Evaluating triggering and causative factors of landslides in Lawnon River Basin, Taiwan

Evaluating triggering and causative factors of landslides in Lawnon River

Basin, Taiwan

Meng-Chia Weng

, Min-Hao Wu

⁎

, Shu-Kuang Ning

, Yeun-Wen Jou

a

Department of Civil and Environmental Engineering, National University of Kaohsiung, Kaohsiung, Taiwan b_{Kaohsiung Division, China Engineering Consultants, Inc., Taiwan}

a b s t r a c t

a r t i c l e i n f o

Article history: Accepted 12 July 2011 Available online 30 July 2011 Keywords:

Lawnon River Landslide Rainfall Earthquake

This study evaluates triggering and causative factors of landslides by comparing their occurrence in the Lawnon River Basin prior to and after rainfall and earthquake events over afive-year period (2005–2009). The landslide ratio in the study area was low (less than 4%) before 2007, and significantly increased in the wake of Typhoon Morakot in 2009. The high accumulation of rainfall was the major triggering factor. In addition, the major seismic activity of March 4, 2008 also contributed to landslide occurrence. The combined influence of rainfall and the earthquake is evaluated based on multi-variable regression analysis. Though no significant co-seismic landslides were found after the March 4, 2008 earthquake, its influence on slope stability has been observed from the apparent growth of landslide ratio in the four rainfall events following the quake. Causative factors include the higher landslide ratios occurring in sedimentary rock zones (especially sandstone formations with intercalations of shale), and dip slopes, which are demonstrably prone to plane failures.

© 2011 Elsevier B.V. All rights reserved.

1. Introduction

Taiwan is an island with abundant rainfall. The torrential rainfall accompanying typhoons frequently causes slope failures such as landslides and debrisflows. Moreover, as the island is located in the Circum-Pacific Seismic Belt, an active mountain-building region, there are frequent earthquakes. The dynamic effects of earthquakes also degrade slope stability. Therefore, slope-land disasters are major natural hazards and threaten both human lives and environmental ecology in Taiwan. One of the most important watersheds on the island, the Lawnon River Basin in southern Taiwan, was selected for study. Previous studies have rarely shown this basin as susceptible to landslides and debrisflows. In early August 2009, Typhoon Morakot attacked Taiwan, causing significant loss of lives and property and an economic loss of over US $5 billion (Ling et al., 2009).. It was the most damaging typhoon to make rainfall in southern Taiwan in half a century. The administration system identified 675 dead and twenty-four persons missing. Along with the typhoon, the southwest monsoon brought torrential rains primarily concentrated in southern Taiwan. Accumulated rainfall in the Lawnon River Basin was up to 2500 mm (Fig. 1). As one of the largest river basins in southern Taiwan, it suffered from the severest slope disasters during this event (Chen et al., 2009). About 1218 landslides occurred and the cumulative landslide area exceeded 133.7 km2_{, 6.6% of the entire}

basin's area, damaging infrastructure and transportation systems. However, before typhoon Morakot, large landslides were rarely observed in this area. Consequently, the potential and triggering factors of landslides in the Lawnon River Basin have seldom been studied or emphasized.

According to previous studies (Keefer, 1984; Schuster et al., 1996; Crosta, 2004; Lee et al., 2008a,b), rainfall and earthquakes are two of the principal mechanisms inducing landslides. Precipitation data such as accumulative precipitation and rainfall intensity are usually applied to establish the thresholds of rainfall-induced landslides (Caine, 1980; Vandine, 1985; Keefer et al., 1987; Wieczorek, 1987; Chen et al., 2005). On the other hand, several studies have identi_{fied characteristics of} co-seismic landslides, particularly those caused by large-scale earth-quakes (Keefer, 1984; Harp et al., 1991; Jibson et al., 1994; Harp and Jibson, 1996; Khazai and Sitar, 2004). Co-seismic landslides may not often accompany quakes, especially those of limited scale. Neverthe-less, some seismic activities influence slope stability for a long period of time. In recent years, more and more researchers have paid attention to landslides triggered by heavy rainfall in a region that has suffered a catastrophic earthquake (Lin et al., 2004, 2006; Ku et al., 2006; Chiou et al., 2007; Chen, 2008). In addition to the triggering factors of landslides, it has been generally accepted that slope failures are related to causative factors such as geomorphology, lithology, geological structure and land cover (Radbruch-Hall et al., 1976; Varnes, 1978, 1984; Carrara, 1983; Hansen, 1984; Cruden, 1993). Quantitative analysis of the landslide record, with triggering and causative factors, is fundamental in digitally investigating landslide hazards. Introduction of the Geographical Information System (GIS) ⁎ Corresponding author. Tel./fax: +886 7 5919 750.

