2
Resolving the era of river-forming climates on Mars
using stratigraphic logs of river-deposit dimensions
3
Edwin Kite1, Alan Howard2, Antoine Lucas3, Michael Lamb4, Kevin Lewis1,5.
4
5
1. Princeton University. 2. University of Virginia. 3. Université de Paris. 4. Caltech. 5. Johns
Hopkins University.
1
6
7
1. Introduction.
8
1.1. Fluvial signatures of climate events on Mars: clearer than on Earth.
9
1.2. Basin-scale mapping distinguishes 102 m-thick river-deposit-hosting
units.
10
11
2. Getting from stereopairs to stratigraphic logs.
12
3. Paleohydrology.
13
3.1. River-deposit scale varies systematically upsection.
14
3.2. Basin-wide paleodischarge variability.
15
3.3. A tool to search for abrupt climate change on Mars.
16
3.4 Timescales.
17
4. Implications for river-forming paleo-climates on Mars.
19
4.1. Both short-term intermittency and long-term intermittency are
required.
20
4.2. Mechanisms for paleodischarge variability.
21
4.3. Global context.
18
22
5. Habitability, taphonomy, and organic matter preservation.
23
6. Conclusion.
24
1
25
Abstract.
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
River deposits are one of the main lines of evidence that tell us that Mars once had a climate
different from today, and so changes in river deposits with time tell us something about how
early Mars climate varied. In this study, we focus in on one sedimentary basin – Aeolis Dorsa –
and use changes in the scale of river deposits with stratigraphic elevation as a proxy for changes
in river discharge. Meander wavelengths tighten upwards and channel widths narrow upwards,
and there is some evidence for a return to wide large-wavelength channels higher in the
stratigraphy. Meander wavelength and channel width covary with stratigraphic elevation, and
show the expected ~10:1 wavelength:width ratio, confirming that preserved-channel width is a
good proxy for original channel width at this location. The factor-of-2 variations in paleochannel
dimensions with stratigraphic elevation correspond to factor-of-3 variability in bank-forming
discharge over timescales of X Kyr (using standard wavelength-discharge scalings and widthdischarge scalings). Taken together with evidence from marker beds for variability at
stratigraphic scales <20m (which are too fine to be resolved by our logging technique), the
variation in the scale of river deposits indicates that bank-forming discharge varied at both ~10m
stratigraphic (10tbd-10tbd yr) and ~100 m stratigraphic (10tbd – 10tbd yr) timescales. Because these
variations are correlated across the basin, they record an externally imposed, basinwide forcing,
rather than internal feedbacks. Changing sediment input leading to a change in characteristic
slopes and/or drainage area could also be responsible, but the most likely forcing is changing
climate (±X W/m2).
45
1. Introduction.
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
There is consensus that many of the now-dry rivers on Mars were once fed by rain or by
snow/ice melt, but physical models for runoff production vary widely (Malin et al. 2010,
Mangold et al. 2004, Irwin et al. 2005, Matsubara et al. 2013, Moore et al. 2003, Malin & Edgett
2003). Possible environmental scenarios range from intermittent <102 yr-duration volcanic- or
impact-triggered transients to >106 yr duration humid greenhouse climates (Andrews-Hanna &
Lewis 2011, Haberle et al. 2012, Kite et al. 2013, Mischna et al. 2013, Segura et al. 2012, Urata
& Toon 2013, Tian et al. 2010, Johnson et al. 2009, Halevy & Head 2014, Wordsworth et al.
2013, Kite et al. 2014, Ramirez et al. 2014). These diverse possibilities have very different
implications for the duration, patchiness and intermittency of habitable environments on Mars.
As a set, they represent an embarrassment of riches for the Mars research community. Lack of
convergence in our understanding of what allowed rain or snowmelt on Mars is not the result of
a lack of model sophistication; rather, what is currently in short supply are paleo-environmental
proxies (ideally, time series), to constrain the models. Weathering rates provide one proxy (e.g.
Olsen & Rimstidt 2007, Hurowitz & McLennan 2007, Berger et al. 2009, Rimstidt et al. 2012,
Elwood Madden et al. 2009, Bandfield et al. 2011, Ehlmann et al. 2011, Loizeau et al. 2014,
Gainey et al. 2014). In principle, different environmental scenarios for river-forming climates
should also leave distinguishable fingerprints on the fluvial record, through the effect of runoff
intensity, intermittency, process, event number, and event duration on sediment transport (e.g.
Devauchelle et al. 2012). Specifically, paleodischarge is constrained by channel width and
meander wavelength (Hajek & Wolinsky 2012), minimum runoff duration is found by dividing
2
66
67
68
69
70
71
72
73
74
75
76
77
78
river-deposit volume by sediment flux (Jerolmack et al. 2004), intermittency during a wet event
is constrained by dividing the duration of runoff by the duration of sediment accumulation (Kite
et al. 2013, Buhler et al. 2014), and the number of wet events is at least equal to the number of
regionally correlatable fluvial packages. For example, many models only permit runoff for brief
intervals (Toon et al. 2010), so Barnhart et al. (2009) and Hoke et al. (2011) use sedimentary
evidence for prolonged sediment transport to argue forcefully that impact-generated hypotheses
for Early Mars cannot be sustained. However, geomorphology usually provides time-integrated
runoff constraints, while constraining climate change requires time-resolved constraints; at a
given patch of Mars’ surface, we see evidence for only one river-forming episode; and
correlation between those patches has to rely on crater counting (Fassett & Head 2008, Hoke &
Hynek 2009, Grant et al. 2012), which has systematic errors that are incompletely understood
and incompletely corrected for. These problems have limited the application of dry-river
evidence to constrain climate change on ancient Mars.
79
80
81
82
83
84
85
86
87
88
89
90
91
92
Aeolis Dorsa, which is a sedimentary basin 10° E of Gale that contains an exceptionally
high number of exceptionally well-preserved river deposits, gets around these problems. At
Aeolis Dorsa, the river deposits are eroding out of mappable stratigraphic units, and thus provide
time-resolved climate constraints (river valleys provide time-integrated constraints). We can
correlate the river deposits within the basin using crosscutting relationships which lack the
ambiguity of crater counts. Aeolis Dorsa records evidence for multiple episodes of climatedriven surface runoff (Burr et al. 2009; Kite et al. submitted). Initial paleodischarge estimates
have been computed using meander wavelength and channel-width scalings (Burr et al. 2010).
Crater-counts and stratigraphic analyses (Zimbelman & Scheidt 2012) correlate the meander
belts to an interval of ≳1-20 Ma (Kite et al. 2013c), with the age of the older rivers that are
currently exposed in outcrop inferred to date to either the central Hesperian or to the NoachianHesperian transition (thought to have been a habitability optimum; Howard et al. 2005).
Therefore, Aeolis Dorsa contains a stratigraphic record of information about runoff-sustaining
Martian climates.
93
94
95
96
97
98
99
100
101
102
103
Our goal in this study is to exploit the exceptional preservation of river deposits in Aeolis
Dorsa to constraint paleodischarge versus time. The key results are set out in Fig. 5 and Table 1.
§1.1 and §1.2 set out the terrestrial roots of the technique and the geologic context of the
stratigraphic sections. §2 introduces our method (Fig. 1), which is elaborated in the
Supplementary Methods; this method is generally applicable to stratigraphic logging from
stereopairs (not just river-deposit dimensions) and may be particularly useful for the HighResolution Stereo Camera (HRSC) on Mars Express and the Colour and Stereo Surface Imaging
System (CaSSIS) on ExoMars Trace Gas Orbiter. §3.1 reports our dataset, and §3.2 our
paleodischarge interpretation. §3.3 reports a new climate change detection method. §4 discusses
implications for climate change, and §5 discusses considerations relevant for landing-site
selection. We conclude in §6.
3
Measure river deposits eroding
out of wind-eroded rock volume:
Q(zs):
Assign stratigraphic elevations zs:
X
! "#$%$&' () *+*! *,*
X
X
Interpolated
structure
contour
Stratigraphic elevation
! -. () *+*" #,*
X
X
λ, w
X
Figure 1. Graphical abstract of this study. We measure the width (w) and wavelength (λ) of Mars
river-deposits eroding out of the rock, assign stratigraphic elevations zs to each measurement, and
convert these to estimates of paleo-discharge Q(zs). Depth of chute cutoff (in left panel) is 2 m.
104
105
1.1.
Fluvial signatures of climate events on Mars: clearer than Earth.
106
107
108
109
Climate events in Earth history are recorded in fluvial sediments through changes in riverdeposit dimensions, channel-deposit proportions, and fluvial styles (Foreman et al. 2012,
Foreman 2014, Schmitz & Pujalte 2007, Lumsden et al. 2010, Zaleha et al. 2013, Macklin et al.
2012, Amundson et al. 2012, Ward et al. 2000).
110
111
112
113
114
115
Mars lacks plate tectonics and has been tectonically quiescent for >3 Ga, making the
fluvial record of climate change clearer than on Earth where the strong effects of base-level
fluctuations and synfluvial tectonics complicate interpretation of the fluvial record in terms of
climate change (Leeder et al. 1998, Blum & Tormqvist 2000, Macklin et al. 2012, Shanley &
McCabe 1994). The weakness of tectonics on Mars makes the planet a natural laboratory for
studying how climate change leaves a signature in river deposits.
