DELTAFORCE Channel switching

Involved institutes and People
PI: Irina Overeem
INSTAAR, Univ. of Colorado, 1560 30^th St., Boulder CO, 80309-0450
phone: (303)492-6631 fax: (303)492-6388

Related Projects

ONR Uncertainty: Seabed Variability and its Influence on Acoustic Prediction Uncertainty.

Objectives

Determine channel morphology, sedimentation rates and channel switching rates in Clyde river delta on Eastern Baffin Island (Fig. 2) as a modern laboratory for fluvio-deltaic dynamics under land-ice conditions.
Refine HydroTrend to generate high-resolution discharge and sediment-load time-series for arctic rivers, including glacial lake outbursts, and test the validity by comparing it to field data.
Advance numerical channel switching routine by feeding it with HydroTrend input and validate the output against field data.

A non-technical summary of this project in Inuhtitut is posted here as pdf file

Introduction
Very little is known about the natural environmental variability in Arctic fluvial-deltaic systems. Ironically, Global Circulation Models (GCMs) predict that future climate change due to antropogenic forcing will be amplified at high latitude and have important feedback on global climate (Dickson, 1999; PARCS, 1999; McCarthy et al., 2001). Sedimentary modeling studies indicate that climate change could have a profound impact on the sediment flux transported into the coastal zone (Syvitski and Andrews, 1994; Syvitski, 2003). Uncertainty in such predictions are related to two main reasons:

(1) Impact of extreme events
The difficulty in geological process studies is that the period of observation is inherently too limited to observe a system from beginning to end. ‘Time is to geology what distance is to cosmology’ (Paola, 2000). Conceptual models in both sciences cannot be tested because the objects or situations under study are too distant in space or time. This is especially true for the high Arctic, where measurement records are short, and fieldwork logistically difficult. Extreme events, e.g. glacial lake outbursts, are of low-frequency (and thus not monitored), but may have a high impact. In earth sciences the classical way out of this impasse is inverse modeling. One infers processes and input from the end product; the paleo-record. Unfortunately, sediments are eroded or transported out of the system boundaries, leaving the record incomplete. Deep time leaves us an observable record of limited quality and resolution. We advocate building forward numerical simulation models of sedimentary systems, that capture the causative factors and dynamic processes. Numerical models allow us to generate results under known and repeatable conditions. They give earth scientists the possibility to formulate quantitative hypotheses on high-impact events, and analyse the effects of turning the 'knobs' of the model.

(2) The need for 3D sediment flux modeling
Arctic rivers and sandurs are three-dimensional; braiding channels develop over a wide sandur area (Fig 1, Church, 1972). The depositional architecture of fluvio-deltaic systems is controlled by a complex interaction of factors: drainage area, discharge and sediment load (i.e. sediment grain-size and density contrasts in the mouth area), sea-level fluctuation, offshore slope, tectonics, and the basinal regime of waves and tides, (Coleman and Wright, 1975; Orton and Reading 1993). These factors vary in space and time and thus require a three-dimensional analysis.

Itirbilung sandur, Baffin Island fjord head delta,

Figure 1. Itirbilung sandur, Baffin Island fjord head delta, showing the inherent three-dimensional character of these high Arctic river systems. Picture taken in the SAFE project, 1982, Canadian Geological Survey, courtesy J. Syvitski.

The deposits of a major outburst event (1996) on the Skeidararsandur, Iceland, show high lateral complexity due to topographic effects and channel dynamics and a high proportion of sediment storage in the sandur/delta plain (Magilligan et al., 2002).
Present 2-D numerical fluvio-deltaic models capture the basic longitudinal patterns of deposition but they cannot describe lateral heterogeneity of fluvial deposits (Rivenaes, 1992; Overeem et al., in press). 2-D models do not allow simulating lateral bypassing of sediment that results from channel switching. Thus the 2-D model is valid as long as it represents a system with a single, stable lobe, which cannot be justified for Arctic river and delta systems (Fig. 1). We hypothesize that in river systems with a relatively gentle gradient and large drainage basin, consequently with a large sediment storage potential, the reconstructions are probably the most unreliable. This is the case for rivers draining a large hinterland land icecap, e.g. the Laurentide Ice-Sheet or present-day Baffin Island ice-caps. Predictions of sediment fluxes to the ocean will become significantly more accurate in these settings, if the channel switching dynamics and storage of sediment in the floodplain can be estimated more precisely.
Our project ties in to the overarching task:

To formulate a three-dimensional model for sediment transfer from land into the coastal zone under land-ice conditions.

