|
Titel |
Quantifying river response to landsliding: experiments in DEM differencing using wide-area, structure-from-motion terrain models. |
VerfasserIn |
Joe James, James Brasington, Simon Cook, Simon Cox, Eliisa Lotsari, Sam McColl, Niall Lehane, Richard Williams, Damià Vericat |
Konferenz |
EGU General Assembly 2017
|
Medientyp |
Artikel
|
Sprache |
en
|
Digitales Dokument |
PDF |
Erschienen |
In: GRA - Volume 19 (2017) |
Datensatznummer |
250146154
|
Publikation (Nr.) |
EGU/EGU2017-10158.pdf |
|
|
|
Zusammenfassung |
Sediment delivery to alpine rivers is characterized by large but infrequent pulses of material
sourced from landslides and debris flows. In extreme cases, when the rate of sediment supply
exceeds the transport capacity of channels, a landslide dam forms; impounding river flows
and creating an inline lake. These rare events play a crucial but weakly understood role in the
evolution of catchment drainage, channel morphology and sediment flux from mountain
catchments to their sedimentary sinks. Until recently, insights into the response of river
systems to such sediment overloading have been based on either localized ground surveys or
expensive airborne lidar campaigns. The recent development of structure-from-motion
photogrammetric methods offers the potential to bridge this scale-cost barrier, but has yet
to be applied over wide-area (101−2 km2) extents which push the boundaries of
traditional SfM workflows based on dense ground-control and low-altitude or terrestrial
imagery.
Here, we present preliminary insights into the response of the braided Dart River, Otago
as it adjusts to a major pulse of sediment supplied by landsliding at Slip Stream (44.59 S
168.34 E) in January 2014. DEM differencing (DoD) is used to develop a sediment budget for
this extreme slope-channel coupling, using wide-area (>80 km2) terrain models derived from
SfM photogrammetry based on aerial helicopter surveys in May 2014 and 2015. Contrasting
camera networks, image density and camera models were used in the two surveys
providing an opportunity to evaluate the sensitivity of the resulting terrain model to
data acquisition strategy. In both cases, georeferencing was based on a network of
ground-control distributed along the 40 km valley floor which was also used to provide
cross-validation tests on horizontal and vertical model reliability. Both models were
subject to inherent systematic bias associated with compensation between the inferred
interior and exterior model geometry. The use of a convergent camera network was
found to reduce systematic bias, giving rise to independent 3D errors of 0.21 m
compared to 0.44 m for models based on nadir only photography. The effects of
systematic bias are, however, most effectively managed by developing an empirical
correction model based on correlation of stable regions beyond the channel margins. In
this paper, we document this workflow and develop a simple error framework to
quantify geomorphic change between the surveys. A preliminary sediment budget,
quantifying the volumes of material delivered by landsliding, in storage on the
debris fan, and then longitudinally redistributed along the Dart River is presented. |
|
|
|
|
|