You are on page 1of 4

Assessment of the Stability for Active Volcanic Small

Island by Use of SAR Interferometry


A Case Study in Nishinoshima Island

Japan Aerospace Exploration Agency (JAXA),


2-1-1 Sengen Tsukuba Ibaraki Japan,
Ryo Natsuaki, Manabu Watanabe, Takeshi Motohka, natsuaki.ryo@jaxa.jp
Masato Ohki and Masanobu Shimada
Earth Observation Research Center (EORC)

AbstractIn this paper, we present the results of the stability vessels and in worst case, a sudden avalanche of the new island
assessment based on Interferometric Synthetic Aperture Radar may cause tsunami and it can hit inhabited islands. Unmanned
(InSAR) for the active volcano. The stability of the lava is a quite devices are required for the safe observation and the satellite
important problem for the risk management. However, there are SAR is one of the most useful instruments. ALOS-2 has 14
limited tools, especially in distant places, to analyze whether the
days revisit orbit and has observed Nishinoshima
lava will continuously spread or it will stop flowing. InSAR has
been used for the deformation detection in the volcanology since approximately monthly since August 2014. Analysis from its
90th century. However, it has not been confirmed that it can be amplitude image is not enough to evaluate the activeness of the
used for this kind of requirement. We present some InSAR volcano. On the other hand, InSAR analysis has a possibility of
results for Nishinoshima island to show that SAR interferometry deformation detection resulting in stability assessment of the
is an effective tool for the stability assessment of the lava by island. Here, we present some InSAR and 3-pass DInSAR
showing its concretation in six months in time series analysis. results for Nishinoshima to show that SAR interferometry is an
effective tool for the assessment for the active volcano.
Index TermsALOS-2, PALSAR-2, Deformation Detection, Experimental results indicate that the south part of the new
Differential SAR Interferometry, Coherence analysis.
island become stable in six months.
I. INTRODUCTION II. METHODOLOGY OF INTERFEROMETRIC ANALYSIS FOR
Interferometric Synthetic Aperture Radar (InSAR) is an NISHINOSHIMA
important method for the landscape observation especially for Figure 1 is the slant-range time series of Nishinoshima
generating Digital Elevation Model (DEM) and deformation taken on August 31 2014, November 9 and March 1 2015
detection (Differential InSAR: DInSAR). Recently, high respectively. The pre-existing island is marked by square at the
spatial resolution and accurate orbit control enable us to create top left part in August 31 image. Their size of SAR images are
high coherence interferogram. One of the latest space borne 900 x 1200 pixels and their spatial resolution is 2.81 m
SARs is the Advanced Land Observation Satellite-2 (ALOS-2) (Azimuth: Top-Bottom) x 1.70 m (Range: Left-Right). Those
launched on 24th May 2014 by Japan Aerospace Exploration three images are taken from the same orbit. Table 1 shows their
Agency (JAXA). The Phased Array type L-band Synthetic scene IDs, baseline perpendiculars and height ambiguities. In
Aperture Radar -2 (PALSAR-2) aboard ALOS-2 is the state- order to create an interferogram, co-registration process is
of-the-art SAR [1] and is expected to observe various important. Especially, Nishinoshima is a small and
geohazards including volcano eruption. continuously expanding island, co-registration should be done
Nishinoshima island in Pacific Ocean has been erupting in the stable part of the island accurately. However, the stable
since November 2013. At first, it created a small new island parts of the island exist discretely and their sizes are quite small.
located in south-east off shore of the pre-existing island. Moreover, most part of the pre-existing island has been
Nowadays the new island rapidly expanded with its lava flow covered by lava in the image by March 1, 2015. In short, it is
and absorbed the pre-existing island. According to the record, difficult to find a stable point in this island. Here, we plot 2,000
the pre-existing island had 760(North-South) x 600(East-West) 64 x 64 pixel windows redundantly for each SAR images and
meter land while the new one had 1550 x 1240 m land on chose approximately 200 tie points for cross correlation
August 26 2014. The size of the island is continuously process. After the cross-correlation based co-registration, we
expanding that it marked 1530 x 1720 m and 1960 x 1800 m on applied the analytical local co-registration method based on
October 26 2014 and February 23 2015 respectively [2]. interferogram's residue and amplitude information [3].
As Nishinoshima is an isolated, uninhabited and small In general, a SAR interferogram contains those three
island, frequent observation is difficult. However, monitoring components except orbital fringes, ionospheric and
the island is very important mission for disaster prevention. atmospheric components.
That is, unexpected eruption may damage nearby fishing

