Environ Monit Assess (2018) 190:483

https://doi.org/10.1007/s10661-018-6869-y

Thermal monitoring and analysis of the large-scale field

earth-dam breach process
Chien-Yuan Chen & Su-Chin Chen & Kuan-Hua Chen &
Zhe-Ho Liu

Received: 6 April 2018 / Accepted: 12 July 2018

# Springer Nature Switzerland AG 2018

Abstract Thermal imaging is a nondestructive testing
method for monitoring internal material changes that are
indicated by changes in an object’s surface temperature.
In this study, field observation using thermographs was
applied to monitor and analyze the breach process of
large-scale earth dams. The earth dam test site was
Landao Creek in Nantou County, Taiwan. Four field
tests were performed to monitor and analyze a single
earth dam and two successive earth dams. Ponding first
occurred at the lowland of the riverbed upstream, and
base seepage occurred at the base of the dam down-
stream; overtopping failure soon followed. Earth-dam
failure mode is affected by the topographic characteris-
tics of the riverbed; specifically, lowland areas are prone
to ponding upstream, which causes piping and seepage
downstream as well as subsequent breaches. Ponding
and piping cause water seepage and soil wetting, which
are reflected in lower surface temperatures recorded on
thermographs. Thermographs can monitor changes in
surface temperature to evaluate the potential failure
Electronic supplementary material The online version of this
article (https://doi.org/10.1007/s10661-018-6869-y) contains
supplementary material, which is available to authorized users.
C.<Y. Chen (*) : K.<H. Chen : Z.<H. Liu
Department of Water and Resources Engineering, National Chiayi
University, 60004 Chiayi City, Taiwan, Republic of China
e-mail: chienyuc@mail.ncyu.edu.tw
S.<C. Chen
Department of Soil and Water Conservation, National
Chung-Hsing University, 402 Taichung City, Taiwan, Republic of
China
modes of dams. Dam surfaces with large temperature
variations may be potential failure areas. If confirmed,
this fact may prove useful for failure-mode prediction.
This paper proposes a monitoring index to reflect the
temperature changes in a given period, and this study
verified it empirically.
Keywords Infrared thermography . Earth dam .
Monitoring . Field test
Introduction
Infrared thermal imaging, based on Planck’s blackbody
radiation law (Planck 1901), involves detecting an object’s
thermographic projection and converting it into a digital
thermogram. Thermal imaging is a noncontact and nonde-
structive screening technique that can quickly detect the
surface-radiation temperature of objects. Temperature can
be easily identified from colorful thermal images.
Thermal imaging was first used for military applica-
tions. It has been extended to daily uses such as machine
maintenance (Boglietti et al. 2009) and medical moni-
toring (Janssens et al. 2017). Thermal imagers can con-
duct nondestructive surface detection quickly for a large
area. They have the camera-like advantage of effective
use at a distance and facilitate both long-term monitor-
ing and short-term inspection of surface-temperature
changes. Thermal imaging has been used to monitor
landslides (Frodella et al. 2014; Martino and Mazzanti
2014; Mineo et al. 2015) and to map tunnel discontinu-
ity (Barla et al. 2016).
483 Page 2 of 17
Fig. 1 Location of the earth-dam
field tests at Landao Creek in
Nantou County, Taiwan (Image
taken on 5 October, 2015 by
Google Earth)
Surface-temperature changes reflect the interior con-
struction quality of spray concrete-protected slopes (Wu
et al. 2005). Thermal images taken in the field show that
surface-temperature changes are larger in hollow areas
than in solid areas. Hollow areas can be detected by
subtracting the lowest temperature from the highest
temperature recorded through thermal imaging. This is
because hollow-area temperatures fluctuate more than
those in structurally robust areas after sun exposure. The
existence of hollow areas inside spray concrete-
protected slopes can be detected and monitored using
thermal images from different periods. Thermal images
can be employed to quickly evaluate rock, crack, and
rockfall distribution over a large area (Baroň et al. 2012;
Gigli et al. 2014; Teza et al. 2015).
Evaluation of natural-dam stability is necessary for
dam breach disaster prevention and mitigation. Natural-
dam formation is dictated by topographic and geologic
conditions. A natural-dam breach can bury buildings
beneath the slope, and secondary effects are caused by
ponding upstream and flooding downstream. An exam-
ple is Xiaolin (Shaolin) Village in Taiwan (Lo et al.
2011; Feng 2012; Wu et al. 2014). Most natural dams
are created by rainfall- and earthquake-induced failure
of bank slopes as well as the subsequent landslides and
debris flows that block rivers. In 2009, Typhoon
Morakot created 18 landslide dams (Forest Bureau
2009), which are difficult to reach to perform stability
analysis. Nondestructive detection at a suitable distance
for preliminary evaluation of stability and potential fail-
ure modes could help in preventing and mitigating dam-
breach disasters.
The failure modes of natural dams include the fol-
lowing (Schuster 1986; Chai et al. 2000; Korup 2005):
Environ Monit Assess (2018) 190:483
1. Overtopping failure
Overtopping failure occurs in dams with low perme-
ability, high shear strength, and large upstream dis-
charge. Water velocity becomes excessively fast, wet-
ting the front dam material; the water level increases,
overtopping occurs, and the dam subsequently fails
through progressive erosion.
2. Slope failure
Slope failure occurs in dams with high permeability
and low shear strength. The ponding water level in-
creases with water from upstream, and the flow overtops
the dam before the wetting front of the phreatic line
reaches the downstream slope face.
3. Progressive erosion failure
Progressive erosion failure occurs when the soil
has higher permeability and weaker shear strength
than those in either of the previously mentioned
failure modes. The rising ponding water level in-
creases with water from upstream, and the flow
overtops the dam before the wetting front of the
phreatic line reaches the downstream slope face.
The piping-induced surface-water flow erodes the
slope toe downstream, causing small landslides and
finally breaching the dam.
4. Complex failure
This failure mode combines two or three of the other
failure modes.
Environ Monit Assess (2018) 190:483 Page 3 of 17 483

