See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/252428628

The investigation of the formation of river delta deposits is an important

subject in geomorphology. Recently a reduced complexity model which
simulates the process of delta format...

Article · April 2009

CITATIONS READS

0 114

5 authors, including:

Peter Molnar Wolfgang Kinzelbach

ETH Zurich ETH Zurich
174 PUBLICATIONS   4,748 CITATIONS    388 PUBLICATIONS   9,801 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Rehabilitation and management strategy for over-pumped aquifers under a changing climate View project

TAHMO Weather Data Analysis View project

All content following this page was uploaded by Wolfgang Kinzelbach on 09 November 2017.

The user has requested enhancement of the downloaded file.

GEOPHYSICAL RESEARCH LETTERS, VOL. ???, XXXX, DOI:10.1029/,

1 The topography of inland deltas: observations,

2 modeling, experiments
1 2 4 3 5
H.J. Seybold, P. Molnar, D. Akca, M. Doumi, M. Cavalcanti Tavares,
3 5 2 1,5
T. Shinbrot, J.S. Andrade Jr., W. Kinzelbach, H.J. Herrmann

H. J. Seybold, Computational Physics, IfB, ETH Zurich, 8093 Zurich, Switzerland (hsey-

[email protected])

1
Computational Physics for Engineering

D R A F T October 27, 2009, 3:46pm D R A F T

X- 2 SEYBOLD ET AL.: THE TOPOGRAPHY OF AN INLAND DELTA

3 The topography of inland river deltas is influenced by the water-sediment

4 balance in the distributary channel system and the local evaporation and seep-

5 age rates. We apply a reduced complexity model to simulate an inland delta

6 and compare the results with the Okavango Delta and with a laboratory ex-

7 periment. We show that water loss through evapotranspiration and infiltra-

8 tion in inland deltas produces fundamentally different dynamics of water and

9 sediment transport than coastal deltas, especially vis a vis deposition asso-

10 ciated with expansion-contraction dynamics at the channel head. These dy-

Materials, IfB, ETH Zurich, 8093 Zurich,

Switzerland

2
Institute of Environmental Engineering,

IfU, ETH Zurich 8093 Zurich, Switzerland

3
Department of Biomedical Engineering,

Rutgers University, Piscataway, NJ 08854,

USA

4
Institute of Geodesy and

Photogrammetry, ETH Zurich 8093 Zurich,

Switzerland

5
Departamento de Fı́sica, Universidade

Federal do Ceará, 60451-970 Fortaleza ,

Ceará, Brasil

DRAFT October 27, 2009, 3:46pm DRAFT

SEYBOLD ET AL.: THE TOPOGRAPHY OF AN INLAND DELTA X-3

11 namics lead to a systematic decrease in the mean topographic slope of the

12 inland delta with distance from the apex that follows a power law with ex-

13 ponent α = −0.69 for both simulation and experiment. In coastal deltas,

14 on the contrary, the slope increases toward the end of the deposition lobe.

D R A F T October 27, 2009, 3:46pm D R A F T

X-4 SEYBOLD ET AL.: THE TOPOGRAPHY OF AN INLAND DELTA

1. Introduction

15 Inland deltas like the Okavango and coastal deltas like the Mississippi are morpho-

16 logically distinct because their deposition patterns are influenced by different dominant

17 fluvial processes. Coastal deltas are dominated by wave and tide action and coastal cur-

18 rents that separate the subaerial and subaqueous parts of the delta lobes. Inland deltas, on

19 the other hand, experience none of these flows, but are dominated by evapotranspiration,

20 infiltration, and the growth of bank- and island-stabilizing vegetation.

21 Inland deltas are less well studied than their coastal counterparts. We investigate

22 geomorphological features of inland deltas, which we compare with coastal ones. We focus

23 on the Okavango Delta as a case study, and we compare its topography with computational

24 simulations using a new reduced complexity model [Seybold et al., 2007, 2009], as well as

25 a small-scale laboratory experiment.

