of 31
Supplementary
Materials
for
Where rivers jump course
Sam Brooke
et al
.
Corresponding author
:
Vamsi Ganti, vganti@ucsb.edu
Sci
ence
37
6
,
9
8
7
(
202
2
)
DOI:
10.1126/sci
ence
.
abm1215
The
PDF
file
includes:
Supplementary Text
Figs. S1 to S12
References
Other Supplementary Material for this manuscript includes the following:
Table S1
Movies S1 and S2
S
ubmitted Manuscript:
Confidential
Template revised February 2021
2
Text S1: Avulsion database
a.
New avulsion events
mapped using satellite imagery
We followed previous work to identify avulsions from the time series of satellite imagery
(
31
,
38
)
. We located avulsion events on alluvial fans and deltas by manually examining planform
change in rivers across the global coastlines, lakes, and mountain fronts. We focused on the lobe
-
5
scale avulsions, i.e., the largest
-
scale avulsion events that occur at
the apex of fans and deltas that
caused a permanent shift in the river course from the apex to the axial river or the shoreline. We
did not consider intralobe avulsions and bifurcations at the river mouths in our study.
The
controls on the location where
intralobe avulsions and bifurcations occur is different than lobe
-
scale avulsions, which set the size of fans and deltas. Thus, it is critical to assess avulsions using
10
a time series of satellite imagery rather than static satellite images where avulsions
can be
conflated with bifurcations.
The time between successive avulsions is inversely proportional to
the in
-
channel sedimentation rates
(
39
)
, whi
ch in turn is set by fluvial sediment supply
(
40
)
.
Avulsion timescales on the order of years to decades that facilitate the direct observation of
avulsions in the satellite record, therefore, occur in the tropics and regions where steep,
15
sediment
-
laden rivers could form del
tas.
We identified avulsions using a combination of the global surface water change masks
(
18
)
and analysis of the
time series of Landsat imagery. In most cases, avulsion locations were
immediately apparent from surface water change masks (Fig. S1), showing clearly where
previous channels had been abandoned in favor of a new channel feeding its new delta or fan
20
lobe.
We also downloaded and curated the full Landsat time series data for each candidate
avulsion event, and turned these images into time
-
lapse animations (see videos S1 and S2 for
delta and fan avulsion time
-
lapse examples), where each cloud
-
free image repres
ented a single
frame. We explored the available NASA/USGS Landsat archive (1972
-
2021) for cloud
-
free
multispectral images, which included Landsat 1
-
4 MSS sensor data with ~60 m pixel resolution
25
(up to 1984 C.E.) and Landsat 5
-
8 TM/ETM+/OLI ~30 m pixel reso
lution data (1984 C.E.
onwards). The advantage of full time
-
lapse animations was the ability to discriminate between
rapid yet continuous migration of river channels across the floodplains from avulsion events.
Further, this method allowed us to distinguis
h lobe
-
scale avulsions from intralobe avulsions
(where the flow rejoins the trunk channel at a downstream location) and bifurcations.
To better
30
S
ubmitted Manuscript:
Confidential
Template revised February 2021
3
highlight changes in surface water, we employed the near
-
infrared (NIR) bands (Fig. S1) and
indices such as the
Normalized Difference Water Index (NDWI) when analyzing the Landsat
imagery (Movie S2).
We used the aforementioned methodology to locate avulsions globally. For each avulsion
event, we recorded the latitude and longitude of the avulsion site (Table S1), i.
e., the location at
5
which the river pathway diversion was initiated. In all cases, the avulsion location was stable
during the relocation process. The mapping of avulsion sites was possible using the Landsat
image taken closest to the avulsion event, with
avulsion sites marked as a point on the channel
centerline. The error of avulsion locations is hence half the width of the channel. To assess
topographic changes at the avulsion sites, we extracted topographic swath profiles across the
10
avulsion sites (see
Text S2). For the delta avulsion sites, we estimated the avulsion length,
L
A
,
which was defined as the streamwise distance from the avulsion site to the river mouth of the
parent channel at the time the avulsion was initiated. For this computation, we chos
e the Landsat
image that best captured the avulsion event and manually measured the streamwise distance
between the avulsion site and river mouth in QGIS.
15
Finally, to mitigate selection bias, we worked as three separate teams to independently record
the l
ocation and timing of avulsion events across the globe. In doing so, we were able to
corroborate each avulsion site and confirm its validity as a lobe
-
scale avulsion event, and
whether the exact timing could be determined from a thorough exploration of the
Landsat
imagery archive. In total, we identified 90 avulsion events in the satellite record of which 13
20
avulsion events were previously reported
(
6
,
31
,
35
,
41
)
.
The other 23 avulsion
events in our
global database were previously reported in the literature and pre
-
dated the satellite record.
b.
Compilation of previously
-
reported historical river avulsions on fans and deltas
We augmented the new observations of avulsions with a compilation
of previously
-
reported
avulsions on fans and deltas. For the delta avulsions, we restricted our analysis to avulsion sites
25
that were either mapped in satellite imagery or on historical maps, which allowed for an
assessment of the avulsion length,
L
A
. In to
tal, we compiled 36 previously
-
reported lobe
-
scale
avulsions on fans and deltas (Table S1). The majority of these avulsions were from large,
lowland delta systems. Our database included 6 avulsions mapped on the Sulengguole River
delta using landsat imager
y
(
35
)
, 7 delta avulsions and 9 fan avulsions on the Huanghe
(
9
,
13
)
,
30