Table 1 Dimensions and purposes of artificial trapezoidal earth dams

Test date Downstream Upstream Purposes

26 September, 2012 Height = 2.5 m None Failure-mode monitoring

Top width = 2 m
Length = 27 m
Side slope = 45°
Volume = 300 m3
Capacity = 855 m3
Riverbed slope = 7°
16 October, 2012 Height = 2.5 m Height = 2 m Failure-mode monitoring under
Top width = 2 m Top width = 2 m flash flood (induced by dam breach)
Length = 27 m Length = 15.5 m
Side slope = 45° Side slope = 45°
Volume = 300 m3 Volume = 125 m3
Capacity = 855 m3 Capacity = 207 m3
Riverbed slope = 7° Riverbed slope = 7°
7 November, 2012 Height = 2.5 m Height = 2 m Comparisons
Top width = 2 m Top width = 2 m
Length = 27 m Length = 15.5 m
Side slope = 45° Side slope = 45°
Volume = 300 m3 Volume = 125 m3
Capacity = 855 m3 Capacity = 207 m3
Riverbed slope = 7° Riverbed slope = 7°
10 April, 2014 Height = 2.5 ma None Monitoring-index verification
Top width = 2 m
Length = 27 m
Side slope = 45o
Volume = 300 m3
Capacity = 855 m3
Riverbed slope = 7°
a
Lower elevation in the middle

Fig. 2 Average surface-

temperature changes of earth
materials over time (Chen and
Chen 2014)
483 Page 4 of 17
Fig. 3 Artificial earth dam and corresponding thermal images. a, b Piping area in the downstream dam (September 21, 2012); the arrow
highlights the piping area. c, d 08:22 on the day of the test (September 26, 2012); the arrow highlights the ponding area
More than half of natural dams are breached through
overflow erosion (Korup 2005); some natural dams have
been breached through piping and slope failure in Taiwan
(Chen and Hsu 2009). The failure modes of natural dams
are determined by the dam material, soil shear strength,
soil permeability, and discharge conditions of source water
upstream (Chen 1999). A series of model tests enhanced
the understanding of the earth-dam breach process (Chen
et al. 2015a), but real-scale field tests remain scant.
Various methodologies are used to monitor the
surface and internal erosion of natural dams, includ-
ing nonintrusive geophysical techniques (Lum and
Sheffer 2005; Lin et al. 2013; Lai et al. 2014), visual
observation (Moffat et al. 2011), and passive seismic
interferometry using ambient seismic noise (Planes
et al. 2016). Using fiber optic-distributed tempera-
ture sensing (Bersan et al. 2018) in dams proved
effective for detecting leakages and internal erosion
(Cuong et al. 2017).
Environ Monit Assess (2018) 190:483
Dam breaches have proved dangerous to downstream
residents and can inundate the upstream reaches during
ponding. Earth embankment failures result from a combi-
nation of flooding, hydraulic weakness, and geotechnical
weakness. Weakness of the embankment induced by
flooding and seepage is a neglected source of hazards
(Polemio and Lollino 2011). The lifetime of natural dams
from formation to breach is generally short and just over
0.01 day at minimum (Korup 2004). Most natural dams are
inaccessibly high and dangerous, and short emergency-
response times are necessary to prevent disaster. A noncon-
tact, long-distance, nondestructive method of evaluating
natural-dam stability could mitigate dam-breach disasters.
Study area and methodology
The test site was Landao Creek at Huisun Forest Farm in
Nantou County, Taiwan (Fig. 1). Landao Creek is a branch
Environ Monit Assess (2018) 190:483 Page 5 of 17 483