26 As we will explain, the geomorphological evolution of the Okavango Delta surface is

27 strongly influenced by the dynamics and transport capabilities of its constituent chan-

28 nels, where interactions with the local slope and properties of the alluvium and surface

29 vegetation play an essential role. Additionally, tectonic movement of the surface can play

30 a role that we briefly discuss.

31 In recent years several field measurements and hydrological models of the Okavango

32 Delta have been produced, including measurements of sediment rating in the delta chan-

33 nels and groundwater fluxes [Bauer et al., 2004; McCarthy, 2006; Gumbricht et al., 2005;

34 Brunner et al., 2007]. The models constructed so far, however, are only capable of short

35 term forecasting, with management horizons of maximally 50 years.

D R A F T October 27, 2009, 3:46pm D R A F T

SEYBOLD ET AL.: THE TOPOGRAPHY OF AN INLAND DELTA X-5

36 In order to describe longer-term dynamics of sediment transport, and consequent de-

37 velopment of the delta, new models must be developed that reduce the complexity of the

38 hydrological and sedimentary equations while maintaining the essential physics [Brasing-

39 ton and Richards, 2007; Crave and Davy, 2001]. In particular, the model proposed by

40 Seybold et al. [2007] has proven to describe coastal delta formation successfully, and we

41 extend this model here to include evaporative water loss and seepage for the study of

42 morphogenesis of inland deltas like the Okavango.

43 To study the processes leading to this rich geomorphology, we present computational

44 model accompanied by a laboratory-scale flume experiment. Flume experiments on delta

45 formation have been carried out in several laboratories, notably the Earthscape Experi-

46 ment laboratory in St. Anthony Falls and the Exxon Mobil laboratories [Hoyal and Sheets,

47 2009; Martin et al., 2009]. Also recently, the formation of alluvial fans caused by rapid

48 water release has been studied by Kraal et al. [2008]. However, experimental work on

49 inland deltas including evaporation is new. We use these experiments as a verification for

50 the modeling and as a tool to understand the interplay between the dominant sedimentary

51 processes.

52 The paper is organized as follows: first we show how the computational model of Seybold

53 et al. [2007, 2009] was modified to include evaporation. Then we describe our experiment

54 mimicking inland delta formation. We then discuss results and analysis in two subsections.

55 First we compare the model results with topographic data of the Okavango, topological

56 structure of the channel system and the influence of evaporation on the sedimentation

D R A F T October 27, 2009, 3:46pm D R A F T

X-6 SEYBOLD ET AL.: THE TOPOGRAPHY OF AN INLAND DELTA

57 process. Second, we study the delta slope distributions in both modeled and observed

58 data and discuss the differences.

2. Computational Modeling

To simulate inland delta formation, we extended the model of Seybold et al. [2007, 2009]

to include evaporation and seepage in the conservation equation for water mass, as follows.

The landscape is discretized on a square grid where the elevation of the topography Hi

and the absolute water level Vi are defined on the nodes. The water flux Iij on a bond

is related to the average water depth σij and the pressure drop Vi − Vj by the following

relation:

Vi − H i Vj − H j
Iij = + (Vi − Vj ). (1)
|
2 {z
2 }
=σij

Conservation of water flows entering and leaving node i is given by:

X
Vi = Vi′ + δt (Iij + Ei ) = 0, (2)
N.N.

where Ei defines the loss of water due to evaporation or infiltration and the sum runs over

the von Neumann neighborhood of a given cell. The evaporation rate is modeled by the

following phenomenological formula

Ei = di Ê (3)

59 where Ê defines the maximum evapotranspiration rate and so determines how far the

60 delta progrades into the domain. The increase of seepage in distal parts of the delta is

61 modeled by the normalized distance of the cell from the inlet, di.