E-mail address:wuminhao@nuk.edu.tw(M.-H. Wu).

0013-7952/$– see front matter © 2011 Elsevier B.V. All rights reserved. doi:10.1016/j.enggeo.2011.07.001

Contents lists available atScienceDirect

Engineering Geology

into landslide research has greatly enhanced the ability to collect and analyze landslide data (Wang and Unwin, 1992; Hansen et al., 1995; Mark and Ellen, 1995; Dikau et. al., 1996; Dai and Lee, 2002). The relationship between landslides and the various factors causing them not only provides an insight into our understanding of landslide mechanisms, but can also form a basis for predicting future landslides and assessing the landslide hazard.

This paper aims to clarify the effects of rainfall and earthquakes on landslide occurrence. In addition, causative factors including lithology and topography prone to slope failures are explored. The study area is the Lawnon River Basin, one of the largest river basins in southern Taiwan. Sixteen satellite images of landslides between 2005 and 2009 differentiate the variations of landslide occurrence prior to and after the typhoon and storm events.

2. Study area

The Lawnon River is 137 km long with a basin area of 1373 km2_.

The upstream of the Lawnon River is typical valley topography, formed along a major thrust fault in southern Taiwan, the Lawnon Fault. The downward and lateral erosion has resulted in riverbank scouring, bank collapse, valley widening and riverbank retreating. The topographic characteristics of this section include river terraces, alluvial fans and steep riverbank slopes.

The study area of this research is located upstream of the Lawnon River Basin, around the Bao-Lai hot spring area with ambit of 130.4 km2

(Fig. 2). The Bao-Lai hot spring region, the most populous area in the neighborhood, suffered a severeflood and landslides during Typhoon Morakot. Geology in the study area contains four rock formations, including Tangenshan Sandstone (Tn), Changchikeng Fm. (Cc), Chaujou Fm. (Co), and Mt. Bilu Fm. (Ep). Among them, Tangenshan Sandstone (Tn) and Changchikeng Fm. (Cc) belong to sedimentary rock areas, and Chaujou Fm. (Co) and Mt. Bilu Fm. (Ep) belong to metamorphic rock areas (Fig. 3). The Lawnon Fault is the boundary line between the sedimentary rock and the metamorphic rock. In addition, there are many hot spring sites within the study area, mostly located in Chaujou Fm. (Co). Table 1 shows the geological ages, lithologies and rock compressive strengths corresponding to the geological formations. 3. Methodology

3.1. Data sources and database 3.1.1. Rainfall

The Water Resource Agency (WRA) and Central Weather Bureau (CWB) in Taiwan supplied digitalfiles of hourly rainfall data for twelve rain observation stations (auto-gauges) in the study area. Extremely heavy rainfall data (those with twenty-four-hour accumulative rainfall

(b) Isohyets of accumulative rainfall

Fig. 2. Domain of study area.

over 130 mm) from 2005 to 2009 were collected from rainfall observation stations within 15 km of the Bao-Lai area. Through pre-analysis of the rainfall data, ten significant typhoon and storm events with heavy rainfall from 2005 to 2009 were selected to evaluate effects on landslides for this study (Table 2), and corresponding rainfall data were collected for further statistical and spatial isohyet analyses.

