Bull Volcanol (2014) 76:857

DOI 10.1007/s00445-014-0857-z

RESEARCH ARTICLE

Erosion and aggradation on persistently active

volcanoes—a case study from Semeru Volcano, Indonesia
Jean-Claude Thouret & Jean-François Oehler &
Avijit Gupta & Akhmad Solikhin & Jonathan N. Procter

Received: 14 March 2014 / Accepted: 3 August 2014 / Published online: 9 September 2014
# Springer-Verlag Berlin Heidelberg 2014

Abstract Erosion processes on active volcanoes in humid The geomorphic response of two of Semeru’s rivers to volca-
climates result in some of the highest sediment yields on nic sediment migration indicates that (1) each river experi-
Earth. Episodic sediment yields after large eruptions have ences alternating aggradation and degradation cycles follow-
been evaluated, but not the long-term and continuous patterns ing PDC-producing eruptions and (2) spatial patterns of sed-
on persistently active volcanoes. We have used high-spatial iment transfer are governed by geomorphic characteristics of
resolution satellite imagery and DEMs/DSMs along with the river reaches. Usually high degradation in the steep source
field-based geologic mapping to assess accurately sediment reach is followed by a long bypassing middle reach. Aggra-
budgets for the active Semeru Volcano in Java, Indonesia. dation predominates in the depositional reaches further down
Patterns of aggradation and degradation on Semeru differ valley on the ring plain. Average sediment yields (103–105 t/
from that of other active volcanoes because (1) both episodic km2/year) at persistently active volcanoes are two to three
pyroclastic density currents (PDC) and continuous supplies of orders of magnitude lower than sediment yields after large
tephra generate pulses of sediment, (2) sediment is transferred and infrequent eruptions, but the continuous and steady sed-
via cycles of aggradation and degradation that continue for iment transfer in rivers removes more sediment on a mid-term
>15 years in river channels after each PDC-producing erup- (10 years) to long-term (30 years) basis. In contrast to the
tion, and (3) rain-triggered lahars remove much greater mate- trend observed on composite cones after large and infrequent
rial than fluvial transport during long, intense rainfall events. eruptions, decay of sediment yields is not exponential and

Editorial responsibility: V. Manville

Electronic supplementary material The online version of this article
(doi:10.1007/s00445-014-0857-z) contains supplementary material,
which is available to authorized users.
J.<C. Thouret (*) : A. Solikhin J. N. Procter
Laboratoire Magmas et Volcans, CNRS UMR6524 and IRD-R163, Volcanic Risk Solution, Institute of Natural Resources, Massey
Clermont Université, Université Blaise Pascal, 5 rue Kessler, University, Palmerston North, New Zealand
F-63000 Clermont-Ferrand, France e-mail: [email protected]
e-mail: [email protected]
A. Solikhin
e-mail: [email protected]
A. Solikhin
J.<F. Oehler Center of Volcanology and Geological Hazard Mitigation
ALTRAN OUEST-ATLANTIDE, Technopôle Brest Iroise, Site du (CVGHM), 57 Jalan Diponogero, Bandung, Indonesia
Vernis – CS 23866, F-29238 Brest Cedex 03, France
e-mail: [email protected]

A. Gupta
School of Earth and Environmental Sciences, University of A. Gupta
Wollongong, Wollongong, Australia Centre for Remote Imaging, Sensing and Processing, National
e-mail: [email protected] University of Singapore, Kent Ridge Road, Singapore
857, Page 2 of 26 Bull Volcanol (2014) 76:857

river channels do not fully recover at steadily active volcanoes but steady sediment from tephra, augmented by episodic
as episodic inputs from BAF eruptions, superimposed on the pulses.
background remobilization of daily tephra, have a greater
cumulative effect. Category 1 Large eruptions result in rapid and major re-
sponses in channel style and sediment yield to
Keywords Erosion . Aggradation . Pyroclastics . Lahar . sudden arrival of voluminous inputs of sedi-
Digital elevation/surface models . Semeru ment. Annual suspended sediment yields fol-
lowing the catastrophic 1980 Mount St. Helens
eruption were as much as 500 times greater than
typical background levels in tephra, PDCs, and
Introduction: scope of the study debris avalanche impacted catchments and de-
clined nonlinearly for more than a decade. With-
Volcanic basins with annual sediment yields of 104 to 107 t/ in only 5 years of the eruption, annual yields
km2 are among the highest sediment producers on Earth plateaued, and average yields remained about
(Major et al. 2000; Gran and Montgomery 2005). Harris ten times (103 t/km2) the background level for
et al. (2006) have distinguished two types of volcanic fluvial 20 years (Major et al. 2000). Gran and
regimes based on (1) transient and (2) persistent variations in Montgomery (2005) reported even higher sedi-
sediment flux. Fluvial response to transient explosive volcanic ment yields from Pinatubo than those from Mount
activity has been documented after the 1980 eruption of Mt. St. Helens, which lasted for almost two decades.
St. Helens (Major 2004; Major and Mark 2006) and the large After the 1991 Mount Pinatubo eruption, very
1991 eruption of Pinatubo (Pierson et al. 1992; Scott et al. high sediment yields (106–107 t/km2/year) oc-
1996; Montgomery et al. 1999; Hayes et al. 2002; Gran and curred in volcanically disturbed rivers due to
Montgomery 2005). The chronology and patterns of exponen- basin-wide sediment loading (Pierson et al.
tial declines in sediment flux and fluvial adjustment have been 1992). Even after major lahars ended in 2004,
examined after the 1902 catastrophic eruption of Santa María sediment yields in some basins remained orders
Volcano (Kuenzi et al. 1979). Readjustment of landscape after of magnitude above the pre-eruption levels.
the large 1.8-ka Taupo eruption is a similar case study Category 2 Moderate-sized eruptions induce short-term re-
(Manville et al. 2009). Erosion, lahar activity, and fluvial sponses, as illustrated by three volcanoes:
transport following smaller eruptions have been noted from Unzen and Mt Yakedake in Japan and Merapi
Japanese volcanoes: Usu, Unzen, and Sakurajima (Suwa and in Indonesia (Suwa and Yamakoshi 1999). The
Yamakoshi 1999; Mizuyama and Kobashi 1996). Their long- specific sediment discharge measured by these
term impacts on rivers and sedimentation, however, are diffi- authors is within the range of published sedi-
cult to assess (Pierson et al. 2011). Long-term monitoring of ment yields of 104–106 t/km2/year for catchment
rain-triggered lahars at Mt. Yakedake in Japan has enabled areas of 101 to 103 km2 on similar volcanoes
Okano et al. (2012) to relate debris flows to rainfall intensity (Lavigne 2004; Gran & Montgomery 2005). In
thresholds. For persistently active volcanoes (Semeru, contrast to large eruptions, moderate ones give
Sakurajima, Santiaguito), however, studies on production of rise to fast increase and decrease in debris flow
sediment flux have focused only on short-term variation pat- frequency. The erosion rate increased abruptly
terns in sedimentation (Lavigne 2004; Lavigne and Thouret to 101 and 102 mm/year after the eruptions and
2002; Harris et al. 2006). On these continuously explosive decreased exponentially to 101 mm/year in less
volcanoes, tephra fall and pyroclastic density currents (PDCs, than 10 years at Merapi and to less than 10 mm/
including pyroclastic flows and surges) deliver a nearly con- year at Yakedake in 32 years (Suwa et al. 2011).
tinuous supply of sediment via lahars. The rapid increase may be due to the accelerated
runoff on the indurated top surface of pyroclastic
Eruptive activity, sediment input, and fluvial system response deposits and bare slopes. Quick decrease may be
promoted by declining runoff due to fragmenta-
Three categories of eruptive activity trigger disturbances in tion of soil crust, soil infiltration, and vegetation
catchments: (1) large but infrequent explosive eruptions pro- recovery, and depletion of the reservoir of early
viding voluminous sediment inputs from tephra fall, PDCs, remobilized material.
lahars, and occasional debris avalanches, (2) moderate-sized Category 3 Recurrent or persistent activities supply steady
explosive eruptions that provide lesser inputs as discrete but relatively small amounts of sediment to riv-
events (such as tephra fall, small PDCs, primary lahars), and ers, but their cumulative effects may be greater
(3) near-continuous eruptive activity that supplies moderate than from episodic large- or medium-sized
Bull Volcanol (2014) 76:857 Page 3 of 26, 857

eruptions. Debris from vulcanian explosions on Java, Indonesia, an excellent site for studying sediment stor-
the bare summit slopes of persistently active age and transfer due to persistent activity in a low-latitude
volcanoes are transported by overland flow. climate with a long rainy season (Figs. 1–4) (Thouret et al.
2007; Solikhin et al. 2012). Deposits of pyroclastic density
We have selected Semeru, one of the few persistently active currents (PDCs) in valleys, mostly block-and-ash flows
composite volcanoes worldwide, to study near-continuous (BAF), are transported downslope by lahars and streamflows,
sediment transfer in river valleys, motivated by two unsolved with headward extension and lateral erosion of the channel
problems: associated with landslides. Tephra deposited on slopes is
eroded by overland flow. Runoff contributes material from
1. The role of near-continuous tephra supply and runoff on the bare summit slopes, but this process is more limited under
lahar activity and channel response is not well understood. dense forest on the interfluves between c. 2,200 and 1,200 m
Waldron (1967) documented erosion of ash deposits from asl or on porous lava flows (Lavigne and Thouret 2002).
Irazu Volcano, Costa Rica between 1963 and 1965, with Lahars, which are the principal process for remobilizing sed-
emphasis on debris flow hazards. Correlations between iment at Semeru, have been analyzed using (1) physical
supplies of debris from a steady dome extrusion, rainfall, characteristics of deposits (Dumaisnil et al. 2010), (2) hydrau-
and lahar activities with the form and behavior of fluvial lic and rheological characteristics measured from in situ flows
systems have not been properly established except for in a river draining the SE flank (Doyle et al. 2010, 2011), and
Santa Maria-Santiaguito Volcano, Guatemala (Harris et al. (3) numerical simulation codes in which input parameters are
2006). The authors used a time series of 21 Landsat calibrated with field observations by means of rheological
Thematic Mapper images between 1987 and 2000 to tests (Dumaisnil et al. 2010). Lahars, which encompass vol-
map, measure, and track temporal and spatial variations canic debris flows and hyperconcentrated flows (Pierson and
in the area of lahar impact and river aggradation. Two Scott 1985; Smith and Lowe 1991) are two-phase flows with
zones of the fluvial system have been covered (Harris poorly known behavior. A lahar wave can vary in character
et al. 2006). In the proximal zone, a positive correlation with time and distance downstream and may comprise one or
has been found between extrusion rate at Santiaguito, more flow types (Vallance 2000). Among these types,
aggradation area 12 months after the eruption, and rainfall hyperconcentrated flows include a turbulent, gravity-driven
during the intervening 12 months. In the medial zone, the flow of water and sediment, typically intermediate in
correlation between extrusion rate and/or area of aggra- suspended sediment concentration between normal
dation is negative due to removal of material in valleys. sediment-laden streamflow and debris flow (Vallance 2005;
2. So far, only short-term sediment yields have been calculat- Pierson 2005).
ed on the flanks of Merapi and Semeru volcanoes, where We describe and map spatial changes in catchment caused
sediment is derived from three sources: PDC deposits, by, and river responses to, bedload-dominated sediment flux
rockfalls from the domes, and material from the riverbed in the Kobokan catchment (108.9 km2) and its tributary sub-
and banks. Lavigne (2004) used two different methods to catchment, the Curah Lengkong (8.23 km2). The Kobokan
estimate sediment yield dominated by rain-triggered lahars belongs to the Rejali basin (Figs. 1 and 2), one of the three
at Merapi and Semeru volcanoes. During the first year major watersheds on the south and east flanks of the volcano.
following the 22 November 1994 Merapi eruption, a sed- Figure 1 outlines in light gray the valleys flooded by lahars on
iment yield of 1·5×105 m3/km2/year was calculated in the the south and east ring plain of Semeru during the twentieth
Boyong River basin based on the volume of sediment century (Thouret et al. 2007). The south and SE drainages, fed
trapped behind check dams. At Semeru, an estimated by the Jonggring-Seloko vent and scar, often carry lahars for
sediment yield of 4×105 t/km2/year was measured for the more than 35 km to the ring plain. We identify three temporal
year 2000 in the Curah Lengkong River. According to patterns of the drainage response to the abundant sediment
Lavigne and Suwa (2004), the volume of suspended ma- load (Fig. 3): (1) the long-term geomorphic changes to the
terial transported in 2000 by 18 debris flows (1–5×105 m3 catchment, (2) the mid-term, decadal cycle of aggradation and
per event) and five hyperconcentrated flows (103–105 m3 degradation as after the 1994–1995 BAF eruption, and (3) the
per event) was much larger than the volume transported by short-term, annual topographic change, illustrated by the
11 streamflow events (103–104 m3 per event). Lengkong 2-km-long bypassing reach. We also examine two
important relationships: (1) aggradation and degradation rates
in rivers with sediment derived from summit tephra, persistent
Study objectives lahars, and intensity duration of rainfall and (2) sediment
supply to rivers with the magnitude and frequency of both
This paper analyzes aggradation and degradation processes at continuous and episodic explosive regimes. Three fundamen-
Semeru Volcano (3,676 m, 8°07′ S and 112°55′ E) in east tal queries underlie these relationships:
857, Page 4 of 26 Bull Volcanol (2014) 76:857

