Detection and Labeling of Sensitive Areas in Hydrological Cartography Using Vector Statistics

The recognition and delineation of hydrological stream lines has, traditionally, been a subjective manual task in cartography. However, digital elevation models (DEMs) are nowadays often employed to extract stream lines automatically, via the use of geographic information systems. Whereas the automatic generation of hydrological networks presents errors, their manual recognition can be almost arbitrary. In this paper, we propose a methodology with which to label potentially sensitive zones in the comparison of hydrological cartographic networks. Two different sources were analyzed: a conventional cartographic stream network, and one automatically extracted from a DEM. The 72 500 vectors of displacement, representing the spatial disagreement (or fit) between the stream networks, were also examined. A number of remarkable distributions of large errors were identified that were a cause for alarm; these errors are here denoted by “warnings” and are classified into six different groups. The displacement vectors were also analyzed in terms of modulus and azimuth, thereby allowing the analysis of the isotropy of the spatial displacements. We propose the use of all of the derived information as metadata for hydrological spatial quality, as well as the extension of the methodology to any other type of cartographic element (roads, cadastral, etc.) for which two different vector format information sources are compared.