**3. Methodology and results**

provided by the Agricultural and Forestry Aerial Survey Institute of Taiwan, Kuo et al. [7] indicated that the major body of the Hsiaolin landslide had the extent of 57 × 104 m2

**Figure 4.** Formosat-II image and aerial photos of the Hsiaolin landslide in the Cishan River watershed, Taiwan.

42 ± 3 m. Moreover, it is 3.2 km long in an E–W direction and 0.8 to 1.5 km wide. The total fall height was 830 m from the top of the head scarp, at an elevation of 1280 m, to the toe of the landslide deposit at 450 m [8]. Comparing the Hsiaolin landslide with the 1:25000 geologic map (provided by CGS, MOEA) shows that this landslide crops the late Miocene to early Pliocene Yenshuikeng Formation composed of mudstone, sandstone, and shale. Strength of sandstone is much greater than mudstone and shale, and its corresponding uniaxial compressive strength is about 15 Mpa [9]. The source area of the Hsiaolin landslide was the dip slope of the east limb of the syncline that could exhibits simple traces of strata on a horizontal crosssection, however, because the strata and slopes on the east region of the Cishan River shows

the similar characteristics, the bedding traces have rather complicated patterns [8].

generation. The landslide dam was estimated to have a volume of 15.4 million m3

deepest deposit was a thick of 60 m. Moreover, this dam drained about 354 km2

A cascade of loose sediment produced by the Hsiaolin landslide deposited on the Cishan River and generated the barrier lake that was suddenly broken about 1 to 2 hours after its

clear that only half of the total sliding mass contributed to the main body of the landslide dam [10]. The maximum elevation produced by this catastrophic landslide is about 475 m, on the west bank of the Cishan River and the maximum water level that can overtop the dam crest is about an elevation of 413 m. The height of the dam was about 44 m and the

of data recorded by continuous river stream gauging from Shanlin Bridge station, the water level of the Cishan river dropped to 118 m nearly at 08:00 AM after the landslide dam formation and rapidly rise up to 126 m at about 9:40 AM [11]. Hence, we can infer that the landslide dam was suddenly flushed out by river water after the dam formation during the

was estimated to have a volume about 24 ± 2 million m3

area and trapped about 9.9 million m3

period of only 2 hours.

18 Environmental Risks

and

. This is

watershed

, distributed at an average depth of

water before overtopping occurred [10]. On the basis

#### **3.1. Rainfall brought by typhoon Morakot**

On August 8, 2009, Typhoon Morakot was "born" at approximately 22.4°N and 133.8°E in the North Pacific Ocean, about 1000 km far from Northeastern Philippines, moving west at a speed of 10–30 km/h towards Taiwan. In retrospect, Typhoon Morakot had not been considered a serious threat before striking Taiwan. However, contrary to the prediction, after landing Taiwan, it caused more damage than any other typhoon because of massive rainfall, especially in Taiwan's southwestern region. **Figure 5** shows the spatial distribution of rainfall in Taiwan for the six-day rainfall during Typhoon Morakot. This "monster "brought significant rainfalls causing severe debris flows, shallow and deep landslides, and debris dam-break in the mountainous areas of the central and southern Taiwan. Consequently, 675 people were dead; 34 people were hurt; the economic loss was up to 164 million NT dollars.

**Figure 6** shows the time series of hourly rainfall data and cumulative rainfall during Typhoon Morakot obtained from Alishan station in central Taiwan. On the basis of this hourly rainfall

**Figure 5.** Spatial distribution of rainfall in Taiwan for Typhoon Morakot during six days.

2000 years (WRA, 2010). Totally, Typhoon Morakot released about 3063 mm rainfall depth during 6 days. This total value is equivalent to about 90% of the annual rainfall in 2009 at Alishan station. In comparison with other typhoon storms, the magnitude of rainfall brought by Typhoon Morakot is quite huge as shown in **Figure 8**. These patterns indicate that this rainfall event can be characterized as high intensity, huge total accumulation and long duration in a large scale, and hence had generated severe hillslope erosion (e.g., landslides) in watersheds.

Taiwan has total landslides of 45,172 after the massive rainfall brought by Typhoon Morakot,

were identified and mapped from the fused Formosat-II images having resolutions of 2 m, with detailed field checking, and then were digitized into a geographic information system (GIS). Via the fused Formosat-II images (in 2008 and 2009), a pair of successive landslide inventories were mapped through the normalized difference vegetation index (NDVI). The NDVI data are helpful in determining the density of green planet presence via the wave-

where NIR is the near-infrared reflection and IR is the infrared reflection. These two reflections can be obtained by observing the different colors in wavelengths reflected by the plants. For landslide areas, the values of NDVI for a given pixel in the Formosat-II satellite images

**Figure 8.** Cumulative rainfall in relation to its corresponding rainfall duration for different typhoons obtained from the

with an average of 12,488 m2

*NIR* − *IR*

(**Figure 9**). These landslides

Landslides Triggered by Typhoon Morakot in Taiwan http://dx.doi.org/10.5772/intechopen.76930 21

*NIR* <sup>+</sup> *IR* (1)

**3.2. Landslides triggered in the Cishan River watershed**

to 3,510,861 m2

*NDVI* <sup>=</sup> \_\_\_\_\_\_

lengths difference. Mathematically, the NDVI is expressed as follows [13]:

ranging from 258 m2

always range from −1 to 0.

records of Alishan rainfall stations.

**Figure 6.** Time series of hourly rainfall amount and its accumulative rainfall from Aug. 6th to the end of Aug. 8th in 2009. This rainfall data is obtained from the records of the Alishan station.

data, the maximum 1-hour, 6-hour, 24-hour, 48-hour and 72-hour rainfall near the headwaters of the Cishan River watershed were 123 mm, 549 mm, 1623 mm, 2361 mm and 2748 mm, respectively. Both the previous 1-hour (113 mm by Typhoon Herb) and 48-hour (1978 mm by Typhoon Herb) rainfall records were broken. The 24-hour and 48-hour rainfall records approximate the world record (1825 mm and 2476 mm, respectively) [12], as shown in **Figure 7**. Moreover, the return periods for 24-hour, 48-hour, and 72-hour rainfall at the Alishan station both exceeded

**Figure 7.** Diagram of rainfall depth-duration for Typhoon Morakot at the Alishan rainfall station near the Cishan River watershed in comparison with the maximum rainfall records in the world.

2000 years (WRA, 2010). Totally, Typhoon Morakot released about 3063 mm rainfall depth during 6 days. This total value is equivalent to about 90% of the annual rainfall in 2009 at Alishan station. In comparison with other typhoon storms, the magnitude of rainfall brought by Typhoon Morakot is quite huge as shown in **Figure 8**. These patterns indicate that this rainfall event can be characterized as high intensity, huge total accumulation and long duration in a large scale, and hence had generated severe hillslope erosion (e.g., landslides) in watersheds.

#### **3.2. Landslides triggered in the Cishan River watershed**

data, the maximum 1-hour, 6-hour, 24-hour, 48-hour and 72-hour rainfall near the headwaters of the Cishan River watershed were 123 mm, 549 mm, 1623 mm, 2361 mm and 2748 mm, respectively. Both the previous 1-hour (113 mm by Typhoon Herb) and 48-hour (1978 mm by Typhoon Herb) rainfall records were broken. The 24-hour and 48-hour rainfall records approximate the world record (1825 mm and 2476 mm, respectively) [12], as shown in **Figure 7**. Moreover, the return periods for 24-hour, 48-hour, and 72-hour rainfall at the Alishan station both exceeded

**Figure 7.** Diagram of rainfall depth-duration for Typhoon Morakot at the Alishan rainfall station near the Cishan River

**Figure 6.** Time series of hourly rainfall amount and its accumulative rainfall from Aug. 6th to the end of Aug. 8th in 2009.