3.1.2. Satellite images

This study uses sixteen images from the FORMOSAT-II satellite (Table 2) from before and after the ten significant typhoon and storm events of 2005 to 2009. FORMOSAT-II satellite is a sun-synchronous

satellite with orbital height of 891 km. It has high-resolution and daily-revisit imagery and four bands with 8-m resolution color mode and 2-m resolution panchromatic mode. Color imaging includes blue, green, red and near-infrared bands corresponding to the wavelengths of 0.45–0.52 m, 0.52–0.60 m, 0.63–0.69 m and 0.76–0.90 m, respec-tively. Since the images identi_{fied landslide areas within the study} domain, they had to be cloud-free. Following the selection of satellite images, preprocessing tasks including geometric correction and radiometric correction were carried out before analysis of the images. 3.1.3. Earthquakes

Records of seismic activities from 2005 to 2009 were collected from Jiasian seismic station (maintained by CWB), the closest seismic monitoring station to the study area. Ten events with detected seismic intensity larger than 2 (larger than 8 gal) were selected. Magnitudes and induced peak ground accelerations at the study area were analyzed.

3.1.4. Geological and topographical data

1:25,000 paper maps published by the Central Geological Survey (CGS) in Taiwan provided geological and topographical data. The data were stored in vector format within the Arc/info GIS software through manual digitization. A digital elevation model (DEM) of 40 × 40 m, obtained from Aerial Survey Office, Forestry Bureau, was constructed from the topographical dataset, and maps of slope-inclination angle and aspect distribution were then derived from the DEM. River networks, roads, and other basic geographical data were also digitized from the basic topographic maps.

3.2. Data digitization and mapping

The accumulative rainfall of each rain observation station was calculated by summarizing hourly records within the whole storm event. The accumulative rainfall data of twelve observation stations within the study area was processed to create the isohyets for ten typhoon and storm events. The spatial analyst function of the GIS was adopted to conduct this work.

Satellite images identified the landslides, and landslide areas were manually digitized with Arc/info GIS software. Comparing pre-event and post-event satellite images, event-based landslide maps could be obtained. Reconnaissance works and in-situ observations were Table 1

Lithological characteristics of the rock in the study area. Formation Geological

age

Lithology UCS (MPa)

Tangenshan (Tn)

Miocene– Pliocene

Thick-bedded to massive graywacke 65.2 Changchikeng

(Cc)

Miocene Light grayfine-grained sandstone with intercalations of shale

41.3 Chaujou (Co) Miocene Black argillite with occasional

inter-beds of sandstone

29.9 Mt. Bilu (Ep) Eocene Black slate and phyllite with occasional

inter-beds of sandstone

33.1

Table 2

Formosat-II satellite images used in the study area.

Year Event Event date Accumulative rainfall (mm) Image Before After 2005 Typhoon Haitang 07/16–07/20 1633 07/10 07/30 Typhoon Talim 08/30–09/01 271 07/30 09/03 2006 06/09 Heavy rainfall 06/06–06/11 1233 04/19 06/16 Typhoon Bilis 07/12–07/15 732 06/16 07/29 2007 Typhoon Sepat 08/16–08/19 1071 01/20 09/02 Typhoon Krosa 10/04–10/07 639 09/02 10/26 2008 07/09 Heavy rainfall 07/08–07/11 155 06/22 07/14 Typhoon Kalmaegi 07/16–07/18 799 07/14 08/21 Typhoon Jangmi 09/26–09/29 507 09/21 10/08 2009 Typhoon Morakot 09/26–09/29 2502 01/08 08/24

Fig. 5. Classification and distribution of landslides (during Typhoon Morakot, 2009). 0 500 1000 1500 2000 2500 3000 0 2 4 6 8 10 12 14 16 18 20

Accumulative rainfall (mm)

Landslide ratio (%)

executed in the study area to understand geological and topographical conditions unseen in the satellite images. Since this study focuses on landslide areas, debris _{flows were not analyzed. To separate} transportation and deposition segments of debris flows from the study area, this study only considered land with slope gradient larger than 25°.Fig. 4illustrates the post-event landslide maps of all storm events for later analysis.

