|
The land cover for Washington State was based on 1991 Landsat TM imagery. The
image data were spectrally clustered into approximately 200 classes per scene, and classes
were grouped by similar spectral values. Areas of similar land cover type were delineated
manually using the clustered TM imagery as a backdrop and a nominal minimum
mapping unit of 100 hectares. Each resulting land cover polygon was given a label that
included ecoregion; vegetation zone; and actual primary, secondary, and tertiary land
cover. Label information was based on visual interpretation of the TM data and available
ground data.
Sixteen TM images from 1991 were acquired from EOSAT (Lanham, MD) and archived
by the Multi-Resolution Land Characterization Project (EROS, United States Geological
Survey, Sioux Falls, SD). All 16 scenes were map-oriented, georeferenced and terrain-corrected by EOSAT in Universal Transmercator (UTM) projection; Clarke 1866
spheroid; 25-meter resample size; with nearest-neighbor resampling.
Ecoregions and vegetation zones: An ecoregion is a contiguous geographic area of
similar climate and geologic history, for example, the Northwest Cascades. We identified
13 ecoregions in Washington. A vegetation zone is an area in which moisture,
temperature, elevation, and other environmental parameters combine to create conditions
that favor similar assemblages of vegetation communities, for example, the Ponderosa
Pine or Alpine/Parkland zones. We identified 31 vegetation zones in the State.
Polygon attributes: Each polygon was labeled with several attributes: ecoregion;
vegetation zone; primary, secondary and tertiary land cover; the respective occupancy
classes of the primary, secondary, and tertiary cover in each polygon; labeling date; the
source of the information for the label if the label was based on information other than the
appearance of the polygon from the satellite data; the person doing the labeling; and
comments (Figure 1). The primary land cover (Table 4; Map 3) was the actual land cover
that occupied the greatest proportion of the area in a polygon, for example, closed canopy
conifer forest or non-irrigated row-crop agriculture. Secondary and tertiary land covers, if
needed, were the actual land covers that occupied the second and third greatest proportion
of area in the polygon. Primary, secondary, and tertiary covers were each assigned one of
six occupancy classes indicating the proportion of the polygon occupied by each. The
occupancy classes were:
| 1: <5% |
Midpoint: 2.5% |
| 2: 5%-25% |
Midpoint: 15% |
| 3: 25%-50% |
Midpoint: 37.5% |
| 4: 50%-75% |
Midpoint: 62.5% |
| 5: 75%-95% |
Midpoint: 85% |
| 6: 95%-100% |
Midpoint: 97.5% |
Polygon delineation and labeling: Areas of similar land cover were delineated on-screen
in ArcEdit using the clustered, colored TM imagery as a backdrop. Roads, hydrology,
and elevation were also displayed as needed to aid in delineation and labeling. Polygons
were drawn with a nominal 100-hectare minimum mapping unit for terrestrial cover types
and a 40-hectare minimum mapping unit for wetlands, i.e., the area of each polygon was
greater than or equal to 100 or 40 hectares, respectively. In practice, some important land
covers smaller than the minimum mapping units were delineated, notably small Pacific
islands important for seabird colonies. Other polygons were considerably larger than the
minimum mapping units because the land cover was uniform over large areas (e.g.,
agricultural polygons in the Columbia Basin) or time constraints prevented further
refinement. Polygons were labeled based on their appearance on the TM scene and with
additional information from maps and ground data when available.
Ground information for polygon labeling: Correct interpretation of satellite data and
subsequent labeling of polygons require some prior knowledge of actual ground cover.
We made use of as much ground information as could be assimilated in the time
available. More information was available but was not used because of lack of time to
either find the data or put it in a usable form. Visitation of areas on the ground by project
personnel was the best source of data, but time was too short to visit more than a few
hundred sites.
Since the land cover occupancy classes were ranges rather than specific values, estimates
of area occupied by each cover class were based on the midpoint of the occupancy class.
Land cover mapping is a continual process of refining and improving spatial resolution
and polygon labels. Because the land cover map changes with each revision, version
numbers were assigned as changes were made. Land cover analysis was based on
Version 6 of the land cover map. Version 5 of the land cover map was used for the
vertebrate models, because we did not have time to rerun the vertebrate models with the
latest version. The differences between the two versions, however, are very minor. Most
changes between Version 5 and Version 6 were in vegetation zone labels and polygon
delineation in Asotin County in extreme southeastern Washington. |
