Understanding the transfer of instream large wood through river networks with graph theory

Instream large wood (LW), downed trees, trunks, root wads, branches, and fragments of wood at least one meter in length and ten centimeters in diameter, contribute to maintaining rivers´ physical and ecological integrity. Additionally, LW can be hazardous to infrastructure, which makes management of LW a priority for local services. Post-analysis of flood events has demonstrated recruitment of LW through several mechanisms such as bank erosion, landslides, and debris flows. Such recruitment and transfer of LW to fluvial networks vary spatially and temporally, which makes predictions of these processes challenging.

The current studies lack a method to express the structure of supply and transfer of LW throughout a catchment. An improvement would be to provide a means to interpret the structural connectivity of LW or the potential impact one recruitment area may have on another. Therefore, we propose to apply the graph theory to construct the network of LW recruitment from its source location to the fluvial network and to a designated outlet. In so doing, we aim to identify critical spatial connections that have previously not been realized, provide a means to view LW recruitment and transfer for an entire catchment wholistically, and explore the application of the graph theory in the realm of LW, which has not previously been done.

The graph theory requires a simple sequence for input data. The sequence is constructed based on physical connections that exist in nature. For instance, a forested region which has been determined to be intersected by a landslide prone area would represent a source of LW and the process which would supply the LW to the channel would be the landslide. Graph theory interprets the source area of LW as well as the location along the river to which the landslide connects as two unique nodes. The mechanism supplying the LW, in this case landslides, would represent an edge that would connect the two nodes. This same principal is then applied to the entire network for all pertinent recruitment mechanisms. The network can then be analyzed for a variety of different purposes including the identification of critical hot spots where multiple processes may be coinciding or locations along the channel where LW recruitment is relatively absent. Each node can additionally be given attributes which can later be extracted, such as the amount of LW available.

Determination of recruitment process prone areas has been based on previously developed methods for landslides, bank erosion, and debris flows. Swiss inventories of the forested area have been applied to delineate the forested area.

This contribution will show preliminary results done in the Vallon de Nant, Canton of Vaud, Switzerland. The catchment is relatively small (approx. 13km²) with active landslide, debris flow, bank erosion, and avalanche processes which supply LW. Current findings indicate landslide and bank erosion processes to be more widespread throughout the catchment with debris flows occurring more intermittently. The constructed LW network will provide a means to relate different recruitment mechanisms to one another and to LW found in rivers.