WEBVTT

1
00:00:13.113 --> 00:00:14.214
- I'm Hannah Dietterich,

2
00:00:14.214 --> 00:00:17.684
and today I'll be talking
about lava-flow monitoring

3
00:00:17.684 --> 00:00:20.653
and forecasting that was
aided by remote sensing

4
00:00:20.653 --> 00:00:24.524
during the 2018 eruption of
Kīlauea Volcano in Hawaii.

5
00:00:25.425 --> 00:00:27.994
This is work I've done with a
number of USGS collaborators,

6
00:00:27.994 --> 00:00:30.130
but it's also greatly benefited from those

7
00:00:30.130 --> 00:00:32.198
who aided in eruption response.

8
00:00:33.400 --> 00:00:37.070
Kīlauea Volcano is in East Hawaii Island,

9
00:00:37.070 --> 00:00:41.741
and was until 2018 erupting
for more than 35 years

10
00:00:41.741 --> 00:00:44.444
from the Pu'u 'Ō'ō vent in
the middle East Rift Zone

11
00:00:44.444 --> 00:00:46.746
of Kīlauea Volcano.

12
00:00:46.746 --> 00:00:47.580
More recently,

13
00:00:47.580 --> 00:00:50.750
there was an active summit
lava lake for 10 years

14
00:00:50.750 --> 00:00:53.853
within the caldera of Kīlauea caldera.

15
00:00:53.853 --> 00:00:56.289
And then in 2018,

16
00:00:56.289 --> 00:00:58.758
magma propagated further down the rift

17
00:00:58.758 --> 00:01:00.593
and erupted in the lower East Rift Zone

18
00:01:00.593 --> 00:01:03.396
within a neighborhood
called Leilani Estates.

19
00:01:05.665 --> 00:01:07.467
The eruption, the effusive eruption,

20
00:01:07.467 --> 00:01:10.203
occurred in sort of three phases.

21
00:01:10.203 --> 00:01:11.905
The first phase, the
first couple of weeks,

22
00:01:11.905 --> 00:01:12.839
were characterized

23
00:01:12.839 --> 00:01:16.342
by the opening of many,
many small fissures.

24
00:01:16.342 --> 00:01:18.745
You can see those here in this thermal map

25
00:01:18.745 --> 00:01:23.483
overlaid on a satellite
image of Leilani Estates.

26
00:01:23.483 --> 00:01:25.718
And these were all very, very small

27
00:01:25.718 --> 00:01:28.388
short-lived fissure eruptions producing

28
00:01:28.388 --> 00:01:31.691
very small pāhoehoe lava
flows for the most part,

29
00:01:31.691 --> 00:01:33.259
but starting in mid-May,

30
00:01:33.259 --> 00:01:36.729
the effusion rates increased
quite dramatically.

31
00:01:36.729 --> 00:01:38.832
The flows were larger,

32
00:01:38.832 --> 00:01:39.999
they were longer-lived,

33
00:01:39.999 --> 00:01:42.168
and the magma that was erupting

34
00:01:42.168 --> 00:01:45.271
was hotter and lower viscosity.

35
00:01:45.271 --> 00:01:48.208
So these flows were able to
travel much, much further

36
00:01:49.142 --> 00:01:51.277
and influence areas downhill

37
00:01:51.277 --> 00:01:54.380
from the neighborhood of Leilani Estates.

38
00:01:54.380 --> 00:01:55.281
And then in late May,

39
00:01:55.281 --> 00:01:57.717
activity really focused
just on one fissure,

40
00:01:57.717 --> 00:01:59.552
fissure number eight,

41
00:01:59.552 --> 00:02:03.990
and produced a very long,
long-lived channel system

42
00:02:03.990 --> 00:02:05.792
that persisted for more than two months

43
00:02:05.792 --> 00:02:08.228
until the end of the eruption in August.

44
00:02:09.195 --> 00:02:11.865
So, what you can see from
the chronology is the,

45
00:02:12.799 --> 00:02:13.967
some of the challenges

46
00:02:13.967 --> 00:02:17.337
in monitoring and
forecasting this eruption.