116
117
118
119
120
121
122
Terrestrial experience also directs us to study river deposits on Mars (rather than just
river valleys). Summarizing terrestrial research, Whittaker (2012) states "If topography forms a
non-unique or difficult-to-decode record of past climate […], it is likely that the sedimentary
record, if and where complete, forms the best archive of landscape response to past climate." The
basin-wide spatial scale and ~100m stratigraphic scale of the changes we report in this study
probably puts them beyond the grasp of “stratigraphic shredding” by autogenic threshold
processes (Sadler & Jerolmack, 2014, and references therein).
123
124
125
126
In summary, the terrestrial record points up the usefulness of dry rivers in reconstructing
climate change, especially when the record is preserved as deposits (as at Aeolis Dorsa) and
when it is uncontaminated by large-amplitude externally-driven tectonic uplift (as at Aeolis
Dorsa).
4
a)
b)
Figure 2. a) Our study area (red rectangle) in global context. Background is shaded relief Mars
Orbiter Laser Altimeter (MOLA) topography, illuminated from top right. Colored contours join
locations with the same relative frequency of years with seasonal surface liquid water from a
seasonal melting model (Kite et al. 2013a: average of 88 different orbitally-integrated thinatmosphere simulations). The black line marks the border of the area of recently-resurfaced
terrain, and approximately corresponds to the hemispheric dichotomy. This figure is modified
from Figure 16d in Kite et al. 2013b. b) Zooming in to our study area, showing locations of
transects (black rectangles). Transect 1 consists of two nearby non-contiguous areas. Green line
shows trace of R-1/R-2 contact – the “zero level” for our measurements of zs. Grey contours are
topography from MOLA, at 200m intervals. Purple dashed line outlines an old crater. Cyan
corresponds to large meander belts, which disappear beneath and reappear from underneath
smooth dome-shaped outcrops of R-2. Purple asterisks correspond to locations where the
“marker bed” (ref to Jacobsen is) detected, in purple.
5
127
1.2. Basin-scale mapping distinguishes 102 m-thick river-deposit-hosting units.
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
The large-scale stratigraphy of Aeolis Dorsa shows four unconformity-bounded rock packages,
with the top three probably requiring near-surface liquid water to make rivers, make alluvial
fans, and cement layers (Kite et al., submitted). Erosion on the unconformities suggests long time
gaps, and at one unconformity the density of craters eroding out at the unconformity requires a
>108 yr time gap (Kite et al., submitted). Smooth deflation at the unconformities suggests aeolian
erosion, but the surface liquid water available during rock-package deposition would trap the
clasts needed for wind-induced saltation abrasion and shut down erosion. Therefore, the erosiondeposition alternations suggest wet-dry alternations (or basinwide variations in sediment supply)
at the stratigraphic scale of hundreds of meters to kilometers. Although the total stratigraphic
thickness in Aeolis Dorsa exceeds 3 km (Kite et al., submitted), this study focuses on ~300m of
stratigraphy bracketing the contact (green line in Fig. 2b) between two river-deposit-containing
units that show dramatically different erosional expression. These units are termed R-1 (low
lying) and R-2 (high-standing). In both units, river deposits are outcropping in plan view, quality
similar to terrestrial sparker data (Reijenstein et al. 2011, Hubbard et al. 2011, Wood et al. 2007,
Miall 2002). R-1 eroded to form yardangs and is very rough at km scale, not retaining
recognizable craters; river deposits in R-1 formed large meanders and broad meander belts. R-2
is topographically high-standing, smoothly eroding, associated with numerous aeolian bedforms,
and retains many craters. River deposits in R-2 left narrow channel deposits and formed tight
meanders. Our goal in this paper is to quantify these differences. We log proxies for
paleohydrology versus stratigraphic elevation relative to the geologic contact between R-1 and
R-2.
149
2. Getting from stereopairs to stratigraphic logs.
150
151
152
153
154
155
156
Aeolis Dorsa is littered with the deposits of ancient rivers, which form a stratigraphic record of
paleohydrology. We want to obtain constraints on river discharge Q as a function of stratigraphic
elevation zs, but the rocks contain only proxy data (such as river-deposit width w and wavelength
λ) as a function of 3D spatial position (latitude, longitude, and elevation). Therefore, we need to
(1) convert 3D spatial position to zs, and (2) use a transfer function to convert proxy-data
averages to Q(zs) (Hajek & Wolinsky 2012). (The following is a brief discussion; for more
details, see the Supplementary Materials.)
157
158
159
160
161
162
163
164
165
166
The stratigraphic interval of interest is exposed over an area of Abasin ~ 2 × 104 km2. In terrestrial
fieldwork we would tackle such a large outcrop area by defining transects and measuring a
stratigraphic section along each transect (e.g. Macdonald et al. 2012). The analogous procedure
to defining a transect for Mars orbiter data is to select high-resolution stereopairs (each HiRISE
stereopair covers an area of ~100 km2 << Abasin), each chosen to span a wide stratigraphic range.
Each stereopair is converted to a digital terrain model (DTM) and accompanying orthorectified
images using BAE Systems SOCET SET software. We use 5 stereo DTMs in total, defining
three transects that collectively span the area where R-2 is exposed (Fig. 1; Supplementary Table
1). Three of the DTMs come from an area of exceptionally good preservation and HiRISE image
coverage near 153.6°E 6°S (Howard 2009), which we term Transect 1.
6
167
168
169
170
171
172
173
174
175
For each transect, we first define the contact surface between R-1 and R-2 by interpolating a
surface between points where the large meander-belts characteristic of R-1 were drowned or
blanketed by smoothly-eroding R-2 deposits. A quadratic polynomial interpolation is used;
similar results were obtained using other deterministic interpolation methods. We then define
structure contours by subtracting this interpolated contact surface from the topography. The
layers are nearly flat-lying, so the elevation difference is also the stratigraphic offset. Next, we
pick channel centerlines, channel banks, areas of channel deposits, and areas of evidence for
lateral migration of channels during aggradation (in ArcGIS). Every vertex from these picks is
tagged with a stratigraphic elevation.
176
177
178
179
180
181
182
183
In order to measure w and λ as objectively as possible, we extract these parameters semiautomatically, and pay some attention paid to blinding for the human-in-the-loop part of the
process. The main subjective step is the initial selection of channel-centerlines and bank-pairs for
tracing, and the assignment of a quality score to each (low-quality “candidate” data are excluded
from the fit). The HiRISE DTMs were valuable for this step because they allow us to identify
places where channels appeared to be continuous in plan-view HiRISE images, but had large
stratigraphically offsets. We used terrestrial compilations to determine the range of acceptable
along-channel stratigraphic offsets (Gibling 2006).
184
185
186
187
188
189
190
191
192
193
To get λ, channel centerlines are picked 3 times by hand and loaded into MATLAB, and points
are uniformly interpolated between the vertices of the resulting polylines (Fig. 3) (Howard &
Hemberger 1991). We then convert into coordinates of along-centerline distance s and direction
θ and identify half-meanders using inflection points (zero-crossings of ∂θ/∂s). Sinuosity is
defined as along-centerline distance between inflection points divided by straight-line distance
between inflection points. Half-meanders with sinuosity <1.1 are rejected. For half-meanders
with sinuosity ≥1.1, we log λ as twice the distance between inflection points. To get w, we
interpolate equally-spaced points along each bank, and then find the closest distance from each
point to the opposing bank. w is then defined as the median bank-to-bank distance. Q-λ scalings
are an independent crosscheck on w-derived Q (Irwin et al. 2008, Irwin 2011).
194
195
196
197
198
199
200
201
202
203
204
205
206
207
We quantify error as follows. There are four principal contributors to the overall error: zs error,
measurement error, preservation biases, and sampling error. Stratigraphic error is quantified
using the root-mean-square error on the interpolated contact surface (The scatter of the data
points in z_strat?). Measurement error is quantified (for λ) using the variance of λ extracted
from repeated, independent traces on the same channel centerlines (Fig. 3) and (for w) by using
the variance of bank-to-bank width measured along the channel trace. Preservation bias refers to
the possibility that w measured from orbit may be wider (or narrower) than the original w. In
practice, we found preservation bias to be minor (§3.2-§3.3). Sampling error is quantified by
bootstrapping. Bootstrapping accounts for the fact that we can only gather data for a subset of
river deposits that are exposed on the modern topography, whereas river-deposits are dispersed
within a three-dimensional rock volume. Bootstrapping accounts for this error, assuming no
correlation between modern topography and channel-deposit scale. When applying the bootstrap,
we need to take account of non-independence between measurements, otherwise the error bars
from the bootstrap will be too small. To avoid understating the bootstrapped sample error, we
7
208
209
210
211
therefore sample (or exclude) all of the data points collected from the same channel thread
together. Our final stratigraphic logs take account of zs error, measurement errors, and sampling
errors. Because w and λ are scale parameters (in the statistical sense – implying a flat log prior),
we log-transform all data before calculating errors and averages.
Figure 3. Showing example of wavelength (λ) extraction. Points are uniformly interpolated
along repeated, independent centerline traces (colored disks). Straight-line distances (thin lines)
link inflection points (asterisks) extracted from the centerline traces. Color corresponds to the
modern topographic range (standard deviation of 4m in this case). The low-sinuosity halfmeander (arrowed) is excluded.
