Methodology

Landsat Thematic Mapper Image

To the right is a thematic mapper (TM) image cut from a larger scene and comprising parts of the USGS topographic quadrangles that make up the Blackwater National Wildlife Refuge on the Eastern Shore of Chesapeake Bay. This refuge (and associated conservation areas) holds most of the tidal wetlands of the Chesapeake Bay area and is the basis for Dr. Kearney's model of wetlands deterioration. Historical maps of the area show increasing rates of open water.

The Coastal Marsh Project (CMP) methodology requires the image to be loaded into PCI image processing software and georeferenced to UTM coordinates to correspond to USDA National Wetlands Inventory (NWI) digitized quadrangles.

Bands 4, 5, and 7 are used in this analysis.

National Wetlands Inventory Digitized Maps

The digitized data from the National Wetlands Inventory is used to create polygon coverages in the ARC/INFO geographic information system (GIS) software. This is the same area, clipped to reflect the borders of the Blackwater marshes. The NWI classifications have been simplified into their larger classes.

The grey area is non-wetland, the red is palustrine, the blue is open water, and the green is the area of interest, coded "Subtidal estuarine" by NWI and comprising the salt marshes of the area.

The CMP methodology revealed that the NWI coverages gave a good definition to the upland boundaries of the marshes, whereas the satellite images gave a more realistic water/wetland interface.

Coastal Marsh Project Mixture Model

Because this is not a photograph, but a sensor image, it shows only a value for the amount of reflected light of a certain waveband from each pixel. A pixel (picture element) in TM represents about 30 meters square on the ground, so the pixel can actually be made up of soil, vegetation, and water in unknown proportions.

A mixure model tries to calculate the actual proportions of each type of land cover by solving simultaneous equations in three variables:

• x% soil + y% vegetation + z% water = pixel reflectance value.

The image at the right shows the results of the CMP mixture model run on this area and categorized by percentage of water per pixel. The colors indicate increasing percentages of water; green is solid marsh, and yellow to red to blue indicate an increasing danger of degradation due to waterlogging.

Marsh Surface Condition Index

Although the data are calculated by pixel size, the memory required to store and display a map showing the values of each pixel in a thematic format is enormous. Thus, the CMP has produced output data in ArcView and ARC/INFO GIS formats. The data has been formatted in a grid which is four hectares per grid cell. The result is a coarse resolution, which for small areas of coastal salt marsh would not be useful, but for large overviews it provides an efficient and effective format.

This file can be loaded into the user's GIS and employed as a layer in a more detailed analysis. The finer resolution is still stored in the attribute tables, but only those areas that are of interest need to be drawn out in such detail.

Historical Change Analysis of the Blackwater Wildlife Refuge

From aerial photos, 1938

From aerial photos, 1989

From Landsat TM image, 1993

This series of images demonstrates the use of the data contained in the Coastal Marsh Project Surface Condition Index files.

The top image, from 1938, was digitized from aerial photos taken by the USDS Soil Conservation Service at that time. This is the most detailed and presumably accurate type of landcover analysis. It is very labor-intensive and can only cover a small region at a time.

The middle image, from 1989, was also digitized by hand, as part of a student thesis (Rizzo, Eric, M.A., UMD Geography, 1995) that demonstrated the trend of wetland loss in the Blackwater area over time.

The third image was done using CMP techniques. The area was clipped out of the much larger area that was produced (the Chesapeake Bay) and enlarged to demonstrate that this technique can overlay existing analyses. It can be updated easily by repeatedly running the model with the latest satellite images. This allows a larger area to be analyzed regularly and updated with minimal cost and human intervention