This rainfall data is obtained from the records of the Alishan station.

20 Environmental Risks

watershed in comparison with the maximum rainfall records in the world.

Taiwan has total landslides of 45,172 after the massive rainfall brought by Typhoon Morakot, ranging from 258 m2 to 3,510,861 m2 with an average of 12,488 m2 (**Figure 9**). These landslides were identified and mapped from the fused Formosat-II images having resolutions of 2 m, with detailed field checking, and then were digitized into a geographic information system (GIS). Via the fused Formosat-II images (in 2008 and 2009), a pair of successive landslide inventories were mapped through the normalized difference vegetation index (NDVI). The NDVI data are helpful in determining the density of green planet presence via the wavelengths difference. Mathematically, the NDVI is expressed as follows [13]:

$$\text{NDVI} = \frac{\text{NIR} - \text{IR}}{\text{NIR} + \text{IR}} \tag{1}$$

where NIR is the near-infrared reflection and IR is the infrared reflection. These two reflections can be obtained by observing the different colors in wavelengths reflected by the plants. For landslide areas, the values of NDVI for a given pixel in the Formosat-II satellite images always range from −1 to 0.

**Figure 8.** Cumulative rainfall in relation to its corresponding rainfall duration for different typhoons obtained from the records of Alishan rainfall stations.

**Figure 9.** Location of landslides in Taiwan mapped after Typhoon Morakot. These 45,124 landslides range from 258 m2 to 3,510,861 m2 , with an average of 12,488 m2 .

assessment and management [14]. Based on the analysis of landslide inventories in worldwide regions, frequency distributions and PDFs (probability density functions) exist in landslide magnitude with heavy tail for medium and large landslides. The mathematical expression of this algebraic PDF right tail decay is generally explained as a linear fitting for power-law scaling [15–18, 45, 46]. Here we characterize the noncumulative frequency-area relations to

**Figure 10.** Landslide inventories mapped from a pair of Formosat-II satellite images in 2008 and 2009, respectively, for

<sup>−</sup>*<sup>β</sup>* (2)

Landslides Triggered by Typhoon Morakot in Taiwan http://dx.doi.org/10.5772/intechopen.76930 23

is the number of landslides with area between *AL*

and the *NLT* is the total number of landslides. The constants *C* and *β* are

) of 0.99. This result was derived from the scaling of the landslide area

obtained from fitting medium and large landslides in order to detect the right heavy tailed

, so that bin widths are approximately equal in logarithmic coordinates. In the CRW, the noncumulative relation of the total 2338 landslides caused by Typhoon Morakot in 2009

power law frequency-area scaling with exponent, *β* = 1.69, intercept, *C* = 60, and a determi-

) is a PDF (probability density func-

with increasing area

, the landslides displayed

clarify the landslide behavior in the CRW.

*NLT* <sup>Δ</sup> *<sup>N</sup>*\_\_\_\_*<sup>L</sup>* Δ *AL*

and 2.48 km2

where *AL*

*AL*

and *AL* + △*AL*

tion) equivalent to \_\_\_1

nation coefficient (*r2*

between 645 m2

*p*(*AL*) = *CAL*

The present paper now considers the noncumulative relation as:

the Cishan River watershed. (a) Pretyphoon image in 2008. (b) Posttyphoon image in 2009.

is the magnitude of landslide area, the *p*(*AL*

, and the △*NL*

based on Eq. 1 is given in **Figure 12**. Above the cutoff at 645 m<sup>2</sup>

.

decay of PDF through a power-law. We increase our bin width △*AL*

As shown in **Figure 9**, the landslide areas after Typhoon Morakot mainly concentrate on southern Taiwan that contains the Cishan River watershed (CRW). **Figure 10** shows the landslide area (see red colored regions in **Figure 8**) before (in 2008) and after (in 2009) Typhoon Morakot. In contrast, the total landslide area increased from 7.83 km2 to 33.98 km2 and the landslide density increased from 5.7 × 10−3 no./ha to 28.4 × 10−3 no./ha (total number of landslide/survey area) after the disturbance of Typhoon Morakot. On average, the mean landslide area is 0.012 km2 , ranging from a smallest area of 0.12 m2 to a maximum value of ~2.49 km2 . These patterns indicate that Typhoon Morakot caused severe disturbances on the Cishan River watershed and could lead to significant changes in the geomorphic systems of this region, which we discuss later.

In addition, landslide rank against cumulative landslide area can exhibits the contribution of different landslide magnitude to erosion processes. In this investigation region, landslides were ranked in a series of numbers from the smallest area (No. 1) to the largest area (No. 2338), as shown in **Figure 11**. As a consequence, the largest 10 landslides (0.43% of total number) account for 27% of total landslide area, the largest 58 landslides (2.5% of total number) account for 50% of total landslide area, and the largest 649 landslides (27% of total number) account for 90% of total landslide area. This result indicated that large landslides can majorly dominate denudation processes in the CRW.

These landslide data were further analyzed by using a fractal model. The dependence of landslide frequency on landslide magnitude is an aspect of critical importance to hazard risk

**Figure 10.** Landslide inventories mapped from a pair of Formosat-II satellite images in 2008 and 2009, respectively, for the Cishan River watershed. (a) Pretyphoon image in 2008. (b) Posttyphoon image in 2009.

assessment and management [14]. Based on the analysis of landslide inventories in worldwide regions, frequency distributions and PDFs (probability density functions) exist in landslide magnitude with heavy tail for medium and large landslides. The mathematical expression of this algebraic PDF right tail decay is generally explained as a linear fitting for power-law scaling [15–18, 45, 46]. Here we characterize the noncumulative frequency-area relations to clarify the landslide behavior in the CRW.

The present paper now considers the noncumulative relation as:

As shown in **Figure 9**, the landslide areas after Typhoon Morakot mainly concentrate on southern Taiwan that contains the Cishan River watershed (CRW). **Figure 10** shows the landslide area (see red colored regions in **Figure 8**) before (in 2008) and after (in 2009) Typhoon

**Figure 9.** Location of landslides in Taiwan mapped after Typhoon Morakot. These 45,124 landslides range from 258 m2

landslide density increased from 5.7 × 10−3 no./ha to 28.4 × 10−3 no./ha (total number of landslide/survey area) after the disturbance of Typhoon Morakot. On average, the mean landslide

These patterns indicate that Typhoon Morakot caused severe disturbances on the Cishan River watershed and could lead to significant changes in the geomorphic systems of this

In addition, landslide rank against cumulative landslide area can exhibits the contribution of different landslide magnitude to erosion processes. In this investigation region, landslides were ranked in a series of numbers from the smallest area (No. 1) to the largest area (No. 2338), as shown in **Figure 11**. As a consequence, the largest 10 landslides (0.43% of total number) account for 27% of total landslide area, the largest 58 landslides (2.5% of total number) account for 50% of total landslide area, and the largest 649 landslides (27% of total number) account for 90% of total landslide area. This result indicated that large landslides can majorly

These landslide data were further analyzed by using a fractal model. The dependence of landslide frequency on landslide magnitude is an aspect of critical importance to hazard risk

to 33.98 km2

to a maximum value of ~2.49 km2

and the

.

Morakot. In contrast, the total landslide area increased from 7.83 km2

.