3.3. Landslide categorization

To better determine landslide characteristics, landslides recog-nized in each satellite image were firstly classified into existing landslides and incremental landslides. The difference in images before and after a storm event, showing landslides caused by a speci_{fic event} between images, indicates incremental landslides. Existing landslides

6/1/04 6/1/05 6/1/06 6/1/07 6/1/08 6/1/09

Date

PGA

(gal) & Landslide ratio(%)

0 2 4 6 8 10 12 14 16 18 20 0 500 1000 1500 2000 2500

Accumative rainfall (mm)

(a)

(b)

represent overlap in landslide images before and after the storm event. In order to assess how many landslides are expanded from existing ones, incremental landslides are further classi_{fied into new} landslides and enlarged landslides. New landslides can only be ob-served in the new image, and enlarged landslides have expanded from existing landslides in the previous image.Fig. 5illustrates the distri-bution of existing landslides, enlarged landslides and new landslides caused by rainfall during Typhoon Morakot.

3.4. Attribute assignment and combined database of landslides Assigning various attributes to individual digitized landslides is critical to quantitative analysis of the relationship between landslides and their causative factors. Gathering and handling various attributes of landslides is a time-consuming task, usually requiring muchfield reconnaissance work. GIS makes managing digital databases faster and more accurate. Connecting the landslide database with other internal or external databases is an ef_{ficient way of assigning} causative factor values to individual landslides.

3.5. Regression analysis of landslides and the triggering factors In this study, the spatial overlay function of the GIS helps understand the relationship between landslides and triggering factors (rainfall and earthquake) and causative factors (lithology, geological structure and topography). Furthermore, it is relevant to evaluate the

associated influence of the triggering factors on the landslides. Since landslides are affected by both storm events and earthquakes, a multi-variable regression analysis proposed byJeng et al. (2004)was used in this research.

As landslides were inter-affected by several triggering factors (e.g., x1, x2,…, xn) such that a function F exists, this function could be

expressed as:

F = f xð1; x2;…; xnÞ: ð1Þ

The function form of F was unknown; however, for the sake of simplicity, it was assumed that F has the following form:

F≅ f1ð Þfx1 2ð Þ…fx2 nð Þ:xn ð2Þ

Each function on the right-hand side of Eq.(2) represents the magnitude of the influence of each factor. Empirical functions of f(xi)

could possibly be determined one by one using a single-variable regressive analysis of actual data. The degree of influence varies from one factor to another. The goal was to find the most influential function, followed by the second-most-influential function, and so on. Determination of the influential functions continued until the remaining factors had no meaningful influence, meaning the increase of r2b0.05 when more factors are considered. Then the function forms of f(x1), f(x2),…, f(xn) needed to be further modified by the iteration

process until the coefficient of correlation stopped increasing. Using the aforementioned regressive process, accumulative rain-fall and peak ground acceleration (PGA) are selected as the major triggering factors in this research.

0 4 8 12 16 20 0 4 8 12 16 20

Pr

edicted landslide ratio (%)

Actual landslide ratio (%)

R2_=0.955 0 50 100 150 200 250 0 500 1000 1500 2000 2500

PGA

(gal)

Accumlative rainfall (mm)

(a)

(b)

Fig. 8. Comparison of actual landslide ratio with predicted landslide ratio. The contour curves in (b) represent the empirical function of landslide ratio defined by Eq.(6), and the actual value of landslide ratio is marked near each symbol.

0.0 1.0 2.0 3.0 0 500 1000 1500 2000

Accumulative rainfall (mm)

Landslide ratio (%)

Accumulative rainfall (mm)

Landslide ratio (%)

(a)

(b)

Fig. 9. Comparison of incremental landslide ratio with respect to different geological formations.