212
213
3. Paleohydrology.
214
3.1 River-deposit scale varies systematically upsection.
215
216
λ tightens upwards across the contact (0 m on the y-axes in Figure 4a). Small meanders are
rare/absent below 0 m, and common above 0 m. From visual inspection of the data, the most
8
217
218
219
220
natural break in the data (i.e. the paleohydrologically-inferred contact) is ~20m below the
geologically inferred contact. [Anderson-Darling test gives Xm, footnote and KolmogorovSmirnov test gives XXm] (Buchhave et al. 2014) From 100,000 Monte Carlo trials, in no cases
did (zs > -20 m) exceed (zs < -20 m). This indicates a statistical significance of >4.4σ for the
221
222
detection of (zs < -20 m) >
(zs < -20 m). [Footnote: The corresponding significance for a
break-point at 0m is XXXX]. The trends are consistent between Transect 1 and Transect 2.
9
Figure 4. Width (w) and wavelength (λ) results. The black symbols show the measurements;
crosses for Transect 1, circles for Transect 2, and diamonds for Transect 3. The error bars in w
and λ differ for each measurement and are shown individually. The median error bars in zs are
shown. The stratigraphic RMS error is X m for all Transect 1 data points; Y m for all Transect 2
data points; and Z m for all Transect 3 data points. The red error bars show the mean and
standard deviation of the data in 40m-wide zs bins (40m is approximately ~2x the median
stratigraphic error). The blue line shows a running median (for a 40m range of stratigraphic
elevations), taking into account all errors via bootstrapping. The blue line is dashed to show extra
uncertainty at zs with few data points or correlation problems. The gray lines correspond to the 2σ errors on the running median from the same bootstrap, which are equivalent to the (2.1% 97.9%) bounds from the bootstrap. Supplementary Figure XYZ blows up the section so every
data point can be seen.
223
10
224
225
226
227
228
229
230
w also narrows upwards across the contact (0m on the y-axis in Figure 4b). Narrow channels are
rare/absent below 0m, and common above 0m. Similar to the wavelengths, visual inspection of
the data suggests that the most natural break in the data (i.e. the paleohydrologically-inferred
contact) is ~20m below the geologically inferred contact [Anderson-Darling statistics?]. From
100,000 Monte Carlo trials, in 1 case the median w above -20m exceeded the median w below 20m. This indicates a statistical significance of 4.4σ for the detection of (zs < -20 m) > (zs >
-20 m). [Footnote: The corresponding significance for a break-point at 0m is XXXX].
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
w and λ are poorly constrained for zs > 130 m. There is some evidence for a return to bigger
rivers at higher zs (shown by the dashed line in Figure 4). While there are relatively few data
points here (all from Transect 3), the trend is statistically significant and visually obvious.
However, the problem is the uncertain stratigraphic placement of Transect 3 (Kite et al.,
submitted). We adopt a stratigraphic uncertainty of 50 m in setting points in Figure 4, but in fact
there is a qualitative stratigraphic uncertainty, as follows. The placement that defines the dashed
line in Figure 3 assumes that the topographic offset between the large rivers in Transect 3 and the
large rivers in Transect 1 (Fig. 2b) corresponds to a stratigraphic offset. Similar to the Grand
Staircase, Utah, USA, this placement assumes we see multiple layers of ancient river deposits
exposed by modern erosion (Blakey & Ranney 2008). An alternative possibility is that the riverdeposits measured in Transect 3 are separated from the river-deposits measured in Transects 1
and 2 by a drainage divide. In this case, the topographically-offset Transect 3 deposits could be
stratigraphically equivalent to the big rivers plotted around -50m in Fig. 4. A terrestrial analog
for a large topographic offset between nearby rivers formed at the same time is the modern
Eastern Continental Drainage Divide, Blue Ridge Scarp, North Carolina, USA (Willett et al.,
2014).
a)
b)
Figure 5. Summary of results. a) Crossplot of width (w) & wavelength (λ) (show inferred
11
discharges …). b) auxiliary data (show “poor exposure” again).
247
248
249
250
251
Internal crosschecks show that our measurements of w have comparable reliability to our
measurements of λ (Fig. 5a). This is surprising, because inverted-channels can easily be
erosionally narrowed relative to the original w (and valleys can easily be erosionally widened
relative to the original w), while it is difficult for erosion to significantly modify λ.1
252
253
254
255
256
To support our scale measurements, we map the area of visible channel deposits and of areas of
lateral accretion deposits (e.g., point bars) and report both as an areal fraction of “all terrain” at
that zs (Fig. 5b). Channel-deposit proportion (CDP) falls off sharply at zs ≈ 0 m. This is
unsurprising because the contact is defined using the disappearance of (high-CDP) meander
belts. We do not break down our results by transect, but these are very similar in both transects.
257
258
259
All these metrics are self-consistent, as expected: larger width rivers should make proportionally
larger meander bends (with w:λ ≈ 10, as observed), and also should migrate laterally at faster
rates (all other things equal).
260
3.2. Basin-wide paleo-discharge variability.
261
264
265
266
The trends across the contact in both median meander wavelength ( ) and in median channel
width ( ) that are apparent in both Transect 1 and Transect 2 and which are quantified in Figure
4 are also visually obvious in the HiRISE DTMs for the transects (Fig. S_XYZ). Beyond the area
covered by our HiRISE DTMs, the reduction in
and
across the R-1/R-2 contact is also
visually obvious in CTX and HiRISE throughout the study area. We conclude that the trends
shown in Figure 4 are representative of a basin-wide change in paleo-hydrology.
267
268
269
Motivated by the coherence of our paleohydrologic proxies – between metrics and between
transects – we conclude that R-2 records a distinct episode of runoff from R-1. We now turn to
quantifying the change in discharge across the contact.
262
263
1
In spite of our expectation that preserved w would be a poor paleodischarge proxy, we find
that:- the ensemble : ratio is within terrestrial expectations; there is no trend for inverted
channel w to be narrower than negative-relief w [CHECK]; the standard deviations of w for a
given zs are comparable to those for λ (Fig. 4); for those λ for which a w measurement was also
taken, most have a w:λ ratio within range of terrestrial expectations; and, again for λ for which a
w was also taken, there is no evidence for w:λ ratio changing with zs. With hindsight, this
consistency is due to the fact that measurements were only made of the best-preserved stretches
of channel in each DTM. From that subset of measured channels, those that showed signs of
relatively more severe erosion were rejected as “candidates.” This left a fairly restricted set of
relatively high-quality measurements. If we had naively measured all channel stretches, then w
measurements would probably have been less reliable.
12
270
271
272
273
274
275
276
[This paragraph needs work]. Terrestrial scaling laws show that bank-full discharge, Q, scales
approximately as w2 and as λ2. The physical basis for the quadratic dependence of bank-full Q on
w is that channel depth h tends to increase in proportion with w, whereas water velocity u
increases only sluggishly with w. Since Q = whu, Q ∝ w2 [Mike, please can you check/refine
this?] (Knighton et al. 1998). Theory suggests that initially-straight rivers are most vulnerable to
sinuous instabilities with λ ~10w (bar-bend theory) (Blondeaux & Seminara 1995); as the
meander amplitude grows, the wavelength of the initial instability is frozen-in to the growing
277
meander . Therefore λ ∝ w.
278
279
We adopt the Eaton et al. (2013) scaling for w and the Burr et al. (2010) scaling for λ. [Spell out
what these scalings are.] We report ‘nominal’ bank-full discharges in Table 1.2
280
281
282
283
Mechanistic explanations for the constant of proportionality empirically obtained for these
terrestrial scalings all include a gravity term, so the scaling must be modified for application to
Mars. We follow Williams et al. (2009) and Burr (2010) in reducing the paleo-discharges
calculated for Mars by ~30% from the corresponding Earth paleo-discharges.
(m)
Bankfull Q from
(m)
(m3 s-1)
Bankfull Q from
(m3 s-1)
Interpolated zs
100m > zs > 0m
zs < 0m
Unit assigned from mapping (no interpolation)
R-2
R-1
284
Table 1. Paleodischarge Q versus stratigraphic elevation zs from transects in S. Aeolis Dorsa.
285
[Maybe an accompanying figure in the style of Foreman et al. (2012), Figure 3?]
286
287
These conclusions are supported by comparing the paleodischarges estimated using ‘handassigned’ R-1 vs. R-2 (Table 1).
288
Evidence for paleodischarge variability at finer stratigraphic scales:
2
More sophisticated approaches to estimating paleo-discharge require slope information.
Unfortunately in S. Aeolis Dorsa the present-day slopes of the river deposits are unlikely to be a
safe guide to the slopes when the rivers were flowing (Lefort et al. 2012, Engelder & Pelletier,
2013).
13
289
Figure 6. Show REma1a and REma1b (but from the DTM!)
290
291
292
293
294
295
296
297
298
299
R-1: A common observation within R-1 (Jacobsen & Burr 2013, Cardenas & Mohrig 2014) is
the superposition of a less-sinuous inverted channel without cutoffs over a more sinuous
meander-belt deposit with numerous cutoffs / meander generations (locations include those
shown by purple asterisks in Fig. 2b). Because this pattern is never seen at more than 1
stratigraphic level locally, has a common erosional expression basinwide, joins at confluences,
and has a consistent elevation just below the R-1/R-2 contact, we interpret this as a stratigraphic
marker of a basinwide event. Because the upper is preserved as an inverted channel, and follows
the general trend of the lower belt, the following sequence of events is required to explain these
observations:
300
301

Highly sinuous first-generation channels developed. There was enough time for the firstgeneration channels to develop cutoffs, primarily neck cutoffs.
