Watershed Decomposition

Motivation

Often it is not necessary to study the entire terrain or river network at once; frequently interest is only on regions that are downstream of a particular river, or the upstream areas that contribute flow to a particular river. By decomposing the terrain into a set of disjoint hydrologic units (or watersheds) — regions where all water within the region flows towards a single, common outlet — areas of interest can quickly be identified without having to examine the entire terrain. Since the desired size of a unit can vary from application to application, it is advantageous to have a hierarchical decomposition of the terrain into nested units of arbitrary small size. Furthermore, it is useful if the (smallest) units are assigned a unique label that also encodes topological properties such as upstream and downstream neighbors; thus making it possible to automatically identify hydrological units of interest based on the label alone.

Watershed Decomposition as implemented in TerraSTREAM

The implemented Pfafstetter watershed labeling method, as proposed by Verdin and Verdin, can be used to hierarchically divide a terrain into arbitrarily small hydrological units, each with a unique label, such that these so-called Pfafstetter labels encode topological properties. Given a (main) river R in a terrain and its basin, that is, the contiguous area that drains into R, the method divides the basin into nine disjoint hydrological units: four basins and five interbasins. The four basins are the four largest basins of the rivers flowing into R; they are labeled 2, 4, 6, and 8 in the order they appear upstream along R. The interbasins are then defined by "breaking" R at the places where the four rivers with largest basins flow into R; they are labeled 1, 3, 5, 7, and 9, in the order they appear upstream along R. To obtain a hierarchy of hydrological units, the method is then applied recursively on each basin and interbasin, and labels of the subdivided regions are appended to the existing label of the original region. Refer to the figures below.

For more detailed information about the pfafstetter method see also, e.g.,  http://gis2.esri.com/library/userconf/proc97/proc97/to350/pap311/p311.htm,  http://gis2.esri.com/library/userconf/proc01/professional/papers/pap1008/p1008.htm, and  http://eg.or.kr:8080/abstractII/I0321801001.html .

Usage

This paragraph contains a short description of the usage and the options provided with the TerraSTREAM Watershed Decomposition module as available in the beta release version 0.2.

Input

The watershed decomposition module requires as input information for every cell in the terrain about (1) where flow goes from that cell and (2) the amount of water flowing through that cell. The Flow Routing and Flow Accumulation modules compute this information, resp.

Caveat

The original Pfafstetter watershed labeling is properly defined only on flow terrain models without any inner sinks and with single flow direction per terrain cell. Consequently, we suggest that Watershed Decomposition input should be produced by first Hydrologically Conditioning the terrain with the threshold parameter omitted and then Flow Routing the conditioned terrain using single flow directions.

Label output formats

The primary output of a Watershed Decomposition is a pfafstetter label for each cell in the terrain and can be output as a grid, as a TPIE file or as a ASCII file, in which each line of the file holds the labels of a grid line of the terrain. The format in which labels are output by the Watershed Decomposition module can be controlled in order to ease interpreting and visualizing the output (see below). By default, labels are be augmented through with ones up to the length of the longest label; therefore labels can be interpreted and compared as integers and hence allow also for a meaningful color ramp visualization of the output based on the integer order of the labels. This padding of labels can be turned off. Furthermore, the user can specify, how many digits per labels should (at most) be output, where that is to which level of the watershed hierarchies.

Labeling modes

Several command line options are provided to control how the terrain is to be labeled. By default, all watersheds will be labeled individually. Watersheds on the highest hierarchy level, which do not flow to the same outlet will be labeled independently: Each one of them gets assigned a unique identifier, which semantically corresponds to the outlet, the watershed drains to. As a supplementary output, specified by options --ouput-basin-ids (-tpie/-ascii) a grid/stream/ascii ffile can be generated which holds for each cell in the terrain its watershed id, i.e. to which outlet the respective cell flows to.