A B

Fig. 1 A map showing the valleys flooded by lahars on the south and Indonesian archipelago, C location of Semeru volcano in eastern Java,
east ring plain of the Semeru volcano during the twentieth century (after and D the Tengger-Semeru volcanic massif. Rectangle in D indicates the
Thouret et al. (2007)). Inset sketch maps: B geodynamic setting of the extent of A

& Do persistently active volcanoes provide more sediment since 1967 by the Volcanological Survey of Indonesia (VSI),
over the long term (>10 years) than large but infrequent now Centre of Volcanology and Geologic Hazard Mitigation
eruptions? (CVGHM). On average, Semeru has produced tephra falls
& Do large eruptions lead to a more efficient fluvial adjust- daily and PDCs every 5 years since 1884 (Thouret et al.
ment than a constant sediment supply from moderate but 2007, Table 1 and references therein). Two types of erupted
persistent eruptive activity? products and sediment supply can be determined from the
& To what temporal extent do episodic PDC-producing available literature and our work on Semeru: (1) tephra fall
eruptions favor sediment yield with respect to constant deposits, which are coarse blocks and bombs of andesite
tephra supply on persistently active volcanoes? around the summit, whose size fines down to scoriaceous
lapilli and coarse ash in river channels. The thickness of tephra
is estimated to be <2 m on steep lava flow surfaces to the east
(from climber’s reports) but at least 10 m thick on the flattish
summit west of the vent, as shown by the SPOT5 image of
A persistently active volcano 2003, draping the digital elevation model (DEM) of Semeru’s
cone (Fig. 4 in Thouret et al. 2007). We estimate a thickness
Eruption chronology of Semeru and pyroclastic sediment range of 5–10 m for the loose tephra deposit, calculating a
supply maximum volume in the range of 12–23×106 m3 available for
lahars; (2) PDCs, mostly block-and-ash flow (BAF) deposits,
Semeru has been reportedly active at least since 1818 (Van which contain semi-vesicular and dense clasts with abundant
Bemmelen 1942). Its daily eruptive activity has been recorded scoriaceous lapilli and coarse ash. Over the 30-year-long
Bull Volcanol (2014) 76:857 Page 5 of 26, 857

Fig. 2 The Curah Lengkong and the Kobokan Rivers on Semeru’s SE between reach K2-L2 and K3-L3 marks overbank flows in 1994–1995
flank (Google Earth Pro 2010) with proximal source and medial reaches. across the interfluve between catchments. The GNSS and TLS survey
Reach boundaries are indicated (with scale). Note the SE-trending sum- area in reach L3 includes two stations for lahar measurements, rain
mit scar, the pattern of the Lengkong and the Kobokan tributaries, and the gauges, and other elements
location of the 1994–1995 BAF fronts in both channels. ‘Overbank’

study period (Fig. 3), episodes of sizeable PDCs (3–5× elevation, does not branch off from the scar. Consequently,
106 m3) occurred twice and larger volumes of 5.5–6.3× PDCs flow towards the Lengkong head valley only when high
106 m3 three times (after Table 1 in Thouret et al 2007). columns collapse for >1-km distance over the scar rim to-
The supply of pyroclastic debris is larger towards the main wards the east. Additionally, PDC debris can reach the lower
Kobokan Valley than the Lengkong sub-catchment. Tephra Lengkong valley on rare occasions, e.g., in 1994 and 1995,
falls occur daily on the summit slopes, on the heads of both when voluminous BAFs spilled over the narrow interfluve
Lengkong and Kobokan valleys, but are dispersed to longer between the Kobokan and Lengkong and travelled c. 9 km
distances towards the west and SW slopes than the east and from the summit to the SE (Figs. 1 and 2).
north slopes. The uppermost Lengkong sub-catchment is Figure 3 shows Semeru’s eruption chronology, pyroclastic
blanketed by tephra fall within 2 km of the summit when the deposits, and the study’s focus on three time intervals phased
dispersal lobe is oriented to the east or SE. Pyroclastic debris is over 30 years at three spatial scales: catchment, sub-catch-
also delivered by episodic PDCs channeled towards the south ment, and valley reach. First, the 1981–1990 time interval
valleys and the Kobokan headwall, which branches off at c. followed the 1981 eruption, which emplaced PDC deposits
2.85 km from the elongated and deep scar cutting down the with an approximate volume of c. 6.2 million m3 (references
Semeru’s SE flank (Fig. 1). In contrast, the Lengkong in Thouret et al. 2007, Table 1). Changes between 1981 and
headwall, which cuts the east summit slope at a higher 1990 have been identified from two sets of aerial photographs
857, Page 6 of 26 Bull Volcanol (2014) 76:857

Fig. 3 Eruption chronology,

PDCs, and tephra volumes (from
Table 1 in Thouret et al. (2007)
and references therein). The
volume of tephra supply has been
estimated at 12–23×106 m3 (see
text in “Data acquisition and
processing”), an estimated one
sixth of this is captured by the
Kobokan and the Lengkong
catchments. This figure also
shows selected time intervals,
spatial scales which narrow with
time, and tools and steps of data
processing for this study

at the scale of the Kobokan and Lengkong catchments. Sec- reaches are filled by BAF deposits. The basin extent affected
ond, the 1994 and 1995 eruptions emplaced c. 6.3 and 5.5 by tephra deposition is 6.2 % for the large Kobokan catchment
million m3 of BAF deposits, respectively, at least two thirds of and 29.4 % for the small Lengkong sub-catchment. Excep-
which filled both catchments under study (based on VSI maps tionally extensive ash deposits can mantle the proximal and
referred to in Thouret et al. 2007). We examined the fluvial medial catchments down to 1,500 m asl, i.e., approximately
readjustment following these eruptions by interpreting SPOT 12 % of the Kobokan and 64 % of the Lengkong catchments.
satellite images taken in 1996, 2001, and 2003, with an Hillslopes contributing to channels have a range of 10–25°
increase in resolution from 20 to 2.5 m. Third, we mapped and 9–10° on average, more than twice as much (>10°) for the
and estimated volumes of PDC and lahar deposits stored and/ proximal (source) reaches than for the middle (transport)
or bypassing the Kobokan and Lengkong Valley reaches reaches (5°). Below the uppermost bare slopes of the summit
based on 2001 SPOT4 and 2003 SPOT5 satellite images. between 2,400 and 1,200 m, the catchments are covered by a
Finally, we computed aggradation and degradation rates be- high, dense forest, while below 1,200–1,000 m, savanna
tween 2005 and 2011 in the Lengkong reach L3 and a 2-km- mixed with crops are widespread on the lower slopes and rice
long segment within L3 (Figs. 1–4) from low-altitude photo- terraces below 800 m asl. Fresh granular tephra and PDC
graphs and annual DEMs with sub-meter resolution (see deposits are relatively quickly recolonized by vegetation in
“Digital elevation and surface models”). the tropical monsoon climate below 2,400 m. The 1994–1995
coarse-grained BAF deposits were colonized by grass after
Characteristics of two catchments on the Semeru volcano 4 years and bush after 9 years, and scattered trees 3–5-m high
were found after 15 years.
Tables 1 and 2 display the characteristics of the two selected The catchment of the Kobokan can be divided into five
catchments. The Kobokan basin is 13 times larger and three reaches (Fig. 4; Table 1): the steep source reach K1; two
times longer than its steep Curah Lengkong tributary (Fig. 4). middle reaches, K2 & K3, with the gradient decreasing below
Table 1 reveals that approximately 50 % of the channel the SE flank’s break-in-slope; and two gently sloping
Bull Volcanol (2014) 76:857 Page 7 of 26, 857

Fig. 4 A Catchment of the Kobokan River and sub-catchment of the Kobokan (K) and the Lengkong (L) rivers per reach, whose areas and
Lengkong River on the Semeru TopoDEM of Semeru (after Solikhin slopes are displayed in the table at the upper right corner. C Hypsometric
et al. (2012)) including five reaches (proximal source 1, middle storage curves and hypsometric integral HI for the more “mature” Kobokan’s
and bypass 2 and 3, and deposition 4 and 5) and valley cross sections at catchment (0.14) and the steeper Lengkong’s sub-catchment (0.3)
specific locations. B Breakdown of the longitudinal profiles of the

deposition reaches, K4 & K5, including the low delta on the 2,000–2,100 m between the bare steep summit and the forest-
coast. The Lengkong sub-catchment is 12.61-km long be- ed flanks, then at ~1,200 m between the volcano flanks and
tween its source at 3,150 m asl and its confluence with the the volcaniclastic aprons, and finally at ~600 m between the
Kobokan at c. 700 m asl (Fig. 4; Tables 1 and 2). The aprons, hills, and low-relief plain surrounding the volcano.
Lengkong comprises three reaches: a narrow gully scouring Predominant geomorphic processes are the best method for
the steep volcano’s summit slope, the source of the debris identifying different reaches. Proximal reach 1 coincides with
source reach L1; a wider bypassing reach L2 across the daily debris supply and removal on the highly dissected, bare
relatively steep lower SE flank and below the break-in-slope uppermost slopes >24°. In the middle reach 2, lahars pass
at c. 1,000 m; and a narrow, lower bypassing reach L3 on the through a dense network of steep gullies on >7° of slope. In
moderately sloping volcaniclastic apron above the confluence. the middle reach 3, sediment waves are stored and/or bypass
A large fan with changing geometry since 1981 has formed at through less entrenched channels across slopes ~3°. The distal
the confluence of the two rivers. We use the word “bypassing” reaches 4 and 5 are fan-like depositional areas created by
after Manville et al. (2009; Fig. 12) to describe the middle wide, braided river floodplains. Usually, reaches upstream of
reaches, where sediment is stored temporarily between the reach 2 are dry outside the rainy seasons, whereas springs
upper erosional and lower depositional reaches before being related to aquifers feed perennial streams below 1,200–
transferred towards the confluence or the sea. 1,000 m between reaches 2 and 3.
Reach boundaries have been determined on the basis of Gravel-braided river channels typically occur with multiple
geomorphic and hydrological parameters (Figs. 2 and 4; Ta- terraces on volcanic slopes (Figs. 1, 2 and 4–6, mapped from
bles 1 and 2). Breaks in the Kobokan slope appear first at images and field observations). In both catchments, terrace T0
857, Page 8 of 26 Bull Volcanol (2014) 76:857

