Vol. 5, No. 3

December 2004

In this Issue

Basics		Using remote sensing to locate and monitor wetlands
Image Data Corner		U.S. Fish and Wildlife Service Wetlands Mapper
Tech Tip		Tips for selecting imagery
New Advancements		Advancements in Image Calibration
Case Studies		University of Hanover, Germany
Hot Sites		Web sites related to wetlands and remote sensing
AAI News		Announcing New Hires - AAI is pleased to announce the addition of a new staff member Upcoming Presentations Recent Presentations Sample Projects
AAI Corporate Overview		About AAI

From the Editor

This issue examines the use of remote sensing for mapping and monitoring wetlands. The term 'wetlands' refers to those areas "that are inundated or saturated by surface or ground water at a frequency and duration sufficient to support a prevalence of vegetation typically adapted for life in saturated conditions" (33 CFR 328.3(b); 1984). These areas, commonly referred to as marshes, bogs, and swamps, perform a wide variety of critical functions such as filtering pollutants and excess nutrients from surface runoff, maintaining streamflow during dry periods, reducing flood damage, minimizing shoreline erosion and providing wildlife habitat.

Less than half of the 220 million acres of wetlands that are thought to have existed in the lower 48 states in the 1600s still exist today. Major losses occurred during the mid-1950s to the mid-1970s as many wetlands were drained and converted to other uses. Although the rate has decreased, it is estimated that approximately 60,000 acres continue to be lost annually (EPA 843-F-01-002d, September 2001, http://www.epa.gov/owow/wetlands/pdf/threats.pdf ).

Remote sensing provides developers and local zoning entities with a valuable tool with which potential wetlands can be located. It provides a cost-effective means of identifying potential wetlands, thereby reducing the required coverage area to be further validated with field work and/or photo interpretation. As wetlands continue to come under pressure of development, the ability of remote sensing to reliably support field evaluations will become increasingly important.

We welcome your comments, ideas, and suggestions!  We are always looking to learn of new remote sensing applications with which our readers have been involved.  Please send email to Brenda.
 

Basics

NASA Research Shows Wetland Changes Affect Florida Freezes
November 19, 2004

Using satellite data, land cover change history, computer models, and weather records, the researchers found a link between the losses of wetlands and more severe freezes in some agricultural areas of south Florida. In other areas of the state, changes in land use resulted in slightly warmer conditions. They concluded, based on the study, the conversion of wetlands by itself may be enough of a trigger to enhance damage inflicted upon agriculture in these areas of south Florida during freezes events.

The Landsat 5 satellite was constructed and launched by NASA, and its data are processed and distributed by the USGS. The researchers studied three freeze events and simulated the conditions with a computer climate model, using weather and land cover change records. The freezes took place on December 26, 1983, December 25, 1989, and January 19, 1997.

The study, authored by Curtis Marshall and Roger Pielke of Colorado State University (CSU), Fort Collins, Colo., and Louis Steyaert of the USGS and NASA's Goddard Space Flight Center, Greenbelt, Md. appeared in a recent issue of the American Meteorological Society's Monthly Weather Review.

The researchers found a strong tie between areas that were converted from wetlands to agriculture use during the 20th century, and those that experienced colder minimum and longer duration subfreezing temperatures in the current land-use scenario. Water typically doesn't cool as quickly as the land at night, which may explain why when wetlands are converted to croplands the area freezes faster and more severely.

"The conversion of the wetlands to agriculture use itself could have resulted in or enhanced the severity of recent freezes in some of the agricultural lands of south Florida," Marshall said. He noted some other areas also experienced warming from land changes.

The study focused on "radiation freeze events" that occur at night, frequently under calm wind conditions and when there is little or no cloud cover. At night much of the warmth absorbed by the land during the day escapes into the atmosphere, cooling the ground.

This study of wetland changes is very important to Florida and the rest of the country. Over the past 150 years, the citrus industry has been moving further south. As a result more wetlands are being transformed into agricultural lands, and these changes are causing temperatures to change. Ironically, as the industry moves further south to avoid freezes, the land changes create the conditions the industry is trying to avoid.

The scientists analyzed land cover changes over the past 100 years in Florida based on reconstructed pre-1900 natural vegetation data and land cover data derived from early 1990s Landsat images. They input land cover and weather record data into the CSU Regional Atmospheric Modeling System and re- created the conditions for the three freezes.