, ranging from a smallest area of 0.12 m2

area is 0.012 km2

to 3,510,861 m2

22 Environmental Risks

region, which we discuss later.

dominate denudation processes in the CRW.

, with an average of 12,488 m2

$$p(A\_{\downarrow}) = CA\_{\downarrow}^{\to} \tag{2}$$

where *AL* is the magnitude of landslide area, the *p*(*AL* ) is a PDF (probability density function) equivalent to \_\_\_1 *NLT* <sup>Δ</sup> *<sup>N</sup>*\_\_\_\_*<sup>L</sup>* Δ *AL* , and the △*NL* is the number of landslides with area between *AL* and *AL* + △*AL* and the *NLT* is the total number of landslides. The constants *C* and *β* are obtained from fitting medium and large landslides in order to detect the right heavy tailed decay of PDF through a power-law. We increase our bin width △*AL* with increasing area *AL* , so that bin widths are approximately equal in logarithmic coordinates. In the CRW, the noncumulative relation of the total 2338 landslides caused by Typhoon Morakot in 2009 based on Eq. 1 is given in **Figure 12**. Above the cutoff at 645 m<sup>2</sup> , the landslides displayed power law frequency-area scaling with exponent, *β* = 1.69, intercept, *C* = 60, and a determination coefficient (*r2* ) of 0.99. This result was derived from the scaling of the landslide area between 645 m2 and 2.48 km2 .

This phenomenon is described by self-organized criticality (SOC). Bak [19] indicates that if a system is well described by power-law scaling over a large portion of its event magnitude range, the system may be in a quasi-static state with SOC. In the CRW, landslides caused by Typhoon Morakot satisfies (2) with *β* = 1.69 and suggest that landslide triggering has SOC in this region. **Table 1** lists the values of *β* based on (2) for the frequency distribution of landslide area derived from the worldwide regions. In comparison, the value of *β* in the CRW is smaller than 2.5 in this study, among the lowest observed worldwide [27], showing our study area has

Landslides tend to occur in groups. Occurrence of landslides on hillslopes is therefore influenced by the geomorphic characteristics of subaerial surface because some geomorphic condition can provide suitable environments for the development of landslide. Hence, the investigation of landslide abundance and its corresponding spatial distribution in relation to the factors in situ is necessary for landslide hazard assessment. Here we used 5 × 5 m digital elevation model (DEM) (provided by Water Resources Agency, WRA) to analyze different geomorphic settings such as elevation, slope, aspect, surface residual, surface curvature, and fractal dimension in the CRW. Also, the distribution of lithologic complexes was obtained from the geological map provided by the Center of Geological Survey, Taiwan. These above mentioned factors were

higher societal risk caused by the higher occurrence probability of large landslides.

examined with the location of landslide triggering in the CRW, as tabulated in **Table 2**.

**Region α β** Japan [20] 0.89 1.89 Western Southern Alps, New Zealand (Hovius et al., 1997) 1.1 2.1 Challana Valley, Bolivia [21] 1.6 2.6 Alameda, USA [21] 2.3 3.3 Umbria, central Italy [22] 1.5 2.5 California, USA [22] 1.3 2.3 Lombardy, northern Italy [23] 0.85 1.85 Southern California, USA [24] 1.1 2.1 Fiordland, New Zealand [25] 1.07 2.07 Val di Fassa, Italy [26] 1.56 2.56 TWR and CYL watersheds, Taiwan [18] 0.65 1.65 CRW, Taiwan 0.69 1.69

divided by the watershed area A, i.e., *Ca* = *ALT*/A. Ca expresses the overall extent of damage caused by landslides done to the land [18, 28] and this proportional damage indicator is used

**Table 1.** Exponent values of fractal model for landslide frequency-area distribution collected form the worldwide

) for a watershed is defined as the total landslide area ALT

Landslides Triggered by Typhoon Morakot in Taiwan http://dx.doi.org/10.5772/intechopen.76930 25

The percentage of landslide area (*Ca*

observations.

**3.3. Landslide triggering response to geomorphic characteristics on hillslopes**

**Figure 11.** Cumulative percentage of landslide area against to the ranking numbers of 2338 landslides identified after typhoon Morakot in 2009 in the CRW. These landslides were ranked in a series of numbers from the minimum area (no. 1) to the maximum area (no. 2338). The cumulative area of the first 58 largest landslides accounts for 50% of the total landslide area while the first 649 largest landslides for 90% of the total landslide area.

**Figure 12.** Noncumulative landslide frequency–area distributions in the CRW, based on landslide data after typhoon Morakot in 2009. The landslides of area larger than 645 m2 display a fractal model with *β* = 1.69.

This phenomenon is described by self-organized criticality (SOC). Bak [19] indicates that if a system is well described by power-law scaling over a large portion of its event magnitude range, the system may be in a quasi-static state with SOC. In the CRW, landslides caused by Typhoon Morakot satisfies (2) with *β* = 1.69 and suggest that landslide triggering has SOC in this region. **Table 1** lists the values of *β* based on (2) for the frequency distribution of landslide area derived from the worldwide regions. In comparison, the value of *β* in the CRW is smaller than 2.5 in this study, among the lowest observed worldwide [27], showing our study area has higher societal risk caused by the higher occurrence probability of large landslides.

#### **3.3. Landslide triggering response to geomorphic characteristics on hillslopes**

Landslides tend to occur in groups. Occurrence of landslides on hillslopes is therefore influenced by the geomorphic characteristics of subaerial surface because some geomorphic condition can provide suitable environments for the development of landslide. Hence, the investigation of landslide abundance and its corresponding spatial distribution in relation to the factors in situ is necessary for landslide hazard assessment. Here we used 5 × 5 m digital elevation model (DEM) (provided by Water Resources Agency, WRA) to analyze different geomorphic settings such as elevation, slope, aspect, surface residual, surface curvature, and fractal dimension in the CRW. Also, the distribution of lithologic complexes was obtained from the geological map provided by the Center of Geological Survey, Taiwan. These above mentioned factors were examined with the location of landslide triggering in the CRW, as tabulated in **Table 2**.

The percentage of landslide area (*Ca* ) for a watershed is defined as the total landslide area ALT divided by the watershed area A, i.e., *Ca* = *ALT*/A. Ca expresses the overall extent of damage caused by landslides done to the land [18, 28] and this proportional damage indicator is used


**Table 1.** Exponent values of fractal model for landslide frequency-area distribution collected form the worldwide observations.

**Figure 12.** Noncumulative landslide frequency–area distributions in the CRW, based on landslide data after typhoon

**Figure 11.** Cumulative percentage of landslide area against to the ranking numbers of 2338 landslides identified after typhoon Morakot in 2009 in the CRW. These landslides were ranked in a series of numbers from the minimum area (no. 1) to the maximum area (no. 2338). The cumulative area of the first 58 largest landslides accounts for 50% of the total

landslide area while the first 649 largest landslides for 90% of the total landslide area.

24 Environmental Risks

display a fractal model with *β* = 1.69.

