There-dimensional change detection in coastal cliffs using UAV and TLSYuichi S. Hayakawa* & Hiroyuki Obanawa

*Environmental Earth Science, Hokkaido University

2020.05.08 fri

coastal

ßprotection

ß

35°19’11.4”N

140°

04’3

6.6”

E

1 km

Kuju

kuri

Sand

y C

oa

st

Suzume-Jima Island

Cape Taito-Saki

The Isumi

Riveràà

Tokyo

a rare opportunity in Japan: natural rapid processes of coastal erosion

out of the coastal protection oceanic waves directly attacks the bedrock cliff

rapidly shrinking for decades1974 2014

Obanawa et al. 2015

33 m

35 m

35 m

20 m

32 m a.s.l.

1967 1972 1977 1982

1987 1992 1997 2002

2007 2012 • constant erosion• shore platform formation• reported erosion rates around this

site: 1 m/y• (1960–1966 by 1:1,000 topo maps;

Horikawa and Sunamura 1967)

study site: rapidly eroded small coastal island

2

TLS #1: TOPCON GLS-1500a medium-range scanner– max. distance: 500 m– max. frequency: 30,000 pts/s– range accuracy: 4 mm @150 m– weight: 16 kg (body) + batteries

TLS #2: Trimble TX5a short-range scanner– max. distance: 120 m– max. frequency: 900,000 pts/s– range accuracy: 2 mm @25 m–weight: 5 kg

methods

DJI Phantom 2 + NIKON COOLPIX A; Phantom 3 Professional/Advanced; Phantom 4; Mavic Pro; Mavic 2 Pro;

post processing• SfM-MVS photogrammetry

by PhotoScan• 500-1,000 photos for each time• positioning accuracy 1: camera-mounted GNSS, >1 m• positioning accuracy 2: GCP by PPK-GNSS (fix solution)

(Trimble GeoXH), 13.4–14.9 mm (14.4 mm RMS)

• further aligned by ICP– UAS dense cloud à TLS cloud– CloudCompare / Trimble RealWorks– cloud-based registration (ICP), errors: 25.1–39.7 mm 10 m

TLS UAV

TLS + UAV

point cloud

1

TLS

GNSS

point cloud

2

point cloud

N

point cloud

3

...

georeference

ICP

ICP

ICP ......

UAS-SfM

point cloud

1

point cloud

2

point cloud

3

.........point cloud

N

ICP ICP ICP ICP ICP

mes

hing

......mesh 

dataset1...N

differentiating

DVE

3methods – workflow

160223160618

pre-point cloud

post-point cloud

calculate cloud-to-cloud distance

invert normals for post dataset

extract changed areas(detectable distance: 

10 cm or more)

merge

Poisson surface reconstruction

3D mesh polygons of changed areas

fill holes, repair error faces

make solid

manual removal of erroneous polygons

workflow of Differential Volume Estimate

4results – quantification of volumetric changes was successful!

interval changes 2014.06.24 àà 2019.10.02 I) 140624 àà 141031

II) 141031 àà 150211

III) 150211 àà 150618

IV) 150618 àà 151023

IV) 151023 àà 160223

V) 160223 àà 160618

(southeast) (northeast) (west)

VI) 160618 àà 161029

VII) 161029 àà 170218

VIII) 170218 àà 170702

IX) 1707028 àà 171007

X) 171007 àà 180127

XI) 180127 àà 180916

(southeast) (northeast) (west)

XII) 180916 àà 190311

XIII) 190311 àà 191002

5results – quantification of volumetric changes was successful!

total changes 2014.06.24 àà 2019.10.02

total volume loss: 1,980 m3

average volume loss per month: 30 m3

current volume of the island: 11,300 m3

estimated time to fully eroded: ca. 30 years

6discussion – effects of waves

temporal variations: frequency of high tidal waves and erosion mass volume

• mass volume varies, 10.6 – 527.7 m3 per 4-7 months• equivalent annual erosion rates: 0.03 – 0.63 m/y• cf. approx. projected area of bedrock: 1,436 m2

< Ty

pho

on

Faxa

i>

volume

erosionrate y = 5.128x + 19.056

R² = 0.1136

0

100

200

300

400

500

600

0 10 20 30 40 50

ero

de

d v

olu

me

in a

pe

riod

(m

3 )

number of observation of waves over 3 m high

spatial variations: differences in wave attacks

refraction

ßß shore platform

N S

E

W

different directions of wave attacks for the cliff faces of the island could cause differences in the amount of erosion

although the relation is not clear, attacks by high tidal waves could have affected the amount of erosion