Boundary and initial conditions are needed to close the problem. The landscape is

initialized with an inclined plane, distorted by random perturbations. Open boundary

D R A F T October 27, 2009, 3:46pm D R A F T
SEYBOLD ET AL.: THE TOPOGRAPHY OF AN INLAND DELTA X-7

conditions are applied at all boundaries except in the nodes closest to the inlet where no-

flow boundaries are applied to keep the water flowing into the domain. Water and sediment

are injected into the system by defining an input flux I0 of water and sediment s0 at an

entrance node. The landscape is initialized with a given water level below the ground,

and runoff is produced when the water level exceeds the surface. The sedimentation-

erosion rate dSij is modeled by a phenomenological deposition-erosion law with a common

constant c for erosion and deposition [Seybold et al., 2009]. The deposition-erosion law

depends only on the magnitude of the flow,

dSij = c(I ⋆ − |Iij |), (4)

62 and a threshold I ⋆ which determines whether erosion or deposition occurs between nodes

63 i and j. After the sedimentation-/erosion process the sediment Jg es is distributed to the

64 outflow directions according to their relative magnitudes of the corresponding water flux,
P
65 e.g. Jkl = Ikl / i Iil . Sediment transport and topography update is done in the same way

66 as in Seybold et al. [2009].

67 Two types of channel ends need to be included in inland deltas: newly forming channels

68 where dI = I(t + δt) − I(t) > 0 do not show sedimentation at the front, and channels

69 that are drying with high deposition rates at their terminal ends. These are included by

70 distinguishing ends with dI > 0 and dI < 0. If dI < 0, deposition is applied according to

71 Eq.4, while if dI > 0, no deposition is applied in the final node of a channel end.

3. Experimental Modeling

72 To validate the computations, we performed a laboratory-scale flume experiment includ-

73 ing evaporation and seepage. Our experiment consists of a 1m by 1m aluminum basin

D R A F T October 27, 2009, 3:46pm D R A F T

X-8 SEYBOLD ET AL.: THE TOPOGRAPHY OF AN INLAND DELTA

74 that is fixed at an inclination of about 6 degrees running along the basin diagonal. An

75 initial surface is created using a uniformly sloped sediment layer with an height of 5cm at

76 the inlet diminishing to zero over about 1.1m (Fig. 1).

77 As sediment, we use crushed glass with diameter 50 to 120 microns, with a bulk density

78 of ̺ = 2.2g/cm3. The sediment is continuously mixed with water by a marine-type

79 impeller in an upstream tank, and is injected steadily into the basin using a peristaltic

80 pump. The volumetric sediment concentration was approximately 0.05 and the inflow

81 was 1000ml/h. Water is continuously pumped out of the basin at the bottom so that no

82 standing water accumulates. Water is evaporated by an array of fifteen 300W heat lamps

83 that are fixed 15cm above the surface.

84 The experiment was run as follows: a water/sediment mixture was injected into the

85 flume over 45 minutes, followed drying over 2:15 hours. In the following we will call one

86 period of injection and drying an “epoch”, where epoch 0 stands for the initial condition.

87 A photo of the initial setup is shown in Fig. 1b. After complete drying, the surface

88 topography is scanned using a Breukmann OptoTOP-SE 3D scanner. Complete drying

89 is necessary to avoid specular reflections produced by wet Sand that would disturb the

90 scanning procedure. The scanning technique is based on a stereoscopic measurement,

91 in which regular fringes are projected onto the surface and the stripes’ deformation is

92 measured using a CCD camera. From the deformation of these lines the topography can

93 be reconstructed with an accuracy of 100 microns [Burke et al., 2002; Akca et al., 2007].

94 Due to the limited field-of view of the scanner (40cm × 31cm), several scans are combined

95 into a co-registered mosaic of the entire surface, using a least square matching method

D R A F T October 27, 2009, 3:46pm D R A F T

SEYBOLD ET AL.: THE TOPOGRAPHY OF AN INLAND DELTA X-9

96 described elsewhere Gruen and Akca [2005]. We use an invariant reference point outside

97 of the sedimentation domain to co-register the different sediment layers so that we can

98 obtain temporal and spatial distributions of sediment during the experiment.

99 The total deposited volume of each epoch including pore space can be obtained by

100 subtracting the co-registered surfaces of two successive topography scans. For the four

101 injection periods we obtained V0,1 = 573cm3 , V1,2 = 904cm3 , V2,3 = 614cm3 and V3,4 =

102 740cm3 , where the indices denote the epochs before and after the deposition run.