The adequacy of numerical models depends on the knowledge of initial conditions and timedependent input, and on the degree to which the processes and their responses are understood.
The selection of time-dependent variations in input parameters and the appropriate choice of process descriptions should ideally be founded on a thorough understanding of the system and the available measurements. Therefore, it is recommended, if not imperative, to confront modellers with ‘real world’ data. In this study both field observations and numerical models will be used. Both have their own shortcomings, but they are complementary. As the geological record becomes more accurate, the models will become better at explaining the forces driving change and thus at predicting the future.

Clyde river: a unique fluvial-deltaic system draining a hinterland land-ice cap
We study Clyde River in Baffin Island for a number of reasons (Fig. 2).

Itirbilung sandur, Baffin Island fjord head delta

Figure 2. TERRA satellite image of the Clyde River drainage basin and Barnes Ice Cap on East-Baffin Island.

The Clyde river drains the Barnes Ice Cap, which is considered to be a last remnant of the Laurentide Ice-Sheet (Bryson et al., 1969). This modern setting on East Baffin Island forms a unique example of a hinterland land-based ice-sheet. Hydrological regimes of rivers draining land-ice are relatively less well-studied as compared to rivers that drain actual active mountain glaciers. The Barnes Ice Cap is located approximately 70 km upstream from the river mouth, which is comparable to the distance of the Laurentide Ice sheet front to the coast line at for example the New Jersey margin under glacial conditions.
The ice-sheet has retreated significantly over the last 300-400 years (Ives, 1962; Loken & Andrews, 1966). Retreating ice-conditions are generally considered as the period of higher impact on the margin stratigraphy, due to drainage of melt water and exposure of fresh paraglacial sediment.
Baffin Island has a history of glacial lake outbursts. A proglacial lake, Generator Lake, exists in the upstream area of the Clyde river drainage basin (fig 3). Glacial lake outburst deposits have been recognized in Coronation Fjord on East Baffin Island (Syvitski and Blakeney, 1983) and McBeth Fjord (Harrison, 1966). A Jökulhlaup, which drained 5.9 × 106 m3 in 30 hours, was actually monitored in the South River in the Ekalugad Fjord (Church and Gilbert, 1975).

Figure 3. Drainage basin Clyde River.
The fact that the relief in the interior of Baffin Island is relatively moderate is of critical importance. The maximum elevation is ~500-600 m. This makes it an unique analogue for the shallow shelf area as compared most other arctic rivers, which have significant relief and associated mountain glaciers in their drainage basin.

Sedimentary and glaciological numerical models
Hydrotrend
No coupling exists between current glaciological and stratigraphic models. Advanced glaciological modeling currently does provide river discharge estimates based on ice-melt rates (Marshall, 1999; 2000). A major drawback is the low time-resolution of the glaciological model output. If we apply a similar large time-step in our stratigraphic model, we even out the peak flooding events, which have an important impact on the stratigraphic variability.

As an alternative approach we use a climate-driven hydrological model, HydroTrend (see Syvitski et al., 1998; Syvitski & Morehead, 1999), to generate both discharge and sediment-load time series with some stochastic variance for our stratigraphic simulation model SedFlux. The model predicts the discharge (Q) and sediment load (Qs) on a daily basis based on drainage basin properties (such as drainage area, A and relief, H) and hydrological parameters (temperature, T, precipitation, P, evapotranspiration) (Fig 4., Syvitski et al., 2000). HydroTrend has successfully been used for discharge and sediment load estimates in the Arctic (e.g. in a pilot-study for the MacKenzie river). Recently, Syvitski (2003) predicted changes in the sediment load of Arctic rivers under a global warming scenario. A glacier model is incorporated in HydroTrend adding ice-melt to the water mass balance or storing precipitation in the ice depending on the temperature trend. However, this routine is designed for thin, mountain glaciers and has not been tested for a land-ice sheet regime, nor for glacial lake outbursts. Syvitski and Andrews (1994) presented scenarios of sediment flux changes as a result of two different global warming scenarios for the Arctic area, but noted the neglection of glacial lake outbursts as a major limitation.
This improvement in HydroTrend will allow us to realize more realistic predictions of sediment fluxes towards the coastal zone under scenarios of climate change.