Table 1 Geometric characteristics of the catchments and reaches of the Kobokan and the Lengkong rivers. Breakdown by five reaches for the Kobokan
and three reaches for the Lengkong. The area of each basin and reach mantled by BAF deposits is indicated in km2 and percent

Basin Surface area (km2) Length (km) Slope (°) Gradient Perimeter (km) Area of BAF deposits

Kobokan (K) 108.9 37.032 5.42 Source 3,472 m 72.91 5 % (0.447 km2)
Lengkong (L) 8.234 12.61 10.20 Source 3,142 m 24.09 2.9 % (3.12 km2)
Reach Surface area Length Slope (°) Gradient Perimeter Area of BAF deposits (%)
K | L (km2) K | L (km) K|L K|L K | L (km) K | L | K+L
K | L (%) K | L (%)
Source 1 6.67 | 2.46 4.94 | 3.51 24.20 | 26.13 0.4494 | 0.4906 11.08 | 7.31 17.63 | 23.55 | 19.21
Proximal 6.13 | 29.44 13.33 | 9.51
Medial 2 5.68 | 2.88 2.22 | 4.43 9.33 | 7.73 0.1643 | 01357 10.43 | 8.63 11.69 | 10.64 | 11.34
Storage and byp 5.21 | 34.96 6.00 | 11.98
Medial 3 13.30 | 2.89 8.04 | 4.67 3.87 | 3.93 0.0677 | 0.0687 19.01 | 12.29 14.62 | 12.59 | 14.26
Storage and byp 12.21 | 35.10 21.71 | 19.43
Deposition and bypassing 4 44.12 | – 10.46 | – 1.98 | – 0.0346 | – 27.50 | n.a. n.a.
40.51 | – 28.25 | –
Deposition 5 39.13 | – 11.37 | – 0.77 | – 0.0134 | – 28.45 | n.a. n.a.
Distal delta 35.93 | – 30.71 | –

K Kobokan, L Lengkong, byp bypassing

is about 0.25-m thick. The T1 (~1 m) and T1′ (~2 m) terraces left overbank deposits on the T3 terraces forming the inter-
are built several times each season by small-volume lahars. fluve between the Kobokan and Lengkong (Fig. 2 and 6). This
The T2 terrace (>3-m high) is covered by medium-sized lahars overbank area coincides with the break-in-slope between the
once a year. The forested or cultivated T3 terrace, 10- to 20-m middle reach L2 and distal reach L3.
high and not covered by historical lahars, has been formed by Historical lava flows and local faults have induced changes in
a succession of pyroclastic and lahar deposits of mid Holocene the Kobokan and Lengkong valleys as interpreted from the air
age, based on 14C dates (14C 3,770±40 years BP=Cal B.P. photos and SPOT5 image (Figs. 5 and 6). The 1875 and 1941
3,986 to 4,283) with 95.2 % confidence interval (charcoal Gif- lava flows cut off the upper Lengkong watershed and subtracted
12,290, ∂13C=−26.39‰; calibration after Reimer et al. 2009), a large area from its reach L2. The Lengkong River took a
obtained at a depth of 0.5 m on the T3 terrace in the middle shorter path from ENE to east towards the Kobokan, ahead of
reach 2 (Figs. 2 and 4). The 1994–1995 block-and-ash flows the 1941 lava flow front (Fig. 6). The lowermost Lengkong has

Table 2 Morphometric indices

of the Koboan and the Lengkong Morphometric parameter Kobokan Lengkong
catchments. These indices were
used to compute the hypsometric Area (A) 108.9 km2 8.23 km2
curves and breakdown the valley Height of basin (H bv) 3,635 m 3,610–725=2,885 m
reaches Height of river (h rv) 3,472 m 3,142–725=2,417 m
Basin length ( L b) 33,653.6 m 11,421.40 m
Perimeter (P) 72.91 km 24.09 km
Relief ratio (Rh=H/Ln) 0.108 0.253
Relative relief (Rhp=H/p) 0.05 0.12
Relative basin height (v=h/H) 0.955 0.838
Relative basin area (x=a/A) 1/3 of E and SE catchments 0.076
Length area (L=1.4 A0.6) 23.35 km 4.96 km
Basin shape (Rf=A/Lb2) 0.1 0.06
Basin area ratio (Ra=An/An−1) 13.23
ΣL sum of channel lengths ~ 290 km 37.02 km
Drainage density (D=ΣL/A) 2.66 4.49
N number of drains ~120 18
Stream frequency (Fs=N/A) 1.10 2.18
Average stream length avL=avL1RL 0–1 37,032.42 m 12,612.43 m
Length ratio (RL =avLn/Ln +1) 0.341
Bull Volcanol (2014) 76:857 Page 9 of 26, 857

not yet recovered from this disruption. As a result, the c.100– response can be complex and lahars can overlap in the middle
150-m-wide, braided channel between terraces of the transfer reaches owing to the interference between the two tributaries
reach L2 contrasts with the narrow (<40 m), single-thread chan- at their headward valley as shown in Fig. 2.
nel of the lower reach L3 (Figs. 2 and 4–6). Eroded in lava flows
with a regular slope (3°), the straight channel of reach L3, which Rainfall and lahars at Semeru
parallels a N110-trending fault (Figs. 5 and 6), is entrenched
within a set of three terraces T1–T3. Abandoned meanders on Semeru is one of the most magmatically active volcanoes on
the left margin (Fig. 5 and 6) indicate that a wider river flowed Earth that also produces huge volumes of lahars. Persistent
earlier on a gentler slope before the formation of T2. explosive activity results in daily tephra falls, and since 1884,
Hypsometric curves (Fig. 4c) suggest that the Kobokan’s PDCs have been recorded to occur every 3.5–5 years on
catchment (hypsometric index, HI, 0.14) is closer to equilib- average (Thouret et al. 2007) with an exceptional PDC runout
rium than the Lengkong’s sub-catchment (HI 0.3). The of 11 km. This combines with a high annual rainfall of 2.20 to
Kobokan’s hypsometric curve, however, exhibits two convex 3.70 m. As a result, pyroclastic debris on the steep slopes (24–
segments (Fig. 4c): the upper one is the gorge between reaches 26° in the upper reach, 15° on average) are remobilized
K3 and K4, suggesting the effect of recent uplift and faulting through a dense network of channels (Figs. 2 and 4) with
of the Tertiary hills (Fig. 3 and 6c; Solikhin et al. 2012); the high-gradient proximal and medial reaches (7 to 3 %) and
second one reflects the large and thick depositional area long, gentle depositional distal ones (≤2°; Table 1). Lahars can
upstream of the delta. The knickpoint is now located between flood widespread areas, as in 1909 and 1981, totaling approx-
the confluence and the wide Kobokan check dam (Figs. 2, 4 imately 350 km2 in the ring plain (Fig. 1).
and 6). The steep and straight Lengkong channel responds The climate at Semeru, east Java (8°S), is subequatorial
faster to both large and modest sediment supply events than with 7–8 months of monsoon rain and nearly 5 months of dry
the long, meandering Kobokan. However, the Lengkong period. The middle reaches of the Kobokan and the

N C 1990
N

A 1981 187
5 la w
1875 lav
19 va va flo
a flow
41 flow 941 la
t by 1
lav nel cu
LE
194
1 la LE
af
low Chan
N GK
va fl
ow KO NG
O
NG
BO KO
KA NG B. Smut
Lengkong N River
R
tributaries
ch ectil
an ine
Re B. Smut A ne ar
Kobokan cti l
lin Abandoned
tributaries ea Old
rc
A ha meanders meanders
Hi nn
st
or
ic
el
al
la B
va Historical Gorge
flo B Confluence lava flow Tributary
w
area Abandoned Conjugated
Gorge channels N130 and
N45 faults
Confluence
0 500m
area
Scale

Scale
0 500m

B D CAPTION:

CAPTION: Lava Flow 1941

112°58'20"E 112°59'10"E 113°0'0"E
112°58'20"E 112°59'10"E 113°0'0"E Lava Flow 1941 Hydrography:

¯ 1990
Hydrography:
1981 Hydrographic network
Hydrographical network

Active channel
8°9'10"S

8°9'10"S

Active channel Temporary channel

8°9'10"S

8°9'10"S

Temporary channel Inactive channel

Inactive channell
Lenkong Incision level of sediment
Incision level of sediment River deposition (channel area):
deposition (channel area):
SD level 1
Lenkong SD level 1
River SD level 2
SD level 2
SD level 3
SD level 3
Incision level of actual
Incision level of actual riverbanks:
8°10'0"S

riverbanks:
Terrace level 1
8°10'0"S

Terrace level 1
Terrace level 2 Terrace level 2
8°10'0"S

8°10'0"S

Kobokan Terrace level 3 Kobokan Terrace level 3

River River
Terrace level 4 Terrace level 4

Incision level of riverbanks Incision level of riverbanks

(inactive channel): (inactive channel):
0 0.4 0.8 1.6 km OT1 OT1
0 0.4 0.8 1.6 km
OT2 OT2
112°58'20"E 112°59'10"E 113°0'0"E Miscellaneous:
112°58'20"E 112°59'10"E 113°0'0"E Miscellaneous:
Farmland
Farmland
Forest
Forest

Fig. 5 Mosaics of air photos taken a in 1981 (scale 1:25,000) and c in and incision, and the channel evolution of the Kobokan catchment and the
1990 (scale 1:50,000) and b, d derived geological maps (outlined in a and Lengkong sub-catchment (see “Major geomorphic changes between
c) showing the hydrographic network, the areas of sediment deposition 1981 and 1990” for explanation
857, Page 10 of 26 Bull Volcanol (2014) 76:857

Fig. 6 Geomorphic changes in the confluence area of the Kobokan and terraces in 1981 and 1990, will be filled by the 1994–1995 BAF deposits
the Lengkong. a 1981 air photograph. b 1990 air photograph. In the area as shown in c. c 23 October 2003 SPOT5 image (2.5-m resolution)
3 and 3” (a, b), the dashed line indicates sediment accumulation in 1990 showing the middle reaches of the Kobokan and the Lengkong, the main
and 2003. Area 2 is more eroded in 1990 than 2” in 1981 (a, b) due to two geomorphic characteristics, and the 1-km-wide overbank area mantled by
active channels. The T2 and T3 terraces, which were 3- to 10-m high in the 1994 and 1995 BAF deposits (about 6.3×106 and 5.5×106 m3,
the area 2, receded after 1981 and have been eroded in stepped banks by respectively). From the overbank BAF deposits mantling the interfluve
the Lengkong lahars, which spilled over the east margin of the confluence between both rivers (Fig. 2; c), avulsion has led material into the
zone. The area 3” (b), which was eroded by a multi-thread river cutting Lengkong on the north side and into a secondary channel to the south side

