We were interested to learn how well our existing roads
coverages represented actual roads in the watershed. Of
course, the best way to do this would be to survey all
the roads in the watershed. Since this was impractical,
we relied on the interpretation of the digital orthophotos
(DOQ) available for the watershed. We assumed that we
could see and correctly identify all the roads in the
DOQ and that the roads have not changed since the photography
was taken.
We created a grid that uniquely identified each hectare
in the study area
. There were 375,000 hectares
in the study area. We then randomly selected approximately
1 % of all hectares and visually examined the DOQs to
see if there were roads present within that sample grid
cell. We avoided areas for which we did not have DOQs.
We used the same grid to sample the
100K
and
24K roads layers so that
results could be directly compared.
This resulted in road frequency (number of ha with roads
present) rather than
road
densities.
For the DOQs we found that 56.4% of
the sampled hectares had roads, the 24K roads layer 32.7%
with roads, and the 100K roads layer had 17.4% with roads.
See extrapolated
road frequency analysis.
These data are in a spatial format and
can be summarized by landowner, soil type or watershed position.
This demonstrates how important data
scale is when producing a summary. These results show that
the 100K roads layer may miss almost 40% of the roads and
the 1:24K roads may miss 23% of the roads visible on the
DOQs.
Please address questions or comments
to:
Dr. Ralph Garono
Principal Investigator
Earth Design Consultants, Inc.
rgarono@earthdesign.com
Laura Brophy
Principal Investigator
Green Point Consulting
brophyl@peak.org |
homepage 