In all three cases, the most densely cultivated areas were colder and also experienced subfreezing conditions for a longer period of time. Those areas include south and southwest of Lake Okeechobee and other key agricultural areas in the Kissimmee River valley.

For more information and images about the study on the Internet, visit:
http://www.nasa.gov/vision/earth/environment/wetland_freeze.html

Top of Page

Image Data Corner

U.S. Fish and Wildlife Service Wetlands Mapper
http://wetlands.fws.gov/mapper_tool.htm

Top of Page

Subpixel Classifier Tech Tip

Tips for Selecting Imagery for use with IMAGINE Subpixel Classifier

IMAGINE Subpixel Classifier software, developed by AAI, analyzes digital imagery and extracts useful information at the subpixel level, thus increasing the effective spatial resolution of a spectral sensor.

IMAGINE Subpixel Classifier is designed for use with data from any multispectral or hyperspectral sensor provided it meets the following two basic requirements:

1. There are at least three spectral bands
2. A pixel covers the same area in each of the bands

Multispectral imagery is frequently available in several formats. Certain formats provide distinct advantages with respect to the quality of derived signatures and the discrimination of classifications. The following is a list of general guidelines users should keep in mind when selecting imagery for use with IMAGINE Subpixel Classifier:

• Pixel Size
  Landsat Thematic Mapper (TM) data is available in several pixel size formats. The sensor samples the terrain with a 30-meter ground sampling distance. To produce smaller pixel sizes, pixels are artificially duplicated and inserted. These extra pixels can degrade spectral quality and affect signature performance. Therefore, the 30-meter format is recommended for use with IMAGINE Subpixel Classifier.

  Do not mix bands that have different spatial resolutions, such as combining MODIS Bands 1 and 2 (250m resolution) with MODIS Bands 3-7 (500m) to create a seven-band image. It may be possible to degrade the higher resolution bands to the lower resolution pixel sizes, as long as the degraded pixels cover the identical area (pixel field of view) as the coarse resolution bands.

• Geometric Correction
  Geometrically-uncorrected data generally produce superior signature quality and classification performance than geometrically-corrected formats. The highest spectral integrity for Landsat TM data is the geometrically-uncorrected, radiometrically-corrected format option. For SPOT imagery, however, processing levels 1A, which is not geometrically corrected, and 2A, which is geometrically-corrected, have identical radiometric corrections that maintain spectral integrity. In this case, the choice of the processing level depends upon the application.

• Resampling Methods
  The three most commonly used resampling techniques are nearest neighbor (NN), bilinear interpolation (BI), and cubic convolution (CC). NN resampling provides consistently superior signature quality and discrimination performance since it provides only radiometrically corrected raw measurements, whereas both BI and CC resampling methods perform spectral averaging with neighboring pixels.

IMAGINE Subpixel Classifier is commonly used with Landsat imagery, primarily for the purpose of broad area search. However, with the recent processing challenges posed by Landsat 7 Enhanced Thematic Mapper (ETM+) Scan Line Corrector off (SLC-off) imagery (July 2003 to present), users may wish to make use of alternative imagery. The versatility of IMAGINE Subpixel Classifier provides a valuable advantage as it allows for easy transition to numerous sensors.

Top of Page

New Advancements in Remote Sensing at AAI

Advancements in Image Calibration

AAI recently upgraded its Image Calibrator software to include the creation of a reflectance image in addition to a digital number (DN) image. Display of the reflectance image can be extremely useful for visual analysis as it allows for automatic color balancing and optimal stretching, as well as haze removal.

Features within the unstretched true color Landsat TM image of Kansas, shown in Figure 1, are difficult to discern. However, the reflectance image, shown in Figure 2, is automatically optimized for visual display. Note the more natural colors and sharper contrast.

Figure 1: Landsat TM of Kansas (bands 3,2,1 in RGB)		Figure 2: Reflectance Image from Image Calibrator (bands 3,2,1 in RGB)

Likewise, features within the Landsat TM image of the Patuxent River in Maryland, shown in Figure 3, are distorted beneath haze, but are clearly visible in the reflectance image, shown in Figure 4. Note that although Image Calibrator can successfully remove haze from an image, clouds, such as those in the lower right portion of Figures 3 and 4, persist.