(a) (b)

(c) (d)

(e) (f)

Fig. 4 Dam breach process during the test on September 26, ponding side; the arrows highlight the discharge overtopped areas.
2012. a, b At 9:36, base seepage and piping occurred at the e, f At 9:57, the dam was breached in the middle and on the initial
downstream face; the arrows highlight the piping area. c, d At ponding side; the arrows highlight the dam breached areas
9:55, discharge overtopped the dam in the middle and on the initial

of the Peikong Stream in the upper stream of the Wuxi that involved observing the downstream inundation area
River. Its watershed area is 536.33 km2. The test is the first following breach. Landao Creek is prone to debris flow,
earth-dam breach test with adjustable upstream discharge with a length of 2792 m and a basin slope of between 44°
483 Page 6 of 17 Environ Monit Assess (2018) 190:483

(a) (b)

(c) (d)

(e) (f)

(g) (h)
 Environ Monit Assess (2018) 190:483
R Fig. 5 Breach process of the upstream dam on October 16, 2012.
a, b At 10:04, ponding water occurred in front of the upstream
dam; the arrow highlights the ponding area. c, d At 10:52, the dam
started to breach on the initial ponding side; the arrow highlights
the dam breached area. e, f At 11:00, water flow overtopped the
dam. g, h At 11:03, the dam breached on the initial ponding side
and in the middle
and 55°. Debris flows occurred in 2001 during Typhoon
Toraji; two bridges across the creek were buried under
masses of debris. Another debris flow occurred in 2004
during Typhoon Mindulle; debris masses buried the brid-
ges, blocked the outreach road, stopped the creek, diverted
the stream flow, and damaged nearby buildings.
The construction sites of dams were located both up-
stream and downstream of the bridge for effective visual
monitoring using a thermal imager. The earth dams were
constructed using an excavator to block the creek. The
distance between the two dams was 100 m. The dam sizes
are listed in Table 1. Field tests were conducted on Sep-
tember 26, October 16, and November 7, 2012, as well as
on April 10, 2014. The dam tested on September 26, 2012
was constructed on September 21, 5 days before the test.
The others were constructed on the test date. The con-
structed dams were unstable (Yang et al. 2015). Thermal
imaging began following formation of the earth dams and
continued through to the release of water from upstream
until breach. The stream discharge was sourced from an
irrigation ditch upstream with a controllable gate.
The single-dam breach test was designed to explore
modes of failure caused by upstream impoundment. The
two subsequent tests monitored changes in the failure
mode of downstream dams under flash floods induced
by breaching of an upstream dam. Dam materials were
sourced from the riverbed. The dam surface was lined
with several grids for monitoring. The riverbed sedi-
ments are cohesionless soils with the following particle
size distribution: 10th percentile diameter d10 = 0.9 mm,
median diameter d50 = 30 mm, and 90th percentile di-
ameter d90 = 95 mm (Chen et al. 2015b).
An NEC G100D infrared thermal imager was used for
analysis. Its measurable range is from − 4 to 500 °C with
precision to 0.01 °C. The visible spectrum was imaged at
the same time as the thermogram for comparison.
Postprocessing software was used for thermogram anal-
ysis and image fusion in dual mode. Zonal temperature
was analyzed to indicate the average, maximum, and
minimum regional temperatures in the images. The mon-
itoring index T/Δt for changes in temperature (T) during
the time period (Δt) was used to monitor potential piping
Page 7 of 17 483
areas of the earth dams (Chen et al. 2016; Chen and Liu
2017). The calculation is performed by first subtracting
the temperature on the thermograph at a given time (T1)
from the temperature at the next monitoring time (T2) to
obtain T; subsequently, the monitoring index is T (T2 −
T1) divided by the time period Δt (t2 − t1).
Thermograph characteristics of earth materials
Surficial thermal images of earth materials were cap-
tured to enhance the characteristics of earth materials
with different particle sizes in dry and wet conditions.
Coarse aggregate more than 4.75 mm in diameter and
fine aggregate between 150 μm and 4.75 mm in diam-
eter were mixed with soil and clay. Potted earth mate-
rials were exposed to sunlight, and thermal images were
captured every 30 min to observe the temperature
changes over time (Fig. 2).
The fine aggregate exhibited the highest surface-
temperature changes, and wet soils exhibited the lowest
surface-temperature changes. This was attributed to the
small void ratio and large surface area of the fine aggregate
after exposure to sunlight. Wet soils have higher water
content than do dry soils and show a lower surface tem-
perature because of the existence of water. Figure 2 shows
the surface-temperature changes and environmental-
temperature changes for the captured images. Surface-
temperature changes over time and the monitoring index,
T/Δt, could be used for long-distance monitoring to dis-
tinguish wet soils from dry soils by water content.
A preliminary test of earth material was conducted
after the construction of a downstream dam was com-
pleted on September 21, 2012. The different surface
temperatures of dry and wet soils and ponding water
exhibited different thermal characteristics in the thermo-
graph (Fig. 3). Ponding was caused by the lower topo-
graphic characteristics on the left side of the dam. The
surface temperature on the left side of the dam was low
because of the high water content.
Results
Single earth dam (September 26, 2012)
The earth dam was constructed on September 21, 2012
downstream from the bridge for clear visual observa-
tion. The thermal imager was implemented on the
bridge and moved to both sides of the dam to capture
483 Page 8 of 17 Environ Monit Assess (2018) 190:483

