived meteorological data (Jones/Dubayah).
4) Development and testing of "simplified" method of ET estimation directly from satellite data (Jones).
5) Comparison of developed methods and evaluation of their utility in water balance modeling (Jones).

FY 1999
1) Acquisition of soil-vegetation-atmosphere transfer model (Jones).
2) Model calibration using subset of ground-based meteorological data (Jones).
3) Application of model driven by satellite derived meteorological fields (Jones/Dubayah).
4) Model/data evaluation and testing (Jones/Dubayah/other modelers).

Periphyton Pilot Timeline:
FY 1998
1) Extend vegetation mapping to concentrate on macrophytic vegetation associated with periphyton (Jones, Desmond).
2) Collect existing AVHUS hyperspectral data flown over South Florida from NASA (Jones). 3) Calibrate hyperspectral data (Jones/NASA collaborator).
4) Derive absorbance curves at the pixel level using hyperspectral imagery previously collected through NASA funded investigations Jones/NASA collaborator).
5) Assess atmospheric and water column impacts on absorbance curves (Jones).
6) Develop adequate mitigation techniques for affects of atmosphere on the image data (Jones).
7) Develop techniques for comparing features generated from absorbance curves with those created through antecedent pigment analysis of field and laboratory samples to determine whether significant discriminating features present in both data sets (Jones/WRD collaborator).

FY 1999
1) Evaluate methods used to account for heterogeneity within pixels (e.g., mixture
modeling and deconvolution through derivative spectral analysis) (Jones).
2) Solicit additional hyperspectral flight missions and the acquisition of satellite hyperspectral data (from system to be launched in June, 1997) from NASA (Jones).
3) Search for relationships between features in absorbance curves generated from hyperspectral data and information in Digital Multispectral Videography, thematic mapper, and SPOT data collected coincidently (or during similar time periods) through on-going efforts under task 1. 10. This will determine if extrapolation of periphyton characteristics beyond the bounds of the hyperspectral imagery is possible (Jones/Desmond/Lemeshewsky).
4) Generate maps of periphyton cycles and distribution (Jones).

Planned Deliverables/Products:
Vegetation Mapping Planned Products:
Vegetation classification maps (type and density) and reports on advanced techniques used to generate them will be the primary products to be delivered under this component. A preliminary example of resolution enhanced TM thermal-infrared image; approx. 50mi. sq. miles around site NESRS3 (South of Cooper town Quad.) should be available at end of 4th quarter FY 1997. For the same site, texture based segmented SAR image maps should be available in 1st quarter FY 1998. Additional work on applied techniques using data fusion for improved classification of multisensor data will result in reports, vegetation map(s), and image products over a broader region.

ET Planned Products:
This component will generate data sets for use in method comparison and regional hydrologic modeling (first method results for review 10/97), software for operational output of useful fields (pending review of results - to final clients/general circulation 9/98), and additional visualization and educational products regarding ET related physical processes (review and release will be on-going) and our developed methods.

Periphyton Planned Products:
The type and amount of products generated by this component will depend on the success of techniques developed in each stage. That is, in addition to reports regarding progress and results, preliminary maps of periphyton distribution may be produced from the limited, available hyperspectral data (available for review 9/98 and release 12/98). If these products prove useful and methods for fusing other data products for broader (spatial and temporal) mapping of periphyton are developed, additional map, image, visualization, and report products will be generated (data products out for review third quarter FY 99).

Planned Outreach Activities: (for Land Characterization)
Presentation of techniques and results from each component will be made at numerous national-level conferences (e.g., AGU, GISALIS, etc..) pending normal application and review of abstracts and manuscripts. Seminars and lectures will also be provided in University and other public forums on watershed and ecosystem restoration and management. All derived data products will be documented and advertised via the database project and the South Florida Ecosystem web site.

New Directions, Expansion of Continuing Project (if applicable): See especially component 3 (Periphyton Mapping Feasibility Component) and the FY 1999 timelines of previous component descriptions under work plan above.

ACCOMPLISHMENTS, OUTCOMES, PRODUCTS, OUTREACH

Accomplishments and Outcomes, Including Outreach:
Outcomes:
Numerous image data sets corresponding to ground collection campaigns have been acquired and georeferenced in preparation for additional pre-processing and analysis.

NMD supplied project expertise has been expanded through the reassignment of a researcher to the Mapping Applications Center. This individual will devote the majority of his time to the Land Characteristics from Remote Sensing project.

Papers/Reports:
Reflectance Calibrated Digital Multispectral Video: A Test-bed for High Spectral and Spatial Resolution Remote Sensing. Highlight Article in Photogrammetric Engineering and Remote Sensing. 3/97. Authors: John E. Anderson, Gregory B. Desmond, George P. Lemeshewsky, and Donald R. Morgan.

Mapping Sawgrass Densities in the Florida Everglades using Spectral Data and Digital Multispectral Video. 16th Biennial Workshop on Color Photography and Videography in Resource Assessment - ASPRS Conference 4/97. Authors: John E. Anderson, Donald R. Morgan, and Gregory B. Desmond

Technique for spatial resolution enhancement of TM thermal band data. USGS Open-File Report (6/97). Author: George P. Lemeshewsky. Vegetation Density Classification from Clustered Multispectral and Resolution Enhanced Thermal Data using a Neural Network. Poster paper for the South Florida Ecosystem Study Symposium, 8/97. Author: George P. Lemeshewsky. Preliminary Spatially Distributed Estimates of Evapotranspiration using In-situ and Satellite Remote Sensing Data. Poster paper for the South Florida Ecosystem Study Symposium, 8/97. Author: John W. Jones

Outreach:
Full color front cover of Photogrammetric Engineering and Remote Sensing showing multispectral image of the impacts of the Value Jet crash on Everglades vegetation and water quality. 3/97.

Paper presentation at the ASPRS Biennial Workshop (4/97)

NEEDS

Required Expertise: Expertise in remote sensing, image processing, land cover mapping, plant geography, radiative transfer modeling, satellite hydrology, geographic information systems, and scientific visualization are all required to accomplish the objectives of this task.

Names of Key Project Staff:
1998 to completion
Greg Desmond, John W. Jones, George Lemeshewsky