302
303
304
305
306
307
308
309

Fluvial activity ceased, or was routed away from the basin, or was inadequate to
transport sediment. The area of the channel belt became infilled with ~10m of nowremoved sediment, which infilled the channel but incompletely muted the topography of
the channel belt. Because this sediment is no longer present, we do not know what
processes transported it. However, in Aeolis Dorsa, areas that were once wet (e.g.,
meander belts, channels) are preferentially preserved against modern wind erosion. The
fact that this channel-belt infill has been completely removed by wind erosion is
consistent with the idea that this was dry airfall (e.g., ash or dust).
310

The channel belt infill was re-incised by the less-sinuous valley.
14
311
312

The river shut down again – after a shorter period of fluvial activity than for the firstgeneration channels.
313

Further deposition followed by erosion to present-day topography.
314
Overall, this is consistent with a wet-dry-wet sequence (Metz et al. 2009).
315
316
317
318
319
320
321
322
323
324
In our DTMs the stratigraphic offset between the less-sinuous and more-sinuous “marker
layer” is only ~10m (which is too small to be resolved in our stratigraphic logs). However, we
can make progress by hand-tagging λ and w measurements in our DTMs corresponding to the
two layers (which are visually distinctive). We find λ = (n = ) for channel 1 and λ = (n = ) for
the less sinuous layer. We find w = of (). Bootstrapping shows that these results are not
significantly different. We conclude that there is no evidence in our DTMs that the event that
formed the less-sinuous channels had smaller bank-forming discharges than the event that
formed the more-sinuous underlying meander belt. However, the two wet events were
presumably not the same: the wet episode that made the less sinuous layer was probably not as
long-lasting as the wet episode that was responsible for the broad meander belts.
325
R-2: Evidence for cut-and-fill from channel intersections.
326
//
327
328
329
330
331
332
333
334
(x) Constrain origin of a candidate regional marker bed. We will document the regional extent
of this marker bed; regional coherence would support allocyclic dynamics. We will then use the
regional extent and variability to evaluate multiple working hypotheses for the origin of this
multilevel channel belt. These hypotheses include:- autocyclic dynamics, with regional
correlation maintained by base-level control; allocyclic dynamics, with draping deposits (ash or
distal ejecta) stochastically resetting the system; and allocyclic dynamics with shut down and
restart of fluvial deposition on a 1-10m stratigraphic scale, consistent with orbitally-paced runoff
on Early Mars.
335
3.3 A tool to search for abrupt climate change on Mars.
336
337
338
339
340
341
342
343
The Monte Carlo bootstrapping used to produce the error bars shown in Fig. 4 can also be used
to search for paleohydrologic evidence of stratigraphically abrupt climate change on Mars. We
do this by repeating the Anderson-Darling test for each of the bootstrapped distributions of data.
If abrupt, then breakpoint would be repeatedly found at same stratigraphic elevation (Fig. 7a).
However, if gradual, then … (Fig. 7b). To demonstrate that our dataset is detailed enough to
detect an abrupt climate change in the data if it existed, we inject a sharp break in artificial proxy
(synthetic data but with real errors) to show that we can recover it from the bootstrap. We
recover it in λ and w (Fig. 7c). Finally we use the real data (Fig 7d). We find …
344
345
346
347
We conclude that our stratigraphic logs do not require stratigraphically abrupt climate change
during the era of river-forming climates on Mars. However, it is possible that abrupt climate
change is recorded by fluvial deposits on Mars at finer scales than investigated in this study. This
might be picked out by rover investigation.
15
a)
b)
c)
d)
CONCEPT – ABRUPT
CONCEPT – GRADUAL
APPLICATION – SYNTHETICS
APPLICATION – DATA.
Figure 7. Abrupt climate change detection figure – to show no abrupt climate change detected.
348
……
349
3.4. Timescales and intermittency.
350
351
352
Challenges for the origin and persistence of life on Mars include short timescale and intermittent
availability of surface liquid water. To quantify these challenges [we really need cosmogenics,
but in the meantime can make do with geomorphics].
353
Tau from volume is OK
354
355
356
I enjoyed reading your new paper on the Atacama! The sequence you describe (rivers, overlain
by alluvial fans, overlain by a big pile of salty dust) is remarkably similar to the Aeolis-Zephyria
basin on Mars, a few hundred km E of Gale Crater.
357
tempting to interpret the 2-3m banding in the point bars as annual floods
358
359
360
For this interval Kite et al. report a depositional interval of >(4-20) Myr from the number of
embedded impact craters. This is a lower limit, and does not constrain whether deposition was
steady or pulsed. Sediment transport calculations provide an estimate of the intermittency.
361
Process interpretations combined with the terrestrial data compilation of Miall [2014] …
362
363
364
365
[Nature of infill]: However there are channel belts preserved - could be all channel? or some overbank
366
367
368
whether the entire landscape, not just the obvious channels, is a fluvial deposit, but is possible based on earth
analogy and transport capacity of these rivers (calculate transport capacity). and the many fragmentary channels
suggest this.
369
370
Using fluvial-deposit volumes and the runoff intensity (paleodischarge) for each fluvial unit
(calculated in §2.4.3.2), we apply existing sediment transport models to obtain runoff durations
component? <-- Nonsense! Layers can be traced continuously into the channel fill, therefore the interfluves ARE
the overbankd deposits by definition. e.g. ESP 19104, px coords 6900 31250 (jp2 quicklook nomap). It is unclear
16
371
372
373
374
375
376
377
378
(Kite 2012). Specifically, we use analytic models (reviewed in Kleinhans 2005, Kleinhans 2010),
scaled to Mars gravity (similar approach to Burr et al. 2010), as well as simple 1D diffusive
models of watershed-scale sediment transport (Armstrong et al. 2011), to obtain minimum flow
timescales (Kite 2012). There is a range of discharge – sediment transport relations in the
literature, and we will report the sensitivity of our results to this range as well as to variations in
modal grainsize (0.2 mm vs. 20 mm) and paleoslope (10-5 – 10-2). Intermittency will be
estimated by dividing flow timescales by the total sediment-accumulation timescale obtained
from embedded-crater frequency (§2.4.3.1; Kite et al. 2013c).
379
Assuming that the drainage density Implied total channel length
380
381
- do the calculation for number of channels contained within and compare to total length of channels in hynek
database (which is 781,370 km, from Arc in Global Gis on "Hynek_Valley_Networks", or 781,394 km using
382
383
384
"Hynek_Valleys" shapefile- just say 8x10^5km to be safe) Nominal: Spacing is ~1km near 150.73E 4.3S (if anything,
less). "Core" area is ~7x10^4 km^2 (similar to size/shape of aeolis dorsa in usgs database). If 300m thick with 10m
stratigraphic spacing (reasonable), length is 2.1e6 km, more than double the hynek database).
385
386
- OOM estimate: supposing strat sep is orbital, 100 Kyr, then 3 Myr, in agreement with embedded-crater lower
limit.
387
4. Implications for river-forming paleo-climates on Mars.
388
We also need error bars to avoid overfitting the climate models to the data.
389
390
4.1. Both short-term intermittency and long-term
intermittency are required.
391
392
4.2. Mechanisms for paleodischarge variability.
393
394
Bullet-pointed list of mechanisms discussing the consistency of each with the
observations.
395
[Follow “Rivers” notebook error discussion.
396
397
-
Increase in slope
398
-
Decrease in drainage area due to “red noise”.
399
-
Shift in drainage divide.]
400
401
402
Using drainage integration as a proxy has two caveats. First, the Opportunity rover has
documented stream deposits that are invisible from orbit. Using orbital imagery might cause us
to miss the deposits of small streams, but not rivers. Second, the record is incomplete and biased
17
403
404
towards the waning stages of wet intervals. Peak runoff (or minimum sediment supply) might
correspond to fluvial-incision unconformities, not deposits.
405
Walther’s Law?
406
407
408
409
410
411
412
The variability in river discharge versus time isn’t enough to exclude orbital forcing. There are
multiple threshholds in the chain of processes linking insolation to carving a river – for example
melt initiation at 273K, overland flow only when melt exceeds infiltration capacity, threshold to
initiate sediment transport (t*c of 0.03-0.045). Because threshholds can nonlinearly ampligy
input perturbations (rectification: Huybers & Wunsch, ), small differences in the peak of
successive cycles can lead to large differences in the amplitude and duration of sediment
transport.
Figure 8. “Mike-Lamb-type” phase diagram. Maybe amplitude of climate change vs. amplitude
of sediment input change? (timescales, amplitudes?)
413
414
- apparently large number of confluences in F2 may mean "small drainage area, arborescent" or it may mean
"rapid switches in drainage direction in an aggrading system."
415
Flexure?
416
417
418
It has been noted that climate cycles have weak effect (Armitage et al. 2014) and that
distinguishing sedimentary imprint of rare vs common events may be extremely difficult
(Sambrook Smith et al. 2010).
419
4.3. Global context.
420
421
422
Other investigations constrain the total duration of wet conditions as a function of wet-event
magnitude (e.g. Hoke et al. 2011), or the duration of individual, late-stage wet events from
sedimentary deposits (Armitage et al. 2011).