(a) (b)

(c) (d)

(e) (f)

Fig. 6 Breach process of the downstream dam on October 16, area downstream. e, f At 11:12, the dam breached on the initial
2012. a, b Initial state of the upstream face at 9:50; the arrows ponding side; the arrow highlights the dam discharge overtopping
highlight the lower temperature zones. c, d At 11:07, piping breached area
occurred at the downstream face; the arrow highlights the piping

the images successively. Upstream discharge was re- corresponding to the initial ponding side (Fig. 4a, b).
leased 5 days after the dam was constructed, and the The upstream face showed a higher surface temperature
test started at 08:21 on September 26, 2012. Piping and than the downstream face because seepage occurred
subsequent erosion occurred at the downstream face downstream.
Environ Monit Assess (2018) 190:483 Page 9 of 17 483

(a) (b)

(c) (d)

(e) (f)

(g) (h)
 483 Page 10 of 17
R Fig. 7 Breach process of the upstream dam on November 7, 2012.
a, b At 10:03, ponding occurred in front of the dam; the arrow
highlights the ponding area. c, d At 11:27, base seepage occurred
downstream; the arrow highlights the base seepage area down-
stream. e, f At 11:31, the dam breached on the initial ponding side;
the arrow highlights the discharge overtopping area. g, h At 11:58,
the breached dam reached its final state; the arrow highlights the
dam breached area
With the increase in discharge, the amount of seepage
water increased, and the erosion area increased. At
09:55, discharge overtopped the dam in the middle and
on the initial ponding side (Fig. 4c, d). The water level
rose faster than the wetting front of the dam material;
with the increased water level, overtopping occurred,
and the dam was breached through progressive erosion.
The dam was breached in the middle and on the initial
ponding side at 09:57 (Fig. 4e, f).
The dam withstood 96 min of upstream discharge
from 08:21 until the breach at 09:57. The dam failed in
the overtopping mode with base seepage and piping.
Two earth dams (October 16, 2012)
In the next test, two dams were tested immediately after
construction on October 16, 2012. The test started at
10:04 following the release of upstream discharge.
Ponding water soon formed in front of the upstream
dam (Fig. 5a, b). At 10:52, the ponding area started to
breach the dam with the increasing water level (Fig. 5c,
d). The water flow overtopped the dam at 11:00 (Fig. 5e,
f), and at 11:03, the dam was breached on the initial
ponding side and in the middle (Fig. 5g, h). The dam
endured for 59 min from 10:04 until 11:03.
Ponding water was present in front of the downstream
dam, and the thermograph showed lower temperatures in
zones 1 and 3 (Fig. 6a, b). The downstream ponding level
increased quickly at 11:04 upon the breach of the upstream
dam. At 11:07, piping occurred at zone 3 of the down-
stream face (Fig. 6c, d). The dam failed in the overtopping
mode on the initial ponding side of the dam and was
breached at 11:12 (Fig. 6c). The downstream dam lasted
for 68 min from the release of upstream water at 10:04
until it was breached at 11:12 (Fig. 6e, f).
Two earth dams (November 7, 2012)
The third test was similar to the second test (October
16) with two earth dams constructed. Ponding oc-
curred in front of the upstream dam at 10:03
Environ Monit Assess (2018) 190:483
(Fig. 7a, b). The surface temperature was lower on
the ponding side than on the other side of the dam
top. Upstream water was released at 11:20, and base
seepage was shown downstream at 11:27 (Fig. 7c,
d). At 11:31, the dam was breached on the initial
ponding side (Fig. 7e, f). The upstream dam failed
in the overtopping mode after only 11 min.
Figure 7g, h shows the final state of the breached
dam at 11:58.
Following the breach of the upstream dam at 11:31,
the ponding water level increased at 11:34 at the down-
stream dam (Fig. 8a, b). One minute later at 11:35, base
seepage occurred downstream (Fig. 8c, d). At 11:36,
piping and progressive erosion occurred downstream
(Fig. 8e, f). At 11:38, creek discharge overtopped the
dam, and it was breached shortly after at 11:41 in the
middle (Fig. 8g, h). The dam was breached in the
overtopping mode.
The downstream dam was breached 21 min after
upstream discharge was released at 11:20 and only
10 min after parts of the upstream dam were breached
at 11:31.
Discussion
Dam breach process
The dam breach processes during the three tests are
summarized in Table 2. Statistics indicate that 89% of
landslide dams fail by overtopping, 10% by piping, and
only 1% by other modes (Costa and Schuster 1988). The
five earth dams in the three field tests all failed in the
overtopping mode. Phases of the dam breach process
were observed. At the beginning of the test, ponding
first occurred at the lowland of the riverbed upstream,
and base seepage occurred at the base of the dam down-
stream. Shortly after, overtopping failure occurred.
The topographic characteristics of the riverbed
caused ponding at the dam upstream. The piping down-
stream was attributed to the lowland ponding upstream,
as indicated in Fig. 3. The ponding side of the dam
upstream in zone 1 (Fig. 3d) and at the top of the dam
(Fig. 4a) showed a low temperature on the thermo-
graphs. The lower temperature indicates higher soil
water content on the initial ponding side. The dam was
shown to be breached first where ponding occurred on
September 26, 2012. Similarly, during the tests on Oc-
tober 16, 2012, the dam was first breached on the
Environ Monit Assess (2018) 190:483 Page 11 of 17 483