978-1-4673-7297-8/15/$31.00 2015
c IEEE 841
TOTAL= TOPO+ DEFORM+ NOISE (1)

where TOTAL is the total phase, TOPO represents


topographic phase, DEFORM represents deformation toward
slant range direction and NOISE represents the noise
component respectively.
Unfortunately, there is no DEM for Nishinoshima island
though we have to subtract topographic fringes TOPO for
deformation detection. On the other hand, Nishinoshima has
100 m height in maximum according to the Japan Coast Guard
measurement [2]. That is, if the baseline perpendicular is
enough small, we can ignore the topographic fringes. Actually,
PALSAR-2 images for Nishinoshima have small baseline
perpendicular as shown in Table 1. We can say that the
maximum topographic fringe is smaller than 1.3 radians in the
interferogram made from the pair of Aug. 31 and Nov. 9. In
this interferogram, height ambiguity calculated from baseline
perpendicular of 108 m is -465 m. If the highest point in
Nishinoshima is 100 m, TOPO is calculated as

TOPO = 2pi * (altitude)/ (height ambiguity) (2)

and its value is approximately 1.3 radians in this case. If we


do not subtract TOPO, this will be the deformation error of 2.5
cm. As a result, there was no problem for not subtracting TOPO
because, the summit of the volcano had no coherency and we
could not find the fringe which represents the topography.
We firstly compare the two 2-pass interferograms and then,
analyze the interferometric coherence images to evaluate how
the coherency of those images changed.

TABLE I. LIST OF SCENE IDS AND THEIR PARAMETERS FOR


INTERFEROGRAM

ALOS-2 PALSAR-2 Scenes


Obs.
2014 Aug. 31 2014 Nov. 9 2015 Mar. 1
Date
ALOS20147105 ALOS202506053 ALOS20416205
Scene ID
30-140831 0-141109 30-150301
Baseline
108 m 35 m
Perp.
Height
-465 m 1,453 m
Amb.

III. ANALYSIS AMONG INTERFEROGRAMS


Figure 2 shows the 2-pass interferograms of those three
images and 3-pass DInSAR image. Figure 3 is the
interferometric coherence images of them. The window size for
the coherency calculation was 5 x 5 pixels. The top and the
middle images are respectively Aug. 31 Nov. 9 pair and Nov.
9 Mar. 1 pair. The bottom one is the 3-pass DInSAR image
and is 4 x 4 multilooked in order to reduce the noise
component. Among two 2-pass images, phase value is shifted
so that the averaged phase value of the pre-existing island
becomes the same. Because of the lack of the ground truth,
Fig. 1. Time series of Nishinoshima island taken by ALOS-2. Top: Aug. 31
2014, Middle: Nov. 9, Bottom: Mar. 1 2015. Squared part of the Aug. 31
those interferograms are all relative value.
image is the pre-existing island.

842 2015 IEEE 5th Asia-Pacific Conference on Synthetic Aperture Radar(APSAR)