Lengkong, on the lower SE flank between 700 m and 1,000 m 1,600 m asl (Fig. 7b) indicates that 60 % of the annual rain
asl, receive an average annual rainfall of 2.75–3.25 m occurs between October and March. January and February are
(Fig. 7a). Rain gauges at higher elevations are scarce, but usually the wettest months, but rain may arrive any time on the
Fig. 7a suggests that as much as 2.25–2.5 m rain may fall SE flank of Semeru. Ninety-two percent of lahars occurred in
annually on the flanks of the volcano between 1,000 and daytime, mostly during afternoon rains. Heavy rainfall capa-
2,500 m asl. The monthly rainfall for Gunung Pakis at ble of triggering lahars occur when wet air masses from the
Bull Volcanol (2014) 76:857 Page 11 of 26, 857

C
A

Fig. 7 Rainfall data recorded around Semeru. a Isohyet map showing the installed in the Lengkong catchment in 2003–2005 and 2010, respective-
25-year-long (1976–2000) average annual rainfall in the river basins ly. Black dots indicate daily rainfall measurements (in mm/minute) car-
draining Semeru southward and eastward (source: Hydrological offices ried out during two rainy seasons. Regression coefficients for intensity (I)
of the Lumajang and Malang Prefectures, Mt. Semeru Project). b Corre- and linear correlation R2 are indicated for each of the rain gauge stations. c
lations of rainfall intensity (I in mm/min averaged over 10-min bins) with Rainfall at Gunung Pakis (1,600 m asl) and occurrence of flows in the
duration (D in minutes) for two rain gauge stations located on the Semeru Lengkong River for the years 1999–2000 (from Lavigne and Suwa 2004)
TopoDEM: Gunung Leker (1,400 m asl) and Sumberwuluh (750 m asl)

ocean 25 km south of Semeru move in and merge with local intensity has been recorded at Gunung Leker (1,330 m) be-
thick cloud cover that characterizes the volcano flanks be- tween 2003 and 2005, Gunung Pakis (1,600 m) in 2000
tween 1,500 m and 2,500 m asl. As a result, thunderstorms are (Lavigne and Suwa 2004), and Sumberwuluh (750 m) near
triggered during the afternoon every few days during the the lowermost reach of the Lengkong in 2010 (Figs. 2 and7a).
wettest months of January–March (Fig. 7a). Rain falls daily Intensity (I in mm/min averaged over 10-min bins) and dura-
during the rainy season, at times in very intense storms. A few tion (D in min) are correlated (Fig. 7b) as I=1.675 D−0.3735 for
hours before the voluminous 14 May 1981 debris flows, the Gunung Leker and I=2.095 D−0.4711 for Sumberwuluh. The
now-abandoned Tawonsongo observatory (at Besuk Sat, correlation improves for rain gauges located at higher eleva-
800 m asl on the east flank) recorded c. 300 mm of rain in 3 tion and closer to the Lengkong upper reach. We computed the
h. The monthly discharge of the Mujur River, whose catch- threshold rainfall for triggering lahars as between 6.9 and
ment size (130 km2) is comparable to that of the Kobokan 8.4 mm in 10-min rainfall events (Fig. 7b). Such rain amounts
12 km to the east, is 23.75 m3 on average (IAHRI 2008). Thus, to 20 mm/h on average (n=30 events). The average max-
both daily rainfall and episodic peaks in rain are crucial. imum intensity measured is 14 mm for 10-min rainfall, but
Rainfall duration and intensity have been analyzed using rainfall measuring up to 36 mm/h can last as long as
data from four rain gauge stations on the east and SE flanks of 102 min (n =25 events). Rainfall exceeding 8 mm in
Semeru and two temporary stations installed by us near the 10 min and lasting at least 70 min is the most favorable
Lengkong in the rainy seasons (Figs. 2 and 7). Rainfall for large debris flows.
857, Page 12 of 26 Bull Volcanol (2014) 76:857

Relatively low infiltration rates of 33–50 mm/min mea- Fig. 8 a The middle Kobokan reach K2 and Lengkong reach L2 mapped„
sured by us in the ash and lapilli size deposits in the Lengkong from the 2001 SPOT4 image (above) and 2003 SPOT5 image (below)
and derived geologic maps. b The Kobokan and Lengkong middle
valley prevent lahar generation outside intense long-term rain-
reaches K3 and L3 mapped from the 2001 SPOT4 image (above) and
fall, but two factors favor lahar activity at Semeru. The per- the 2003 SPOT5 image (below) derived geologic maps. Mapping BAF
sistent tephra activity introduces ash to the lapilli matrix and deposits is based on a rough and somewhat dry surface with high
cobble-sized BAF deposits, and long-term (>3 h) rainfall reflectance, whereas lahar deposits appear smoother and wetter with
low reflectance. The interpretation is sometimes difficult where valleys
intensity in between several days of sustained rainfall en-
cross farmlands or inhabited areas with their own reflectance signatures
hances a subsurface perched water table in the deposits. We
have been unable to tie rainfall thresholds at high altitude with
the actual site of lahar initiation as this process occurs in deep and 2008, 15-cm resolution) taken from an ultralight paraglid-
gullies eroded into the bare volcano flanks above 2,000 m asl. er, one SPOT4 image (27 March 2001; 20-m resolution), and
Access to the high, steep south slopes close to the active one high-spatial resolution SPOT5 image (24 October 2003,
summit cone is impossible, but inspection of high-spatial 2.5 m resolution) (Figs. 6 and 8–11). Selected airphotos from
resolution satellite images reveals the initiation area of lahar 2005 and 2008 over the 2 km-long middle reach L3 of the
deposits in steep, bare drainage channels between 1,900 m and Lengkong valley (Figs. 2, 8 and 9; App. Tab A) were ortho-
2,500 m asl. Gullies capture debris from unconsolidated rocks rectified, mosaicked, and geo-referenced using photogramme-
forming the bare, dissected slopes in the source reach above try techniques (Pierrot-Deseilligny and Paparoditis 2006;
2,000 m asl (Fig. 2). Within 3–4 km of the summit, terraces of Pierrot-Deseilligny 2007). The mosaics were used to outline
likely lahar deposits can be observed approximately at 2,000– areas covered by BAF and lahar deposits confined in valleys
1,900 m asl in entrenched, narrow river channels. or deposited overbank on interfluves (Figs. 8–11). The 2001
In sum, initiation and mobilization of lahars are related to SPOT4 and 2003 SPOT5 scenes were similarly used to map
heavy rainfall and loose debris from pyroclastic deposits that and separate the BAF deposits with high reflectance from the
rest on steep river gravel beds. Small-scale lahars (<105 m3) smoother and wetter lahar deposits (Figs. 8–11 and App.
occur several times each rainy season, medium-scale lahars Figs. A-E). The difference between our maps and those drawn
(0.1–1.5×106 m3) at least once a year, and large-scale lahars from automatic segmentation process proved to be small
(1.5–6×106 m3) every 6 years on average (Table 2 in Thouret (about 10–15 % in calculated volumes), as illustrated by one
et al. 2007). Post-eruption lahars have two origins: First, they subscene of the Kobokan check dam and the confluence area
may start by remobilizing tephra and PDC deposits emplaced with the Lengkong from the 2001 SPOT4 image (App.
as a result of accelerated dome extrusion rates every 3.5– Fig. A). This difference is deemed insignificant, and mapping
5 years. Eruptions and dome collapse both produced PDC has been done manually.
deposits between 5 and 7×106 m3 in volume in 1972, 1977, Volumes of deposits have been calculated on the basis of
1981, 1994, 1995, and 2002–2003 (Thouret et al. 2007). The surface areas deduced from our maps and the average thickness
1994 and 1995 eruptions emplaced about 11.8 million m3 of of BAF (2 m) and lahar terraces (T0 0.25 cm, T1 1 m, and T1′
BAF deposits as far as 9.2 to 11.55 km from the source in the 2 m) from field observations (Fig. 9). Budgets of BAF, lahar
Kembar, Kobokan, and Lengkong valleys; second, lahars may deposits and total deposits, aggradation and degradation rates,
also start as large but infrequent debris flows triggered by and sediment mass balances have been estimated (Ta-
landslides in heavy rain. Such lahars exceeded 6 million m3 ble 3, Figs. 12 and 13, App. Tables B-F). Table 3 and
in 1909 and 1981 on the east flank (Siswowidjoyo et al. 1997). App. Tables C-D cover the Kobokan and Lengkong reaches
while App.Tables E-F present the data from two smaller
representative segments from the two rivers. The linear aggra-
dation (+ sign in Table 3, Fig. 13 and App. Tables C-F) and
Data acquisition and processing degradation (− sign in Table 3, Fig. 13 and App. Tables C-F)
rates (m3 per meter of channel length) express deposition or
A stepwise analysis of the Kobokan and Lengkong catch- incision over a given time interval. The term “sediment yield”
ments has been carried out using data from satellite images, is the mass balance (m3/km2/year) between storage (+in Ta-
aerial photographs, field mapping, and DEMs/digital surface ble 3, Fig. 13 and App. Tables C-F) and removal (− in Table 3,
models (DSMs) for 30 years of data (1981–2011). Fig. 13 and App. Tables C-F) of deposits over a time interval,
normalized to the length of reach or the entire channel. The
Geologic mapping and computing aggradation/degradation computation of the linear aggradation and degradation rates
and sediment yields takes into account the geometric charac-
Geologic mapping (Figs. 5–6 and 8) for this study was based teristics of each reach as summarized in Table 1. We could not
on 1981 and 1990 air photographs (scale 1:25,000), two sets measure sediment from mass wasting or water erosion pro-
of recent low-altitude (200–300 m) airphotos (August 2005 cesses from the non-channelized parts of the landscape.
Bull Volcanol (2014) 76:857 Page 13 of 26, 857
857, Page 14 of 26 Bull Volcanol (2014) 76:857

Digital elevation and surface models the average density and coverage of the datasets, only 3-m
DEMs were computed from 2006, 2007, and February 2008
The topography of the Lengkong valley along the 2-km-long D-GNSS and 2010 TLS data and 50-cm DEMs from 2011
segment from the uppermost check dam to the confluence TLS data with the best resolution (Fig. 12a). In addition, two
(Figs. 1 and 3) was determined and measured with Differential 50-cm DSMs were computed from August 2005 and 2008 air
Global Navigation Satellite System (D-GNSS) and a Terres- photos and ground control points selected inside the D-GNSS
trial Lidar Scanner (TLS) since 2006 (App. Table B). Given available database by photogrammetry (Pierrot-Deseilligny

Fig. 9 Orthorectified mosaic of low-altitude air photos taken in August overbank on interfluves. Sediment mass balances after 2005 have been
2008 (a) and derived geologic map (b) outlining terraces, channels in the computed on the basis of low-altitude photographs, this geologic map,
Kobokan K3 and the Lengkong L3 reaches, including the confluence. and GNSS and TLS surveys (Figs. 10–12). Rectangles indicate the
The 2005–2008 mosaics (1-m resolution) were used to carefully outline mapped areas of Figs. 10 and 11. The contour outlines the 2-km-long
areas covered by BAF and lahar deposits confined in valleys or deposited channel surveyed for annual DEMs/DSMs (Fig. 12)
Bull Volcanol (2014) 76:857 Page 15 of 26, 857

Table 3 Aggradation and degradation mass balance between 2005 and and thresholded with 1 m minimum limit of detection (minLoD) difference
2011 in the 2-km-long Lengkong segment and reach 3, calculated from models (see Figs. 12 and 13c and App. Tables E-F for the values shown
D-GPS-based DEMs (2007 and 02/2008), photogrammetric DSMs (08/ here)
2005 and 08/2008), and TLS-based DEMs (2011) for non-thresholded