Morakot in 2009. The landslides of area larger than 645 m2


in the evaluation of other hazard [29]. Landslide density *DL*

equivalent to ratio of landslide number in the region to total watershed area. *RL*

percentage of landslide area in the region in relation to the corresponding region area.

accounting for 13.98% with landslide density of 8.19 of 2.89 No./km<sup>2</sup>

of each geomorphic characteristic to the extent of geomorphic characteristics. These landslide variables can help us to clarify the landslide triggering in response to diverse geomorphic and

**Classification** *C<sup>a</sup>*

2.4~2.6 2.02 10.58 4.91 2.6~2.8 0.08 0.42 0.66 2.8~3 0.003 0.02 0.09

Fractal dimension 2.2~2.4 2.47 12.96 6.43

is the percentage of landslide area in the region in relation to total watershed area. *DL*

**Table 2.** Landslide distribution in response to geomorphic and geologic settings in the CRW.

As shown in **Table 2**, landslide distributed in the different geomorphic and geologic settings cropping out CRW after Typhoon Morakot. The largest elevation group in the CRW is the region lower than 500 m that extends 44% of total watershed area. This region only cropped

) of 2.89 No./km<sup>2</sup>

slides. The smallest unit is the region that its elevation is between 500 m and 1000 m, only

10.51%. In addition, the region of elevation higher than 1000 m has landslide density of 14.89

On the basis of the Soil and Water Conservation Technical Guide issued from Taiwan government, the terrain gradient in the CRW can be categorized into seven classes as well as shown in **Table 2**. In this watershed, the hillslopes are most widely mantled by the sixth-class slope gradient (between 55% and 100%, i.e., between 28.8° and 45°; see **Table 1**), with 32.9% of the

and landslide rate (7.45%), respectively. In other words, the smallest unit is the gradient of hillslopes less than 5% (i.e., 2.86°) and therefore has the lowest values for landslide density of 0.07

(**Table 2**). It is clear that there are no evident differences in each hillslope aspect that accounts for about 8–15% of the total watershed area. The largest unit is west-facing slopes that extends

suggest that there are no significant variations in landslide distribution in diverse hillslope

The above mentioned geomorphic characteristics are usually used to be as landslide triggers, and to examine landside occurrence in response to geomorphic systems. However, those

and *RL*

, *DL*

to 3.91 No./km<sup>2</sup>

aspects, indicating the aspect has low effect on landslide triggering.

) and landslide rate (4.68%). Of particularly, landslides in the CRW after Typhoon Morakot

and landslide rate of 0.09%. Aspects of hillslopes were also classified into eight groups

and landslide rate of 6.47%. Results show that landslides tend to occur at the region

) and has the greatest values for landslide density (13.67 No./km<sup>2</sup>

CRW area, with the quite small values of landslide density (3.5 No./

in eight aspect groups, ranging from 0.47 to

and from 3.43 to 5.26%, respectively. These patterns

slide to the watershed area. Landslide rate *RL*

of 0.52%, landslide density (*DL*

of elevation higher than 500 m on hillslopes.

total watershed area (842 km2

15.33% of the 843 km2

0.7%, from 2.62 No./km<sup>2</sup>

cropped out the similar values of *Ca*

geologic settings.

out *Ca*

*Ca*

No./km<sup>2</sup>

No./km<sup>2</sup>

km2

is the ratio of numbers of land-

is the landslide rate equivalent to the

**, %** *D<sup>L</sup> R<sup>L</sup>*

Landslides Triggered by Typhoon Morakot in Taiwan http://dx.doi.org/10.5772/intechopen.76930

**,**

27

is the landslide density

and is covered 1.16% of total by land-

and has landslide rate of

)

is the percentage of landslide area in the region


*Ca* is the percentage of landslide area in the region in relation to total watershed area. *DL* is the landslide density equivalent to ratio of landslide number in the region to total watershed area. *RL* is the landslide rate equivalent to the percentage of landslide area in the region in relation to the corresponding region area.

**Table 2.** Landslide distribution in response to geomorphic and geologic settings in the CRW.

**Classification** *C<sup>a</sup>*

500~1000 1.47 8.19 11.74 >1000 2.67 14.89 6.92

Northeast 0.51 2.83 6.75 East 0.62 3.43 6.64 Southeast 0.70 3.91 5.95 South 0.57 3.15 4.43 Southwest 0.68 3.76 4.91 West 0.63 3.50 4.26 Northwest 0.49 2.76 3.56

5% < S < =15% 0.08 0.42 1.06 15% < S < =30% 0.31 1.72 2.49 30% < S < =40% 0.46 2.55 4.29 40% < S < =55% 1.11 6.21 6.89 55% < S < =100% 2.45 13.67 8.05 S > 100% 0.23 1.30 7.07

Slate 0.13 0.75 25.45 Interbedded slate and sandstone 0.08 0.43 3.05 Interbedded sandstone and shale 4.07 22.67 6.00 Hard sandstone 0.13 0.73 4.46 Hard shale 0.17 0.92 3.35 Gravel 0.00 0.02 0.09

Flat 0.08 0.42 1.57 Concave 2.60 14.50 5.71

0.3~0.6 2.03 10.81 5.19 0.6~0.9 0.80 4.28 4.46 0.9~1.2 0.23 1.25 4.45 1.2~1.5 0.03 0.15 3.17 1.5~1.8 0.00 0.02 6.54 2~2.2 0.00 0.03 1.61

Elevation 0–500 0.52 2.89 1.17

26 Environmental Risks

Aspect North 0.47 2.62 4.65

Slope gradient S < =5% 0.01 0.07 0.09

Geologic settings Loam 0.08 0.45 0.76

Curvature Convex 1.98 11.03 4.40

Surface roughness 0~0.3 1.50 7.99 4.71

**, %** *D<sup>L</sup> R<sup>L</sup>*

**,**

in the evaluation of other hazard [29]. Landslide density *DL* is the ratio of numbers of landslide to the watershed area. Landslide rate *RL* is the percentage of landslide area in the region of each geomorphic characteristic to the extent of geomorphic characteristics. These landslide variables can help us to clarify the landslide triggering in response to diverse geomorphic and geologic settings.

As shown in **Table 2**, landslide distributed in the different geomorphic and geologic settings cropping out CRW after Typhoon Morakot. The largest elevation group in the CRW is the region lower than 500 m that extends 44% of total watershed area. This region only cropped out *Ca* of 0.52%, landslide density (*DL* ) of 2.89 No./km<sup>2</sup> and is covered 1.16% of total by landslides. The smallest unit is the region that its elevation is between 500 m and 1000 m, only accounting for 13.98% with landslide density of 8.19 of 2.89 No./km<sup>2</sup> and has landslide rate of 10.51%. In addition, the region of elevation higher than 1000 m has landslide density of 14.89 No./km<sup>2</sup> and landslide rate of 6.47%. Results show that landslides tend to occur at the region of elevation higher than 500 m on hillslopes.

On the basis of the Soil and Water Conservation Technical Guide issued from Taiwan government, the terrain gradient in the CRW can be categorized into seven classes as well as shown in **Table 2**. In this watershed, the hillslopes are most widely mantled by the sixth-class slope gradient (between 55% and 100%, i.e., between 28.8° and 45°; see **Table 1**), with 32.9% of the total watershed area (842 km2 ) and has the greatest values for landslide density (13.67 No./km<sup>2</sup> ) and landslide rate (7.45%), respectively. In other words, the smallest unit is the gradient of hillslopes less than 5% (i.e., 2.86°) and therefore has the lowest values for landslide density of 0.07 No./km<sup>2</sup> and landslide rate of 0.09%. Aspects of hillslopes were also classified into eight groups (**Table 2**). It is clear that there are no evident differences in each hillslope aspect that accounts for about 8–15% of the total watershed area. The largest unit is west-facing slopes that extends 15.33% of the 843 km2 CRW area, with the quite small values of landslide density (3.5 No./ km2 ) and landslide rate (4.68%). Of particularly, landslides in the CRW after Typhoon Morakot cropped out the similar values of *Ca* , *DL* and *RL* in eight aspect groups, ranging from 0.47 to 0.7%, from 2.62 No./km<sup>2</sup> to 3.91 No./km<sup>2</sup> and from 3.43 to 5.26%, respectively. These patterns suggest that there are no significant variations in landslide distribution in diverse hillslope aspects, indicating the aspect has low effect on landslide triggering.