Example of river runoff estimates.

Figure 4. Schematic diagram of the modeling sequence (input-processes-output).

SedFlux
Our numerical process simulation model predicts the varying impact of the sedimentary processes on the stratigraphic architecture. SedFlux simulates fluvio-deltaic stratigraphy by including the following processes (for a more detailed description we refer to Syvitski & Hutton, 2001):

Stochastic channel avulsion.
Floodplain sedimentation.
Spreading of coarse bedload as river mouth deposits.
Dispersal of a suspended sediment plume from a river through either hypopycnal or hyperpycnal plumes.
Dispersal and sorting of sea-floor sediment by ocean storms, failure and subsequent transport as sediment gravity flows (turbidity currents or debris flows).

SedFlux has been tested over a range of conditions (e.g. the Eel Margin (Morehead & Syvitski, 1999) and the Knight Inlet in British Columbia (Morehead et al., 2001)). Figure 5 shows a stratigraphic profile of a SedFlux simulation, representative for the shallow East Coast margin over the last glacial-interglacial cycle. The main deltaic sediment wedges match the architecture as reconstructed from geoaccoustic data. The profile shows variation in the grain-size distribution with depth, but the uncertainty in the ice-sheet sediment output dynamics limits the reliability of this model results.

Example of a 2D Sedflux simulation

Figure 5. Example of a 2D Sedflux simulation that mimicks the sedimentary development over the last ~30,000 years of the New Jersey margin along a longitudinal profile from the present coast to the margin drop-off.
* This coarse wedge located further offshore has been deposited at low sea level and with river input from a large drainage basin which includes the Laurentide ice sheet.
* During the later phase of the melting of the large ice-sheet the discharge and sediment-load pulses were probably more erratic (as reflected from the fine-grained matrix, with distinct coarse deposits resulting from peak events).
*the Holocene deposits are finer and sedimentation rates are lower.

We have added a subaerial delta routine to Sedflux3D to examine the coupled delta/basin deposition record. The model employs a nested approach, in which first the dynamics of the main channel belt are calculated and subsequently sediment is distributed laterally. Sedimentation curves for the lateral direction are error functions, because the channel belt or delta plume axis is the zone of the most active sedimentation and the effects tend to decay spatially away from it (Overeem et al., 1999; Overeem & Weltje, 2001; Stewart et al., 2002).
Preliminary results show the importance of the frequency of channel switching on stratigraphic architecture. For a low-gradient setting channel switching appears to be a dominant control on the sedimentary patterns. The progradation pattern of the delta front depends heavily on the rate of channel switching and to a lesser extend on sea-level change and sediment supply (Fig. 7). This influence of channel pattern in the floodplain on architecture in the delta front zone applies for development of incised valleys along the delta front as well. Incision is seen to be more important if the channel belt is confined than if the channels avulse frequently (Overeem, 2002).

Theoretical 3D model runs showing the importance of channel switch frequencies.

Figure 6. Theoretical 3D model runs showing the importance of channel switch frequencies (A vs. B-scenarios) for the stratal architecture, as compared to changes in sea level (A2 vs. B1) or sediment supply (B1 vs. B2).

Recent work of Magilligan et al., (2002) on the deposits of a major outburst event (1996) on the Skeidararsandur, emphasizes the limitations in classic sedimentological models of sandur deposits (e.g. Bluck, 1974; Boothroyd and Nummedal, 1978). It was found that lateral complexity due to topographic effects and channel dynamics during the peak event dominates over longitudinal trends, which underlines the importance to formulate 3D numerical models.

Methodology
Task 1 Review existing data
Existing data will be assembled to place the Clyde river delta setting in a context of time and spatial variation of sedimentation rates. The sparseness of datable material in the arctic environment of Baffin Island makes exact unraveling of sedimentation rates virtually impossible. However, careful comparison of the existing unpublished data can provides us with a range of sedimentation rates over different time-scales. A relatively nearby fjord-head delta that does not drain the Barnes Ice Cap is the McBeth river delta, which has been investigated in the framework of the Sedimentology of Arctic Fjords (SAFE) project (Syvitski and Blakeney, 1983; Syvitski, 1984; Syvitski and Praeg, 1987).

