|
Titel |
Metrics for assessing the performance of morphodynamic models of braided
rivers at event and reach scales |
VerfasserIn |
Richard Williams, Richard Measures, Murray Hicks, James Brasington |
Konferenz |
EGU General Assembly 2017
|
Medientyp |
Artikel
|
Sprache |
en
|
Digitales Dokument |
PDF |
Erschienen |
In: GRA - Volume 19 (2017) |
Datensatznummer |
250149045
|
Publikation (Nr.) |
EGU/EGU2017-13359.pdf |
|
|
|
Zusammenfassung |
Advances in geomatics technologies have transformed the monitoring of reach-scale (100-101
km) river morphodynamics. Hyperscale Digital Elevation Models (DEMs) can now be
acquired at temporal intervals that are commensurate with the frequencies of high-flow events
that force morphological change. The low vertical errors associated with such DEMs enable
DEMs of Difference (DoDs) to be generated to quantify patterns of erosion and deposition,
and derive sediment budgets using the morphological approach. In parallel with reach-scale
observational advances, high-resolution, two-dimensional, physics-based numerical
morphodynamic models are now computationally feasible for unsteady, reach-scale
simulations. In light of this observational and predictive progress, there is a need to identify
appropriate metrics that can be extracted from DEMs and DoDs to assess model
performance. Nowhere is this more pertinent than in braided river environments,
where numerous mobile channels that intertwine around mid-channel bars result in
complex patterns of erosion and deposition, thus making model assessment particularly
challenging.
This paper identifies and evaluates a range of morphological and morphological-change
metrics that can be used to assess predictions of braided river morphodynamics at
the timescale of single storm events. A depth-averaged, mixed-grainsize Delft3D
morphodynamic model was used to simulate morphological change during four
discrete high-flow events, ranging from 91 to 403 m3s−1, along a 2.5 x 0.7 km
reach of the braided, gravel-bed Rees River, New Zealand. Pre- and post-event
topographic surveys, using a fusion of Terrestrial Laser Scanning and optical-empirical
bathymetric mapping, were used to produce 0.5 m resolution DEMs and DoDs. The pre-
and post-event DEMs for a moderate (227m3s−1) high-flow event were used to
calibrate the model. DEMs and DoDs from the other three high-flow events were used
for model assessment using two approaches. First, “morphological” metrics were
applied to compare observed and predicted post-event DEMs. These metrics include
measures of confluence and bifurcation node density, bar shape, braiding intensity, and
topographic comparisons using a form of the Brier Skill Score and cumulative
frequency distributions of rugosity. Second, “morphological change” metrics were
used to compare observed and predicted morphological change. These metrics
included the extent of the morphologically active area, pairwise comparisons of
morphological change (using kappa and fuzzy kappa statistics), and comparisons
between vertical morphological change magnitude and elevation distribution. Results
indicate that those metrics that assess characteristic features of braiding, rather than
making direct comparisons, are most useful for assessing reach-scale braided river
morphodynamic models. Together, the metrics indicate that there was a general
affinity between observed and predicted braided river morphodynamics, both during
small and large magnitude high-flow events. These results thus demonstrate how
high-resolution, reach-scale, natural experiment datasets can be used to assess the
efficacy of morphological models in predicting realistic patterns of erosion and
deposition. This lays the foundation for the development and assessment of decadal
scale morphodynamic models and their use in adaptive river basin management. |
|
|
|
|
|