103 To compare with the wet delta case, we remove the heat lamps and change the boundary

104 conditions at the downstream end of the flume to preserve a constant water level, while

105 keeping the other parameters unchanged.

106 Although the experimental results cannot be directly compared with natural Deltas due

107 to large scale variations, the deposition processes and the resulting patterns are similar

108 to those observed in nature.

4. Analysis of inland delta formation

4.1. Modeling and observations of delta channels

109 Visually, the computational inland delta model produced deposition structures and

110 channels similar to natural deltas. The topological similarities of the different delta chan-

111 nel systems are quantified by estimating the fractal dimension of simulated and observed

112 networks with the box counting technique [Feder , 1989; Turcotte, 1997].

113 A least squares fit of a power law, N ∼ s−D to the data yields a fractal dimension of

114 D = 1.85 ± 0.05 for the Okavango Delta, as compared with the simulation result of D =

115 1.84 ± 0.05. The pattern of the flooded area of the Okavango was extracted through the

D R A F T October 27, 2009, 3:46pm D R A F T

X - 10 SEYBOLD ET AL.: THE TOPOGRAPHY OF AN INLAND DELTA

116 vegetation by a combined analysis of high resolution aerial photos from GoogleEarthTM and

117 NOAA satellite measurements. The added processes of evaporation and infiltration lead to

118 complex dynamics of channel extension by erosion and contraction due to deposition at the

119 channel heads during low flows which we could not observe in wet deltas. Furthermore,

120 the model reproduces the development of bank levees by lateral deposition on channel

121 margins (Fig. 1c). The natural formation of bank levees by overbank deposition occurs in

122 the Okavango and in many natural dryland rivers [McCarthy et al., 1988] due to processes

123 such as vegetation that are notoriously difficult to simulate. This riparian vegetation

124 [McCarthy et al., 1992] modulates overbank deposition in the Okavango delta and affects

125 not only deposition but also evapotranspiration and infiltration rates. In the model,

126 natural levee formation is due to lateral topographic gradients and the balance of sediment

127 supply and transport.

4.2. Delta slope distributions

A useful topographic metric to quantify the shape of a delta is the mean slope as a

function of downstream distance from the delta apex. We define the mean topographic

slope S(d) averaged over circular arcs at a distance [d, d + dr] from the delta apex. In

order to compare the different observed, numerical and experimentally modeled slopes,

we normalize the results with the overall spatial mean,

1
S(d) = hSid+dr . (5)
hSi

128 The averages are computed over a spatial domain which contains the whole delta surface.

129 S(d) is a useful measure of the delta form for two reasons. First, it is an integral measure

130 (over the whole domain) of the processes of deposition in space at an equal distance from

D R A F T October 27, 2009, 3:46pm D R A F T

SEYBOLD ET AL.: THE TOPOGRAPHY OF AN INLAND DELTA X - 11

131 the apex. Second, topographic slope is a fundamental variable for sediment transport in

132 transport capacity-limited conditions such as delta distributary channels.

Most landscape evolution models relate sediment transport capacity Qs and the ero-

sion/deposition rate to upstream drainage area A and local slope Qs ∼ Am S n , where

m and n are fundamental exponents which reflect the erodibility of the surface and the

erosivity of the flow (e.g., Davy et al. [2009]; Niemann et al. [2001]. Delta distributary

systems are conduits without an increase in drainage area, and it is therefore reasonable

to assume that the specific sediment transport rate will scale as,

Qs ∼ S n . (6)

133 Therefore, S(d) is an indication of the radial distribution of the potential sediment

134 transporting capacity of the delta system, where the specific water q and sediment Qs

135 transport rate and the downstream change in total distributary channel width W (d)

136 together determine the total sediment transporting capacity of the system.

137 In order to compare Okavango and simulated topography more directly, we rescaled

138 the horizontal extents of the simulation to fit the experimental domain. A comparison of

139 S(d) for the modeled surface, the experiment and the Okavango DEM surface [Gumbricht

140 et al., 2005] is shown in Fig. 2.