The above mentioned geomorphic characteristics are usually used to be as landslide triggers, and to examine landside occurrence in response to geomorphic systems. However, those geomorphic characteristics are simple and using those characteristics (i.e., elevation, slope and aspect) could not represent complex geomorphic systems in natural environments, e.g., convergent slopes and surface roughness. To consider the complexity of geomorphic conditions, we employed curvature, surface roughness and fractal dimension, comparing with the location of landslide occurrence to find out the 'hotspot' or 'prone area' in the CRW.

where *h* is the lag between measured cells and n is the number of pairs considered. Landserf calculates *D* within a moving window around each cell across the raster. This calculation can represent how the surface roughness/complexity changes over the study area. *D* is computed at different window sizes for n ≥ 9 and its value is between 2 (flat surface) and 3 (a space filling rough surface). In the CRW, the fractal dimension of land surface was digitalized in Landserf, then comparing with the landslide inventory after Typhoon Morakot (**Table 2**). Classification of fractal dimension in the CRW is defined as 2.0–2.2, 2.2–2.4, 2.4–2.6, 2.6–2.8 and 2.8–3. Results show that the most regions have the values of fractal dimension from 2.2 to 2.4 and from 2.4 to 2.6, accounting for 38.45% and 40.54%, respectively. Meanwhile, percentage of landslide area, landslide densities and landslide rates are quite larger in these regions (*D* values between 2.2 and 2.4 and between 2.4 and 2.6) than that in other regions. In particular,

In general, Geologic settings determine the strength of rocks that can further influence sediment production caused by weathering processes on hillslopes. Soft rock can usually lead to weak resistance to erosion. Based on geologic map provided by Central Geological Survey, MOEA, Taiwan, we can find that the largest geologic unit is interbedded sandstone and shale, cropping out about 72% of the total CRW area. We also examined landslide distribution in

are 4.07%, 22.67 No./km<sup>2</sup>

region of interbedded sandstone and shale and these values are much greater than that in other regions, as shown in **Table 2**. Other regions crop out geologic units of gravel, loamy sand, slate, interbedded slate and sandstone, hard sandstone, and hard shale, respectively. These geologic units are usually recognized as hard rocks and structural settings. Although some of them are composed of loose materials that are tended to landslide triggering (loamy sand and gravel), but just crops out only 10.83 and 4.52% of total watershed area in the CRW. The geologic setting crops out interbedded sandstone and shale that has weak, low resistance to erosion and lies most regions of the CRW, favoring the generation of landslides on hillslopes. The above landslide inventory shows that landslide triggering can be influenced by geomorphic and geologic settings. The hillslope with elevation of 500 m, slope between 28.8° and 45°, convex slopes, surface roughness index from 0 m to 0.6 m, fractal dimension from 2.2 to 2.6 and geologic unit composed of sandstone has high potential for the development of

Denudation processes play a very important role in the fluvial systems of a watershed. A large amount of sediment materials produced by landslides could be entrained downslope from hillslopes into river channels, influencing the evolution of river morphology. Typhoon Morakot led to severe hillslope erosion in the CRW (**Figure 8**), and its consequent generation of sediment materials could deposit on hillslopes and river channels, changing the geomorphic response of the CRW. Here we compared the river borne suspended sediment and river bathymetry of the Cishan River to show the aftermath effect of Typhoon Morakot on the CRW.

response to geologic settings. Results show that percentage of landslide area *Ca*

in the regions of *D* values ranging from

Landslides Triggered by Typhoon Morakot in Taiwan http://dx.doi.org/10.5772/intechopen.76930

and 5.66%, respectively, in the

, landslide

29

total landslide density is very high with 23 No./km<sup>2</sup>

, and landslide rate *RL*

**3.4. Aftermath of typhoon Morakot in the CRW**

density *DL*

landslides.

2.2 to 2.6, indicating that landslides tend to occur in this region.

Curvature is the changing rate of slope along *x* and y direction and can be categorized into three groups of divergent, flat and convergent based on curvature values of greater than zero, equal to zero and smaller than zero, respectively [47], as shown in **Table 2**. In comparison, divergent and convergent slopes individually lie the most regions of the total watershed area (about 47 and 48%). On the divergent slope, landslide triggering after Typhoon Morakot accounted for 1.98% of the total watershed area with 11.03 landslides per km2 and landslide rate of 4.21%. In other words, the convergent slopes have 2.6% percentage of landslide area comparing with the area of the CRW, landslide density of 14.5 No./km<sup>2</sup> and landslide rate of 5.4%. Results show that both of divergent and convergent slopes can provide suitable condition for the development of landslides. Divergent slopes can lead to landslide triggering due to the effect of gravity along slope. Convergent slopes can concentrate water flow leading to increase in groundwater level and pore water pressure, therefore generating landslides on hillslopes. However, **Table 2** still indicates that landslides slightly tend to occur on convergent slopes in comparison with divergent slopes in the CRW. In addition, the flat region is not appropriate to landslide triggering and hence strictly has 0.42 landslides per km2 .

Here we define the surface roughness is a residual topography that can be derived from the cell by cell subtraction of original 5 × 5 m DEM and the mean of this DEM. Mean DEM was created by averaging elevation values within a 3-cell moving window. The raster of residual topography was then calculated as the cell-by-cell difference between the original DEM and the mean DEM. In the Cishan River watershed, surface roughness (i.e., residual topography) was calculated as the above mentioned flow and further categorized into six groups by 0–0.3, 0.3–0.6, 0.6–0.9, 0.9–1.2, 1.2–1.5 and 1.5–1.8. Based on this classification, the largest unit of surface roughness is the region of the values from 0.3 to 0.6 that covers about 39.32% of the total CRW area, with 10.81 landslides per km<sup>2</sup> respect to this region. Note that there is no significant difference between landslide rates in each surface roughness settings, ranging from 3.07% to 6.14%. These patterns suggest that landslides tend to be triggered in group in the region of surface roughness between 0 to 0.6. Above this cutoff, the landslide density and landslide rate evidently decrease as well as shown in **Table 2**.

In other words, natural landscapes generally have fractal characteristics. Terrain can be considered self-similar in the two horizontal directions and self-affine in cross-section [30]. This leads to the discussion of the relationships of landslide triggering with respect to the selfsimilarity or fractals of topography be necessary. Fractal dimension in the CRW was calculated using the Landserf software on the basis of 5 × 5 m DEM, comparing with the location of landslide occurrence. Landserf implements the variogram method (e.g., [31]), which we use in this study. The variogram is calculated as:

$$D = \frac{1}{2n(h)} \sum\_{i=1}^{n} \sum\_{j=1}^{n} \left(\mathbf{z}\_i - \mathbf{z}\_j\right)^2 \tag{3}$$

where *h* is the lag between measured cells and n is the number of pairs considered. Landserf calculates *D* within a moving window around each cell across the raster. This calculation can represent how the surface roughness/complexity changes over the study area. *D* is computed at different window sizes for n ≥ 9 and its value is between 2 (flat surface) and 3 (a space filling rough surface). In the CRW, the fractal dimension of land surface was digitalized in Landserf, then comparing with the landslide inventory after Typhoon Morakot (**Table 2**). Classification of fractal dimension in the CRW is defined as 2.0–2.2, 2.2–2.4, 2.4–2.6, 2.6–2.8 and 2.8–3. Results show that the most regions have the values of fractal dimension from 2.2 to 2.4 and from 2.4 to 2.6, accounting for 38.45% and 40.54%, respectively. Meanwhile, percentage of landslide area, landslide densities and landslide rates are quite larger in these regions (*D* values between 2.2 and 2.4 and between 2.4 and 2.6) than that in other regions. In particular, total landslide density is very high with 23 No./km<sup>2</sup> in the regions of *D* values ranging from 2.2 to 2.6, indicating that landslides tend to occur in this region.