Figure 3: Landsat TM of Maryland (bands 3,2,1 in RGB)		Figure 4: Reflectance Image from Image Calibrator (bands 3,2,1 in RGB)

For more information about Image Calibrator or other software being developed at AAI, feel free to contact us at 978-663-6828 or visit our Web site at http://www.discover-aai.com/company/contact.htm

Subpixel Classifier Case Studies

A project conducted at the Institute for Photogrammetry and GeoInformation at the University of Hanover investigated the use of IMAGINE Subpixel Classifier to map different tree species in forested areas southeast of Hanover, Germany using Landsat 7 ETM+ data. A subset of this scene is displayed in
Figure 5.

Figure 5. Subset of Landsat 7 ETM+ of Hanover, Germany (bands 4,3,2 in R,G,B)

While traditional landcover classification approaches do not account for differences within heterogeneous classes, this investigation tried to differentiate tree species within forested areas at a subpixel level. Aside from fir and beech trees that are regionally wide spread, the investigation also covered oak, alder, sycamore and ash trees. The results from IMAGINE Subpixel Classifier were compared to those from using the Maximum Likelihood method. Both methods yielded comparable results once the training sets were of homogeneous nature. The subpixel analysis, however, allows for additional conclusions about the density and purity of stands (populations), which can not be obtained from traditional classifiers.

Figure 6: IMAGINE Subpixel Classifier results (Yellow to Red indicate Fir trees, light to dark green indicate Beech trees.)		Figure 7: Comparison of Maximum Likelihood Classifier (MLC on left side) to IMAGINE Subpixel Classifier (right side)

For more information on this project, visit
http://www.caf.dlr.de/caf/anwendungen/projekte/projekte_nutzung/landsat/landsat_projekte/
projekt_lohmann/projekt_lohmann.htm

If you have an interesting application utilizing IMAGINE Subpixel Classifier that you'd like to share, we'd love to hear about it and highlight it in our newsletter. Feel free to contact Brenda via email or call 978-663-6828 ext. 239.

Top of Page

Hot Sites

Remote Sensing of Wetlands: Procedures and Considerations
http://www.vims.edu/rmap/cers/tutorial/rsecol.htm

This paper examines several methods by which remote sensing can be utilized to evaluate impacts to wetlands. Advantages and limitations of the following procedures are discussed: detecting signs of vegetation stress and pollution, monitoring changes in vegetation composition, and monitoring encroaching land development and erosion.

Subpixel Classification of Bald Cypress and Tupelo Gum Trees in Thematic Mapper Imagery
R. Huguenin, M. Karaska, D. Van Blaricom, and J. Jensen; Photogrammetric Engineering & Remote Sensing Vol. 63, pp. 717-725, June 1997.
http://www.discover-aai.com/technology/papers/WP-3.PDF

This paper examines the use of a subpixel spectral analytical process to classify Bald Cypress and Tupelo Gum wetlands in Landsat TM imagery of Georgia and South Carolina. Two hundred pixels were field verified indicating that both tupelo and cypress trees were successfully classified when they occurred as pure stands and as mixed stands. The subpixel classification yielded improved results over traditional classification techniques such as ISO-DATA clustering, maximum likelihood, and minimum distance.

Remote Sensing and Geospatial Application for Wetland Mapping, Assessment, and Mitigation
http://www.isprs.org/commission1/proceedings/paper/00079.pdf

Restoring and Preserving Wetlands and Riparian Areas
http://www.csc.noaa.gov/crs/rs_apps/issues/ifsar_scwrp.htm

Wetlands Violators Caught From Air
Allen, Scott. "Wetlands Violators Caught from Air." Boston Globe 11 Dec. 2003: B1.
http://www.boston.com/news/local/articles/2003/12/11/wetlands_violators_caught_from_air?mode=PF

Top of Page

AAI News 

New Hires at AAI

Jill Kukis joined the AAI team in September and will assist in a variety of roles including Accounting, Contract Administration, and Office Manager. Jill was most recently employed at Navisite, Inc. in Andover, MA.

Upcoming Presentations by AAI

Join AAI at the 4th Annual Coastal GeoTools Conference sponsored by the National Oceanic and Atmospheric Administration (NOAA) to be held March 6 to 10, 2005 in Myrtle Beach, SC. The conference will focus on real-life coastal applications of geographic information systems (GIS), remote sensing, and decision-support tools, as well as the development of the National Spatial Data Infrastructure. Dr. Scott Stoodley, AAI's Director of Environmental Programs, will be presenting the results of a comparison of several water quality parameters of the Patuxent River as measured via an IKONOS satellite sensor to temporally coincident field data collected by NOAA. We hope to see you there! For more information, visit the conference Web site at http://www.csc.noaa.gov/geotools/index.htm

Recent Presentations by AAI

Dr. Scott Stoodley, AAI's Director of Environmental Programs, recently presented "The Use of Remote Sensing to Locate and Characterize Areas of Dense Growth of Eurasian watermilfoil" at the North American Lakes Management Society (NALMS) Conference held November 3-5, 2004 in Victoria, British Columbia.