141 The modeled surface shows a gradual decrease in S(d) downstream as the sediment

142 transporting capacity in smaller (but more numerous) channels decreases, and the delta

143 becomes flatter as a consequence. Furthermore the transport capacity decreases due to the

144 loss of transporting water by evaporation and seepage. Local variations may be associated

145 with the varying heads of individual distributary channels which may be actively eroding

D R A F T October 27, 2009, 3:46pm D R A F T

X - 12 SEYBOLD ET AL.: THE TOPOGRAPHY OF AN INLAND DELTA

146 and so can be expected to have a higher local slope. Fig. 2a displays the statistical average

147 over nine simulation runs with the same evaporation rate and boundary conditions, but

148 different random perturbations to the initial surface. The errorbars indicate the statistical

149 variability. The parameters for this simulation have been chosen to be I0 = 1 × 10−3 , I ⋆ =

150 −7.5 × 10−6 , s0 = 0.0025, c = 0.1 [Seybold et al., 2009], and the evaporation rate in this

151 case is set to Ê = 5 × 10−8 .

152 As shown in the inset of Fig. 3, the local slopes obtained from the dry delta experiments

153 as well as from model simulations clearly follow power-law behavior, S = a(d − d0 )α .

154 As one would expect, the least-squares fit to the data sets of this function indicates

155 that both the a and d0 depend on the particular experimental/simulation conditions. In

156 particular, the parameter d0 assumes negative values in the two cases, which is consistent

157 with the physical condition of divergence-free slope profiles for positive distances. More

158 strikingly, however, is the fact that we obtain identical power-law exponents α ≈ −0.69

159 for experimental data and model simulations.

160 As shown in the main plot of Fig.3, this is corroborated by the data collapse obtained

161 by rescaling S to S ∗ = S/a and plotting it against d∗ = d − d0 .

162 The Okavango surface shows a more complex behavior affected strongly by local geology

163 and tectonics. During the first 100km the Okavango is confined between the fault lines

164 forming the a confined area, called the Panhandle.

165 Outside the Panhandle the delta surface is almost totally flat with only small local

166 variability around the constant slope of the fan. The increase in mean slope at the

167 bottom end of the delta is a consequence of the Kunyere and Thamalakane fault lines.

D R A F T October 27, 2009, 3:46pm D R A F T

SEYBOLD ET AL.: THE TOPOGRAPHY OF AN INLAND DELTA X - 13

168 The downstream distribution of slope also highlights the fundamental difference between

169 wet (coastal) and dry (inland) deltas. The Mississippi Delta Balize Lobe profile from

170 bathymetric data in [Seybold et al., 2009; Divins and Metzger , 2006] shows that in coastal

171 deltas the mean slope increases downstream as the distributary channels enter the ocean

172 and sediment deposition becomes limited by the settling velocity of particles and their

173 advection by currents and tides (Fig. 4). In the DEM data of the Balize Lobe the slope

174 is increasing exponentially with a characteristic inverse length of τ = 0.026 at the head of

175 the delta (Fig. 4b). Furthermore it can be seen that with time the Mississippi River has

176 adjusted the average slope of its fluvially accessible area to the optimal transport capacity

177 of the stream and therefore we observe a constant slope on the coastal plain.

178 A strongly increasing slope toward the end of the lobes is also observed in the reduced

179 complexity model simulations of Seybold et al. [2009] but the functional behavior of the

180 increase at the delta head could not be verified from the data (Fig.4a). The average slope

181 of the birdfoot delta simulation presented in Fig. 4a is averaged over 5 different samples

182 with the same set of parameters but different random noise in the initial conditions of

183 the surface. The parameters have been chosen similar to Seybold et al. [2009], namely

184 I0 = 0.00017, I ⋆ = −4 × 10−6 , s0 = 0.00025, c = 0.1. In the simulation the river first

185 adjusts the slope imposed by the initial conditions to its transport capacity (Fig. 4a-I)

186 until it flows at an almost constant slope in the newly formed lobe (Fig. 4a-II between

187 d = 80 − 120km). The decay of the slope in the initial part can be fitted by a power law

188 with an exponent around α ≈ −0.4.