In general, Geologic settings determine the strength of rocks that can further influence sediment production caused by weathering processes on hillslopes. Soft rock can usually lead to weak resistance to erosion. Based on geologic map provided by Central Geological Survey, MOEA, Taiwan, we can find that the largest geologic unit is interbedded sandstone and shale, cropping out about 72% of the total CRW area. We also examined landslide distribution in response to geologic settings. Results show that percentage of landslide area *Ca* , landslide density *DL* , and landslide rate *RL* are 4.07%, 22.67 No./km<sup>2</sup> and 5.66%, respectively, in the region of interbedded sandstone and shale and these values are much greater than that in other regions, as shown in **Table 2**. Other regions crop out geologic units of gravel, loamy sand, slate, interbedded slate and sandstone, hard sandstone, and hard shale, respectively. These geologic units are usually recognized as hard rocks and structural settings. Although some of them are composed of loose materials that are tended to landslide triggering (loamy sand and gravel), but just crops out only 10.83 and 4.52% of total watershed area in the CRW.

The geologic setting crops out interbedded sandstone and shale that has weak, low resistance to erosion and lies most regions of the CRW, favoring the generation of landslides on hillslopes.

The above landslide inventory shows that landslide triggering can be influenced by geomorphic and geologic settings. The hillslope with elevation of 500 m, slope between 28.8° and 45°, convex slopes, surface roughness index from 0 m to 0.6 m, fractal dimension from 2.2 to 2.6 and geologic unit composed of sandstone has high potential for the development of landslides.

#### **3.4. Aftermath of typhoon Morakot in the CRW**

geomorphic characteristics are simple and using those characteristics (i.e., elevation, slope and aspect) could not represent complex geomorphic systems in natural environments, e.g., convergent slopes and surface roughness. To consider the complexity of geomorphic conditions, we employed curvature, surface roughness and fractal dimension, comparing with the

Curvature is the changing rate of slope along *x* and y direction and can be categorized into three groups of divergent, flat and convergent based on curvature values of greater than zero, equal to zero and smaller than zero, respectively [47], as shown in **Table 2**. In comparison, divergent and convergent slopes individually lie the most regions of the total watershed area (about 47 and 48%). On the divergent slope, landslide triggering after Typhoon Morakot

rate of 4.21%. In other words, the convergent slopes have 2.6% percentage of landslide area

5.4%. Results show that both of divergent and convergent slopes can provide suitable condition for the development of landslides. Divergent slopes can lead to landslide triggering due to the effect of gravity along slope. Convergent slopes can concentrate water flow leading to increase in groundwater level and pore water pressure, therefore generating landslides on hillslopes. However, **Table 2** still indicates that landslides slightly tend to occur on convergent slopes in comparison with divergent slopes in the CRW. In addition, the flat region is not

Here we define the surface roughness is a residual topography that can be derived from the cell by cell subtraction of original 5 × 5 m DEM and the mean of this DEM. Mean DEM was created by averaging elevation values within a 3-cell moving window. The raster of residual topography was then calculated as the cell-by-cell difference between the original DEM and the mean DEM. In the Cishan River watershed, surface roughness (i.e., residual topography) was calculated as the above mentioned flow and further categorized into six groups by 0–0.3, 0.3–0.6, 0.6–0.9, 0.9–1.2, 1.2–1.5 and 1.5–1.8. Based on this classification, the largest unit of surface roughness is the region of the values from 0.3 to 0.6 that covers about 39.32% of the

nificant difference between landslide rates in each surface roughness settings, ranging from 3.07% to 6.14%. These patterns suggest that landslides tend to be triggered in group in the region of surface roughness between 0 to 0.6. Above this cutoff, the landslide density and

In other words, natural landscapes generally have fractal characteristics. Terrain can be considered self-similar in the two horizontal directions and self-affine in cross-section [30]. This leads to the discussion of the relationships of landslide triggering with respect to the selfsimilarity or fractals of topography be necessary. Fractal dimension in the CRW was calculated using the Landserf software on the basis of 5 × 5 m DEM, comparing with the location of landslide occurrence. Landserf implements the variogram method (e.g., [31]), which we use

> <sup>2</sup>*n*(*h*) ∑ *i*=1 *n* ∑ *j*=1 *n*

(*zi* − *zj*)

and landslide

and landslide rate of

.

respect to this region. Note that there is no sig-

<sup>2</sup> (3)

location of landslide occurrence to find out the 'hotspot' or 'prone area' in the CRW.

accounted for 1.98% of the total watershed area with 11.03 landslides per km2

appropriate to landslide triggering and hence strictly has 0.42 landslides per km2

comparing with the area of the CRW, landslide density of 14.5 No./km<sup>2</sup>

total CRW area, with 10.81 landslides per km<sup>2</sup>

28 Environmental Risks

in this study. The variogram is calculated as:

*D* = \_\_\_\_\_ <sup>1</sup>

landslide rate evidently decrease as well as shown in **Table 2**.

Denudation processes play a very important role in the fluvial systems of a watershed. A large amount of sediment materials produced by landslides could be entrained downslope from hillslopes into river channels, influencing the evolution of river morphology. Typhoon Morakot led to severe hillslope erosion in the CRW (**Figure 8**), and its consequent generation of sediment materials could deposit on hillslopes and river channels, changing the geomorphic response of the CRW. Here we compared the river borne suspended sediment and river bathymetry of the Cishan River to show the aftermath effect of Typhoon Morakot on the CRW.

#### *3.4.1. Changes in river-borne suspended sediment concentrations*

River-borne suspended sediment is an important feature of the global denudation system and is often adopted as a measure of terrestrial erosion rates and the intensity of erosion processes in watersheds [32, 33]. Here we quantitative the impact of Typhoon Morakot on the sediment loads of Cishan River by using the measurement of suspended-sediment discharge of Shenlinbridge river stage station from 1987 to 2005 and 2010 (no measurement from 2007~2009). The water discharge is daily recorded, and the sediment concentration is measured fortnightly using a USDH-48 suspended-sediment sampler from the Water Resources Agency of Taiwan. Evaluating the amount of sediment loads can use the formula of rating curve based on water discharges. Fitting the plots of suspended-sediment discharge on log-log scale using least square can obtain an equation as

$$\mathbf{C}\_s = a\mathbf{Q}^\flat \tag{4}$$

where *C*<sup>s</sup> is sediment concentration (ppm), *a* is the unit sediment concentration and *b* is the sediment mobilization capacity of water discharges. By fitting model curves based on (5) to both pretyphoon (from 1987 to 2005) and posttyphoon (in 2010) data, while keeping the exponent fixed to permit comparison between models, we obtained model coefficients *a*pre and *a*post. The ratio of *a*post to *a*pre denoted by △*a* that can be used to estimate the influence of Typhoon Morakot on the fluvial system of the CRW. If △*a* > 1, it means unit sediment concentration increased after the typhoon events.

