Downscaling of runoff using river models with graphs
No Thumbnail Available
Insinööritieteiden korkeakoulu | Master's thesis
Master’s Programme in Geoinformatics
AbstractA graph representation of river networks is not a new area of research. The complex dendritic structure of river networks is most suitably represented as graphs. Many of the algorithms and structural properties of graphs are transitive to rivers networks like topological sorting, connectivity and clustering coefficient. Downscaling is a method for obtaining high-resolution climate or climate change information from relatively coarse-resolution global climate models (GCMs). Non-statistical downscaling uses a high-resolution model (a regional climate model, or RCM) driven by boundary conditions from a GCM to derive smaller-scale information. A traditional river flow model relies on digital elevation model (DEM) to derive the slope, flow direction and watershed delineation. This thesis examines an alternative to DEM by using vector river network. It models river networks as graphs. The thesis performed sorting, querying algorithms, and graph traversals to determine the flow direction. It utilizes the topological information available at the GIS attribute table. The study chose three parameters (river length, Strahler stream order, and equal weight) to study the variation in the flow of river inside a cell. The research is based on with freely available data, and ArcGIS is used for the analyses in conjunction with Python. The study correctly estimated the Strahler order and the flow direction of the river network. The Strahler order and equal weight have shown a high correlation of weights for the first stream orders. The river length weight parameter has the highest variance but as a physical attribute ought to be given more emphasis. The actual river flow estimates are grossly estimated and need further study and investigation. The initial results seem promising, and this method could serve as an alternative method of replacing and complimentary for DEM models, but further development should be done.
Thesis advisorKallio, Marko
hydrology, downscaling, graph theory, non-statistical