The
cartographic model outlines in great
detail the process by which the spatial analysis took place. Much of it
will not be covered here, however, two of the more critical sections will
be discussed in detail.
The Boolean MCE ties together all of the constraints on urban stream daylighting
- those considerations which can - to a great extent - be expressed as
boolean images (either "yes" or "no".) Many considerations - such as slope
- are not reducible to a boolean image, and these are treated in the fuzzy
MCE, which is also outlined here.
Boolean MCE
The Boolean MCE (shown on the cartographic model) is where all of the boolean
constraints on daylighting urban streams are processed. Here we see an
analysis which takes into account existing land use, soil properties, parklands,
and bikeways. A weighted MCE was used here to process these images and
perform an overlay. This weighted MCE was based on four boolean images:
parksb (whether or not there is a park present), vandirtb (whether or not
the soil is suitable), vanluseb (whether or not the existing land use was
suitable), and bikewayb (whether or not the area was within 300 metres
of a bikeway.)
These boolean images were derived as follows: (also refer to the cartographic
model and text in previous section)
Parksb:
The parks dataset consisted of patches of like raster cells - each patch
representing a park and with a unique identifier, set against a background
of zero values. This was reclassed to show all parks as 1, and all else
as 0.
Vandirtb:
As outlined in the previous section, names of soil classes were linked
to new numeric values by SQL queries. For example a matrix such as, [soilclass]
= "15" where [description] = "sandy clay" might be used. These numeric
values were then selected and reclassed according to their suitability
for stream daylighting. Below is the result: the vandirtb map.
Vanluseb:
This image was derived by a simple reclass operation performed on the GVRD
land use map. The result is shown below.
and, following the reclass operation, we see:
Bikewayb:
This image was derived from a simple BUFFER operation performed on the
bikeway raster image.
The
MCE itself was weighted in the following manner, as derived by the WEIGHT
module:
vanluseb = 0.3636
vandirtb = 0.3636
parksb = 0.1818
bikewayb = 0.0909
These values were chosen after a long, iterative process, and are believed
to reflect the true weights that each of these would have on the selection
of an area (stream) to be daylighted. Land use and soil type would be the
most important, as it was felt necessary to eliminate undesirable land
uses from the solution set (eg. heavy industrial uses). Soils were seen
as being equally important, as soil types would dramatically affect the
actual physical daylighting process, as well as the effectiveness of the
stream itself in performing its hydrological function. Parks were included
as a lesser weight, but still a significant one, to reflect the positive
effect that a park would have on the daylighting process, making it easier
from a land tenure standpoint to perform. A park would also provide a much
greater riparian zone for the new stream. Bikeways were included as a minor
weight, only to say that areas within 300 metres of a bikeway should be
given some additional consideration. (Bikeways are covered more fully in
the fuzzy MCE - which follows.)
The results of this first MCE are shown below, mcetest.
The Fuzzy MCE Analysis
The fuzzy MCE analysis ties together all of the spatial considerations
that cannot be reduced easily to boolean images. This includes slope, aspect,
distance from water (eg. a shoreline or an existing stream), distance from
a bikeway, and distance from a major road or highway. These were operationalized
using the FUZZY module, before a weighted MCE was performed on them.
The fuzzy MCE was based on five images, which were derived as follows (again,
see the cartographic model.)
Vanslope
Vanslope was derived from a SLOPE analysis of the vandem image. The slope values were calculated in percent. Below is the vanslope image, before processing. Note that main streets have been added for reference here.
The FUZZY moduile was invoked, and a user-defined curve was applied to
determine set membership. The curve was defined as follows, on slope values
in percent:
0, 0
2, 0.7
3.5, 1
8, 0.9
11, 0.6
16, 0
These values were chosen to give prominence to slopes in the range of 3
- 8%. Anything lower would have caused water to stagnate and would have
been difficult to engineer initially. Areas higher than 16 were often "errors"
on the dem, such as the edge of the map, which in some areas exhibited
slope values exceeding 50%. The value for 11 was chosen so as to negate
there being a strictly linear relationship between (8,0.9) and (16,0),
which would unnecessarily exclude some values.