4. Analysis results

4.1. In_{fluence of rainfall on landslides}

Rainfall, especially from storms, is the dominant factor inducing landslides. Combining the accumulative rainfall data and landslide ratio of each storm event (Fig. 6), landslide regions of the study area are observably growing in the pastfive years, though some small recovery appeared between two consecutive events. Herein, landslide ratio is expressed as a percentage, yielded from total landslide area divided by overall area within a specific geological zone. Before 2008, the variation of rainfall within the study area was not obvious, so that the variation of landslide areas was not significant (Fig. 6). The extreme rainfall brought by Typhoon Morakot in 2009 remarkably increased the scale of landslides. Comparing rainfall data among events, Typhoon Morakot has the highest accumulative precipitation (2502 mm on average), followed by Typhoon Haitang (1882 mm on average). However, it is significant that the average landslide ratio induced by Typhoon Haitang was only 1.29%, much less than those induced by Typhoons Jangmi and Kalmaegi in 2008.

4.2. Influence of seismic activities on landslides

In order to investigate the influence of other landslide-triggering factors, further examination of the seismic events around the study area is warranted.

Seismic activity records were collected from Jiasian seismic station (Fig. 7b). From 2005 to 2009, ten earthquake events with detected

seismic intensity larger than 2 (PGA≧8 gal) were selected. The peak ground acceleration of each event is plotted in Fig. 7(a) and the distributions of the epicenters with magnitudes for the events are shown in Fig. 7(b). As observable from Fig. 7(a), a remarkable earthquake with PGA near 140 gal occurred on March 4, 2008, which is much larger than the other quakes. Looking back atFig. 6, we can see that the landslide ratio increases since 2008. The seismic event might play an important role in this transition. Therefore, to evaluate the combined influence of the rainfall and the earthquake, a multi-variable regression analysis is conducted in the following section to identify the relationship between landslides and their triggering factors.

4.3. Combined effect of triggering factors on landslides

In the procedures inSection 3.5, it was assumed the landslide ratio (R), could be approximately expressed as:

R = f ARð ; PGAmÞ ≅ f1ð ÞfAR 2ðPGAmÞ ð3Þ

where AR represents the accumulative rainfall and the unit is mm; PGAmrepresents the recorded maximum PGA within one year and the

unit is gal.

The functions f1(AR) and f2(PGAm) could be determined by

iterations of the regression analysis and are found to be: f1ð Þ = 3:03 × 10AR 7ð ÞAR 2 5:58 × 105ð Þ + 0:7AR ð4Þ

Tn

Accumulative rainfall (mm)

Landslide ratio (%)

Enlarged landslide after 2008 New landslide after 2008

Cc

Accumulative rainfall (mm)

Landslide ratio (%)

Enlarged landslide after 2008 New landslide after 2008

(a)

Tangenshan (Tn) formation

(b)

Changchikeng (Cc) formation

Co

Accumulative rainfall (mm)

Landslide ratio (%)

Enlarged landslide after 2008 New landslide after 2008

Ep

Accumulative rainfall (mm)

Landslide ratio (%)

Enlarged landslide after 2008 New landslide after 2008

(c)

Chaujou (Co) formation

(d)

Mt. Bilu (Ep) formation

f2ðPGAmÞ = 1:3 exp 0:012 PGAð mÞ: ð5Þ

Substituting Eqs.(4) and (5)into Eq.(3), the landslide ratio (R) could be expressed in terms of AR and PGAmas:

R = f1ð ÞfAR 2ðPGAmÞ = 1:3 ½3:03 × 107ð ÞAR 2

5:58 × 105ð ÞAR + 0:7 exp 0:012 PGAð mÞ

ð6Þ

where the accumulative rainfall (AR) ranges from 150 mm to 2500 mm and the PGAmranges from 13 gal to 140 gal.