423
424
5. Habitability, taphonomy, and organic matter preservation.
18
425
426
427
428
429
430
431
432
433
[Erosion rates and potential for organic-matter preservation:] We conclude that the potential
for organic matter to be buried is low compared to long-lived paleolake sites. Therefore the
HiRISE transects investigated in detail for this study are a poor choice for Mars 2020 landing
site, given the focus of Mars 2020 mission objectives on organic matter and the existence of
more attractive alternatives. However the rivers suggest a habitable environment, the relatively
high deposition rates favor preservation against early degradation, and the relatively high erosion
rate favors preservation against degradation during exhumation. Therefore the Aeolis rivers
might make a good landing site for a mission focused on habitability and unable to land at a
long-lived paleolake site.
434
435
436
437
438
439
440
441
442
443
There does seem to be an area of low crater density near 153.6E, 5.6S. Could be this was covered by Y2 more
recently, or just that F1 here is retreating downward faster. This is the area of exceptionally good preservation.
Normally it is dangerous to pick on an area of low crater density and calc retreat rates , because o fthe lookelsewhere effect. However in this case wehad previously identified area as exceptional preservation, so ok. ~5000
km^2 with 1 crater at d=1 → N~200. About MA/LA boundary, about 300-400 Ma.Could be withdrawal of a blanket
~300-400 Ma, or in steady state 1km crater is 100m deep, so wind erosion rate is 0.3-0.4 microns year, not that
much really. (but strength controlled so ~1 micron year). Organic preservation implications? Pleasing internal
consistency in that only other area with obviously low N(1), 150.5E, 2.4S, looks very eroded - big holes chewed in
it. Can back out erosion rates for other sub-terrane,s but they would be very slow.
444
//
445
446
447
448
449
(ix) Oxbow lake sedimentation predictions from nature of cutoffs. We will document the diversion
(opening) angles for all cutoffs, and compare with interpreted cutoff mechanism (chute or neck cutoff).
Opening angles regulate sedimentation and thus the potential for preservation of organic matter in oxbow
lakes (Constantine et al. 2010), which is higher for floodplain deposits than for aeolian, alluvial-fan, or
fluvial-channel deposits (Summons et al. 2011).
450
451
//
452
Ch 7 in taphonomy 2nd edition (Allison & Bottjer)
453
454
“The relationship between continental landscape evolution and the plant-fossil record: Long term hydrologic controls
455
//
456
Integrate with Pavlov radiation damage models (and Allen et al. …)
on preservation Robert A. Gastaldo & Timothy M. Demko”
457
458
6. Conclusions.
459
Show Data and Models.
460
Workflow can be used for CASSIS.
19
461
462
Acknowledgements.
463
464
465
466
467
468
469
470
471
472
473
474
To the extent we understand Aeolis Dorsa, it is thanks to conversations with the following
scientists, whose diverse analytic approaches and skills we have relied heavily on in thinking
through this problem: Frederik Simons, Alexandra LeFort, Laura Kerber, Misha Kreslavsky, Tim
Michaels, Caleb Fassett, Noah Finnegan, Lynn Carter, Nicolas Mangold, Susan Conway,
Sanjeev Gupta, Matt Balme, Sam Harrison, Maarten Kleinhans, Leif Karlstrom, Michael Manga,
Bill Dietrich, Ross Irwin, Jeff Moore, Christian Braudrick, Devon Burr, Robert Jacobsen,
Bethany Ehlmann, Noah Finnegan, Jonathan Stock, Ross Irwin, Gary Kocurek, Richard
Heermance, Paul E. Olsen, Katie Stack, Roman DiBiase, Ajay Limaye, Joel Scheingross,
Woody Fischer, Or Bialik, and Brian Hynek. E.S.K. was supported by a Harry Hess fellowship.
We thank the HiRISE team for maintainin the HiWish program, which supplied multiple images
essential for this study. We thank Ross Beyer, Sarah Mattson, Audrie Fennema, Annie
Howington-Kraus, and especially Cynthia Phillips for help with DTM generation.
475
476
References.
477
478
479
480
Malin, Michael C.; Edgett, Kenneth S.; Cantor, Bruce A.; Caplinger, Michael A.; Danielson, G.
Edward; Jensen, Elsa H.; Ravine, Michael A.; Sandoval, Jennifer L.; Supulver, Kimberley D,
(2010), An overview of the 1985-2006 Mars Orbiter Camera science investigation, International
Journal of Mars Science and Exploration (Mars Journal), 4, 1-60.
481
482
Blakey, R.C., & W. Ranney, 2008, Ancient landscapes of the Colorado Plateau, Grand Canyon
Association.
483
484
Hoke, Monica R. T.; Hynek, Brian M.; Tucker, Gregory E. (2011), Formation timescales of
large Martian valley networks, Earth and Planetary Science Letters, Volume 312, Issue 1, p. 1-12
485
486
487
Matsubara, Yo; Howard, Alan D.; Gochenour, J. Parker (2013), Hydrology of early Mars: Valley
network incision, Journal of Geophysical Research: Planets, Volume 118, Issue 6, pp. 13651387.
488
489
490
Mangold, Nicolas; Quantin, Cathy; Ansan, Véronique; Delacourt, Christophe; Allemand, Pascal,
(2004), Evidence for Precipitation on Mars from Dendritic Valleys in the Valles Marineris Area,
Science, 305, 78-81.
491
492
493
Peter B. Buhler, Caleb I. Fassett, James W. Head III, Michael P. Lamb (2014), Timescales of
fluvial activity and intermittency in Milna Crater, Mars, Icarus, Volume 241, October 2014,
Pages 130–147.
20
494
495
496
Lesli J. Wood, Quantitative Seismic Geomorphology of Pliocene and Miocene Fluvial Systems
in the Northern Gulf of Mexico, U.S.A., Journal of Sedimentary Research September 2007
77:713-730; published online September 26, 2007, doi:10.2110/jsr.2007.068
497
498
499
500
501
Geoffray Musial, , , Jean-Yves Reynaudb, Murray K. Gingrasc, Hugues Fénièsd, Richard
Labourdettee, Olivier Parizef (2012) Sedimentary Geology Volume 279, 20 November 2012,
Pages 156–172 Modern and Ancient Tidal Depositional Systems: Perspectives, Models and
Signatures , Subsurface and outcrop characterization of large tidally influenced point bars of the
Cretaceous McMurray Formation (Alberta, Canada), DOI: 10.1016/j.sedgeo.2011.04.020
502
503
P. Blondeaux and G. Seminara (1985). A unified bar–bend theory of river meanders. Journal of
Fluid Mechanics, 157, pp 449-470 doi:10.1017/S0022112085002440
504
505
Irwin, Rossman P., III; Craddock, Robert A.; Howard, Alan D. (2005), Interior channels in
Martian valley networks: Discharge and runoff production, Geology, 33, 489-493.
506
507
508
Cardenas, B. T.; Mohrig, D (2014), Evidence for Shoreline-Controlled Changes in Baselevel
from Fluvial Deposits at Aeolis Dorsa, Mars, 45th Lunar and Planetary Science Conference, held
17-21 March, 2014 at The Woodlands, Texas. LPI Contribution No. 1777, p.1632
509
510
511
512
Jacobsen, R. E.; Burr, D. M. (2013), Local-Scale Stratigraphy of Inverted Fluvial Features in
Aeolis Dorsa, Western Medusae Fossae Formation, Mars, 44th Lunar and Planetary Science
Conference, held March 18-22, 2013 in The Woodlands, Texas. LPI Contribution No. 1719,
p.2165
513
514
Ehlmann, B.L. et al., 2011. Subsurface water and clay mineral formation during the early history
of Mars. Nature 479, 53–60
515
516
O. Devauchelle, A. P. Petroff, H. F. Seybold, and D. H. Rothman (2012), Ramification of stream
networks, PNAS, 109, 51, 2012
517
518
519
Barnhart, Charles J.; Howard, Alan D.; Moore, Jeffrey M. (2009), Long-term precipitation and
late-stage valley network formation: Landform simulations of Parana Basin, Mars, Journal of
Geophysical Research, Volume 114, Issue E1, CiteID E01003.
520
521
522
Moore, Jeffrey M.; Howard, Alan D.; Dietrich, William E.; Schenk, Paul M. (2003), Martian
Layered Fluvial Deposits: Implications for Noachian Climate Scenarios, Geophysical Research
Letters, Volume 30, Issue 24, pp. PLA 6-1, CiteID 2292, DOI 10.1029/2003GL019002.
523
524
Bandfield, Joshua L.; Deanne Rogers, A.; Edwards, Christopher S. (2011), The role of aqueous
alteration in the formation of martian soils, Icarus, Volume 211, Issue 1, p. 157-171.
525
526
Gainey, S. R.; Hausrath, E. M.; Hurowitz, J. A.; Milliken, R. E. (2014), Nontronite dissolution
rates and implications for Mars, Geochimica et Cosmochimica Acta, Volume 126, p. 192-211
527
528
Macdonald, F.A., et al., 2012. Early Neoproterozoic basin formation in the Yukon. Geoscience
Canada: Paul Hoffman Series, 39, pp.77-99.
21
529
530
Berger, G. et al., 2009. Evidence in favor of small amounts of ephemeral and transient water
during alteration at Meridiani Planum, Mars. Am. Mineral. 94, 1279–1282.
531
532
Elwood Madden, M., Madden, A., Rimstidt, J., 2009. How long was Meridiani Planum wet?