Task 2 Aerial photograph and satellite data interpretation
Delta progradation as mapped from aerial photographs approximates sedimentation rates. Channel system dimensions and dynamics (e.g. avulsions) can be mapped from the photos as well. In addition, the extent of the sediment plume in the fjord yields an indication of the travel distance of the suspended load. Different sets of aerial photos taken for NAPL trimetrigon are available from 1958 and 1960 (Falconer, 1962). More recent photos have been taken for the SAFE project (1982-1985) and for the aerial triangulation database of the Canada Land Surveyors on a 1:60,000 scale (1994). We suspect that we can add to this database with even more recent satellite imagery (e.g. RADARSAT image of 29 July 1997, available from the Canadian Ice Service).

Task 3 Field mapping of sedimentary architecture (3 weeks)
The sedimentary architecture of the Clyde river delta will be mapped along longitudinal and cross-sectional transects. The work will have an exploratory character and serve as a ‘ground truthing’ of the aerial photograph interpretation. Comparisons to data collected by Church (1972) for the small drainage areas of the Lewis river and the Ekalugad sandur (both on Baffin Island) can be used to place the Clyde data into a regional perspective. Specific parameters that are vital to the comparison with numerical model output include:

Downstream grain-size distribution trends The modeling output shows stratigraphy along a 2D longitudinal profile. Field estimates of grain-size distributions will be made. Additionally, we’ll sample a limited number of key locations for sedimentological laboratory analysis.
Distance that coarse-grained river mouthbars extend offshore The spreading distance of bedload from the river mouth is a user-defined parameter in SedFlux, field data will give us better grip on this parameter.
Channel dimensions and channel stability Channel dimensions, i.e. channel cross sectional area, yield rough estimates of bankfull discharges, Q. These can be compared with the HydroTrend estimates. The channel stability is important to define the avulsion rates in our model.

Task 4 Exploratory numerical modeling
The HYDRO1k Digital Elevation Model (DEM) of Baffin Island will be used to derive characteristic of the Clyde river drainage area.
Regional meteorological data are available from Clyde’s weather station. The annual temperature record dates back to 1948, which is well within the limits of the temporal coverage of the aerial photographs.
Satellite image interpretation and data from the automated weather station on the Barnes Ice Cap (available from Atmospheric Environmental Services, Canada) will form the input dataset for glacier dynamics in the hinterland. Andrews and Miller (1972) presented data on the summerfreezing line and glacier equilibrium-line altitudes on Baffin Island.
Based on this data, HydroTrend generates a discharge and sediment input dataset for SedFlux (Fig. 4). The model estimates may be compared to the measured channel dimensions. The dataset on Baffin rivers compiled by Church (1972) is probably the most complete to test against our model results.
Subsequently, we can simulate river and delta deposition with Sedflux over an approximately 50 year timespan. Sensitivity tests will be interactively matched to the field data base. Key calibration parameters are the progradation rate and the grain-size trends, which can be directly compared with field data. Avulsion rate and the floodplain sedimentation are the first-order parameters we will quantify. These examining experiments should identify the limitations in our present models.

References
You can find the referred literature under this link: references

Results

In the framework of this project we worked on the development of our 3D numerical models, AquaTellUS and SedFlux. These models have not yet been fully applied to the specific case-study of Clyde River. But we are publishing the descriptions of these models in a SEPM special Issue in 2005. Here you can find a PDF of the draft paper: Three-dimensional modeling of deltas.

Overeem, I., Briner, J.P., and Kettner, A.J., 2004. How dynamic are fluvial-deltaic systems draining land-ice? A case-study of Clyde River, Baffin Island. 34th Annual International Arctic Workshop, Boulder, Colorado, USA. Abstract | poster (17Mb!)

Here a short photo impression of our field season in July-August 2003. A couple of pictures illustrate the geological fieldwork; shell sampling, sedimentological descriptions and terrace elevation mapping. Some additional photos show the typical way of traveling the fjord in August, the fieldcamp and our interaction with the people of Clyde River.