If the landslide ratio expressed by Eq. (6) represented the empirical landslide ratio, it could be compared to the actual ratio (Fig. 8a). In general, the actual R and the empirical R were closely related, with a correlation coefficient square (r2

) of 0.955. As a result, Eq.(6)provided a practical relationship for evaluating the influence of the rainfall and the earthquake. Moreover, Fig. 8b illustrates the variation of landslide ratio with two triggering factors, and the contour lines indicate the empirical landslide ratio and numbers near each symbol mark the actual ratio. Before the earthquake of March 4, 2008 (PGA close to 140 gal), the landslide ratios under various accumulative rainfall conditions fell below 4% (Fig. 8b). After the earthquake, the landslide ratios increase remarkably. Therefore, the combined effect of triggering factors on landslides is significant.

4.4. Influence of lithology on landslides

In addition to triggering factors, Fig. 9 shows the variations of landslide ratio with respect to each geological zone, and the corre-sponded accumulative precipitation record is also plotted. The incre-mental landslide ratio of each geological zone generally remained below 2% before 2008 (Fig. 9a). The landslides frequently occurred in formation Co. However, after the 2008 earthquake the landslide ratios of sedimentary formations (Cc and Tn) surpassed those of metamorphic rock areas (Co and Ep) (Fig. 9b).Fig. 10illustrates the variations of enlarged and new landslides with respect to different geological formations. No matter the lithology, the failure ratios of enlarged landslides is larger than those of new landslides, which means most landslides came from regeneration and expansion of existing landslides. In other words, as a landslide occurs, it will precipitate subsequent failure nearby.

4.5. Influence of dip slope on landslides

A dip slope means a topographic surface where the dip is consistent with that of the underlying strata. Dip slopes are commonly found in cuesta and vale topography. Usually, dip slopes are quite prone to landslides, due to plane failure along persistently weak planes. In the Lawnon River Basin, dip-slope topography is commonly seen in the slope land. According to the definition of Central Geological Survey (CGS) of Taiwan, dip slopes are the slopes with

difference of slope direction and dip direction of weak planes being less than twenty degrees. The distribution of dip slopes in the study area was digitized on the GIS platform (Fig. 11).Fig. 12illustrates the variations of landslide ratios in the dip-slope zone, and it indicates that dip-slope areas indeed possess higher landslide ratios, whether before or after the earthquake in 2008.

5. Conclusion

This study investigates the triggering and causative factors of the storm-induced landslides in the Lawnon River Basin from 2005 to 2009. The analysis led to the following conclusions:

1. From 2005 to 2009, the landslide events and scales increased year by year and reached the peak value at the storm event of Typhoon Morakot 2009, which aroused attention to the landslide behavior in this area. Undoubtedly, rainfall is one of the primary triggering factors causing the slides. However, in comparison to rainfall data among the events except Typhoon Morakot, while Typhoon Haitang in 2005 had higher cumulative rainfall (1882 mm in average) than the other storm events, its landslide area is smaller. The patterns of landslide scale variation and rainfall record seem to be identical after 2008, which implies the existence of other factor (s) triggering the transition of landslide sensitivity.

2. Based on the results of multi-variable regression analysis, the combined influence of the rainfall and the earthquake is evaluated. Before the earthquake of March 4, 2008 (PGA close to 140 gal), the landslide ratios under various accumulative rainfall conditions fell below 4%. After the earthquake, the landslide ratios increase

remarkably. Therefore, the combined effect of triggering factors on landslides is significant.

3. The lithological effect on landslide sensitivity is not apparent before the March 4, 2008 earthquake. After the quake, the sedimentary formations (Cc and Tn) are more prone to the increment of land-slide ratios than the metamorphic rock areas (Co and Ep). 4. Existing landslide areas are more sensitive to landslide

enlarge-ment or developenlarge-ment. The failure ratios of enlarged landslides is larger than those of new landslides, which indicates as a landslide occurs, it will more easily precipitate subsequent failure nearby. 5. The results of this study prove that dip slopes have more

pronounced landslide ratios compared to other slope land both before and after the earthquake in 2008.