47
00:02:17.337 --> 00:02:19.739
It was characterized by many, many vents,

48
00:02:19.739 --> 00:02:22.108
they were changing all the time.

49
00:02:22.108 --> 00:02:26.846
The eruption occurred over
a seven kilometer stretch

50
00:02:26.846 --> 00:02:28.448
of the rift zone

51
00:02:28.448 --> 00:02:31.918
from the most up-rift to
most down-rift fissures,

52
00:02:31.918 --> 00:02:34.420
and fissures would erupt, stop erupting,

53
00:02:34.420 --> 00:02:36.055
and then reactivate.

54
00:02:36.055 --> 00:02:40.426
Also, overflows and levee failures

55
00:02:40.426 --> 00:02:43.163
within the flows produced flow branching,

56
00:02:43.163 --> 00:02:47.300
so there were many sort of
active sources to keep track of

57
00:02:47.300 --> 00:02:49.202
over the course of the eruption.

58
00:02:49.202 --> 00:02:52.105
As you can see, there were
also evolving effusion rates,

59
00:02:52.105 --> 00:02:55.542
more than two orders of magnitude
changes in effusion rates,

60
00:02:55.542 --> 00:02:58.945
as well as rheology that
produced a really wide range

61
00:02:58.945 --> 00:03:00.813
of erupted behavior.

62
00:03:00.813 --> 00:03:03.016
And so to forecast flows,

63
00:03:03.016 --> 00:03:05.451
fundamentally lava flows
are gravity currents,

64
00:03:05.451 --> 00:03:06.519
they flow downhill

65
00:03:06.519 --> 00:03:08.721
and if we have a good
sense of the topography,

66
00:03:08.721 --> 00:03:13.560
we can sort of figure out where
flows are most likely to go,

67
00:03:13.560 --> 00:03:15.395
the path they're most likely to take.

68
00:03:15.395 --> 00:03:18.031
But the complex dynamics of these flows,

69
00:03:18.031 --> 00:03:19.766
those changing effusion rates,

70
00:03:19.766 --> 00:03:22.936
the evolution from very
small pāhoehoe lava flows

71
00:03:22.936 --> 00:03:25.438
to very large channelized 'a'ā flows,

72
00:03:25.438 --> 00:03:28.641
really limits our ability to
actually forecast advance rates

73
00:03:28.641 --> 00:03:29.676
during this eruption.

74
00:03:31.377 --> 00:03:34.013
So traditionally, lava-flow
forecasting in Hawaii

75
00:03:34.013 --> 00:03:37.183
has been done using lines
of steepest descent.

76
00:03:37.183 --> 00:03:39.452
These are the blue lines on this map,

77
00:03:39.452 --> 00:03:43.623
and they are just the
centers of drainages,

78
00:03:43.623 --> 00:03:46.259
so equivalent to a watershed.

79
00:03:46.259 --> 00:03:50.630
These are the drainages
that lava flows might follow

80
00:03:50.630 --> 00:03:52.832
when they follow the terrain.

81
00:03:52.832 --> 00:03:54.801
But of course, they don't incorporate

82
00:03:54.801 --> 00:03:57.003
any sort of uncertainty in the terrain,

83
00:03:57.003 --> 00:03:58.605
and they're somewhat simplified

84
00:03:58.605 --> 00:04:02.141
and don't start at an eruption source

85
00:04:02.141 --> 00:04:04.177
like an active vent, say.

86
00:04:04.177 --> 00:04:05.578
So in this eruption,

87
00:04:05.578 --> 00:04:08.314
we wanted to have a tool to forecast

88
00:04:08.314 --> 00:04:11.317
from specific active new vents,

89
00:04:11.317 --> 00:04:13.519
active overflows or flow fronts,

90
00:04:14.587 --> 00:04:16.422
but still incorporating the terrain

91
00:04:16.422 --> 00:04:19.692
so we employed the downflow model

92
00:04:19.692 --> 00:04:23.596
from Favalli et al. at INGV Pisa,

93
00:04:23.596 --> 00:04:28.268
which allows you to
simulate many, many lava,

94
00:04:28.268 --> 00:04:30.837
steepest-descent lines
from a given location

95
00:04:30.837 --> 00:04:33.873
with stochastic perturbation
of the topography

96
00:04:33.873 --> 00:04:38.444
that allows the flows
to spread a little bit

97
00:04:38.444 --> 00:04:43.449
to overcome small topographic obstacles

98
00:04:43.783 --> 00:04:48.321
and therefore behave a little
bit more like a lava flow.

