9371

• UCREFRP Catalog Number: 9371
• Author: Levee Removal Evaluation Group.
• Title: Green River levee Removal and Floodplain Connectivity Evaluation.
• USFW Year: 1998.
• USFW - Doc Type: CAP-6 LR,
• Copyright Material: NO

1

1
1

1
We derived two base maps from different data sources: a 1993 Landsat thematic mapper
(TM) image at a resolution of 30 meters, and low-altitude aerial photographs at a resolution of 2
meters. Base maps serve as points of reference for identifying and quantifying changes in the
size and distribution of vegetation types. Because digital analysis is more efficient, and often
more accurate, than manual interpretation (Snook et al. 1987), aerial photographs were scanned
and processed for digital analysis. Rectification of the digital photographs against 1:24,000
orthophotographs corrected for areal distortion. Due to the data-intensity of high-resolution
photographs, only floodplain restoration sites were included in the analysis. TM imagery was
obtained in digital form and was pre-processed to adjust for atmospheric distortion. Analysis of
the TM data included the area between the Bonanza and Ouray bridges. Both data types were
classified with an unsupervised classification algorithm, which divides reflectance data into
classes that are spectrally similar and makes no a priori assumptions about the reflectance data.
Field data were collected to identify the vegetation types represented by each spectral
class. Between July 1996 and August 1998, ground-truth data were collected from 235 plots
within the study area. Variables included geographic location, plot size and configuration,
relative soil brightness, a list of species present at greater than 5% cover, and ocular estimates of
average height and percent cover for each species. Ocular estimation methods were chosen for
two reasons. First, they require much less time than traditional transect or quadrat methods.
Second, unlike more traditional survey methods, ocular methods do not imply that vegetative
cover is static and known to the nearest percentage (Daubenmire 1959). Vegetative cover may
change significantly within the growing season, and from one season to the next.
Eighty-five of the 235 plots were used for classification of the TM data. These plots were
at least 90 by 90 meters to ensure that the sampling frame was large enough to be accurate at the
resolution of the TM data. All 235 plots were used to develop the map based on aerial
photographs. To identify the vegetation type represented by each spectral class, the spectral
classes created by the unsupervised classification algorithm were summarized with field plot
locations in a geographic information system. Each pixel was classified into one of the
following eight cover types based on the dominant vegetation type:
1) Water
2) Emergent vegetation (Carex and Typha sps.)
3) Native tree (Populus sps.)
4) Exotic tree (Elaeagnus angustifolia)
5) Native shrub (Salix sps., Rhus trilobatum)
6) Exotic shrub (Tamarix ramosissima)
7) Native terrestrial herb (many species)
8) Exotic terrestrial herb (primarily Lepidium latifolium, Centaurea repens)
The TM data was further classified into 17 vegetation types (see Table 3.1). The overall '
accuracy of each classification was assessed by generating 100 random points, and comparing
the classified vegetation type with the actual vegetation type at those points. All digital analy
were performed in ERDAS/Imagine 8.2.
a
22