However, we can see the continuous slant range extension
in the east part of the interferograms compared to the pre-
existing island. Especially in the east end of the interferogram
of Nov. 9 Mar 1 pair, we can see two cycles of the fringes,
i.e., there are more than 23.8 cm slant range extension between
the satellite and the island. As the incidence angle is
approximately 60 degree, this can be regarded as 40 cm
subsidence if this deformation does not contain horizontal
movement but is only caused by the subsidence. This
subsidence can be explained by the weight of the new lava
flow. That is, the old lava is pressed by the new one. The
northern part of the island had no coherency because the lava is
continuously flowing in this part.
The bottom interferogram in Figure 2 is the 3-pass DInSAR
image. In general, one of the 3-Pass DInSAR pair must be pre-
seismic (or no deformation) pair. However, there is no such
pair for this island, so we evaluated the deformation tendency
between two pairs by 3-Pass DInSAR. The result indicates that
the speed of the line-of-sight displacement is different between
southwest and southeast part of the island. If the deformation
speed of the southwest part is stable, the southeast part has
decreased between two interferograms.
We compared those coherencies for three areas in the island
marked as red squares in the top image of Figure 3. Those
areas represent A: Part of the pre-existing island which havent
been covered by lava, B: Relatively old lava which stopped
flowing by May 21 airborne observation, and C: Relatively
new lava which stopped flowing by August 26 airborne
observation [2]. There are 20 pixels in area A and 400 pixels in
B and C. In area A, there were limited numbers of pixels that
can be confirmed that those pixels represent the pre-existing
island.
The average coherency value of area A was 0.58 in Aug. 31
- Nov. 9 pair and 0.54 in Nov. 9 - Mar. 1 pair. That is, pre-
existing part keeps high coherency among three observations.
On the other hand, the average coherency was increased in area
B from 0.43 to 0.55 and in area C from 0.30 to 0.46
respectively. As the lava in area C is 3 months younger than
that in area B, we can say that the difference of coherency
represents the history of concreting lava. That is, the fresh (0
3 months old) lava in this volcano has coherency value of 0.30
and concreted (3 6 months old) lava has 0.43 0.46.
Of course these values are not same for all volcanoes.
However, if this hypothesis is true, we can predict the stability
of the lava from the coherency value. For example, in area C,
the coherence value will increase to 0.5 by June 2015 as long
as no new lava covers the area C. In short, if the lava becomes
stable, the coherency in the interferogram increases and the
increasing value is unique for one volcano.

Fig. 2. Interferogram images of Nishinoshima. Top: Interferogram between


Aug. 31 and Nov. 9 Middle: Nov. 9 and Mar. 1 2015, and Bottom: 3-
pass DInSAR among three images. Note that 3-pass DInSAR is 4 x 4
multilinked and the others are single look images.

2015 IEEE 5th Asia-Pacific Conference on Synthetic Aperture Radar(APSAR) 843


TABLE II. COHERENCE COMPARISON AMONG LEFT TWO
INTERFEROGRAMS IN FIGURE 3

Interferometric Coherence Values


Mark in Fig. 3 A B C
Old lava (stopped New lava (stopped
Description of Pre-existing
flowing by May flowing by Aug.
area island
21) 26)
Aug. 31 - Nov.
0.58 0.43 0.30
9 pair
Nov. 9 - Mar. 1
0.54 0.55 0.46
pair

IV. CONCLUSION
In this paper, we performed SAR interferometry, 3-pass
DInSAR and analyzed interferometric coherency for the active
volcano island Nishinoshima. By increase of coherency, we
found that the south part of the new island has been stable for
more than six month. On the other hand, the east part of the
island seems uplifting by lava flow. 3-pass DInSAR result
indicates that the south west part of the island has not moved
for 6 months but, south east part is uplifting. These results
proved the possibility of SAR interferometry for stability
assessment of the active volcano.

REFERENCES
[1] Yukihiro Kankaku, Masakazu Sagisaka and Shinichi Suzuki,
PALSAR-2 Launch and early orbit status IEEE Geoscience
and Remote Sensing Symposium 2014, pp. 3410 - 3412.
[2] Japan Coast Guard website Nishinoshima Volcano 5 Mar.
2015.
http://www1.kaiho.mlit.go.jp/GIJUTSUKOKUSAI/kaiikiD8/kai
yo18-e1.htm
[3] Ryo Natsuaki and Akira Hirose, "Performance improvement of
InSAR local co-registration with multiresolution interferogram,"
Asia-Pacific Conference on Synthetic Aperture Radar, pp.
WE2.R3.1, 2013.

Fig. 3. Coherence images of Figure 2. Note that Bottom coherence image for
3-pass DInSAR is 4 x 4 multilooked and the others are single look
images. Average coherence values are compared among three red
squares in the top figure. A: Pre-existing island, B: Old lava and C: New
lava

844 2015 IEEE 5th Asia-Pacific Conference on Synthetic Aperture Radar(APSAR)

You might also like