Automatic recognition of drainage patterns in river networks and its application to generalization

Abstract

In both GIS and terrain analysis, drainage systems are important components. Owing to local topography and subsurface geology, a drainage system achieves a particular drainage pattern based on the form and texture of its network of stream channels and tributaries. The drainage pattern can reflect the geographical characteristics of a river network to a certain extent, because it depends on the topography and geology of the land. Although research has been done on the description of drainage patterns in geography and hydrology, automatic drainage pattern recognition in river networks is not well developed. In addition, whether in cartography or GIS, hydrography is one of the most important feature classes to generalize to produce representations at various levels of detail. There are many methods for river network generalization, but few of them consider the drainage pattern in the first place, and the generalized results are always inspected by expert cartographers visually. Therefore, this research focuses on the drainage pattern and its application to map generalization. First of all, this thesis introduces a new method for automatic classification of drainage systems in different patterns. The method applies to river networks and the terrain model is not required in the process. A set of geometric indicators describing each pattern are presented and the membership of a network is defined based on fuzzy logic. For each pattern, the fuzzy set membership is given by a defined IF-THEN rule composed of several indicators and logical operators. The method was implemented and experimental results are presented and discussed. Second, this thesis proposes a method that evaluates the quality of a river network generalization by assessing if drainage patterns are preserved. This method provides a quantitative value that estimates the membership of a river network in different drainage patterns. Assessing the quality of a generalization is done by comparing and analyzing the value before and after the network generalization. This assessment method is tested with several river network generalization methods on different sets of networks. Finally, this thesis proposes a solution to deal with multiple factors at same time during the river network generalization. The multi-objective optimization problem is settled by the genetic algorithm with consideration of the drainage pattern. According the characteristic of each drainage pattern, the factors, such as drainage pattern membership, stream order and tributary balance, are considered and built into objective functions. In the multi-objective model, different weights are used to aggregate all objective functions into a fitness function. Then, the generalization is implemented by a designed genetic algorithm.