99
00:04:48.321 --> 00:04:51.157
And we can display the
results probabilistically,

100
00:04:51.157 --> 00:04:55.995
in terms of how frequently
a given area is,

101
00:04:58.097 --> 00:05:00.900
(laughs) has lava flow paths,

102
00:05:00.900 --> 00:05:02.935
forecasts lava flow
paths going through it.

103
00:05:02.935 --> 00:05:07.940
So, areas that more of
the flow paths traverse

104
00:05:07.940 --> 00:05:11.244
are more likely to be
the route the flow takes.

105
00:05:11.244 --> 00:05:12.779
And so, the results look like this,

106
00:05:12.779 --> 00:05:17.250
where the main drainage is
of course highlighted in both

107
00:05:17.250 --> 00:05:20.620
the downflow model and
the steepest descent lines

108
00:05:20.620 --> 00:05:21.621
shown in blue.

109
00:05:22.855 --> 00:05:26.392
And if we compare the, where the flow went

110
00:05:26.392 --> 00:05:29.529
to where the flow was forecast to go,

111
00:05:29.529 --> 00:05:33.032
we can see that in general flows went

112
00:05:33.032 --> 00:05:34.867
where we thought they might.

113
00:05:34.867 --> 00:05:35.935
In this example,

114
00:05:35.935 --> 00:05:37.203
which is the initial forecast

115
00:05:37.203 --> 00:05:40.106
of the fissure eight reactivated flow,

116
00:05:40.106 --> 00:05:42.775
we sort of had routes to
the North and to the east

117
00:05:42.775 --> 00:05:46.145
and sort of two out of three
of these were forecast.

118
00:05:46.145 --> 00:05:47.113
The one to the North,

119
00:05:47.113 --> 00:05:51.250
actually there were large
ground cracks in this area

120
00:05:51.250 --> 00:05:53.219
that may have caused the flow to stall

121
00:05:53.219 --> 00:05:55.321
in that direction very locally.

122
00:05:56.989 --> 00:05:58.658
We can see that the downflow model

123
00:05:58.658 --> 00:06:00.860
has some advantages over
the steepest descent model,

124
00:06:00.860 --> 00:06:04.964
not just in that it's
originating at the lava source,

125
00:06:04.964 --> 00:06:09.135
but also the flow paths can traverse areas

126
00:06:09.135 --> 00:06:12.038
between the steepest-descent lines.

127
00:06:12.038 --> 00:06:15.508
So, allowing the flow to
overcome some topography

128
00:06:15.508 --> 00:06:18.177
allows it to enter multiple drainages

129
00:06:18.177 --> 00:06:20.213
as lava flows tend to do.

130
00:06:20.213 --> 00:06:23.149
And they can also show the degree to which

131
00:06:23.149 --> 00:06:26.185
the topography is confined or not,

132
00:06:26.185 --> 00:06:27.720
so areas where the flow is more likely

133
00:06:27.720 --> 00:06:29.522
to spread or narrow,

134
00:06:29.522 --> 00:06:31.791
so showing you a little bit

135
00:06:31.791 --> 00:06:36.262
about what flow width
might look like as well.

136
00:06:36.262 --> 00:06:39.932
But we can see that there
is also a flow path,

137
00:06:39.932 --> 00:06:42.034
potential flow path here to the South

138
00:06:42.034 --> 00:06:44.804
that goes straight over the
flows that were already in place

139
00:06:44.804 --> 00:06:46.839
which are shown in dark gray.

