Geoscience Reference
In-Depth Information
Once the width of each section has been calculated, each section can be
classified according to the target data model, following the IGMI specifi-
cations; if due to model generalization the geometry of a river section has
to be collapsed from area to line, the algorithm simply deletes the area and
keeps the relative edge.
The choice of averaging the width measures in each section could be ques-
tioned: indeed the search for a minimum or maximum value could be a
valid alternative too and may provide some more insightful information
(e.g. the minimum width is related to the maximum allowed boat size).
Anyway our guess was that the average would be a more reliable measure,
being robust to eventual local minima or maxima in which our simple
algorithm could fall (see
Figure 2
); furthermore since each river is divided
into a new section at every confluence, the size of each section will be
more or less constant (the water flowing in or out of it is the same), with
the average width then being not too far away from the real minimum and
maximum width.
Figure 2:
measuring the width of an areal river section may incur in some errors: on the
left the normal projected from a point on an “hits” the boundary in two points very far
away (dark dots); on the right the intersection points (dark dots) are local minima and
maxima; averaging the measures of every section mitigates the influence of these errors
on the measured width.
Moreover some errors in our measurement could be tolerated, as our aim
was to calculate a reference measure to classify the rivers and not to calcu-
late their exact width.
The simple approach that we set up gives the best approximation of the
width of each river section when the shape of its area is mostly regular -i.e.
when the river banks run parallel to the graph edge-. Results can be worse
for odd shaped sections or sections that are small compared to the para-
meter
d
. For the former of these issues our guess was that averaging the width