**Figure 13** shows the river-borne suspended-sediment rating curves of the Cishan River on the basis of data recorded by the Shanlin Bridge gauge. The power-law relation of measurement data before Typhoon Morakot (from 1987 to 2005; see red line) fitted by least-square regression is expressed as

$$\mathcal{C}\_s = 16 \, Q^{\mu \tau} \tag{5}$$

Morakot for the Cishan River by the rating curve regression employing 4 successive daily water discharges at the beginning of 6 Aug. Note that all the records of daily water discharges

**Figure 13.** Suspended-sediment rating curves for the Cishan River. Red circles show measurements made before Typhoon Morakot; blue circles show measurements after the typhoon storms. Dashed lines are power-law relations fitted to pretyphoon data using log-transformed least-squares regression; solid lines are power-law relations fitted only

Landslides Triggered by Typhoon Morakot in Taiwan http://dx.doi.org/10.5772/intechopen.76930 31

**Figure 14** shows the mean annual sediment load, *Qsy*, the annual sediment load in 2009, *Qs09* and the sediment load brought by Typhoon Morakot, *Qsm* for the Cishan River. On average, the annual river-borne suspended-sediment loads were ~ 1.06 Mt./y for the Cishan River. The total mass flushed out from the Cishan River watershed in 2009 was about 0.71 Mt. and, 0.64 Mt. of which was mobilized by Typhoon Morakot. It shows that Typhoon Morakot determined erosion processes on hillslopes, removing the most sediment materials from the CRW and accounted for 91% of total river-borne suspended-sediment during the period of 4 days. Only about 10% of sediment materials (~0.06 Mt) was transported by the Cishan River during the other period in 2009. Moreover, four-day suspended-sediment loads generated by

In general, in mountain belts, sediment produced by storm-triggered landslides usually rapidly transfers into the fluvial systems and could not taking a long time to storage on hillslopes. However, we have shown a large amount of sediment materials deposited on hillslopes at the end of 2010 after Typhoon Morakot in the above section. Landslides caused by Typhoon Morakot supplied magnificent sediment to the Cishan River and could lead to sedimentation in the river channels. Here we collected the elevation data for the 62 cross section of

were obtained from the Shanlin Bridge gauge.

to posttyphoon data.

Typhoon Morakot reached 61% of the decadal mean annual value.

*3.4.2. Changes in the riverbed elevation of the Cishan River*

and the relation of that after the typhoon events (in 2010; see black line) fitted by least-square regression is expressed as

$$\mathbf{C}\_s = 160 \,\mathrm{Q}^{0.7} \tag{6}$$

According to (5) and (6), change in unit sediment concentration for the Cishan River after Typhoon Morakot is equivalent to 10 (i.e., △*a* = 10). Suspended-sediment concentrations in the Cishan River after the disturbance of Typhoon Morakot are as much as 10 times greater than decadal background value. This elevated posttyphoon erosion rate has resulted from the abundant landslide triggered by the typhoon events, and led to rich sediment supply from the hillslopes into river channels.

Here we calculated the annual river-borne suspended-sediment loads via the estimates of the rating curve for the Cishan River that combines the relationship of suspended-sediment concentration with respect water discharges, with the instrumental records of daily water discharges. In addition, we estimated the suspended-sediment loads driven by Typhoon

**Figure 13.** Suspended-sediment rating curves for the Cishan River. Red circles show measurements made before Typhoon Morakot; blue circles show measurements after the typhoon storms. Dashed lines are power-law relations fitted to pretyphoon data using log-transformed least-squares regression; solid lines are power-law relations fitted only to posttyphoon data.

Morakot for the Cishan River by the rating curve regression employing 4 successive daily water discharges at the beginning of 6 Aug. Note that all the records of daily water discharges were obtained from the Shanlin Bridge gauge.

**Figure 14** shows the mean annual sediment load, *Qsy*, the annual sediment load in 2009, *Qs09* and the sediment load brought by Typhoon Morakot, *Qsm* for the Cishan River. On average, the annual river-borne suspended-sediment loads were ~ 1.06 Mt./y for the Cishan River. The total mass flushed out from the Cishan River watershed in 2009 was about 0.71 Mt. and, 0.64 Mt. of which was mobilized by Typhoon Morakot. It shows that Typhoon Morakot determined erosion processes on hillslopes, removing the most sediment materials from the CRW and accounted for 91% of total river-borne suspended-sediment during the period of 4 days. Only about 10% of sediment materials (~0.06 Mt) was transported by the Cishan River during the other period in 2009. Moreover, four-day suspended-sediment loads generated by Typhoon Morakot reached 61% of the decadal mean annual value.

#### *3.4.2. Changes in the riverbed elevation of the Cishan River*

*3.4.1. Changes in river-borne suspended sediment concentrations*

square can obtain an equation as

increased after the typhoon events.

sion is expressed as

regression is expressed as

hillslopes into river channels.

where *C*<sup>s</sup>

30 Environmental Risks

River-borne suspended sediment is an important feature of the global denudation system and is often adopted as a measure of terrestrial erosion rates and the intensity of erosion processes in watersheds [32, 33]. Here we quantitative the impact of Typhoon Morakot on the sediment loads of Cishan River by using the measurement of suspended-sediment discharge of Shenlinbridge river stage station from 1987 to 2005 and 2010 (no measurement from 2007~2009). The water discharge is daily recorded, and the sediment concentration is measured fortnightly using a USDH-48 suspended-sediment sampler from the Water Resources Agency of Taiwan. Evaluating the amount of sediment loads can use the formula of rating curve based on water discharges. Fitting the plots of suspended-sediment discharge on log-log scale using least

*Cs* = *aQb* (4)

sediment mobilization capacity of water discharges. By fitting model curves based on (5) to both pretyphoon (from 1987 to 2005) and posttyphoon (in 2010) data, while keeping the exponent fixed to permit comparison between models, we obtained model coefficients *a*pre and *a*post. The ratio of *a*post to *a*pre denoted by △*a* that can be used to estimate the influence of Typhoon Morakot on the fluvial system of the CRW. If △*a* > 1, it means unit sediment concentration

**Figure 13** shows the river-borne suspended-sediment rating curves of the Cishan River on the basis of data recorded by the Shanlin Bridge gauge. The power-law relation of measurement data before Typhoon Morakot (from 1987 to 2005; see red line) fitted by least-square regres-

*Cs* = 16 *Q*0.7 (5)

and the relation of that after the typhoon events (in 2010; see black line) fitted by least-square

*Cs* = 160 *Q*0.7 (6)

According to (5) and (6), change in unit sediment concentration for the Cishan River after Typhoon Morakot is equivalent to 10 (i.e., △*a* = 10). Suspended-sediment concentrations in the Cishan River after the disturbance of Typhoon Morakot are as much as 10 times greater than decadal background value. This elevated posttyphoon erosion rate has resulted from the abundant landslide triggered by the typhoon events, and led to rich sediment supply from the

Here we calculated the annual river-borne suspended-sediment loads via the estimates of the rating curve for the Cishan River that combines the relationship of suspended-sediment concentration with respect water discharges, with the instrumental records of daily water discharges. In addition, we estimated the suspended-sediment loads driven by Typhoon

is sediment concentration (ppm), *a* is the unit sediment concentration and *b* is the

In general, in mountain belts, sediment produced by storm-triggered landslides usually rapidly transfers into the fluvial systems and could not taking a long time to storage on hillslopes. However, we have shown a large amount of sediment materials deposited on hillslopes at the end of 2010 after Typhoon Morakot in the above section. Landslides caused by Typhoon Morakot supplied magnificent sediment to the Cishan River and could lead to sedimentation in the river channels. Here we collected the elevation data for the 62 cross section of

**Figure 14.** Mean annual sediment loads, *Qsy*, annual sediment loads in 2009, *Qs09* and sediment loads brought by Typhoon Morakot, *Qsm* in the Cishan River derived from rating curve estimates.

the Cishan River, from Jiashian weir to Erren-Yuemei weir and the river reach is shown in **Figure 15**, which was reported by Water Resources Planning Institute [34].

