Adaptive river network extraction method using local characteristics of
water system
Abstract
River network extraction is the basis of hydrological analysis and
related issues. The accuracy of river extraction directly affects the
accuracy of watershed related research. The key to extracting the river
network by using hydrological analysis method is to determine the
drainage area threshold. At present, there are existing methods for
determining the drainage area threshold, which have problems such as
inaccurate extraction with a single threshold and difficulty in
obtaining data with multi-threshold. In view of this, based on DEM data
of Jiuyuangou watershed in the Loess Plateau, combined hydrological
analysis and digital terrain analysis methods, based on the principle of
river network extraction from slope runoff, and taking into full
consideration the hydrological characteristics of the terrain of the
watershed, this paper proposes a threshold determination method based on
multi-threshold constraints of local characteristics of the water system
and compares the river network accuracy between the river network
extracted by the threshold determined by this method and the single t
value determination method and the river network extracted by the
threshold determined by the river network density method. The results
show that among the two river network quantitative indicators including
average branch ratio and average length ratio, the corresponding values
of the extracted river network by the threshold determined by the
multi-threshold constraint method are 4.94 and 9.90, which are the least
different from the real river network (4.36 and 9.60), and the other two
methods are quite different. The research results show that the river
network extracted by the threshold determined by the multi-threshold
constraint method can more realistically express the characteristics of
the water system, and requires less data, which provides a new idea for
determining the optimal the drainage area threshold for the DEM water
system.