D R A F T October 27, 2009, 3:46pm D R A F T

X - 14 SEYBOLD ET AL.: THE TOPOGRAPHY OF AN INLAND DELTA

189 The same phenomenon is observed in the experiment (Fig. 4c). We find a decreasing

190 slope in the initial part of the flume where the stream adjusts the base slope due to erosion

191 and deposition (Fig. 4c-I). In a middle part (Fig. 4c-II), the coastal slope is adjusted to

192 the stream transport capacity and this is constant over a certain range. Toward the head

193 of the delta the slope increases strongly again and then drops to the base slope of the

194 basin (Fig. 4c-III). Although the average slope for the experiment is noisy, the three parts

195 can be identified. The functional behavior of the increasing part could not be determined

196 from the data. We have removed the very initial part of the flume from the average slope

197 calculation because this section is strongly influenced by details associated with the slurry

198 injection.

5. Conclusions

199 In this paper we adjusted and applied a reduced complexity model which was originally

200 developed for coastal deltas [Seybold et al., 2009], to an inland delta, using elevation- and

201 slope-based metrics to describe its shape and change. The Okavango Delta was used as a

202 reference to compare with the model, together with a small-scale laboratory experiment

203 to verify the modeling results and understand the time evolution of the delta system.

204 The chief finding of this study is that water loss through evapotranspiration and in-

205 filtration in an inland delta combine to produce a fundamentally different dynamics of

206 water and sediment transport compare to coastal deltas. In particular, the expansion-

207 contraction dynamics at the channel heads and the deposition associated with these dy-

208 namics lead to a consistent decrease in the mean topographic slope in inland deltas with

209 distance from the apex. The decrease of the slope in the experimental as well as the mod-

D R A F T October 27, 2009, 3:46pm D R A F T

SEYBOLD ET AL.: THE TOPOGRAPHY OF AN INLAND DELTA X - 15

210 eled systems shows a clear power law behavior. By rescaling the variables it is possible to

211 collapse the two curves on a single power law with exponent α = −0.69.

212 Inland deltas are transport capacity-limited systems. The systematically decreasing

213 topographic slope from the delta apex is indicative of a drop in specific sediment trans-

214 porting capacity along the delta channel system, which, combined with the number and

215 width of the distributary channels, governs the sediment transport capacity of the delta

216 as a whole and the sedimentation within the fan. This simple topographic measure also

217 highlights the difference between inland and coastal deltas insofar as topographic slope in

218 the latter case increases dramatically at the land-ocean interphase.

219 Acknowledgments. This work was funded by the Swiss National Foundation Grant

220 NF20021-116050/1.

References

221 Akca, D., F. Remondino, D. Novàk, T. Hanusch, G. Schrotter, and A. Gruen (2007),

222 Performance evaluation of a coded structured light system for cultural heritage applica-

223 tions, in Videometrics IX, Proc. of SPIE-IS&T Electronic Imaging, pp. 64,910V–1–12,

224 San Jose (California), USA.

225 Bauer, P., G. Thabeng, F. Stauffer, and W. Kinzelbach (2004), Estimation of the

226 evapotranspiration-rate from diurnal groundwater level fluctuations in the okavango

227 delta, botswana, Journal of Hydrology, 288 (3-4), 344–355.

228 Brasington, J., and K. Richards (2007), Reduced-complexity, physically-based geomor-

229 phological modelling for catchment and river management, in Geomorphology, special

D R A F T October 27, 2009, 3:46pm D R A F T

X - 16 SEYBOLD ET AL.: THE TOPOGRAPHY OF AN INLAND DELTA

230 issue, vol. 90, pp. 171–177, Elsevier, doi:10.1016/j.geomorph.2006.10.028.