Applying a jarosite stopwatch to constrain the duration of diagenesis. Geology 37, 635.
533
534
Haberle, R.M., et al. (2012), LPSC 43, 1665. A Cloud Greenhouse Effect on Mars: Significant
Climate Change in the Recent Past?
535
Kite, Edwin S.; Lucas, Antoine; Fassett, Caleb I. (2013), Pacing early Mars river activity:
536
537
Embedded craters in the Aeolis Dorsa region imply river activity spanned ≳(1-20) Myr, Icarus,
Volume 225, Issue 1, p. 850-855
538
539
Hurowitz, J.A., McLennan, S.M., 2007. A 3.5 Ga record of water-limited, acidic weathering
conditions on Mars. Earth Planet. Sci. Lett. 260, 432–443.
540
541
542
Johnson, Sarah Stewart; Pavlov, Alexander A.; Mischna, Michael A. (2009), JGR Planets. Fate
of SO2 in the ancient Martian atmosphere: Implications for transient greenhouse warming,
Journal of Geophysical Research, Volume 114, Issue E11, CiteID E11011.
543
544
545
Williams, Rebecca M. E.; Irwin, Rossman P., III; Zimbelman, James R. (2009), Evaluation of
paleohydrologic models for terrestrial inverted channels: Implications for application to martian
sinuous ridges, Geomorphology, Volume 107, Issue 3, p. 300-315
546
547
548
549
Halevy, Itay, and J.W. Head III, Climatic and chemical consequences of episodic eruptions on
Early Mars, in The Fifth International Workshop on the Mars Atmosphere: Modelling and
Observation, held on January 13-16 2014, in Oxford, U.K. Edited by F. Forget and M. Millour,
id.4204.
550
551
552
553
Paper No. 19-9
, Presentation Time: 4:15 PM-4:30 PM, ANCESTRAL MISSISSIPPI
RIVER: DISCHARGE EVIDENCE FROM THE UPLAND COMPLEX, LUMSDEN, David
N.1, COX, Randel Tom1, VANARSDALE, Roy2, and PARKS, Alan3, North-Central Section
(44th Annual) and South-Central Section (44th Annual) Joint Meeting (11–13 April 2010)
554
555
556
Session No. 19, Advances in the Chronology, Correlation, and Stratigraphy of Pre-Wisconsinan
Glacigenic Sediments, Geological Society of America Abstracts with Programs, Vol. 42, No. 2,
p. 69.
557
558
559
Loizeau, Damien; Carter, John; Mangold, Nicolas; Poulet, François; Rossi, Angelo; Allemand,
Pascal; Quantin, Cathy; Bibring, Jean-Pierre (2014), Widespread Surface Weathering on Early
Mars, EGU General Assembly 2014, held 27 April - 2 May, 2014 in Vienna, Austria, id.4651.
560
561
Olsen, A., Rimstidt, J., 2007. Using a mineral lifetime diagram to evaluate the persistence of
olivine on Mars. American Mineralogist 92 (4), 598–602.
562
563
Rimstidt, J. Donald; Brantley, Susan L.; Olsen, Amanda A. (2012), Systematic review of
forsterite dissolution rate data, Geochimica et Cosmochimica Acta, Volume 99, p. 159-178.
22
564
565
Tian, F., et al. (2010), Photochemical and climate consequences of sulfur outgassing on early
Mars, Earth and Planetary Science Letters, Volume 295, Issue 3-4, p. 412-418
566
567
Howard, A. D., (2009), How to make a meandering river, Proceedings of the National Academy
of Sciences, vol. 106, issue 41, pp. 17245-17246, 10/2009
568
569
Howard, Alan D.; Hemberger, Allen T. (1991), Multivariate characterization of meandering,
Geomorphology, Volume 4, Issue 3, p. 161-186
570
571
572
573
Mischna, M. et al. (2013), JGR 118, 560. Effects of obliquity and water vapor/trace gas
greenhouses in the early martian climate, Journal of Geophysical Research: Planets, Volume
118, Issue 3, pp. 560-576
574
575
576
Toon, O.B., et al. (2010), AREPS 38, 303, The Formation of Martian River Valleys by Impacts
Annual Review of Earth and Planetary Sciences, vol. 38, p.303-322 (Annual Reviews
Homepage), DOI:10.1146/annurev-earth-040809-152354
577
578
579
Urata, Richard A.; Toon, Owen B (2013), Simulations of the martian hydrologic cycle with a
general circulation model: Implications for the ancient martian climate, Icarus, Volume 226,
Issue 1, p. 229-250.
580
581
Wordsworth et al. (2013), Icarus, 222, 1. Global modelling of the early martian climate under a
denser CO2 atmosphere: Water cycle and ice evolution, Icarus, Volume 222, Issue 1, p. 1-19.
582
583
584
Zaleha et al. Cret. Res. 2013 Paleochannel hydraulics, geometries, and associated alluvial
architecture of Early Cretaceous rivers, Sevier Foreland Basin, Wyoming, USA. Cretaceous
Research, Volume 45, October 2013, Pages 321–341, DOI: 10.1016/j.cretres.2013.06.003
585
586
Segura, T.L., McKay, C.P., and Toon, O.B., 2012, An impact-induced, stable, runaway climate
on Mars, Icarus, Volume 220, Issue 1, p. 144-148.
587
588
589
Lefort, A., Burr, D. M., Beyer, R. A., Howard, A. D. 2012. Inverted fluvial features in the
Aeolis-Zephyria Plana, western Medusae Fossae Formation, Mars: Evidence for post-formation
modification. Journal of Geophysical Research (Planets) 117, 3007.
590
591
592
Andrews-Hanna, J.C., & K.W. Lewis 2011, Early Mars hydrology: 2. Hydrological evolution in
the Noachian and Hesperian epochs, Journal of Geophysical Research, Volume 116, Issue E2,
CiteID E02007.
593
594
595
Ramirez, Ramses M.; Kopparapu, Ravi; Zugger, Michael E.; Robinson, Tyler D.; Freedman,
Richard; Kasting, James F., Warming early Mars with CO2 and H2, Nature Geoscience, Volume
7, Issue 1, pp. 59-63
596
597
598
Kite, Edwin S.; Williams, Jean-Pierre; Lucas, Antoine; Aharonson, Oded, 2014, Low
palaeopressure of the martian atmosphere estimated from the size distribution of ancient craters,
Nature Geoscience 7(5), 335-339.
23
599
600
Eaton, B.C. (2013), Chapter 9.18: Hydraulic Geometry, in Wohl, E.E. (ed.), Treatise on
Geomorphology, vol. 9, Fluvial Geomorphology, Elsevier, Oxford, UK.
601
602
Fassett & Head 2008, The timing of martian valley network activity: Constraints from buffered
crater counting, Fassett, Caleb I.; Head, James W., Icarus, Volume 195, Issue 1, p. 61-89.
603
604
605
Hoke & Hynek 2009, Roaming zones of precipitation on ancient Mars as recorded in valley
networks, Hoke, Monica R. T.; Hynek, Brian M., Journal of Geophysical Research, Volume
114, Issue E8, CiteID E08002
606
607
608
Grant et al. 2012, The timing of alluvial activity in Gale crater, Mars, Grant, John A.; Wilson,
Sharon A.; Mangold, Nicolas; Calef, Fred; Grotzinger, John P., Geophysical Research Letters,
Volume 41, Issue 4, pp. 1142-1149
609
610
Blum, M. D. and Törnqvist, T. E. (2000), Fluvial responses to climate and sea-level change: a
review and look forward. Sedimentology, 47: 2–48. doi: 10.1046/j.1365-3091.2000.00008.x
611
612
613
Leeder, M. R., Harris, T. and Kirkby, M. J. (1998), Sediment supply and climate change:
implications for basin stratigraphy. Basin Research, 10: 7–18. doi: 10.1046/j.13652117.1998.00054.x
614
615
Willett, S.D., et al., 2014, Dynamic Reorganization of River Basins, Science 343, doi:
10.1126/science.1248765.
616
617
//
618
619
Good: Invited review Tributary, distributary and other fluvial patterns: What really represents the
norm in
620
the continental rock record? Christopher R. Fielding
621
622
//
623
Nice McMcMurray formation seismic e.g.
624
625
http://www.worldoil.com/October-2008-Data-integration-improves-oil-sandscharacterization.html
626
627
628
Seismic geomorphology and sedimentology of a tidally influenced river deposit, Lower
Cretaceous Athabasca oil sands, Alberta, Canada Stephen M. Hubbard, Derald G. Smith, Haley
Nielsen, Dale A. Leckie, Milovan Fustic, Ronald J. Spencer, and Lorraine Bloom: - aapg bulletin
629
//
630
631
632
Sulfate-Rich Eolian and Wet Interdune Deposits, Erebus Crater, Meridiani Planum, Mars, Metz,
J. M.; Grotzinger, J. P.; Rubin, D. M.; Lewis, K. W.; Squyres, S. W.; Bell, J. F., Journal of
Sedimentary Research, vol. 79, issue 5, pp. 247-264\
24
633
634
Anatomy of an avulsion NORMAN D. SMITH*, TIMOTHY A. CROSS?, JOSEPH P.