The Darrin Freshwater Institute (DFWI) is tasked with the responsibility for monitoring water quality and invasive aquatic exotic species on Lake George, NY. Eurasian watermilfoil, an exotic aquatic plant, is recognized as a major threat to native plant and animals populations in Lake George. A total of 26 areas are designated as large, dense growth populations of Eurasian watermilfoil whose size exceeds the current resources for management. Mapping of existing bottom coverage and rates of expansion are critical for planning purposes for future management efforts. Very limited resources are available to locate new areas of dense growth. Visual inspection of the 3000 acres of littoral zone by divers trained to identify Eurasian watermilfoil is not a practical consideration. Remote sensing offers the possibility of surveying large areas of the littoral zone in an efficient and cost-effective manner.

Applied Analysis Inc. has developed new software, called Supervised Anomaly Finder (SAF), and has demonstrated that it can map, classify, and identify submerged aquatic vegetation (SAV) under certain conditions within multispectral data. SAF detects and identifies anomalous materials in image pixels. It does so without prior knowledge of the signature of the material, and it can find very small occurrences.

DFWI acquired IKONOS multispectral imagery for the spatial extent of Lake George. Additionally, they provided AAI with shapefiles of known beds of Eurasian watermilfoil. Results from mapping the Eurasian watermilfoil via remotely sensed data were presented at this conference along with a comparison of mapped vegetation versus known locations. For more information about this conference, visit http://www.nalms.org/symposia/victoria/

Sample Projects at AAI

Based on the above initial analysis, AAI identified specific problem areas of interest, and will analyze them in greater detail using high resolution IKONOS satellite data.

Top of Page

AAI Corporate Overview 

Applied Analysis Inc. (AAI) is a world leader in automated analysis of remotely-sensed multispectral and hyperspectral imagery for both government and industry. AAI is a small business headquartered in Boston, MA since 1984. We provide innovative, leading-edge solutions for worldwide remote monitoring and analysis (including totally inaccessible locations) without having to resort to time-consuming and expensive surface-based data collection.

Founded by Dr. Robert Huguenin: Former MIT professor, Sloan Fellow, and founder of the University of Massachusetts Remote Sensing Center, Dr. Huguenin and his team developed the advanced technologies that comprise the core of AAI's powerful suite of software tools for automated image processing/analysis.

Primary Markets: AAI offers world-class expertise in remote sensing technology applied to two primary markets: military intelligence and environmental services.

• We originally developed our advanced-technology software tools for the military intelligence market and have achieved great success addressing a broad range of remote sensing applications for reconnaissance and surveillance, including: automated target recognition, target characterization, and change detection of scenes or objects.

• Then we transitioned our proven defense technology to address the environmental market including: 1) water quality analysis of oceans, lakes, rivers, streams, and estuaries, and 2) land cover analysis of rural, urban, and natural areas.

Powerful Software Tools: AAI's unique software advances the state-of the-art in remote sensing of water quality, land cover, and other objects of interest.

• Our water quality analysis software derives pixel-resolution, depth-integrated, remotely-sensed measurements of water quality and depth over wide geographic areas without need for ground truth data

• Our land cover/object analysis software overcomes the inherent limitations of a sensor's pixel-level spatial resolution by extracting useful information at the subpixel level. This allows detailed remote monitoring and analysis, and when cost is critical, our tools can often achieve required results with less expensive, lower resolution, remote imagery.

Industry Partners: AAI's relationships with virtually all of the major satellite vendors and several airborne vendors, allow us to approach each project with an unbiased perspective and then customize a solution to each project.

Invaluable Lessons-Learned: AAI's 20+ years of experience in remote sensing research, development and applications support can directly benefit new projects in cost, schedule and performance improvements.

One-Stop Shopping: Our big-picture awareness offers clients a total integrated solution to their remote sensing needs.
Top of Page

Visit AAI's Web site for additional information

Material Copyright © 2004 Applied Analysis Inc. All Rights Reserved