(a) (b)

(c) (d)

(e) (f)

(g) (h)
 483 Page 12 of 17 Environ Monit Assess (2018) 190:483

R Fig. 8 Breach process of the downstream dam on November 7,
2012. a, b At 11:34, the ponding water increased upon the up-
stream dam being breached; the arrows highlight the lower tem-
perature zones. c, d At 11:35, base seepage occurred downstream;
the arrows highlight the base seepage area downstream. e, f At
11:36, piping and progressive erosion occurred downstream. g, h
At 11:41, the dam was breached in the middle; the arrows high-
light the dam piping downstream and breached areas
ponding side of the upstream dam (Fig. 5a) and on the
ponding side and piping area of the downstream dam
(Fig. 6a). The test on November 7, 2012 exhibited the
same failure mode at the upstream dam (Fig. 7a). These
results show that the failure mode of earth dams is
affected by the topographic characteristics of the river-
bed; lowland areas are vulnerable to ponding upstream,
which causes piping downstream and subsequent
breaches.
Soil erodibility is a critical parameter of the breach
process (Zhong et al. 2018). The breach processes of the
dams were all similar in the tests. Base seepage occurred
quickly with piping and subsequent progressive erosion,
leading to dam breaching or overtopping. However, the
breach locations were different. Most of the dams were
breached at one location, but some failed at two loca-
tions. In the test on September 26, 2012, the upstream
dam was breached in the middle and on the initial
ponding side of the dam. Others were breached only
on the initial ponding side. In the November 7 test, the
downstream dam was breached in the middle. This was
attributed to the fact that it was constructed on an
existing check dam (Fig. 5a), and the soil–concrete
interface was prone to separation after seepage of
ponding water. Additionally, the creek width was con-
fined by the check dam, and leading water flow was
concentrated to the creek center. This dam was breached
first in the main channel and where ponding occurred.
Piping of two types is common for earth dams: piping
through the body (internal erosion) and piping through
the foundation (base seepage) (Omofunmi et al. 2017).
Internal erosion failures are most common in areas
where hydraulic fracturing can occur (Flores-Berrones
et al. 2011). Base seepage on September 26, 2012 and
on November 7, 2012 was captured by the thermo-
graphs. The downstream dam test on November 7,
2012 showed a low temperature on thermographs at
the dam base downstream (Fig. 8a). The tests show that
low temperatures at the dam surface indicate vulnerabil-
ity to seepage downstream and piping, which could lead
to dam breaching.
Time parameters are critical to numerical models of
dam breaches (Wahl 2004; Froehlich 2008). In an anal-
ysis of 204 landslide dams, 87% were shown to be
breached in less than 1 year, 83% in less than 6 months,
71% in less than 1 month, 51% in less than 1 week, and
34% in less than 1 day (Peng and Zhang 2012). The
upstream dam on November 7 was breached after only
11 min, which is far shorter than the time for the dam
tested on October 16 (59 min). This is attributed to the

Table 2 Comparisons of the dam breach processes

Date Location Lifetime Failure mode /breach Breach process Thermograph changes
(min) area

26 Upstream – – – –
September, Downstream 96 Overtopping/initial Base seepage and piping, Lower temperature on ponding
2012a ponding side then overtopping failure side of the dam (Figs. 3d and 4b)
16 October, Upstream 59 Overtopping/initial Dam was breached through Lower temperature on ponding
2012 ponding side progressive erosion on the side of the dam (Fig. 5b)
and middle initial ponding side first
and failed through overtopping
Downstream 68 (8b) Overtopping/initial Piping first then breach on Lower temperature on ponding
ponding side the initial ponding side of the side and in piping areas of the
dam dam (Fig. 6b)
7 November, Upstream 11 Overtopping/initial Base seepage, then overtopping Lower temperature on ponding
2012 ponding side failure side of the dam (Fig. 7b)
Downstream 21 (10b) Overtopping/middle Base seepage and piping, then Lower temperature at the dam base
overtopping failure downstream (Fig. 8b)
a
Test was 5 days after construction
b
Lifetime after the upstream dam breach
Environ Monit Assess (2018) 190:483 Page 13 of 17 483