DUFFICY* and STEPHEN R. CLOUGH* --> some styles of splay can have no sand
635
Figure 14 of "anatomy of an avulsion" - sudden abandonment could leave
636
more sand on the banks than in the channel
637
figure 17 is interesting
638
639
640
641
642
Seismic geomorphology and high-resolution seismic stratigraphy of inner-shelf fluvial, estuarine,
deltaic, and marine sequences, Gulf of Thailand Hernán M. Reijenstein, Henry W. Posamentier,
and Janok P. Bhattacharya - shows incised tributuary valleys feeding an otherwise normallooking meander belt
643
644
645
"Architecture of vertically stacked fluvial deposits, Atane Formation, Cretaceous, Nuussuaq,
central West Greenland MARIA A. JENSEN" sed 'ology 2010 has ideas - basically, sitting
within incised valley
646
647
648
Quantitative Seismic Geomorphology of Fluvial Systems of the Pleistocene to Recent in the
Malay Basin, Southeast Asia* Faisal Alqahtani1 , Christopher Jackson1 , Howard Johnson1 ,
Alex Davis2 , and Ammar Ahmad1
649
650
the Malay Basin, Southeast Asia* Faisal Alqahtani1 , Christopher Jackson1 , Howard Johnson1 ,
Alex Davis2 , and Ammar Ahmad1
651
652
Miall, A.D., 2002, Architecture and sequence stratigraphy of pleistocene fluvial systems in the
Malay Basin, based on seismic time-slice analysis: AAPG Bulletin, v. 88/7, p. 1201-1216
653
654
FLUVIAL RESERVOIR ARCHITECTURE FROM NEAR-SURFACE 3D SEISMIC DATA,
BLOCK B8/32, GULF OF THAILAND by Hathaiporn Samorn
655
656
Lindgren, W., 1911, The Tertiary gravels of the Sierra Nevada: U.S. Geological Survey
Professional Paper 73,
657
658
659
M. G. Macklin, J. Lewin, and J. C. Woodward, Research article: The fluvial record of climate
change, Phil. Trans. R. Soc. A May 13, 2012 370 1966 2143-2172; doi:10.1098/rsta.2011.0608
1471-2962
660
661
Alexander C. Whittaker, How do landscapes record tectonics and climate?, Lithosphere, April
2012, v. 4, p. 160-164, doi:10.1130/RF.L003.1
662
663
664
665
Todd M. Engelder and Jon D. Pelletier (2013), Autogenic cycles of channelized fluvial and
sheet flow and their potential role in driving long-runout gravel progradation in sedimentary
basins, Lithosphere, August 2013, v. 5, p. 343-354, first published on May 10, 2013,
doi:10.1130/L274.1
25
666
667
668
J.J. Armitage, R.A. Duller, and S.M. Schmalholz (2014), The influence of long-wavelength
tilting and climatic change on sediment accumulation, Lithosphere, L343.1, first published on
June 17, 2014, doi:10.1130/L343.1
669
670
671
672
673
Sadler, P.M., & D.J. Jerolmack (2014), Scaling laws for aggradation, denudation and
progradation rates: the case for time-scale invariance at sediment sources and sinks, in Smith, D.
G., Bailey, R. J., Burgess, P. M. & Fraser, A. J. (eds) Strata and Time: Probing the Gaps in Our
Understanding.
Geological
Society,
London,
Special
Publications,
404,
http://dx.doi.org/10.1144/SP404.7
674
675
676
677
Miall, A.D. (2014), Updating uniformitarianism: stratigraphy as just a set of ‘frozen accidents,’
in Smith, D. G., Bailey, R. J., Burgess, P. M. & Fraser, A. J. (eds) Strata and Time: Probing the
Gaps in Our Understanding. Geological Society, London, Special Publications, 404,
http://dx.doi.org/10.1144/SP404.4
678
679
Schmitz, B., and Pujalte, V. (2007), Abrupt increase in seasonal extreme precipitation at the
Paleocene-Eocene boundary, Geology, March, 2007, v. 35, p. 215-218, doi:10.1130/G23261A.1
680
681
682
683
Gregory H. Sambrook Smith, James L. Best, Philip J. Ashworth, Stuart N. Lane, Natalie O.
Parker, Ian A. Lunt, Robert E. Thomas, and Christopher J. Simpson, Can we distinguish flood
frequency and magnitude in the sedimentological record of rivers?, Geology, July 2010, v. 38, p.
579-582, doi:10.1130/G30861.1
684
Supplementary Materials.
685
Supplementary Methods.
686
687
The purpose of these Supplementary Methods is to describe our methods in sufficient detail that
they can be applied to other orbiter-data stratigraphic-logging problems elsewhere on Mars.
688
M1. Stratigraphic-elevation assignments.
689
-
Describe protocol in detail.
690
-
Discuss DTM errors.
691
Stereopair
DTM resolution (m/pixel)
ESP_017548_1740/
Transect
1
ESP_019104_1740
ESP_019038_1740/
ESP_019882_1740
26
PSP_006683_1740/
1
PSP_010322_1740
PSP_007474_1745/
2.5 2
ESP_024497_1745
PSP_002002_1735/
3
PSP_002279_1735
1 (only used for contact
interpolation)
692
693
694
SUPPLEMENTARY Table 1. The HiRISE and CTX DTMs used in this study. All DTMs were
produced by ASL, except for PSP_006683_1740/PSP_010322_1740 which was produced by
ESK.
695
696
697
698
699
700
Not superimposed on rock – cite kite et al. submitted and add:- River taphonomy: A common
pattern seems to be plug exposed at higher elevations, then double ridges, then negative relief
only (widening channel) at lower elevations. This makes sense in terms of wind erosion
unzipping the topography. This is consistent with the channels being near-horizontal and
embedded in stratigraphy. (the erosion pattern for many of the rivers in F2 is (with decreasing background
701
702
703
704
705
How to interpret this pattern? (i) backfill of nonerosive material; early floods flush the lower part of the channel
(dubious because some of these channels are long, so why are we erosionally unzipping just the transition from
empty to backfilled), also does the gradient change much? (ii) Farley et al. Science 2014 style wavelike erosional
unzipping - just happens to be different for river channel fill than background material (begs the question, but
why?) (iii) More resistant “crunchy top” to channel fill (but why?)
706
M2. Paleohydrologic-proxy measurements.
topographic elevation) - first exposed in +ve relief- then exposed as double ridges- finally, exposed in -ve relief.
Supplementary Figure S_ABC showing locations of all measurements. (use colors
for widths and wavelengths and UnitAssign)
707
708
709
710
711
Boxes were defined such that each pixel on the screen corresponded to 1.5 pixels on the
orthorectified image. Screen is 1200 px x 1920 px. Therefore set to 1920x5 meters-per-pixel x1.5
= 14.4 km wide for the sweep. --> 1:29000 on ORTELIUS is a pretty good approximation to this
(though subjectively it *feels* a bit zoomed-out).
712
713
714
715
Channels were picked one box at a time. For the three independent channel-centerline picks, the
view was rotated by 90 degrees between each pick. The stretches on both the DTM and the ortho
were kept at 2 standard deviations for almost all cases. The following attributes were assigned
for every width entry:
27
716
717
718
719
720
721
Size of Area-Searched boxes, stretches applied to both the DTM and the ortho, what is done for
channels near the edges of the DTM (CTX inspection), e.t.c. e.t.c. (in an Appendix). Sometimes
the picks are made on the complementary image, leading to an offset (complementary image is
not orthorectified, it is georeferenced, to save processing time). very narrow (e.g., 3m) doubleridge features tended to be straight and hard to distinguish from fracture fill, so they were not
counted, but they are still tabulated and in the supporting data.
Fig. S_DEF showing Pstyle sketches.
722
723
724
725
726
727
On outer bank of valley bend, typically see a steep cut bank topped by a crisp break-in-slope to
the plateau. On inner bank, typically more gentle. Example, channel on the extreme W edge of
the ESP_017548_1740 DTM, ~800m S of the 200m-diameter infilled crater ("classic" example
used in the Icarus sedimentation rate paper). This is consistent with outward migration of a
channel as it cuts down into the substrate.
728
729
730
For pstyle = 4 and 8, widths are basically lower limits (with exceptions), for pstyle=1 and most
of 2, widths are basically upper limits (left censored and right-censored data, need to do
appropriate statistics …).
731
732
Error bars for points combined in this way are also combined assuming log-Gaussian distribution
of underlying errors.
733
734
735
736
737
Complete blinding is not possible because the human operator can generally tell (from the
erosional expression of the river deposits) what part of the stratigraphic column they are working
in. However, because almost all measurements were done at a zoom level where the margins of
the ‘boxes’ were not visible, the human operator is usually unaware of the absolute scale of the
features being picked.
738
M3. Data reduction.
Figure SXYZ: Blow-up of the main data figures to make sure that every figure can be seen.
739
740
741
742
743
744
745
746
Our data reduction is carried out in MATLAB. First, the paleohydrologic-proxy line picks (bank
pairs and channel centerlines) are converted to widths and wavelengths using
parseMeanderWavelengthMeasurementsMarsRivers and
parseWidthMeasurementsMarsRivers. Similarly, area picks are converted to fractional
areas versus stratigraphic elevation using parseAuxiliaryDataMarsRivers. The output
from these files is passed to aggregateTransects, which calls
bootstrapRiverDepositMeasurementsMarsRivers in order to generate error bars.
Fig. S_GHI. Showing overall script flow.