Acknowledgment

The research is supported by the National Science Council of Taiwan, Grant No. NSC 7 97-2221-E-390-019 and China Engineering Consultants, Inc., Plan No. 99925.

References

Caine, N., 1980. The rainfall intensity-during control of shallow landslides and debris flows. Geografiska Annaler 62, 23–27.

Carrara, A., 1983. Multivariate models for landslide hazard evaluation. Mathematical Geology 153, 403–427.

Chen, C.Y., 2008. Sedimentary impacts from landslides in the Tachia River Basin, Taiwan. Geomorphology.doi:10.1016/j.geomorph.2008.10.009.

Chen, C.Y., Chen, T.C., Yu, F.C., Yu, W.H., Tseng, C.C., 2005. Rainfall duration and debris-flow initiated studies for real-time monitoring. Environmental Geology 47, 715–724. Chen, T.C., Wu, C.C., Weng, M.C., Hsieh, K.H., Wang, C.C., 2009. Slope failure of Lawnon

basin induced by Typhoon Morakot. Sino-Geotechnics 122, 13–20 (in Chinese). Chiou, S.J., Cheng, C.T., Hsu, S.M., Lin, Y.H., Chi, S.Y., 2007. Evaluating landslides and

sediment yields induced by the Chi-Chi Earthquake and following heavy rainfalls along the Ta-Chia River. Journal of GeoEngineering 2, 73–82.

Crosta, G.B., 2004. Introduction to the special issue on rainfall triggered landslides and debrisflows. Engineering Geology 73, 191–192.

Cruden, D.M., 1993. A simple definition of a landslide. Bulletin of the Association of Engineering Geologists 43, 27–29.

Dai, F.C., Lee, C.F., 2002. Landslide characteristics and slope instability modeling using GIS, Lantau Island, Hong Kong. Geomorphology 42, 213–228.

Dikau, R., Cavallin, A., Jager, S., 1996. Databases and GIS for landslide research in Europe. Geomophology 15, 226–239.

Hansen, A., 1984. Engineering geomorphology: the application of an evolutionary model of Hong Kong's terrain. Zeitschrift für Geomorphologie (Suppl. 51), 39–50. Hansen, A., Franks, C.A.M., Kirk, P.A., Brimicombe, A.J., Fung, T., 1995. Application of GIS to hazard assessment with particular reference to landslides in Hong Kong. In: Carrara, A., Guzetti, F. (Eds.), Geographical Information Systems in Assessing Natural Hazards. Kluwer Academic Publishing, Dordrecht, pp. 273–298. Harp, E.L., Jibson, R.W., 1996. Landslides triggered by the 1994 Northridge, California

earthquake. Bulletin of the Seismological Society of America 86, 319–332. Harp, E.L., Schmidt, K., Wilson, R., Keefer, D.K., Jipson, R.W., 1991. Effects of landslides

coseismic fractures triggered by the 17 October 1989 Loma Prieta, California Earthquake. Landslide News 5, 18–22.

Jeng, F.S., Weng, M.C., Lin, M.L., Huang, T.H., 2004. Influence of petrographic parameters on geotechnical properties of Tertiary sandstones from Taiwan. Engineering Geology 73, 71–91.

Jibson, R.W., Prentice, C.S., Borissoff, B.A., Rogozhin, E.A., Langer, C.L., 1994. Some observations of landslides triggered by the 29 April 1991 Racha earthquake, Republic of Georgia. Bulletin of the Seismological Society of America 84, 964–973. Keefer, D.K., 1984. Landslides caused by earthquakes. Bulletin of Geological Society of

America 95, 406–421.

Keefer, D.K., Wilson, R.C., Mark, R.K., Brabb, E.E., Brown III, W.M., Ellen, S.D., Harp, E.L., Wieczorek, G.F., Alger, C.S., Zatkin, R.S., 1987. Real-time landslide warning during heavy rainfall. Science 238, 921–925.