231 Brunner, P., H. Hendricks Franssen, L. Kgotlhang, P. Bauer-Gottweiler, and W. Kinzel-

232 bach (2007), Howremote sensing contribute to ground water modeling, Hydrogeol. J.,

233 15 (1), 5–18.

234 Burke, J., T. Bothe, W. Osten, , and C. Hess (2002), Reverse engineering by fringe

235 projection, Interferometry XI: Applications, SPIE, 4778, 312–324.

236 Crave, A., and P. Davy (2001), A stochastic ”precipiton” model for simulating ero-

237 sion/sedimentation dynamics, Computers & Geosciences, 27 (7), 815 – 827, doi:DOI:

238 10.1016/S0098-3004(00)00167-9.

239 Davy, P., , and D. Lague (2009), Fluvial erosion/transport equation of landscape evolution

240 models revisited, J. Geophys. Res., 114 (F03007), doi:doi:10.1029/2008JF001146.

241 Divins, D., and D. Metzger (2006), Coastal relief model, Central Gulf of Mexico Grids.

242 Feder, J. (1989), Fractals, 4 ed., Plenum Press, New York.

243 Gruen, A., and D. Akca (2005), Least squares 3d surface and curve matching, Journal of

244 Photogrammetry and Remote Sensing, 59 (3), 151–174.

245 Gumbricht, T., T. S. McCarthy, and P. Bauer (2005), Micro-topography of the okavango

246 delta wetlands, Earth. Surf. Proc. Land., 30, 27–29.

247 Hoyal, D., and B. Sheets (2009), Morphodynamic evolution of cohesive ex-

248 perimental deltas, J. Geophys. Res. Earth-Surface, 114 (F02009), 78–89, doi:

249 10.1029/2007JF000882.

250 Kraal, E., M. van Dijk, G. Postma, and M. Kleinhans (2008), Martian stepped-delta

251 formation by rapid water release, Nature, 451, 973–976.

D R A F T October 27, 2009, 3:46pm D R A F T

SEYBOLD ET AL.: THE TOPOGRAPHY OF AN INLAND DELTA X - 17

252 Martin, J., B. Sheets, C. Paola, and D. Hoyal (2009), Influence of steady base-level rise on

253 channel mobility, shoreline migration, and scaling properties of cohesive experimental

254 delta, J. Geophys. Res., 114 (F03017), doi:doi:10.1029/2008JF001142.

255 McCarthy, T. (2006), Groundwater in the wetlands of the Okavango Delta, Botswana,

256 and its contribution to the structure and function of the ecosystem, J. Hydrol., 320,

257 264–282.

258 McCarthy, T. S., K. H. Rogers, I. G. Stanistreet, W. N. Ellery, B. Cairncross, K. Ellery,

259 and T. Gorbicki (1988), Features of channel margins in the okavango delta, Palaeoecol.

260 Africa, 19, 3–14.

261 McCarthy, T. S., W. N. Ellery, and I. G. Stanistreet (1992), Avulsion mechanisms on the

262 okavango fan, botswana - the control of a fluvial system by vegetation, Sedimentology,

263 39 (5), 779–795.

264 Niemann, J., N. Gasparini, and G. Tucker (2001), A quantitative evaluation of playfair’s

265 law and its use in testing long-term stream erosion models, Earth. Surf. Proc. Land.,

266 26 (12), 1317–1332.

267 Seybold, H., J. Andrade Jr., and H. J. Herrmann (2007), Modeling river delta formation,

268 Proc. Natl. Acad. Sci. USA, 104 (43), 16,804–16,809.

269 Seybold, H., P. Molnar, H. Singer, J. Andrade Jr., H. Herrmann, and W. Kinzelbach

270 (2009), Simulation of birdfoot delta formation with application to the Mississippi Delta,

271 J. Geophys. Res. Earth-Surface, 114 (F03012), doi:doi:10.1029/2009JF001248.

272 Turcotte, D. (1997), Fractals and chaos in geology and geophysics, Camebridge University

273 Press.

D R A F T October 27, 2009, 3:46pm D R A F T

X - 18 SEYBOLD ET AL.: THE TOPOGRAPHY OF AN INLAND DELTA

Pump 1

Pump 2 (a) (b) (c)