140
00:06:46.839 --> 00:06:49.842
And so that really highlights
one of the biggest challenges

141
00:06:49.842 --> 00:06:52.044
we had during this eruption,

142
00:06:52.044 --> 00:06:54.447
which was as the eruption went on,

143
00:06:54.447 --> 00:06:56.616
topography was changing more and more

144
00:06:56.616 --> 00:06:59.485
from what our pre-eruptive
surface looked like.

145
00:06:59.485 --> 00:07:01.187
But that of course greatly influences

146
00:07:01.187 --> 00:07:02.955
where the flows are likely to go.

147
00:07:02.955 --> 00:07:04.457
So, if we had run the simulation

148
00:07:04.457 --> 00:07:08.060
over just the pre-eruptive
topography, on the left,

149
00:07:08.060 --> 00:07:10.496
you can see a number of flow
paths going to the South

150
00:07:10.496 --> 00:07:13.766
straight over where lava flows
have already been emplaced.

151
00:07:13.766 --> 00:07:17.770
But we were updating flow
maps, flow extent maps

152
00:07:17.770 --> 00:07:19.939
as frequently as we
could during the eruption

153
00:07:19.939 --> 00:07:23.209
at high resolution, sometimes twice a day.

154
00:07:23.209 --> 00:07:25.511
So, what we would often
do during the eruption

155
00:07:25.511 --> 00:07:28.614
is take that most
up-to-date flow extent map,

156
00:07:28.614 --> 00:07:31.851
add artificial thickness to the terrain

157
00:07:31.851 --> 00:07:34.320
and use that to inform our forecasts

158
00:07:34.320 --> 00:07:36.456
to make them more accurate,

159
00:07:36.456 --> 00:07:39.025
but then we were also trying to collect

160
00:07:39.025 --> 00:07:42.929
as much syn-eruptive
topographic data as we could.

161
00:07:42.929 --> 00:07:45.665
So, this example on the
far right uses a DEM

162
00:07:45.665 --> 00:07:46.766
from single-pass InSAR

163
00:07:48.301 --> 00:07:50.837
to actually quantify the new terrain

164
00:07:50.837 --> 00:07:53.105
where flows have been emplaced

165
00:07:53.105 --> 00:07:58.077
and run simulations over
up-to-date topographic data.

166
00:07:58.177 --> 00:08:01.981
And you can see the effect that this has.

167
00:08:03.182 --> 00:08:06.185
So, in order to collect
up-to-date topographic data,

168
00:08:06.185 --> 00:08:09.155
we relied on a number of
remote sensing techniques.

169
00:08:09.155 --> 00:08:12.024
The one we used very frequently

170
00:08:12.024 --> 00:08:15.862
was aerial surveys with
unoccupied aircraft systems.

171
00:08:15.862 --> 00:08:17.630
These are small UAS,

172
00:08:17.630 --> 00:08:20.933
including hexacopters
and fixed-wing aircraft.

173
00:08:20.933 --> 00:08:23.569
And these would fly photo survey flights

174
00:08:23.569 --> 00:08:26.172
that we could use
structural motion processing

175
00:08:26.172 --> 00:08:28.841
to reconstruct the terrain,

176
00:08:30.510 --> 00:08:33.713
or things like orthophotos
mapping the flow field.

177
00:08:34.780 --> 00:08:36.182
These have somewhat limited extent

178
00:08:36.182 --> 00:08:37.950
but could be flown quite frequently.

179
00:08:37.950 --> 00:08:41.854
So, getting updated topography
as often as 45 minutes.

180
00:08:42.822 --> 00:08:46.626
We also were able to conduct
two syn-eruptive LiDAR surveys,

181
00:08:46.626 --> 00:08:49.962
in particular one in June focused on

182
00:08:49.962 --> 00:08:53.332
the heavily densely-vegetated rain forest

183
00:08:53.332 --> 00:08:55.001
within Leilani Estates

184
00:08:55.001 --> 00:08:58.037
and really captured well a graben feature

185
00:08:58.037 --> 00:09:01.507
where the terrain changed by meters

186
00:09:01.507 --> 00:09:03.442
right above where the dike was emplaced

187
00:09:03.442 --> 00:09:05.211
and these new features were erupting

188
00:09:05.211 --> 00:09:07.313
and we were worried that this might impact

189
00:09:07.313 --> 00:09:11.784
where flows were going to be
going within Leilani Estates.