(a) (b)

1 2
4 5 3
6 7 8
12 9 10 11
13 14 15 16
20 21 17 18 19
22 23
24 25 26 27

(d)
(c)

(f)
(e)

(h)
(g)
 483 Page 14 of 17
R Fig. 9 Dam breach process during the test on April 10, 2014. (a,
b) at 14:18, the earth dam was in its initial state. (c, d) At 14:19,
base seepage occurred; the arrow highlights the base seepage area
downstream. (e, f) At 14:33, piping occurred; the arrows highlight
the piping areas downstream. (g, h) At 14:35, the dam was
breached in the middle
different discharge rate; it took 1 h to overtop the dam on
October 16 and only 10 min on November 7. Water
discharge and flow rate from upstream affect the dam
breach process. A small discharge in a short time from
upstream is prone to causing piping and seepage; a large
discharge at a high rate causes ponding water to rise
faster and water to overtop the dam before the wetting
front of the phreatic line reaches the slope face down-
stream. The area overtopped is first breached through
progressive erosion.
The existence of an upstream dam can affect the
failure process of a downstream dam. The down-
stream dam was breached after 21 min on November
7, which is far shorter than the times for the down-
stream dams on October 16 (68 min) and September
26 (96 min). The downstream dam was breached
only 8 min after parts of the upstream dam were
breached on October 16 and only 10 min after the
upstream breach on November 7. The flash flood
Fig. 10 Surface-temperature
changes on thermographs for the
test on April 10, 2014
Environ Monit Assess (2018) 190:483
induced by the breach of the upstream dam reduced
the time to breach the downstream dam.
Validation of the monitoring index
An earth dam was constructed downstream on April 10,
2014 and tested for validation of the monitoring index
(T/Δt). The dam surface downstream was divided into
27 grids for monitoring (Fig. 9a, b). The upstream water
was released at 14:00, and base seepage first occurred at
14:19 (Fig. 9c, d). Dam surface temperatures dropped
abruptly on the thermograph at 14:33 when piping oc-
curred at grids 7, 9, 15, 17, and 25 in the middle of the
dam (Fig. 9e, f). At 14:35, the dam was overtopped and
breached in the middle (Fig. 9g, h).
The thermographs indicated lower temperatures for
grids 7, 9, 15, 17, and 25 than for others at 14:33, and
they subsequently extended into the progressive erosion
zone at 14:34 (Fig. 10). However, the data shown were
not sensitive to the rate of temperature change. For
continual abrupt changes in temperature, an index to
provide the rate of temperature changes could be useful
for monitoring purposes.
The index T/Δt was used for dam breach process
monitoring. The index was high at 14:21 but
Environ Monit Assess (2018) 190:483
Fig. 11 Variations of the
monitoring index (T/Δt) for the
test on April 10, 2014
suddenly dropped at 14:23 because of rising
ponding water (Fig. 11). No notable changes were
observed until 14:28 and 14:31 when the index
abruptly changed for grids 1 and 25, respectively.
Finally, at 14:33, grid piping occurred and abruptly
reduced the index for grids 7, 9, 15, 17, and 25. The
results showed favorable agreement with the dam
breach process (Fig. 9), and the index can accurately
reflect the trend of temperature changes over time.
The applications of thermal imaging for dam failure
mode monitoring could be limited to sparsely vege-
tated soil surfaces above the water table. Thermal
imaging cannot monitor the internal erosion process-
es of a dam. This would be useful for evaluating the
stability of natural (landslide) dams.
Conclusions
A noncontact, nondestructive, long-distance method
of monitoring the earth-dam breach process using
thermal imaging was tested. In four real-scale breach
tests, the dam breach process started with ponding at
the lowland of the riverbed upstream; piping and
base seepage occurred downstream; and the dam
soon failed by overtopping. The overtopping failure
mode was observed in the main channel location and
on the initial ponding side first. The topographic
Page 15 of 17 483
characteristics of the stream bed, the initial water
content of the soil, and the discharge conditions
affected the failure mode and breach process of the
earth dams in field tests.
Areas with high soil water content display lower