Time interval Aggradation (m3) Degradation (m3) Budget (m3) Linear aggradation/ Sediment mass balance
degradation rate (m3/m) (m3/km2/year)

2-km-long reach Reach 3

Non-thresholded (without uncertainty analysis)

2011–08/2008 +16,550 −21,020 −4,470 −2.6 −4.03.104 −6.21.102
08/2008–02/2008 +8,580 −20,490 −11,910 −7.0 −8.08.105 −9.93.103
02/2008–2007 +10,030 −41,480 −31,450 −18.5 −8.86.105 −1.31.104
2007–2005 +30,800 −1,500 +29,300 +29.9 +3.91.106 +6.10.103
Thresholded with 1 m minLoD
2011–08/2008 +12,080 +15,800 −3,720 −2.2 −3.35.104 −5.17.102
08/2008–02/2008 +5,920 +14,200 −8,280 −4.9 −5.61.105 −6.90.103
02/2008–2007 +7,200 +35,850 −28,650 −16.9 −8.07.105 −1.19.104
2007–2005 +29,720 +970 +28,750 +29.3 +9.58.105 +5.99.103

See Fig. 13c for a graphic display of values

and Paparoditis 2006; Pierrot-Deseilligny 2007). These DSMs threshold was estimated for each appropriate pair of
cover the Lengkong valley channel and the confluence with DEMs/DSMs by subtracting them from a known stable area,
the Kobokan. Difference models between each appropriate here the large persistent T1 terrace on the Lengkong right bank
pairs of annual DEMs/DSMs were calculated (Fig. 12b). The- immediately downstream of the upper check dam (Fig. 10). A
se enable us to estimate directly volumes of debris transported chosen average minLoD of 1 m is a good compromise for all
each year, and a mass balance between aggraded and degraded studied pairs. The only significant elevation differences great-
areas was synthesized in Table 3. We also calculated rates of er than 1 m are thus mapped (Fig. 10–11) and accounted for in
aggradation and degradation and sediment yields at the scale budget calculations (Table 3, Fig. 13, App. Tables C-F). The
of the 2-km-long Lengkong branch and of the reach 2B. differences between threshold and non-threshold budgets,
rates of aggradation and degradation, and sediment yields
Uncertainty analysis for each period of time provide uncertainties for our estimates
that are synthesized in Fig. 13c.
Because of the differences in resolution in various data
sources and the possible error, an uncertainty analysis has
been carried out to test our results. We have taken into account
the standard deviations of the thickness of the deposits (50 cm Results: spatial and temporal patterns of aggradation
for the BAFs and T1′ terraces, 25 cm for T1, and 5 cm for T0) and degradation
and an error of half a pixel for the manual mapping (i.e., 10 m
for the SPOT4 image, 1.5 m for the 2003 SPOT5 image, and Major geomorphic changes between 1981 and 1990
0.075 m for the 2005 and 2008 air photos) to estimate mini-
mum and maximum volumes of BAF deposits, lahar terraces, The 9-year changes (1981–1990) in the Kobokan and
and all formations together for each year. We compiled Lengkong middle reaches and their confluence can be identi-
(Figs. 13 and 14a) the resulting ranges of sediment budgets, fied on air photographs and SPOT images. The 1981 photo-
aggradation and degradation rates, and sediment mass bal- graph (Fig. 5a, b) was taken after a volume of 6 million m3 of
ances for each period of time. We have analyzed the uncer- BAF deposits choked both channels on the SE and E flanks of
tainties in the difference models between annual DEMs/DSMs Semeru on 28–30 March and 3–5 April 1981 (Table 1 in
calculated along the 2-km long Lengkong using the simplified Thouret et al. 2007). The widening of the two channels
approach of the method followed by Wheaton et al. (2010). between 1981 and 1990 formed a set of erosion terraces along
We have defined a minimum limit of detection (minLoD) in the lower transfer reach of both rivers (Figs. 5a, b). Area (1) in
absolute value from which elevation differences are insignif- Fig. 6a shows the 15–20-m-high Holocene terrace with a
icant when they are lower than the sum of errors from local- decreasing number of channels for 1981, 1990, and afterwards
ization, measurements, and interpolation. A single minLoD (Figs. 5 and 6). Between these dates, the large, braided
857, Page 16 of 26 Bull Volcanol (2014) 76:857

Fig. 10 Four pairs of images showing increasing resolution BAF and 2008. Down valley from the dam, degradation was promoted by the
lahar deposit maps near the upper check dam across the Lengkong reach removal of the T1 and T1′ terraces, but lahar deposition on top of the
L2, based on 2001 SPOT4 and 2003 SPOT images and on 2005 and 2008 T1′ terrace between 2001 and 2005 has reduced the degradation rate. The
low-altitude airphotos. Upstream of the dam, BAF deposit was removed channel morphology has not changed drastically between 2005 and 2008,
from the T1′ terrace, followed by the successive deposition and degrada- but river incision revealed the second wall of the dam 25 m down valley
tion of the T1 terrace, while the T0 terrace was buried between 2003 and (arrow), which was buried between 2001 and 2008

Kobokan channel migrated about 250 m southward due to River in 1992, 0.35 km north of the confluence (Fig. 6c),
the growth of the Lengkong fan at the confluence. The induced a storage zone upstream of the site, which holds a
active Lengkong channel migrated about 100 m to the large amount of BAF and lahar deposits. This also happened
south, to the bank of Kobokan, and increased the landslide in the confluence area between the two dams, but all were
activity on this bank in weathered rocks of the Tertiary hills subsequently breached or flattened, allowing debris to by-
(Fig. 6a, b). The check dams at the end of the middle reach pass the obstacles. Regressive erosion propagates from the
have induced changes in deposition/erosion down valley. steep gorge towards the confluence of the Kobokan with
The construction of open check dams across the Kobokan Lengkong (Figs. 4–6).
Bull Volcanol (2014) 76:857 Page 17 of 26, 857

Fig. 11 Four pairs of images showing increasing resolution of BAF and shift of the Kobokan 250-m-large meander and the Langkong fan and bed
lahar deposit maps of the confluence area and the meandering channel of removed a large part of the 1994–1995 BAF deposits from the north bank
the Kobokan reach K3 based on remote sensing data as in Fig. 10. The

Decadal response to the 1994 and 1995 BAF deposition sediment input in 1994 and 1995 was twice as large as in
1981, the observed geomorphic changes are clearer on the
After the emplacement of 1994–1995 BAF deposits, all post-1995 images (Fig. 6c).
reaches showed quick aggradation followed by slow degrada- Reach K1 (source) of the Kobokan catchment on the sum-
tion, but with strong variations in sediment yields (Figs. 8–11 mit cone (Figs. 1 and 4, App. Fig. C) has been outlined from
and App. Tables C-F, App. Figs. B-E). The 1981 and 1994– the pair of SPOT4 and SPOT5 images draped on a DEM (after
1995 eruptions triggered landscape re-adjustment, but their Solikhin et al. 2012). The scar-shaped headwall of Kobokan’s
effects may have also altered drainage channels. As the catchment covered by tephra occupies 2.29 km2, i.e., 46 % of
857, Page 18 of 26 Bull Volcanol (2014) 76:857

Fig. 12 a Seven annual (shaded relief) DEMs/DSMs from 2005 to 2011 (Pierrot-Deseilligny and Paparoditis 2006; Pierrot-Deseilligny 2007).
in the 2-km-long Lengkong middle reach L3, most of them including the DEMs and DSMs incorporate the temporal and spatial aggradation and
confluence area (see Fig. 9 for the survey area). The DEMs were calcu- degradation patterns over seven rainy seasons. Note the geomorphic
lated from D-GNSS (2006, 2007, and February 2008, grid step 3 m) and changes in the uppermost check dam area and the fan-like confluence
LIDAR TLS data (2010 and 2011, grid steps 3 and 0.5 m, respectively). zone. b The color-coded bar shows δz values (in m) with a minimum
The DSMs (August 2005 and August 2008, grid step 50 cm) were derived limit of detection (minLoD) of 1 m. Aggradation is colored in red and
from low-altitude air photographs using photogrammetric techniques degradation in blue

the summit cone. From an estimated annual volume of 12– channel length) and sediment mass balance of −5.7×104 m3/
23×106 m3 of tephra falling on the cone, one sixth (i.e., the km2/year. The aggradation and degradation balance from the
volume falling on the heads of two catchments) can be avail- Lengkong reach L1 (2.46 km2) cannot be estimated because
able for lahar formation in both valleys. How much of this an ash cloud covered this area on the 2001 SPOT image (App.
volume can be remobilized every rainy season is unknown, Fig. C).
but the pair of SPOT4 and SPOT5 images allows us to Reaches K2 and L2 show a negative budget and contrast-
estimate the aggradation and degradation balance over 2 years ing sediment mass balances between 2001 and 2003
between 2001 and 2003 (Fig. 13a, App. Tables C & D). The (Fig. 8a, b and 13a; App. Tables C & D). The total budget in
Kobokan reach K1 (6.67 km2) shows high degradation of reach K2 (App. Table C) (−9.6×105 m3) indicates a net
BAF deposits over 2 years with a loss of 7.6×105 m3 of transfer of debris outside this reach, K2, which is much steeper
material, implying a degradation rate of −154 m3/m (per than reach K3 (Table 2, Fig. 4), funnels the debris from reach
Bull Volcanol (2014) 76:857 Page 19 of 26, 857

