Geography Reference
In-Depth Information
diverse basins in the region. The Willamette River and its
tributaries contain the richest native fish fauna in Ore-
gon, as well as several species listed under the Endangered
Species Act as threatened or endangered (spring Chinook
salmon, steelhead trout, Oregon chub). Forests account
for more than two-thirds of the WRB, including federal
forests in higher elevation uplands and a mosaic of private
forest lands distributed across the lower elevations of the
Cascade Mountains and Coast Range. More than half of
the Willamette Valley has been converted to agricultural
use (43% of the valley area) and urban/rural residen-
tial development (11% of the valley area). Population
growth, social policies, and individual choices have sub-
stantial potential to change the environment and livability
for local communities (Baker et al., 2004). The research
described in this chapter created a framework for quan-
tifying historical resource loss and anticipating potential
future trends while providing several mechanisms for cit-
izen involvement in the assessment and prescription of
their region.
Oregon is relatively unique in the United States because,
in addition to resource management agencies, it has
developed land use zoning regulations for all lands, urban
growth boundaries for all cities, ecosystem restoration-
oriented natural resource management agencies, and
more recently citizen-based watershed councils to imple-
ment local conservation and restoration practices. The
WRB contains 23 active watershed councils comprised of
local citizens and land owners. In addition, the Willamette
Livability Forum, a group of 150 regional community
leaders, critiqued our representations of future scenarios
and the assumptions on which they were based (Hulse
et al., 2004).
The stakeholder advisory team defined plausible
assumptions of future actions and policy decisions over
the period from 1990 to 2050 for three different alterna-
tive future scenarios, each accommodating a doubling of
the human population. The three scenarios encompassed
a range of potential future trajectories (Figure 12.1), with
the Plan Trend 2050 scenario representing the future
landscape if we continue to implement our current
policies and practices. The Development 2050 scenario
represented a shift in current policies and practices that
relaxed environmental regulations to provide greater
flexibility for market forces. At the opposite end of the
spectrum of plausible choices, the Conservation 2050
scenario depicted the future landscape with greater
application of conservation and restoration actions
to insure long-term ecological function. Future land
use and land cover patterns under these three future
scenarios were determined by land allocation models
for agriculture, forestry, urbanisation, rural residential
development, natural habitat processes, and water use.
12.2 Methods
Maps of land use and land cover (LULC) change over
time, created in part from remotely sensed information,
provided an empirical source of evidence for the tra-
jectories of landscape change. The spatial framework of
coupled human and natural systems in the river basin
required an extensive and detailed land cover representa-
tion (Hulse et al., 2004). Each LULC map was represented
by 30 m by 30 m pixels and the 33 million pixels collec-
tively represented the 30,000-km
2
WRB. Several sources
of data were used to define the land cover or the land use
depicted for each pixel. A pre-European settlement land
cover map was created from surveys of the U.S. General
Land Office (GLO) from plat maps and survey notes ca.
1851. LULC maps for ca. 1990 were created initially from
image classification of multi-temporal Landsat Thematic
Mapper (TM) satellite imagery.
Upland portions of the basin in the Coast Range
or Cascade Mountain Ecoregions were mapped from
a single-date 1988 TM data set. The map of the more
heavily populated and agricultural valley region below
315 m elevation was created from an analysis of phys-
iognomic changes in a multi-season data set consisting
of five images from March, May, June, July, and August
1992 (Oetter et al., 2001). Each of the five 1992 images
was georeferenced to the geocoded 1988 TM data set. We
used an automated ground control point selection algo-
rithm and a second-order polynomial, nearest neighbour
resampling (Kennedy and Cohen, 2003). The root mean
square error was less than one pixel for all dates.
Radiometric normalisation was an important consid-
eration. We used a relative normalisation technique,
whereby each date was adjusted to a common image.
The June 1992 image served as the reference, and the
other dates were adjusted to it. First a subset of image
pixels containing water, forest, and urban areas were
identified which were determined to be no-change pixels
throughout all the image dates. The no-change pixels were
selected by multiple isodata clustering of a seven-band
image (six bands of TM differences plus Band 4 of the ref-
erence image). Then those pixels were subsampled using
a random stratified method to represent equally the three
types of land cover. Each band of the uncorrected images
was then transformed using a single regression linear
Search WWH ::
Custom Search