temperatures and could be prone to piping and pro-
gressive seepage. With the aid of thermographs, a
monitoring index (T/Δt) is proposed for monitoring
surface-temperature changes over time. The moni-
toring index T/Δt decreased at grids where soil
piping occurred. This methodology should be ap-
plied to safety monitoring of natural-dam stability
and failure mode prediction for dam breach and
disaster mitigation.
Funding information Parts of this work were supported by the
Ministry of Science and Technology in Taiwan, under contracts
NSC 99-2625-M-415-003-MY3 and MOST 106-2625-M-415-
003.
References
Barla, G., Antolini, F., & Gigli, G. (2016). 3D laser scanner and
thermography for tunnel discontinuity mapping.
Geomechanics and Tunnelling, 9, 29–36.
Baroň, I., Bečkovský, D., & Míča, I. L. (2012). Application of
infrared thermography for mapping open fractures in deep-
seated rockslides and unstable cliffs. Landslides, 36, 265–275.
Bersan, S., Koelewijn, A. R., & Simonini, P. (2018). Effectiveness
of distributed temperature measurements for early detection
483 Page 16 of 17
of piping in river embankments. Hydrology and Earth System
Sciences, 22, 1491–1508.
Boglietti, A., Cavagnino, A., Staton, D., Shanel, M., Mueller, M.,
& Mejuto, C. (2009). Evolution and modern approaches for
thermal analysis of electrical machines. IEEE Transactions
on Industrial Electronics, 56, 871–882.
Chai, H. J., Liu, H. C., Zhang, Z. Y., & Xu, Z. W. (2000). The
distribution, causes and effects of damming landslides in
China. Journal of Chengdu University of Technology, 27,
302–307.
Chen, S. C. (1999). Failure mechanism and disaster mitigation on
landslide dammed lakes. Journal of Chinese Soil and Water
Conservation, 30, 299–312 (in Chinese with English
abstract).
Chen, G. H., & Chen, C. Y. (2014). Applications of thermal
images for highway slopeland monitoring and analysis.
Taiwan Highway Engineering, 40(4), 25–57 (in Chinese).
Chen, S. C., & Hsu, C. L. (2009). Landslide dams induced by
typhoon Morakot and its risk assessment. Sino-Geotechnics,
122, 77–86 (in Chinese with English abstract).
Chen, C. Y., & Liu, Z. H. (2017). Applications of thermal images
for monitoring surficial temperature changes of naked slope.
Proceedings of Engineering Technology and Innovation, 5,
37–40.
Chen, S. C., An, S. P., Hsu, T. Y., & Wang, C. (2015a). High
precision and adjusted discharge sediment, the experimental
station in Landao Creek, Huisun Forest. Journal of Chinese
Soil and Water Conservation, 46, 1–6 (in Chinese with
English abstract).
Chen, S. C., Lin, T. W., & Chen, C. Y. (2015b). Modeling of
natural dam failure modes and downstream riverbed morpho-
logical changes with different dam materials in a flume test.
Engineering Geology, 188, 148–188.
Chen, C. Y., Chen, S. C., & Chen, K. H. (2016). Earth dam
monitoring by using infrared thermography detection.
Proceedings of Engineering Technology and Innovation, 4,
40–42.
Costa, J. E., & Schuster, R. L. (1988). The formation and failure of
natural dams. Geological Society of America Bulletin,
100(7), 1054–1068.
Cuong, B. Q., Thai, N. C., Yihong, Z., & Chunju, Z. (2017).
Estimating seepage in embankment dams based on tempera-
ture measurement: a review paper. International Journal of
Engineering Research and Technology, 6, 106–113.
Feng, Z. Y. (2012). The seismic signatures of the surge wave from
the 2009 Xiaolin landslide-dam breach in Taiwan.
Hydrological Processes, 26, 1342–1351.
Flores-Berrones, R., Ramirez-Reynaga, M., & Macari, E. J.
(2011). Internal erosion and rehabilitation of an earth dam.
Journal of Geotechnical and Geoenvironmental
Engineering, 137(2), 150–160.
Forest Bureau (2009). Council of Agriculture, Executive Yuan,
Luodong Forest District Office, http://luodong.forest.gov.
tw/public/. (Accessed 27 October 2017).
Frodella, W., Morelli, S., Gigli, G., & Casagli, N. (2014). Contribution
of infrared thermography to the slope instability characterization.
Proceedings of World Landslide Forum, 3, 2–6.
Froehlich, D. C. (2008). Embankment dam breach parameters and
their uncertainties. Journal of Hydraulic Engineering,
134(12), 1708–1721.
Environ Monit Assess (2018) 190:483
Gigli, G., Frodella, W., Garfagnoli, F., Morelli, S., Mugnai, F.,
Menna, F., & Casagli, N. (2014). 3-D geomechanical rock