190
00:09:13.719 --> 00:09:18.024
We also used a single-pass
InSAR flown on a jet.

191
00:09:18.024 --> 00:09:19.892
This is the NASA GLISTIN instrument

192
00:09:19.892 --> 00:09:22.295
and work that Paul Lundgren lead.

193
00:09:22.295 --> 00:09:25.331
And this allowed collection of terrain,

194
00:09:25.331 --> 00:09:27.833
both from the summit of
the volcano in the caldera,

195
00:09:27.833 --> 00:09:31.737
as well as throughout the
whole lava flow field,

196
00:09:31.737 --> 00:09:34.774
capturing things like this
series of flow thickness maps,

197
00:09:34.774 --> 00:09:36.175
seen on the right.

198
00:09:38.844 --> 00:09:41.881
So, a nice example of
how updated topography

199
00:09:41.881 --> 00:09:44.517
was really critical in the eruption.

200
00:09:44.517 --> 00:09:47.720
This is an example from May 20th,

201
00:09:47.720 --> 00:09:50.089
I've covered the date with my face,

202
00:09:51.057 --> 00:09:53.793
but the lava flows so far

203
00:09:53.793 --> 00:09:56.362
have been emplaced in the pink area

204
00:09:56.362 --> 00:09:58.331
and everything's gone to the South so far

205
00:09:58.331 --> 00:09:59.765
in the eruption in this area.

206
00:09:59.765 --> 00:10:01.934
All the flows have gone to the South,

207
00:10:01.934 --> 00:10:06.672
but with a new flow reported
on the North side of this area

208
00:10:06.672 --> 00:10:09.642
and now, a lot of vent and lava material

209
00:10:09.642 --> 00:10:12.478
has constructed and changed
the typography here,

210
00:10:12.478 --> 00:10:14.880
the question was whether flows

211
00:10:14.880 --> 00:10:18.150
were going to eventually start
going to the North instead.

212
00:10:18.150 --> 00:10:20.753
And so we wanted to collect
updated terrain data

213
00:10:20.753 --> 00:10:21.587
in this area,

214
00:10:21.587 --> 00:10:23.489
this thickness map and hillshade

215
00:10:23.489 --> 00:10:26.892
from a digital surface
model from a drone flight.

216
00:10:28.094 --> 00:10:31.230
Just of this region around the vent area,

217
00:10:31.230 --> 00:10:32.798
you can see that a lot of topography

218
00:10:32.798 --> 00:10:34.900
has been constructed in this area

219
00:10:34.900 --> 00:10:37.637
that might affect where the flow will go.

220
00:10:37.637 --> 00:10:41.207
So if we run simulations
on the initial terrain,

221
00:10:41.207 --> 00:10:43.576
again the flow paths
largely go to the South.

222
00:10:44.443 --> 00:10:46.746
But we know that there's
now already lava there.

223
00:10:46.746 --> 00:10:48.381
And if we incorporate some of that,

224
00:10:48.381 --> 00:10:50.282
just this little area outlined in red,

225
00:10:50.282 --> 00:10:53.552
where a new DEM has been swapped in,

226
00:10:53.552 --> 00:10:55.254
we can see that there's now enough terrain

227
00:10:55.254 --> 00:10:57.757
that some of the flow
paths maybe expected to go

228
00:10:57.757 --> 00:10:59.191
to the North instead

229
00:10:59.191 --> 00:11:01.494
and that that would be
something to look for.

230
00:11:03.329 --> 00:11:07.166
We can use this sequence
of terrain data as well

231
00:11:07.166 --> 00:11:09.201
to characterize eruptive volume

232
00:11:09.201 --> 00:11:10.803
and the evolution of the effusion rate

233
00:11:10.803 --> 00:11:12.171
over the course of the eruption,

234
00:11:12.171 --> 00:11:14.240
which is also important for keeping track

235
00:11:14.240 --> 00:11:16.208
of evolving hazards

236
00:11:16.208 --> 00:11:19.945
and whether the eruption
is waxing or waning.