Below is the result image: slopefuzzy. Note again that streets have been
included for reference.
Vanaspect
This
image is derived from an ASPECT operation performed on the vandem image.
Aspect was calculated in degrees, and FUZZY was used here by means of a
user-defined set membership function specifically designed to favour south-facing
areas. Remember that aspect is partially intended here as a surrogate for
rainfall, and south-facing areas generally recieve more rainfall in Vancouver
due to the prevailing southwesterlies in this area of the coast.
The set membership function was defined as follows:
0, 0
90, 0.7
180, 1
270, 0.9
360, 0
This gives prominence to slopes generally facing south (180 degrees) or
southwest (225 degrees).
The resulting image, aspectfuzzy, is shown below, again with roads for
reference.
Waterdistance
Distance from water was derived through the application of the
DISTANCE module to the waterdist image. (see cartographic model). FUZZY
was then applied on a more-or-less linear set membership function, defined
below with distance in metres:
0,1
1000, 0.3
3589, 0
(The value 3589 was the maximim distance value shown on the image.) The
resulting image, waterfuzzy, is below, again with roads as a reference.
Bikewaydistance
Urban
stream restoration often serves as a catalyst for the development of walking
and cyling trails. As such, there are two conflicting issues that were
to be considered here. One, proximity to a bikeway would be ideal (as shown
in the boolean MCE), as it would increase accessibility and exposure to
urban streams. However, it would perhaps be better if the stream restored
is far enough away from any bikeway or greenway so as to cause a new one
to be constructed! As such, a user-defined set membership function was
again developed here, but with a trough in its centre, as follows (distance
from bikeway in metres):
0, 0.85
300, 0.4
600, 0.8
900, 1
The resulting image, bikefuzzy, is shown below.
Note that we see the
effects of the set membership function very well here; the areas immediately
adjacent to bikeways are high values, but these values drop off, and then
slowly climb again.
Majorsts
Distance
from a major street was viewed as a good thing, as the traffic noise and
pollution would affect not only the habitat crerated, but also on human
enjoyment of the park environment. DISTANCE was applied to the majorsts
image to produce STDIST, which was then processed with FUZZY on a user-defined
set membership function, as follows: (distance from street in metres)
0, 0
200, 0.7
300, 1
1000, 0.8
Here we see that 200 to 300 metres was seen as a sufficient distance away
from a major road to buffer any significant noise, light (at night), or
pollution. The function drops off after 300 (towards 1000, 0.8) to reflect
the marginal drop in accessibility that the greater distance from a major
road (eg. drivers and transit riders) might imply.
The resulting image, streetfuzzy, is shown below. Roads are obviously not
required here as a reference.
This
new weighted MCE was carried out five times, each with slightly different
weightings. This was to ensure that the weighting scheme decided on (with
the help of the WEIGHT module) closely reflected the actual importance
of these factors. The final weights were decided on as follows:
slope: 0.3841
water: 0.2890
aspect: 0.1090
bike: 0.1090
street: 0.1090
Slope was considered to be the most important, as it often determines the
engineering capacity for stream daylighting. It was also deemed important
to weed out the areas with extremely low and high relief - as noted previously.
Water was ranked second, as it was deemed extremely desirable to have a
daylighted stream link up with another, already daylighted waterway (or
coastline). This was seen as a very important factor, more so than others
except slope itself. It does seem very commonsensical; if you want to "re-create"
a stream, do it in an area where it can be a part of something greater.
Aspect was initially ranked higher, but given the surrogate nature of this
dataset (for rainfall), it was dropped down the list significantly.
Bike and street were last on the list not because they do not matter, but
because these concerns were generally overshadowed by all others - noteably
slope and water. The resulting image, mcefuzzy5, is shown below.
Outcome
Following this, the two MCEs were combined, and final maps were produced. The ideal stream was selected from the suitmap image, which can be seen - with all other pertinent and interesting results - in the following section.