Figure 1. Sketch of the experimental setup (a). Water and sediment are fed from a container

where the sediment is kept in suspension using an electric mixer. The sediment-water suspension

then is injected into the basin using a peristaltic pump (Ismatec ecoline, pump 1). At the end

of the flume remaining water is pumped out of the basin (pump 2). (b) Initial condition of the

experiment. The flume is mounted on an inclined concrete base and an initial conical landscape

of sediment is created in the first third of the domain. (c) The photo shows the deposition

pattern of the dry delta experiment after several cycles of delta formation with braiding streams

and levees confining the channels.

4 1.5
Epoch0
2 Panhandle Fault line Epoch3
3 1.25 Epoch4

1.5
S(d)

S(d)

S(d)

2 1

1
1 0.75
(a) (b) (c)
0.5 0 0.5
25 50 75 100 125 0 50 100 150 200 250 25 50 75 100
d [cm] d [km] d [cm]

Figure 2. Plot of the slope S(d) (a) for the simulation averaged over 9 samples, the errorbars

indicate the statistical error, (b) the Okavango derived from DEM data and (c) the experiment

(initial condition, epoch 3 and 4). Both experiment and simulation show a similar decreasing

slope with distance to the apex. The detailed analysis of the decrease is presented in Fig. 3.

D R A F T October 27, 2009, 3:46pm D R A F T

SEYBOLD ET AL.: THE TOPOGRAPHY OF AN INLAND DELTA X - 19

Figure 3. The inset shows the decay of the slope for both experiment (epoch 5, black

diamonds) and model (red crosses) in dry deltas together with the corresponding least-square

fits of a power-law of the form S = a(d − d0 )α . The fit parameters are a = 17.4 and d0 = −9.9

for the simulation (black dashed line) and a = 23.5 and d0 = −47.4 for the experiment (black

solid line). We obtain the same value for the exponent, namely, α ≈ −0.69 in both cases. In the

main plot we confirm that, by rescaling the slope data of the experiment and the simulation and

plotting S ∗ = S/a against d∗ = d − d0 , both curves can be collapsed onto a single power-law (a

straight line in log-log plot) with exponent α ≈ −0.69 (black solid line). Red stars indicate the

rescaled values of the simulation and black circles are used for the rescaled experimental data.

We also show that the slope of the Panhandle region of the Okavango delta (blue squares) follows

a similar behavior.

D R A F T October 27, 2009, 3:46pm D R A F T

X - 20 SEYBOLD ET AL.: THE TOPOGRAPHY OF AN INLAND DELTA

4 0.6 7 1 1.15
6

Normalized mean slope

0.4 0.5 0.026
1.1

log (S)
log10(S)

3 5

10
0.2 −0.4 0
4 1.05
S(d)

S(d)

2 0 −0.5
3
−0.2 −1 1
1.4 1.6 1.8 2 2.2 0 50 100 150
log10(d) 2 d [km]
1
(b) 0.95
(a) 1 (c)
I II III I II III
0 0 0.9
50 100 150 200 50 100 150 25 50 75 100 125
d [km] d [km] d[cm]

Figure 4. Plot of the slope S(d) from Eq. 5 for wet deltas: Simulation (a), the Mississippi (b)

and the wet experiment (c). The average slope for wet deltas is different from that of dry deltas

in Fig. 2. Simulation and experiment show an initial part with deceasing slope (I) where the

stream adjusts the inclination to its transport capacity followed by a part of almost constant slope

with free flow (II) and a strongly increasing part at the head of the delta (III). The inset of the

simulation (a) shows that the initial slope decreases like a power law with exponent α ≈ −0.4.

In the real DEM data (b) the Mississippi already adjusted its fluvially accessible area in the

coastal plain to its transport capacity and we observe a constant slope at the beginning and an

exponentially increasing slope at the end (inset).

D R A F T October 27, 2009, 3:46pm D R A F T

View publication stats