237
00:11:19.945 --> 00:11:22.448
So we used again all these datasets,

238
00:11:22.448 --> 00:11:26.686
including post-eruptive multi-beam sonar,

239
00:11:26.686 --> 00:11:31.123
and are able to construct
eruptive volume over time.

240
00:11:31.123 --> 00:11:33.926
You can see volume data
for the entire flow field

241
00:11:33.926 --> 00:11:35.194
is somewhat sparse

242
00:11:35.194 --> 00:11:36.996
because the flow field was very large

243
00:11:36.996 --> 00:11:40.933
and beyond the extent of
just drones, drone mapping.

244
00:11:41.867 --> 00:11:43.402
And so, to supplement this

245
00:11:43.402 --> 00:11:46.172
we also used drones to record video

246
00:11:46.172 --> 00:11:49.175
of the channel near the vent over time,

247
00:11:49.175 --> 00:11:53.245
to record regular observations
of velocities in the channel,

248
00:11:53.245 --> 00:11:56.415
to track flow rates and
thus effusion rates.

249
00:11:56.415 --> 00:11:59.185
So, the blue dots here show

250
00:11:59.185 --> 00:12:02.421
the DEM time-average discharge
rates of the eruption,

251
00:12:02.421 --> 00:12:04.790
showing the initial low effusion rates

252
00:12:04.790 --> 00:12:07.493
that popped up to much
higher effusion rates

253
00:12:07.493 --> 00:12:09.095
later in the eruption,

254
00:12:09.095 --> 00:12:11.797
and the drone data is shown in black.

255
00:12:11.797 --> 00:12:14.500
That shows again the
sort of maximum peak flow

256
00:12:14.500 --> 00:12:16.368
of the eruption in mid-June,

257
00:12:16.368 --> 00:12:18.170
and then continued high rates

258
00:12:18.170 --> 00:12:21.040
until right before the
eruption ended in August,

259
00:12:21.040 --> 00:12:26.045
including little details like
post-caldera-collapse events

260
00:12:26.045 --> 00:12:28.848
causing surges in lava effusion rate

261
00:12:28.848 --> 00:12:30.916
further down the channel
that were important

262
00:12:30.916 --> 00:12:34.053
for hazard assessments
at the time as well.

263
00:12:34.053 --> 00:12:37.189
These were associated with many overflows.

264
00:12:37.189 --> 00:12:39.692
So, in conclusion, we can see that simple,

265
00:12:39.692 --> 00:12:42.628
on-demand, very simple
lava-flow forecasting

266
00:12:42.628 --> 00:12:44.330
just based on terrain,

267
00:12:44.330 --> 00:12:48.267
was able to offer up-to-date
and accurate information

268
00:12:48.267 --> 00:12:49.902
to emergency managers.

269
00:12:49.902 --> 00:12:52.238
And this was really
enabled by a collection

270
00:12:52.238 --> 00:12:55.708
of repeat high-resolution
topographic mapping.

271
00:12:55.708 --> 00:12:58.043
Without new topographic data,

272
00:12:58.043 --> 00:13:01.280
our forecasts would not
have been as accurate.

273
00:13:01.280 --> 00:13:03.616
And what we can also see is that

274
00:13:03.616 --> 00:13:05.751
this data informs hazard response,

275
00:13:05.751 --> 00:13:07.853
but it also informs our understanding

276
00:13:07.853 --> 00:13:09.021
of lava-flow emplacement.

277
00:13:09.021 --> 00:13:10.923
The syn-eruptive remote sensing data

278
00:13:10.923 --> 00:13:14.260
offers a lot of insights
into eruption dynamics.

279
00:13:14.260 --> 00:13:17.530
So we hope to learn much more
about lava flow emplacement

280
00:13:17.530 --> 00:13:19.832
and improve our lava-flow forecast models

281
00:13:19.832 --> 00:13:22.835
using eruption data from this eruption.

282
00:13:22.835 --> 00:13:24.003
So, thank you.