Khazai, B., Sitar, N., 2004. Evaluation of factors controlling earthquake-induced landslides caused by Chi-Chi earthquake and comparison with the Northridge and Loma Prieta events. Engineering Geology 71, 79–95.

Ku, C.Y., Cheng, C.T., Chi, S.Y., Yu, S.S., Yang, S.D., Chiao, C.H., 2006. Impact of Chi-Chi earthquake on the occurrence of debrisflows: example from the Da-Chia river watershed. Sino-Geotechnics 100, 83–94 (in Chinese).

Lee, C.T., Huang, C.C., Lee, J.F., Pan, K.L., Lin, M.L., Dong, J.J., 2008a. Statistical approach to earthquake-induced landslide susceptibility. Engineering Geology 100 (1–2), 43–58.

Lee, C.T., Huang, C.C., Lee, J.F., Pan, K.L., Lin, M.L., Dong, J.J., 2008b. Statistical approach to storm event-induced landslide susceptibility. Natural Hazard and Earth System Sciences 8, 941–960.

Lin, C.W., Shieh, C.L., Yuan, B.D., Shieh, Y.C., Liu, S.H., Lee, S.Y., 2004. Impact of Chi-Chi earthquake on the occurrence of landslides and debrisflows: example from the Chenyulan River watershed, Nantou, Taiwan. Engineering Geology 71, 49–61.

Cc

Accumulative rainfall (mm)

Landslide ratio (%)

Incremental landslide - dip slope area Incremental landslide after 2008 Incremental landslide after 2008 - dip slope area

Incremental landslide after 2008 - dip slope area

(a)

Changchikeng (Cc) formation

Co

Accumulative rainfall (mm)

Landslide ratio (%)

Incremental landslide - dip slope area Incremental landslide after 2008

(b)

Chaujou (Co)formation

Lin, C.W., Liu, S.H., Lee, S.Y., Liu, C.C., 2006. Impacts of the Chi-Chi earthquake on subsequent rainfall-induced landslides in central Taiwan. Engineering Geology 86, 87–101.

Ling, H.I., Menq, F.Y., Wu, M.H., Huang, Y.C., Hsu, C.W., 2009. Reconnaissance report of the August 8, 2009 Typhoon Morakot; Taiwan. GEER Association Report No. GEER-018, New York.

Mark, R.K., Ellen, S.D., 1995. Statistical and simulation models for mapping debris-flow hazard. In: Carrara, A., Guzetti, F. (Eds.), Geographical Information Systems in Assessing Natural Hazards. Kluwer Academic Publishing, Dordrecht, pp. 93–106. Radbruch-Hall, D.H., Varnes, D.J., Savge, W.Z., 1976. Gravitational speeding of

steep-sided ridges (“sacking”) in Western United States. Bulletin of the International Association of Engineering Geology 14, 23–35.

Schuster, R.L., Nieto, A.S., O'ouke, T.D., Crespo, E., Plaza-Nieto, G., 1996. Mass wasting triggered by the 5 March 1987 Ecuador earthquakes. Engineering Geology 42, 1–23.

Vandine, D.F., 1985. Debrisflows and debris torrents in the Southern Canadian Cordillera. Canada Geotechnique Journal 22, 44–68.

Varnes, D.J., 1978. Slope movement types and processes. In: Schuster, R.L., Krizek, R.J. (Eds.), Landslides: An Analysis and Control, Special Report 176, Transportation Research Board, National Research Council, National Academy of Sciences, Washington, DC, pp. 11–33.

Varnes, D.J., 1984. Landslide Hazard Zonation: A Review of Principles and Practice. UNESCO, Paris. (63).

Wang, S., Unwin, J., 1992. Modeling landslide distribution on losses soil in China. International Journal of Geographical Information Systems 6, 391–405. Wieczorek, G.F., 1987. Effect of rainfall intensity and during on debrisflows in central

Santa Cruz Mountains, California, Flows/Avalanches: process, recognition and mitigation. Geological Society of America. Reviews in Engineering Geology 7, 93–104.