Geoscience Reference
In-Depth Information
(SigmaPlot, Systat Software Inc., San Jose, CA)
were then used to identify significant relationships
among variables.
With 58 sites, two sample collection dates, 15
stream chemistry variables, 42 land-cover variables
(14 land-cover types, 2001 data, 2006 data, and
the change between 2001 and 2006), 12 variables
from county tax records, and three variables from
census records (already reduced from many more
variables), it was first necessary to reduce the
dataset to a more manageable size for statistical
analysis. Some sites were deleted because of
missing data. The remaining sites were separated by
size into two groups and analysed separately so that
sites upstream and downstream of each other were
never included in the same analysis. This resulted
in a set of 26 small catchments ranging in size from
0.4to14km
2
and a set of 15 larger catchments
ranging from 3 to 45 km
2
. Two intermediate size
sites, the only two sites with significant urban area,
were included in both catchment groups.
Many chemistry variables were correlated.
Cluster analysis was used to identify groups of
chemistry variables that could be represented by
single variables. Principal components analysis
was used to determine which variables were
useful in discriminating among sites. For example,
winter and summer specific conductance were
highly correlated (Figure 8.1), so only summer
values were used. On the other hand, turbidity
measurements from winter and summer were
not correlated (Figure 8.1), so both values were
retained. All nitrogen measurements (NO
3
,NH
4
,
DON, TDN) were generally correlated, so only
summer NO
3
was used. DOC was generally low
for all sites and provided no useful discrimination
among sites. TSS was strongly correlated with
turbidity and therefore not used. The percentage
of organic matter in particles in winter and
summer were not very similar, but data were
missing from many sites in winter owing to very
low concentrations, so only summer values were
used. In summary, six chemistry variables for
analysis were used: summer specific conductance,
summer nitrate concentration, winter and summer
turbidity, summer percentage organic matter in
particles, and total dissolved phosphorus.
Similarly, the number of independent variables
was also reduced. The original 14 land-cover
variables were reduced to four: developed, forest,
scrub, and agriculture (Plate 13). For this study,
all forms of agriculture were put in the one
category. Agricultural land in the basin is primarily
pasture with some hay fields and some small
areas of row crops. Gardens and lawns were also
included as agriculture. The land-use variables that
proved most useful were: parcel area residential,
parcel area commercial, parcel area agriculture,
parcel area untaxed (primarily National Forest),
and average land value in each catchment. From
the tax record data, the estimated rank of potential
real estate activity and the index of land ownership
persistence (LOPI) were used. The census data were
represented by the aggregated rank of demographic
diversity.
Results
These data illustrate the importance of
distinguishing between land cover and land
use (Figures 8.2, 8.3). Agricultural land-cover
and agricultural land-use (parcel area) were well
correlated, but there were some distinct outliers in
the data from the larger catchments (Figure 8.3).
There is no land classified as agricultural in the
Franklin catchment, but 13% of the area shows
up in satellite imagery as agricultural land-cover.
These areas probably represent lawns and small
gardens. Also, several catchments have greater
agricultural land-use than agricultural land-cover
(Jones Creek, Cowee Creek, Hickory Knoll),
probably as a result of abandoned agricultural land.
A similar contrast can be seen when forest
and developed land-cover are compared with
residential and commercial land-use. All but two
catchments have more than 50% forest cover
(Figure 8.2). Even the most urban catchments still
have 30% forest cover. This forest cover is largely
National Forest but also includes private forest land.
Also, even though a large percentage of catchment
area may be listed on the tax records as residential
or commercial, it may not show up as developed