mass characterization for the evaluation of rockslide suscep-
tibility scenarios. Landslides, 11, 131–140.
Janssens, O., Van de Walle, R., Loccufier, M., & Van Hoecke, S.
(2017). Deep learning for infrared thermal image based ma-
chine health monitoring. IEEE/ASME Transactions on
Mechatronics, 23(1), 151–159.
Korup, O. (2004). Geomorphometric characteristics of New
Zealand landslide dams. Engineering Geology, 73, 13–35.
Korup, O. (2005). Geomorphic hazard assessment of landslide
dams in South Westland, New Zealand: Fundamental prob-
lems and approaches. Geomorphology, 66, 167–188.
Lai, S. L., Lee, D. H., Wu, J. H., & Dong, Y. M. (2014). Detecting
the cracks and seepage line associated with an earthquake in
an earth dam using the nondestructive testing technologies.
Journal of the Chinese Institute of Engineers, Transactions of
the Chinese Institute of Engineers, Series A, 37(4), 428–437.
Lin, C. P., Hung, Y. C., Yu, Z. H., & Wu, P. L. (2013).
Investigation of abnormal seepages in an earth dam using
resistivity tomography. Journal of GeoEngineering, 8(2),
61–70.
Lo, C. M., Lin, M. L., Tang, C. L., & Hu, J. C. (2011). A kinematic
model of the Hsiaolin (a.k.a. Shaolin) landslide calibrated to
the morphology of the landslide deposit. Engineering
Geology, 123, 22–39.
Lum, K. Y., & Sheffer, M. R. (2005). Dam safety: review of
geophysical methods to detect seepage and internal erosion
in embankment dams. Tulsa, OK: PennWell Corporation.
Martino, S., & Mazzanti, P. (2014). Integrating geomechanical
surveys and remote sensing for sea cliff slope stability anal-
ysis: the Mt. Pucci case study (Italy). Natural Hazards and
Earth System Sciences, 14, 831–848.
Mineo, S., Pappalardo, G., Rapisarda, F., Cubito, A., & Di Maria,
G. (2015). Integrated geostructural, seismic and infrared ther-
mography surveys for the study of an unstable rock slope in
the Peloritani chain (NE Sicily). Engineering Geology, 195,
225–235.
Moffat, R., Fannin, R. J., & Garner, S. J. (2011). Spatial and
temporal progression of internal erosion in cohesionless soil.
Canadian Geotechnical Journal, 48, 399–412.
Omofunmi, O. E., Kolo, J. G., Oladipo, A. S., Diabana, P. D., &
Ojo, A. S. (2017). A review on effects and control of seepage
through earth-fill dam. Current Journal of Applied Science
and Technology, 22(5), 1–11.
Peng, M., & Zhang, L. M. (2012). Breaching parameters of
landslide dams. Landslides, 9(1), 13–31.
Planck, M. (1901). On the law of distribution of energy in the
normal spectrum. Annalen der Physik, 4, 553–559.
Planes, T., Mooney, M. A., Rittgers, J. B. R., Parekh, M. L., Behm,
M., & Snieder, R. (2016). Time-lapse monitoring of internal
erosion in earthen dams and levees using ambient seismic
noise. Géotechnique, 66, 301–312.
Polemio, M., & Lollino, P. (2011). Failure of infrastructure em-
bankments induced by flooding and seepage: a neglected
source of hazard. Natural Hazards and Earth System
Sciences, 11, 3383–3396.
Schuster, R. L. (1986). Landslide dams: processes, risk, and
mitigation. In Geotechnical Special Publication. New
York: American Society of Civil Engineers.
Environ Monit Assess (2018) 190:483
Teza, G., Marcato, G., Pasuto, A., & Galgaro, A. (2015).
Integration of laser scanning and thermal imaging in moni-
toring optimization and assessment of rockfall hazard: a case
history in the Carnic Alps (northeastern Italy). Natural
Hazards, 76, 1535–1549.
Wahl, T. L. (2004). Uncertainty of predictions of embankment
dam breach parameters. Journal of Hydraulic Engineering,
130, 389–397.
Wu, J. H., Lin, H. M., Lee, D. H., & Fang, S. C. (2005). Integrity
assessment of rock mass behind the shotcreted slope using
thermography. Engineering Geology, 80, 164–173.
Page 17 of 17 483
Wu, C. H., Chen, S. C., & Feng, Z. Y. (2014). Formation, failure,
and consequences of the Xiaolin (a.k.a Shaolin) landslide
dam, triggered by extreme rainfall from typhoon Morakot,
Taiwan. Landslides, 11, 357–367.
Yang, M. D., Chen, W. T., & Huang, K. S. (2015). Establishment
of a 3D on-site model of landslide dam using UAV images.
Journal of Chinese Soil and Water Conservation, 46, 88–95
(in Chinese with English abstract).
Zhong, Q. M., Chen, S. S., Mei, S. A., & Cao, W. (2018).
Numerical simulation of landslide dam breaching due to
overtopping. Landslides, 15(6), 1183–1192.