K1, and may also receive more rainfall (Fig. 7c). The negative eruption years (1995–2003) and also channel changes
budget is due to the loss of BAF deposits exceeding lahar (2001–2008).
aggradation. The degradation rate is very high (−433 m3/m),
and the sediment mass balance exceeds −8.4×104 m3/km2/
Changes in channel forms and terraces
year in the reach K2 (5.68 km2). The Lengkong reach L2
(2.88 km2) also shows a negative budget and BAF loss ex-
Two channel styles evolved in the 13 years after the 1994–
ceeding lahar aggradation (Fig. 13a; App. Table D). However,
1995 eruption: a multi-thread sediment-choked pattern in the
the total budget (−1.7×105 m3) and degradation rate (−38 m3/
Kobokan and a single-thread eroding channel in the Lengkong
m) are five- to tenfold lower than those of the Kobokan. The
below a set of three terraces. The channel pattern and terraces
resulting sediment mass balance (−2.95×104 m3/km2/year) is
in the Kobokan and Lengkong reaches K3 and L3 and their
three times smaller than that of the Kobokan.
confluence (Figs. 8–11) suggest four geomorphic responses to
Reach 3 degradation, which prevails in the long and gently
sediment input:
sloping reaches K3 and L3 of the Kobokan and Lengkong
(Figs. 4, 6 and 13a; App. Tables C & D), is also more conspic-
1. Bypassing of sediment in the 250-m-wide Kobokan reach
uous in the large channel of the Kobokan (13.30 km2) than in
K3 upstream of the check dam. The multi-thread, steep
the Lengkong (2.89 km2). In the Kobokan, degradation de-
(6 %) river bed carried a larger input of sediment from the
creases from reach K2 to reach K3 with rates declining from
1994–1995 BAFs than the Lengkong (Fig. 6).
−433.6 to −173 m3/m and sediment mass balances from −8.4 to
2. Erosion of BAF deposits and formation of a 3-m-high T2
−5.2×104 m3/km2/year. In the Lengkong, degradation rate
terrace in the upper valley (reach L3) near the Lengkong
(−38 to −19 m3/m) and sediment mass balance (−2.9 to
upper check dam. BAF-derived T2 terraces, preserved
−1.5×104 m3/km2/year) both halve from reach L2 to L3. This
150-m down valley of this dam, serve as a reference for
decrease between reaches L2 and L3 in both rivers can be
the top of BAF deposits that filled the upper valley in
explained by a decrease in average slope from 9.3° to 3.9° in
1994–1995 (Figs. 6, 9 and 10).
the Kobokan and from 7.7° to 3.9° in the Lengkong.
3. Increased entrenchment of the narrow, single-thread
Maps of the lahar deposits in the Kobokan deposition
Lengkong in 2008 and formation of the T1′, T1, and T0
reaches (K4 and K5) across the ring plain for three years
terraces (Figs. 8 and 9) and low river banks (<0.25 m) by
allowed the identification of the aggrading reach K4 upstream
past and current lahar deposits.
from the degrading reach K5 downstream (App. Table C;
4. Reduction of the terraces at the confluence (Fig. 11). The
App. Fig. D). Reach K4 is 44.12 km2 in area, starting at the
volcaniclastic fan, visible in 2005 at the confluence of the
break in slope below the gorge across the hills. The deposi-
two rivers, was buried by sediment migrating through the
tional process has been reinforced by construction of dikes in
Kobokan.
the upper part of reach K4. This reach shows weak aggrada-
tion at a very low rate (+1.8 m3/m) and sediment mass balance
(2.1 ×102 m3/km2/year). The long (11.37 km) and large
(39.13 km2) reach K5 is measured from upstream of the sharp Two typical river segments
change in river direction (App. Fig. E). The slightly negative
budget in reach K5, where the river crosses the delta, is due to The complex spatial and temporal changes between 2001 and
lateral dikes constructed between 2001 and 2003 in order to 2008 have been studied in detail for two selected segments on
help flush sediment out to sea. The contrasted patterns can be the Lengkong: the river course straddling the upper check dam
explained by storage in reach K4 induced by its situation in the Lengkong (Fig. 10) and the fan of the Lengkong at the
between two uplifted areas (the Tertiary hills to the north confluence with the Kobokan (Fig. 11). The dam was con-
and the coast to the south). Reach K4 concentrates most structed in 1992 to prevent the passage of sediment down
sediment carried by lahars, helped by dikes. valley. We chose this area because the surface of stored
In brief, the large 1994–1995 sediment input (Figs. 3 and 6) sediment induced by the check dam was used as a reference
triggered very high degradation of the Kobokan source reach for calculations. This was also the farthermost reach impacted
K1 and fast degradation in reaches K2 and L2 of both catch- by the 1994–1995 PDCs in the Lengkong Valley. Further-
ments. Downstream, aggradation still prevails in the Kobokan more, the breach in the upper dam enables sediment to bypass
reach K4, but a slight degradation affected the delta in reach K5. the construction. The second area, at the confluence, was
selected to observe two fluvial responses on gravel river beds.
Mid-term behavior of the Kobokan and Lengkong Interactions between storage and bypassing in the Kobokan
reach K3 (with an open dam) can be compared with bypassing
Both catchments displayed high degradation immediately af- and down-cutting processes that acted in the Lengkong
ter 1995, followed by weak aggradation for the 8 post- reach L3.
857, Page 20 of 26 Bull Volcanol (2014) 76:857

In reach L2, a fast and significant degradation was followed includes the channel and a set of terraces, 420 m up and
by weak aggradation in the Lengkong upper check dam area downstream of the dam. The 1994 BAF deposits reached
(Fig. 10 and 13b; App. Table D). This area (0.031 km2) c.150 m down valley from this dam and formed a 3-m-high
 Bull Volcanol (2014) 76:857 Page 21 of 26, 857

ƒFig. 13 Graphics depicting the computed aggradation and degradation aggradation of BAF and lahar deposits (+4.01×104 m3/km2/
mass balances between 2001 and 2011 in the distinct Kobokan and year) by 2005. This was followed by high degradation
Lengkong reaches (see App. Tables C-F). a Mass balance (from BAF
and lahar deposits) for all the Kobokan and Lengkong reaches between
(−3.85×104 m3/km2/year). In contrast, downstream of the
2001 and 2003. Volumes are estimated on the basis of average thicknesses Kobokan check dam, negative budget and higher degradation
of BAF (2 m) and lahar terraces measured in the field. The average (up to −7.5×104 m3/km2/year) before 2005 decreased to
thickness has been fixed to ~0.25 m for T0, ~1 m for T1, and ~2 m for values similar to upstream after 2005 (−3.05×104 m3/km2/
T1′ lahar terraces. Results for the Lengkong reach L1 are biased because
the area is hidden by an ash cloud on the 2001 SPOT4 image (see App.
year). A large amount of BAF deposits was removed and lahar
Table C-D for the values portrayed here). b Mass balance between 2001 deposits aggraded (App. Table D), but both processes have
and 2008 in the two representative areas (the Lengkong upper check dam been more active downstream. Degradation decreased before
and confluence of the Kobokan and Lengkong) based on geological 2005 (from −86.1 to −0.4 m3/m) but increased again after
mapping using 2001 and 2003 SPOT4 and SPOT5 scenes (Figs. 8, 10,
and 11), 2005 and 2008 low-altitude air photographs, and a geologic map
2005 (−27.1 m3/m) due to the removal of BAF deposits from
(Fig. 9) (see Table E & F for the values portrayed here). c Mass balance the upstream area. Figure 11 indicates that at least two thirds
between 2005 and 2011 in the 2-km-long Lengkong segment (within of the BAF deposit forming a T2-type terrace was removed
reach L3), calculated from D-GPS-based DEMs (2007 and 02/2008), between 2003 and 2008. In contrast, the volume of the T1′ and
photogrammetric DSMs (08/2005 and 08/2008), and TLS-based DEMs
(2011) for non-thresholded and thresholded with 1 m minimum limit of
T0′ terraces increased two to threefold but not enough to
detection (minLoD) difference models (see Table 3 for values portrayed compensate for the loss of BAF material. At the confluence,
here) removal of sediment is probably enhanced by the presence of
the Kobokan check dam, which acts as a knickpoint for the
regressive erosion from the gorge.
In brief, degradation occurred between 2001 and 2008 in
T2-type terrace. The T1 and T1′ terraces in lahar deposits and both areas, but they do not have the same pattern after 2008.
the T0 banks have been formed inside this BAF terrace. The Degradation apparently decreased between 2001 and 2005 for
sediment mass balance was negative between 2001 and 2005 the Lengkong and stopped in 2008. This trend is observed on
(−3.5×105 to −9.3×103 m3) but became positive between both sides of the upper check dam. In the Kobokan, degrada-
2005 and 2008 (+8.9×102 m3). BAF deposits have been tion increased after 2005 but halved between 2003 and 2005,
highly eroded, whereas lahar deposits have generally perhaps due to the increased tephra supply on the summit
remained in depositional areas, especially downstream of the slopes and in reach K1 during the 2002–2003 eruption.
dam. The degradation rate, high between 2001 and 2003,
decreased between 2003 and 2005 (−83 to −22 m3/m), and Mass balance in the Lengkong reach L3 based on annual
erosion nearly stopped after 2005. As a result, the sediment DEMs/DSMs
mass balance up and downstream of the dam changed from
−5.6×105 to −1.5×105 m3/km2/year, but sediment started to We have analyzed the temporal and spatial patterns of aggra-
accumulate after 2005, as illustrated by the expanding T1 dation and degradation along the 2-km-long Lengkong reach
terrace in 2008 (Figs. 9 and 10). Upstream, the high sediment L3 since 2005 using seven high-resolution DEMs/DSMs
mass balance halved between 2001 and 2005 (−2.7 to −1.4× (Fig. 12, App. Table B). We have estimated mass balances
105 m3/km2/year) and was further diminished by two orders of from pairs of DEMs covering the 2005–2011 period in order
magnitude (−5.2×103 m3/km2/year) between 2005 and 2008. to evaluate the response of the channel bed and banks to
Downstream, high degradation, reflected by annual sediment sediment transfer (Fig. 13c, Table 3). Degradation prevailed
mass balances of −1.1×105 and −1.6×105 m3/km2 between for 13 years after the 1994–1995 eruptions of BAFs.
2001 and 2005, was followed by weak aggradation (+1.1× Topographic and geomorphic changes between 2005 and
104 m3/km2/year) until 2008. 2011 are visible from the shaded relief image of seven annual
The confluence of the Kobokan and Lengkong is the DEMs/DSMs in the upper check dam area (see Fig. 10 for
second mapped segment (0.271 km2; Fig. 11). Here morpho- comparison) and in the confluence zone (see Figs. 11–12 for
logical contrasts in channel styles observed between 2001 and comparison, except for the 2006 DEM). The 2005 to 2011
2008 suggest different river responses to sediment migration. DEMs/DSMs show little change in the upper area during this
We have again mapped BAF and lahar deposits and estimated time, as demonstrated by the large T1 terrace on the west side
mass balances between 2001 and 2008 (Figs. 2, 10 and 13b, of the Lengkong down valley from the upper dam. In contrast,
App. Table D). the confluence area was always affected as illustrated by the
The overall mass balance fluctuated but showed less re- changing width of the main active channel between 2005 and
moval of sediment between 2001 and 2008, but areas up- and 2011 (Fig. 11).
downstream of the Kobokan check dam had different rates. The overall mass balance in the Lengkong reach L3
Upstream, the strongly negative pre-2003 budget became (Fig. 13a) shows a marked decline in degradation over 6 years.
positive (from −2.25×105 m5 to +1.4×103 m3) with high Sediment mass balance decreased by at least two orders of
857, Page 22 of 26 Bull Volcanol (2014) 76:857

2003 - 2005 -
2001-2003 2007-2011
2005 2007
ALL REACHES
TWO LENGKONG LENGKONG REACH 3
AND TWO LENGKONG
RIVER SEGMENTS
RIVER REACHES (survey area)
SEGMENTS, 2 AND 3
LENGKONG +9.58 x105
10 6 +1.3 x 10
5
300
5
10
AGGRADATION

250

10 4 +2.1 x 10 2

AGGRADATION
200
2
+8.9 x 10
10 3 150

10 2 100

10 1

LINEAR RATE (m 3 /m)

50
MASS BALANCE
(m3 /km2 /yr)

+1.8 + 0.2 + 29.3 -4.9 -2.2

1
-16.6 -22.1 -0.4 -27.1 -16.9
-10 1 -50
-86.1
-10 2 -83.1 -100

-10 3

DEGRADATION
-150
DEGRADATION

-10 4 -6.8 x
104 -2.4 x
-200
103 -3.6 x -3.35
-10 5 -7.3 x 10 4 x 10 4 -250
102
-1.1 x -1.7 x -1.5 x -5.61x
-10 6 10 5 10 5 105 -300
105 -8.07 x
-5.7 x 105
10 5
-350

