Model Generalization of the Hydrography Network in the CARGEN Project - Advances in Cartography and GIScience - page 442

• to apply the 250 meters length threshold to each watercourse, i.e. to
prune the network of those branches shorter than this measure.
Furthermore, a third objective was set by the IGMI specifications, requir-
ing that:
• in the regions where the hydrography network is too dense the less im-
portant branches should be pruned.
In the following sections we will describe the generalization process of the
river network. Water bodies as lakes, ponds, swamps will not be covered;
even though the generalization process could be applied also to canals, we
will not make explicit reference to them. Also the generalization of ditches
is not included in this paper, as they lend themselves to a different gener-
alization approach (i.e. typification)(Savino et al 2011).
1.3 Related work
When we started to develop our process for the hydrography generaliza-
tion we looked with interest at the experiences done in the past by others.
The objective that we pursued was three-folded: first we had to find a way
to classify the rivers on their width, second prune the network of the short-
est branches and third prune the network of the least important branches in
regions where they were too many. A great deal of work on generalization
of hydrography can be found in literature, that helped us to set up our
generalization process.
(Horton 1945 and Strahler 1952) developed a metric to classify the
branches of a river network using a counter that increases when two
branches meet. This metric, known as Strahler order, is widely used to en-
rich the river data models extending the original classification to prune the
network (Thomson and Brooks 2002; Touya 2007). To prune the river
network many other parameters and thresholds can be used, singularly or
in conjunction: density (Stanislawski 2008), water basin (Ai et al. 2006),
the upstream drainage area (Stanislawski 2009) or other (Zhang 2007),
(Brewer et al. 2009).
As a matter of fact most of the river network generalization algorithms
step first through a process of data enrichment and then prune the river
network. In the following section we will illustrate the generalization process