747
28
748
749
parseMeanderWavelengthMeasurementsMarsRivers.m : -
750
Fig. S_JKL. [Summarize script with boxes and arrows]
751
752
753
754
755
756
757
758
759
760
761
762
763
The workflow for this script is heavily influenced by Hemberger (1986) and Howard &
Hemberger (1991). Channel-centerline picks are ingested from ArcGIS-exported shapefiles.
Candidate (low) quality data are then deleted, and points are interpolated uniformly along the
channel-centerline polyline traces. The distance between interpolated points is 10 m. The
polylines are then converted to coordinates of { s, θ }, where s is distance along the channel and
θ is the direction of. Curvature ∂2θ / ∂2s is then evaluated at each interpolated point on the
channel, and is smoothed using a 10-point moving baseline. Every point where the curvature
changes sign is tagged as a candidate inflection point. Trial half-meanders are defined by joining
the candidate inflection points on a polyline with straight lines. The sinuosity for each trial halfmeander. The minimum sinuousity for acceptance is 1.1. If any candidate inflection points are
the end-points for two half-meanders below the minimum sinuousity threshold, then that
candidate inflection point is removed and the trial half-meanders are re-calculated.
764
765
766
767
768
769
A real half-meander should be identifiable in repeated picks of the same channel. Therefore, we
compare the three picks of each meandering channel to seek half-meanders that are reproduced
on multiple picks. To do this, we require that the median distance between points dotted
uniformly along the straight lines defining half-meanders be less than 0.25 of the straight-line
length of the half-meanders. The error assigned to the replicable half-meander is then the
standard deviation of the straight-line lengths of the reproduced half-meanders on each trace.
770
771
772
773
774
775
Finally, some centerline traces contain multiple replicable half-meanders. Because our centerline
traces are fairly short, meander wavelengths on the same centerline trace are not independent. (If
we artificially altered the wavelength of an upstream-most half-meander on one of our centerline
traces, then the flow pattern for the downstream half-meanders would be greatly affected).
Therefore, we combine the (log-)mean and (log) standard deviation of half-meanders on the
same centerline trace assuming underlying Gaussianity.
776
parseWidthMeasurementsMarsRivers.m : -
777
[Illustrate with boxes and arrows.]
778
Bank-pair picks are ingested from ArcGIS-exported shapefiles.
779
780
M4. Error analysis.
29
781
782
783
784
Near-surface hydraulic conductivity is assumed not to change with time (not unreasonable in an
aggrading system). Finally, river deposits preferentially record the waning stages of wet intervals, and
peak runoff might correspond to fluvial-incision unconformities (which we will document separately).
785
//
786
787
788
- In detail, I don't know if F2 channels were cut at 2% plus slopes into rugged paleotopography that was then
deformed to make F2 channels oblique to contours. (I do know that F2 channels occurred after F1 channels.) This
means that assigning proxies to strat levels *within* F2 is possibly dodgy.
789
790
very narrow (e.g., 3m) double-ridge features tended to be straight and hard to distinguish from fracture fill, so they
were not counted, but they are still tabulated and in the supporting data.
791
792
- For pstyle = 4 and 8, widths are basically lower limits (with exceptions), for pstyle=1 and most of 2, widths are
basically upper limits (left censored and right-censored data, need to do appropriate statistics …).
793
794
795
- River taphonomy: A common pattern seems to be plug exposed at higher elevations, then double ridges, then
negative relief only (widening channel) at lower elevations. This makes sense in terms of wind erosion unzipping
the topography. This is consistent with the channels being near-horizontal and embedded in stratigraphy.
796
797
- Absence of topographic barriers in the reconstructed synfluvial paleolandscape supports the idea that F1/F2
extends hundreds of km E and W along the dichotomy boundary.
798
//
799
nice McMcMurray formation seismic e.g.
800
http://www.worldoil.com/October-2008-Data-integration-improves-oil-sands-characterization.html
801
//
802
Leif points out: gooseneck cutoff versus chute cutoff - note really
803
understood how chute cutoffs form, presumably a curvature effect - chute
804
cutoffs not bedrock! And there is evidence for chute cutoffs in the
805
aeolis meanders: 3777 14475 on NOMAP 22862, for example? (Or is this
806
just a '2nd generation' channel?)
807
//
808
P08_004270_1746_XI_05S207W - meandering channels emerging from beneath
809
?subliming mantle - mantle is cut by 'headwaters' high drainage density
810
channels - to S of ridge, ?sublimation textures
811
//
812
stunning/beautiful examples in B19_016981_1746_XI_05S205W, a litte to
813
the north of the 'eastern trend'
30
814
//
815
email to Woody Fischer and Katie Stack of 16 March 2012
816
//
817
150.63, -4.64 - inside of fan?
818
//
819
on hirise 154.576, -5.322, incised channel is overlain by the point bar
820
deposits of a subsequent channel generation (nice! DTM?)
821
//
822
on hirise 154.6,-5.23 - really beutiful outcrop of 'stuff' that abuts
823
channel
824
ditto 154.57, -5.292 (location is ish for this one)can be traced into
825
the layers underlying the fluvial channel in this case
826
//
827
nice measure of minimum overburden from mesas over rivers: CTX:
828
P05_003136_1746_XN_05S208W
829
//
830
in psp_7474 dtm, we see stage ?2 incising stage 1 deposits
831
the ability to crosscorrelate stage 2 across several hundred km (with an
832
"inferred" bit north of the 14km D crater) implies one of the following:
833
(1) autocyclic, but talking through a node downstream --> diffusion time
834
(2) allocyclic - draping deposit (ash or distal ejecta). Would have to
835
be quite thick (close, big) if distal ejecta.
836
(3) allocyclic - shut down system then restart on 10m distance scale.
837
analog: Rapid and widespread fl uvial response to eustatic forcing in the
838
Lower Mississippi Valley, figure 5c. State of LA has lidar!
839
//
840
there is evidence for 2 generations of fretting.
841
//
842
843
844
In introduction mention GRS data (important because no spectroscopic info). H2O content 5-6 Wt % from H - on
the shoulder of a broad lozenge. Cl is elevated (~0.7%) and elevated along axis. K is moderate. On the shoulder of a
planetary maximum in S (~3 wt%). Together, supports hydrated salts in the upper m.
31
845
//
846
847
"Architecture of vertically stacked fluvial deposits, Atane Formation, Cretaceous, Nuussuaq, central West
Greenland MARIA A. JENSEN" sed 'ology 2010 has ideas - basically, sitting within incised valley
848
//
849
(based on Hajek & Wolinsky Sed Geol 2012)
850
//
851
- true "cold start" requires at least one channel volume removed through the end of the channel system
852
- Sternberg's Law (grain size reduction) probably not visible in T.I.
853
//
854
- w/D crosscheck - should be about 57.
855
//
856
- *Lots* of good stuff Bridge 2003 (p. 202-214, 162-164). - cite Willis
857
'89 ...
858
//
859
- cutoff's and logic from Dunne's group
860
//
861
- constantine gsa-b 2010 lists 3 chute-cutoff mechanisms: swale
862
enlargement, headcut extension, embayment extension (chute cutoffs imply
863
a flow that isn't monotonically decreasing, suggesting that runoff
864
production isn't monotonically decreasing)
865
- chute cutoffs can be a confound for scroll bars where resolution is
866
poor
867
//
868
869
- exception to the "not much tectonics rule" - the wrinkle ridges. They are post-F. They are significantly post-F
based on the crater at 148.6E, 0.7S which postdates F1 and is wrinkle-ridged.. find offsets on these faults.
870
//
871
- I think Kerber & Head picked up on at least the first part of this - must cite
872
--
873
874
nature of channel fill is unknown. could be debris flows from waning stages of wet activity (but woah! debris flows
implies discrete event, wouldn't we expect quiescent flows gradually backing up?)
875
--
32
876
877
878
milliken et al. grl 2014 suppl info Figure A7C (ESP_014397_1745, according to Table S1) is a nice e.g. of channel cut
into bedforms. Analagous to alan howard's meanders region? Or am I just looking at yardangs in near-horizontal
strata? on inspection they look to be horizontal strata. and the orientation makes sense as yardangs.
879
//
880
Valley depth as probe of cut/fill amplitude
881
//
882
883
Quantitative Seismic Geomorphology of Fluvial Systems of the Pleistocene to Recent in the Malay Basin, Southeast
Asia* Faisal Alqahtani1 , Christopher Jackson1 , Howard Johnson1 , Alex Davis2 , and Ammar Ahmad1
884
885
the Malay Basin, Southeast Asia* Faisal Alqahtani1 , Christopher Jackson1 , Howard Johnson1 , Alex Davis2 , and
Ammar Ahmad1
886
887
FLUVIAL RESERVOIR ARCHITECTURE FROM NEAR-SURFACE 3D SEISMIC DATA, BLOCK B8/32, GULF OF THAILAND
by Hathaiporn Samorn
888
//
889
890
Transect 1: Assumption of isotropy underlying kriging is troubling because of preferred wrinkle ridge orientations
(obvious in MOLA, and also Lefort+ JGR 2012).
891
//
892
893
Tag each measurement with mapped unit (obvious!) and with whether or not it is part of a channel deposit. Then
can report in form similar to Foreman et al. 2012 Figure 3.
894
//
895
896
Could argue against 'distal ejecta and ash supply' (sed supply limitation for channel shift) on the basis of interval
between events and (small?) thickness of individual depositional event
33