KOBOKAN - LENGKONG KOBOKAN

LENGKONG -400
REACH 1 REACH 4
Upper CHECK DAM

KOBOKAN - LENGKONG CONFLUENCE AREA

REACH 2 KOBOKAN -450
REACH 5
LENGKONG 2 km-long
- 472 KOBOKAN - LENGKONG REACH 3
REACH 3
-500

Fig. 14 Cycles of aggradation and degradation in the Kobokan and Lengkong reaches. Spatial and temporal distribution (histograms) of mass balance
(m3/km2/year) and aggradation and degradation rate (m3/m) (black line) for each of the Kobokan and Lengkong reaches between 2001 and 2011

magnitude as shown by pairs of DEMs/DSMs: from −106 to from the Lengkong reach. Our field observations showed that,
−104 m3/km2/year for the 2-km-long reach and from −104 to except for the confluence area, degradation has markedly
−102 m3/km2 /year for reach L3. Degradation prevailed along decreased between 2008 and 2011, which we correlate with
the Lengkong transfer reach for 13 years after the emplace- the end of the removal of BAF deposits.
ment of 1994–1995 BAFs. Both degradation rate and sedi-
ment mass balance, which were positive before 2007 (>29 m3/
m and 106 m3/km2/year, respectively), however, decreased by
one and two orders of magnitude between 2008 and 2011 Discussion
(−2.6 m3/m and −4.03×104 m3/km2/year). Aggradation was
predominant along reach L3 (Fig. 11 and 13a, Table 3) be- Constraints and difficulties of the study
tween the two check dams (Figs. 2 and 9) over 3 years,
whereas high degradation prevailed in the confluence zone. It is difficult to correlate the supply of sediment and aggrada-
The sediment mass balance in reach L3 was low (104– tion and degradation rates throughout the fluvial system with
102 m3/km2/year) when normalized to the entire channel the rainfall at Semeru. The technical constraints are due to
(2.89 km2 in area). Following the pre-2007 aggradation, deg- several reasons: (1) diversity of the tools used to map deposits
radation accelerated in 2007–2008 and then slowed down in river channels prevented a homogeneous analysis (Fig. 3,
until 2011, suggesting that sediment has been flushed away Table 3); (2) difficulty in distinguishing lahars and pyroclastic
Bull Volcanol (2014) 76:857 Page 23 of 26, 857

LONG-TERM STUDY MID-TERM STUDY SHORT-TERM STUDY

1000 TWO CATCHMENTS TWO CATCHMENTS +10 6

Pinatubo
LENGKONG REACH LENGKONG
L1, 2 and 3 REACH
L3 and

AGGRADATION
2km-long reach

1991
100 +10 5

Mt. St
Measured cycle of
50 aggradation / degradation
+10 4
-Hel at Semeru
PYROCLASTIC SEDIMENT (PDCs and TEPHRA) SUPPLY

MASS BALANCE m 3 / km2 / yr

en s
+10 3
19

27 +10 2
80

+10 1
Approximate cycle 1 1
24 -10
-10 2
21 Estimated annual tephra -10 3
? supply on summit cone:
18 ? 12-23 x 106 m3 -10 4
Volume x 10 6 m 3

15 ?
-10 5

DEGRADATION
Background flux 16% tephra supply (1.9-3.7x10 6 m 3 ) to Kobokan and Lengkong
V

12

BAFs in S rivers
and Kobokan

in south rivers
headvalley
9 6
-10

Small BAFs
BAFs
6
6.2

6.3
5.5

3
5

0
1981

1985

1988

1990

1994
1995
1996

2001

2003

2005

2008

2011
Year

Period covered by Figs.13 and 14

Fig. 15 Long-, mid-, and short-term cycles of aggradation and degrada- In contrast, two generalized curves suggest the steady, exponential de-
tion promoted by two types of pyroclastic sediment inputs from persistent cline of sediment mass balance for large but infrequent eruptions of Mt.
eruptive activity, exemplified by Semeru over 30 years. The fluctuating St-Helens (1980) and Pinatubo (1991). The gray area indicates the
curve indicates the approximate, estimated cycle before 2001 and the estimated annual tephra supply (see “Data acquisition and processing”
measured cycle after 2001. Aggradation and degradation rates are sum- and Fig. 3 for the volume of tephra supply). BAF volumes stem from
marized at the scale of the Lengkong sub-catchment from Figs. 13 and 14. Table 1 in Thouret et al. (2007) and references therein

debris in channels in wet conditions using low-spatial- sediment per year (93–95 %) than streamflows at Semeru
resolution images available before 2003. Manual and auto- (Lavigne and Suwa 2004). Large floods can also transport
matic methods for identifying BAF and lahar deposits have large amounts of suspended sediment, but they are rare at
been compared (App. Fig. A) and an uncertainty analysis has Semeru.
been carried out (Table 3); we don't need the dot. Only ;. (3) We need to explore the apparent decrease in lahar frequen-
pyroclastic debris comes from two sources, tephra and BAFs, cy and/or magnitude since 2007–2008. This may reflect
at distinct time intervals, but tephra production has varied over changes in rainfall pattern or in sediment supply as the stored
the years and our determination of tephra volumes on the summit volumes of 1995 BAF deposits in the Kobokan has been
cone are only estimates; and (4) conditions for lahar initiation in waning over 13 years. Sediment forming the river beds on
the uppermost reach are unknown as access to such areas is Semeru comprises scoriaceous tephra and coarse and less
impossible. mobile BAF products. Once deposits have been removed,
There are also other limitations: (1) rain gauges are absent coarse sediment forming the bed requires higher shear stresses
above 1,600 m asl. Thus, the rainfall gradient is based on the for removal. However, the ash matrix needs to be winnowed
available isohyet map (Fig. 7). The largest amount of rain falls to promote channel armouring and incision. This will in turn
between 800 and 1600 m asl on reaches K2 and L2, where a promote dispersion of a sediment wave instead of translation
lahar-triggering threshold of c.7–8.5 mm/10 min has been (Lisle et al. 2001). We also need to investigate the aggradation
determined; (2) measurements of river discharges were avail- between 2005 and 2007 in the Lengkong bypassing reach L3
able in the ring plain, but none on sediment carried by normal (Table 3, Fig. 13a). It may reflect the input of pyroclastic
streamflow in the Kobokan. However, frequent lahars (three debris in reach L1 following the 2002–2003 eruption
to four per month during the rainy season) convey much more (Solikhin et al. 2012).
857, Page 24 of 26 Bull Volcanol (2014) 76:857

Magnitude/frequency relationship with sediment yield on average (Fig. 3), they will instantaneously add one to
two thirds of the volume of estimated annual tephra
Mass balance for BAF and lahar deposits was computed at supply (Fig. 15). Hence, another wave of migrating sed-
multiple spatial scales (catchment, sub-catchment, reach, iment can be quickly transported to down valley reaches.
selected river segments) and for variable time intervals Degradation rates and sediment mass balances are
(22 years, 6 years, and annual). Spatial and temporal patterns spatially variable depending on the geomorphic style of
for each reach (Fig. 14) and the relationship, plotted in Fig. 15, the disturbed river channels. Sediment migration is still
show how cycles of aggradation and degradation depend on going on in 2011 in the larger, more impacted Kobokan
constant tephra supply and/or episodic increases in sediment valley 15 years after the 1994–1995 eruption, whereas the
from BAFs at Semeru. The generalized cycle of sediment sediment wave apparently ceased after 12 years in the
availability suggest three sediment inputs (Fig. 2 and 15): smaller, less impacted Lengkong. Remobilization of de-
bris slowed down in 2008 after higher degradation rates
1. The annual background tephra volume of roughly 12– between 2001 and 2007, which enabled the Lengkong
23×106 m3 on the summit cone can be translated into middle course to be almost cleared of BAF and lahar
lahars, while one sixth of this can supply sediment to both debris from the 1994–1995 eruptions.
catchments under study. As a result, the source reach 1 of
both valleys delivers a positive sediment mass balance to
the lower reaches (1.3×105 m3 km2/year).
2. BAF deposits comprise 5 to 6.3×106 m3 per eruption
every 5 years on average, following volumes estimated Conclusion
by Indonesian scientists (Table 2 in Thouret et al. 2007).
Episodic but cumulative increases in pyroclastic debris Geomorphic response of rivers to volcanic disturbances at
thus trigger numerous cycles of aggradation/degradation, Semeru indicates the following sequential landscape adjust-
which contrast with steady declines after large eruptions ment to active volcanism:
(Fig. 15).
3. The rainfall threshold for triggering lahars ranges between 1. The annual (2001–2003) average sediment mass balance
6.9 and 8.4 mm during 10-min-long rainfall events and measured from the Kobokan’s catchment is −104 to
within long-term (>24 h) rainfall periods. The flows can 10 3 m 3 /km 2 and that from the Lengkong’s sub-
incorporate more material available in the middle reaches catchment is −105 to −104 m3/km2. However, the annual
K2 and L2, from volcaniclastic fans, terraces, and figure for the Lengkong reach L3 drops to −104to 102 m3/
overbank PDC deposits. Although lahars are triggered km2 when sediment has bypassed the reach and moved
upstream in the source reaches K1 and L1, the sediment down valley. In general, measured annual sediment mass
input is easily dispersed down valley because the middle balances from the Kobokan and Lengkong catchments are
transport reaches extend below the maximum annual rain- similar to those computed (104 m3/km2) on three similar
fall amount. volcanoes, Yakedake, Merapi, and Unzen (Suwa and
Yamakoshi 1999), but the computed sediment mass bal-
ance for a decade is at least one order of magnitude less
Outcomes from Figs. 14 and 15 indicate: than the published values on similar active volcanoes in
wet climate (e.g. Lavigne 2004). Such discrepancy stems
1. Fast, increased aggradation after each BAF eruption for 2 from the methodological problems listed here. Previous
to 5 years and, subsequently, high degradation rate (190– budgets were estimated from volumes of lahars observed
500 m3/m) and sediment mass balance in the 105 m3/km2/ as immediate post-eruption values over a short time inter-
year range. val at one given site on Semeru; these values were not
2. Long-lasting degradation rates in tens of m3/m and sedi- normalized to the diverse geomorphic valley reaches, and
ment mass balances 104–105 m3/km2/year during the the estimates took little account of complex spatial and
post-event period of 10–13 years are followed by weak temporal patterns of aggradation and degradation for each
degradation and low sediment mass balances (102– reach.
103 m3/km2/year). 2. The Kobokan and Lengkong channels show contrasted
3. Modest but constant sediment mass balance of about records of aggradation and degradation cycles in response
102 m3/km2/year is probably linked to daily tephra fall to sediment loading following the 1994–1995 BAF-
supply. forming eruptions (Fig. 14b). Episodic PDC loading has
4. Where BAF-producing eruptions occur twice back to been added at least five times in 30 years to the constant
back (e.g., in 1994 and 1995) instead of every 5 years tephra fall supply at the summit. The geomorphic
Bull Volcanol (2014) 76:857 Page 25 of 26, 857

response to sediment loading had two characteristics: first, and creates terraces in infilling sediment. This suggests a
rapid aggradation for about 5 years was followed by a 6- fast downstream transport of the sediment wave by atten-
year-long degradation, which slowed down after 2008 in uation and dispersion in small rather than large catch-
the Lengkong channel. This low degradation mass bal- ments.
ance (−102 m3/km2/year) is probably a measure of the Second, near decadal sediment yields are two to three
background value between BAF-producing eruptions on orders of magnitude less than annual, immediate post-
Semeru; second, in contrast, the migration of sediment eruption sediment yields after large but infrequent events.
through the Kobokan continued in 2011 as the input was In contrast to the exponential decay of sediment yields
larger in 1994–1995 and the river gradient less steep. The shown after these large events, average sediment yields at
degradation rate in the Kobokan (area 109 km2) is three to persistently active volcanoes in general and Semeru in
eight times that measured in the Lengkong (area 8.2 km2), particular remain one to two orders of magnitude above
suggesting that size of catchments affects how they inte- background level between eruptions. Episodic inputs
grate rainfall and runoff. from BAF eruptions are superimposed on a background
3. Both catchments show different characteristics regarding flux arising from remobilization of quasi-continuous teph-
the river and channel. The Kobokan valley has not yet ra production. During long-term periods of 10 to 30 years,
recovered from the 1994–1995 eruptions and numerous more sediment will be derived from persistently active
associated PDCs. However, regressive erosion, which volcanoes than those with large but infrequent eruptions.
reached the confluence area between 2008 and 2010, is
being blocked by the large dam across the Kobokan. In
contrast, the entrenchment of the Lengkong middle reach Acknowledgments This research project has been funded by the
French National Research Agency “RiskNat” Project “Laharisk” and
L3 is promoted by regressive erosion and decreasing BAF CNES-TOSCA projects. The HSR images were funded by SPOT Image
and LH deposits since 2008. (CNES) and the STIC-ASIA (French Foreign Office) research project
4. Spatial patterns of aggradation and degradation are with CRISP, National University of Singapore. We thank S.C. Liew for
governed by geomorphic characteristics in all reaches. supporting our work in Singapore. Thanks to N. Jasiak, MaGeo (Lille)
and S. Labbé (UMR Tétis, Montpellier) for the post-treatment of Lidar-
Degradation is stronger in middle reaches K2 and L2 based DEMs. We thank L. Thouret, Mr. Mistari, and the Lengkong
and K3 and L3. Aggradation predominates in the villagers for field support. This is Laboratory of Excellence ClerVolc
Kobokan reaches K1 (from PDCs and tephra), K3, and contribution number 109.
K5. In between these two reaches, deposition occurs at a
low rate in the Kobokan reach K4 which has been artifi-
cially modified. Aggradation apparently continues at a
high rate in the Lengkong reach L1. References
5. Temporal patterns of aggradation and degradation and
sediment movement through storage and bypassing Doyle EE, Cronin SJ, Cole SE, Thouret J-C (2010) The coalescence and
reaches are governed by complex relationships between organization of lahars at Semeru, Indonesia. Bull Volc 72(8):961–
970. doi:10.1007/s00445-010-0381-8
rainfall intensity–duration, infiltration rates, and grain size Doyle E, Cronin SJ, Thouret J-C (2011) Cycles of bulking and debulking
changes in deposits down valley and are related to the in lahars at Semeru, Indonesia. Geol Soc Amer Bull 123:1234–
difference in between the tephra and BAF materials and 1246. doi:10.1130/B30227.1
the response of river channels to such changes. Dumaisnil C, Thouret J-C, Chambon G, Doyle EE, Cronin SJ (2010)
Hydraulic, physical and rheological characteristics of rain-triggered
We may reach two basic conclusions: First, in addition lahars at Semeru Volcano, Indonesia. Earth Surf Proc Landf 35:573–
to constant tephra supply, frequent PDC-producing erup- 590. doi:10.1002/esp.2003
tions at persistently active volcanoes favor high aggrada- Gran KB, Montgomery DR (2005) Spatial and temporal patterns in
tion rates due to repeated waves of migrating sediment fluvial recovery following volcanic eruptions: channel response to
basin-wide sediment loading at Mount Pinatubo, Philippines. Geol
(Lisle et al. 2001). Sediment movement appears to be Soc Amer Bull 117(1–2):195–211
driven more by downstream dispersion than translation Harris AJL, Vallance JW, Kimberly PG, Rose WI, Matías O, Flynn LP,
waves for gravel bed rivers as modeled by Lisle et al. Garbeil H (2006) Downstream aggradation owing to lava dome
(2001). At Semeru, two categories of river responses are extrusion and rainfall runoff at Volcan Santiaguito, Guatemala.
Geol Soc Amer Spec Paper 412:85–104
induced by sediment waves: (1) a slow long-term Hayes SK, Mongomery DR, Newhall CG (2002) Fluvial sediment trans-
(>15 years) response to voluminous sediment inputs tend port and deposition following the 1981 eruption of Mount Pinatubo.
to choke multi-thread and meandering, low-gradient Geomorphology 45:211–224
channels of relatively large catchments, inducing a rela- Indonesian Agroclimate and Hydrology Research Institute (IAHRI)
(2008) Indonesia. In: Proc. Internat. Workshop on Integrated
tively slow dispersion wave and (2) a quick (<10 years) Watershed Management for Sustainable Water Use in a Humid
response to limited sediment input by short, narrow, and Tropical Region. JSPS-DGHE Joint Research Project, Tsukuba,
single-thread channels of steeper sub-catchments erodes October: Bull TERC, Univ Tsukuba, No.8 Suppl 2, 2008: 55-64
857, Page 26 of 26 Bull Volcanol (2014) 76:857

Kuenzi WD, Horst OH, McGehee RV (1979) Effect of volcanic activity Pierson TC, Janda RJ, Umbal JV, Daag AS (1992) Immediate and long
on fluvial–deltaic sedimentation in a modern arc-trench gap, south- term hazards from lahars and excess sedimentation in rivers draining
western Guatemala. Geol Soc Amer Bull 90:827–838 Mount Pinatubo, Philippines. US Geol Survey Water-Res Invest
Lavigne F (2004) Rate of sediment yield following small-scale volcanic Report no. 92-4039: 35 p
eruptions: a quantitative assessment at the Merapi and Semeru stra- Pierson TC, Pringle PT, Cameron KA (2011) Magnitude and timing of
tovolcanoes, Java, Indonesia. Earth Surf Proc Landf 29:1045–1058 downstream channel aggradation and degradation in response to a
Lavigne F, Suwa H (2004) Contrasts between debris flows, dome-building eruption at Mt. Hood, Oregon. Geol Soc Amer Bull
hyperconcentrated flows and stream flows at a channel of Mount 123:3–20
Semeru, east Java, Indonesia. Geomorphology 61:45–58 Reimer PJ et al (2009) INTCAL09 and Marine 09 radiocarbon age
Lavigne F, Thouret J-C (2002) Sediment transportation and deposition by calibration curves, 0–50,000 years Cal BP. Radiocarbon 51(4):
rain-triggered lahars at Merapi Volcano, Central Java, Indonesia. 1111–1150
Geomorphology 49(1–2):45–69 Scott KM, Janda RJ, de la Cruz EG, Gabinete E, Eto I, Isada M, Sexon M,
Lisle TE, Cui Y, Parker G, Pizzuto JE, Dodd AM (2001) The dominance Hadley KC (1996) Channel and sedimentation responses to large
of dispersion in the evolution of bed material waves in gravel-bed volumes of 1991 volcanic deposits on the east flank of Mount
rivers. Earth Surf Proc Landf 26:1409–1420 Pinatubo. In: Newhall CG, Punongbayan RS (eds) Fire and mud:
Major JJ (2004) Posteruption suspended sediment transport at Mount St. eruption and lahars of Mount Pinatubo, Philippines. PHIVOLCS
Helens: decadal-scale relationships with landscape adjustments and and Washington Univ Press, Seattle, pp 971–988
river discharges. J Geophys Res 109, F01002. doi:10.1029/ Siswowidjoyo S, Sudarsono U, Wirakusumah AD (1997) The threat of
2002JF000010 hazards in the Semeru volcano region in East Java, Indonesia. J
Major JJ, Mark LE (2006) Peak flow responses to landscape disturbances Asian Earth Sci 15:185–194
caused by the cataclysmic 1980 eruption of Mount St. Helens, Smith GA, Lowe DR (1991) Lahars: volcano-hydrologic events and
Washington. Geol Soc Amer Bull 118:938–958 deposition in the debris flow-hyperconcentrated flow continuum.
Major JJ, Pierson TC, Dinehart RL, Costa JE (2000) Sediment yield In: Fisher RV, Smith GA (eds) Sedimentation in volcanic settings,
following severe volcanic disturbance—a two decade perspective SEPM Special Publication 45. Tulsa, Soc Sedim Geol, pp 59–70
from Mount St Helens. Geology 28:819–822 Solikhin A, Thouret JC, Harris A, Liew SC, Gupta A (2012) Geology,
Manville V, Segschneider B, Newton EH, White JLD, Houghton BF, tectonics, and the 2002–2003 eruption of Semeru volcano,
Wilson CJN (2009) Environmental impact of the 1.8 ka Taupo Indonesia: interpreted from high-spatial resolution satellite imagery.
eruption: landscape responses to a large-scale explosive rhyolitic Geomorphology 138:364–372. doi:10.1016/ j.geomorph. 2011.10.
eruption. Sedim Geol 220:318–336 001
Mizuyama T, Kobashi S (1996) Sediment yield and topographic change Suwa H, Yamakoshi T (1999) Sediment discharge by storm runoff at
after major volcanic activity. In: Erosion and sediment yield: Global volcanic torrents affected by eruption. Zeit für Geomorph Suppl Bd
and regional perspectives (Proceedings Exeter Symposium), Intern 114:63–88
Assoc Hydrol Sci (IAHS) Publication 236: 295-301 Suwa H, Okano K, Kanno T (2011) Forty years of debris flow monitoring
Montgomery DR, Panfil MS, Hayes SK (1999) Channel-bed mobility at Kamikamihorizawa Creek, Mt Yakedake, Japan. In: Genevois,
response to extreme sediment loading at Mount Pinatubo. Geology Hamilton and Prestinzi, eds., Debris-flow hazards mitigation. Casa
27:271–274 Editrice Universita La Sapienza: 605–613
Okano K, Suwa H, Kanno T (2012) Characterization of debris flows by Thouret JC, Lavigne F, Suwa H, Bambang RE, Surono (2007)
rainstorm condition at a torrent on the Mount Yakedake Volcano, Assessment of volcanic hazards at Semeru, Java, with em-
Japan. Geomorphology 136:88–94 phasis on lahars. Bull Volc 70(2):221–244. doi:10.1007/
Pierrot-Deseilligny M (2007) MicMac, un logiciel pour la mise en s00445-007-0133-6
correspondance automatique d'images dans le contexte Vallance JW (2000) Lahars. In: Sigurdsson H (ed) Encyclopedia of
géographique. Bull Inf Scient Technique IGN 77:1–12 (in French) volcanoes. Academic, San Diego, pp 601–616
Pierrot-Deseilligny M, Paparoditis N (2006) A multiresolution and Vallance JW (2005) Volcanic debris flows. In: Jakob M, Hungr O (eds)
optimization-based image matching approach: An application to Debris-flows hazards and related phenomena. Springer, Berlin, pp
surface reconstruction from SPOT5-HRS stereo imagery. In: 247–272
IAPRS, vol XXXVI-1/W41. In: ISPRS workshop on topographic Van Bemmelen RW (1942) De Semoroe. Open File, Direktorat
mapping from space (with special emphasis on small satellites): Volkanologi, VSI, Bandung 12 p (in Dutch)
Ankara, Turkey, 02-2006 Waldron HH (1967) Debris flow and erosion control problems caused by
Pierson TC (2005) Hyperconcentrated flow – transitional process between the ash eruptions of Irazu Volcano, Costa Rica, US. Geol Survey
water flow and debris flow. In Jakob M, Hunger O (eds) Debris flow Bull 1241-I:1–35
hazards and related phenomena. Springer, Berlin, pp 157–202 Wheaton JM, Brasington J, Darby SE, Sear D (2010) Accounting
Pierson TC, Scott KM (1985) Downstream dilution of a lahar: transition for uncertainty in DEMs from repeat topographic surveys:
from debris flow to hyperconcentrated streamflow. Water Res Res improved sediment budgets. Earth Surf Proc Landf 35(2):
21:1511–1524 136–156. doi:10.1002/esp.1886