The longitudinal profile is a continuous line by the lowest elevations at each stream cross sections. **Figure 16** shows the longitudinal profile of this 22.8 km river reach for the Cishan River, illustrated by the measurement data of each stream cross section during the periods of pretyphoon (in 2005) and posttyphoon (in 2010). Result indicates that the riverbed has a slight scour cumulative distance of below 1 km and above 18 km for the original cross section No. 22, respectively, but other reaches had significant sedimentation. To quantitative the effect of changes in the riverbed on the transport capacity for this river reach (**Figure 15**), the unit stream power [35] was used as

$$
\phi = \rho g Q S / w \tag{7}
$$

Of particular interest, (9) infers that unit stream power could be only influenced by changes in the gradient of riverbed, because water density and water discharges could be treated as the same values in the same stream reach. Hence, based on (9), we can only examine the gradient of riverbed pretyphoon and posttyphoon to describe shifts in unit stream power for the river reach as shown in **Figure 16**. Calculating the gradient of longitudinal profile for the river reach form the measurement data in situ show that the pretyphoon riverbed gradient (in 2005) was 2.8% and the posttyphoon riverbed gradient (in 2010) was 1.3%, respectively. This indicates that unit stream power for the Cishan River had been significantly decreased (about 55%) after the disturbance of Typhoon Morakot and could lead to the lowering of transport capacity for the fluvial system, increasing sedimentation on the

**Figure 15.** Formosat-II images of river reach with 62 cross section measurement for the Cishan River from Jiashiang weir

Landslides Triggered by Typhoon Morakot in Taiwan http://dx.doi.org/10.5772/intechopen.76930 33

In other words, the longitudinal profile for the river reach is further separated into four parts with the cumulative distances of ~5 km, dissecting shifts in the unit stream power the river channels influenced by Typhoon Morakot. **Figure 17** shows the variations in the gradients of the riverbed calculated by the pretyphoon and posttyphoon measurement data in situ. Before the typhoon disturbance in 2005, the gradients of the riverbed were 0.00794, 0.00539, 0.00586 and 0.00705 for the upstream to the downstream, indicating that unit stream power in the river reach gradually decreased along the longitudinal river profile and had an increase in the riverbed gradient approximating the Erren-Yuemei weir. Sedimentation in the river

riverbed.

to Erren-Yuemei weir.

where *ω* is the unit stream power on river bed (W/m<sup>2</sup> ), *ρ* is water density (1000 kg/m3), *Q* is discharge of river channel (m3 /s), *S* is the gradient of river channel and *w* is channel width (m). In Taiwan, the channel width can be represented as [6]

$$w = Q^{0.5} \tag{8}$$

replacing the *w* (7) by Eq. (8), we can obtain as following equation:

$$
\rho = \rho \, Q^{0.5} \,\text{S} \tag{9}
$$

**Figure 15.** Formosat-II images of river reach with 62 cross section measurement for the Cishan River from Jiashiang weir to Erren-Yuemei weir.

the Cishan River, from Jiashian weir to Erren-Yuemei weir and the river reach is shown in

**Figure 14.** Mean annual sediment loads, *Qsy*, annual sediment loads in 2009, *Qs09* and sediment loads brought by Typhoon

The longitudinal profile is a continuous line by the lowest elevations at each stream cross sections. **Figure 16** shows the longitudinal profile of this 22.8 km river reach for the Cishan River, illustrated by the measurement data of each stream cross section during the periods of pretyphoon (in 2005) and posttyphoon (in 2010). Result indicates that the riverbed has a slight scour cumulative distance of below 1 km and above 18 km for the original cross section No. 22, respectively, but other reaches had significant sedimentation. To quantitative the effect of changes in the riverbed on the transport capacity for this river reach (**Figure 15**), the unit stream power [35] was used as

*ϖ* = *gQS*/*w* (7)

*w* = *Q*0.5 (8)

*ϖ* = *ρ Q*0.5 *S* (9)

), *ρ* is water density (1000 kg/m3), *Q* is

/s), *S* is the gradient of river channel and *w* is channel width (m).

**Figure 15**, which was reported by Water Resources Planning Institute [34].

where *ω* is the unit stream power on river bed (W/m<sup>2</sup>

Morakot, *Qsm* in the Cishan River derived from rating curve estimates.

In Taiwan, the channel width can be represented as [6]

replacing the *w* (7) by Eq. (8), we can obtain as following equation:

discharge of river channel (m3

32 Environmental Risks

Of particular interest, (9) infers that unit stream power could be only influenced by changes in the gradient of riverbed, because water density and water discharges could be treated as the same values in the same stream reach. Hence, based on (9), we can only examine the gradient of riverbed pretyphoon and posttyphoon to describe shifts in unit stream power for the river reach as shown in **Figure 16**. Calculating the gradient of longitudinal profile for the river reach form the measurement data in situ show that the pretyphoon riverbed gradient (in 2005) was 2.8% and the posttyphoon riverbed gradient (in 2010) was 1.3%, respectively. This indicates that unit stream power for the Cishan River had been significantly decreased (about 55%) after the disturbance of Typhoon Morakot and could lead to the lowering of transport capacity for the fluvial system, increasing sedimentation on the riverbed.

In other words, the longitudinal profile for the river reach is further separated into four parts with the cumulative distances of ~5 km, dissecting shifts in the unit stream power the river channels influenced by Typhoon Morakot. **Figure 17** shows the variations in the gradients of the riverbed calculated by the pretyphoon and posttyphoon measurement data in situ. Before the typhoon disturbance in 2005, the gradients of the riverbed were 0.00794, 0.00539, 0.00586 and 0.00705 for the upstream to the downstream, indicating that unit stream power in the river reach gradually decreased along the longitudinal river profile and had an increase in the riverbed gradient approximating the Erren-Yuemei weir. Sedimentation in the river

*3.4.3. Widening of the Cishan River*

Cishan River, leading to its consequent river width widening.

Sediment sources produced by Typhoon Morakot was richly supplied from hillslopes to the fluvial system, elevating the riverbed of the Cishan River. This elevated riverbed lowered the unit stream power of the river channel and could lead to water flow centrally erode the riverbanks. To investigate changes in the riverbanks for the Cishan River, we used a pair of Formosat-II images to digitalize the pretyphoon (in 2008) and posttyphoon (in 2009) edges of the riverbanks within GIS. Then, HEC-RAS was used to extract the river widths per 100 m along the river channel before and after the disturbance of Typhoon Morakot. **Figure 18** shows the box-whisker plot of the river widths estimated from pretyphoon and posttyphoon data for the Cishan River. The second and third quartiles of the pretyphoon river widths are ~110 and 190 m, respectively, with a maximum of 728 m, a minimum of 16 m, and an average of 144 m. However, after the typhoon disturbance, the Cishan River widths were evidently widened and had the second and third quartiles of 260 and 471 m. Its maximum and minimum river widths were shifted into 2090 and 17 m, with an average of 342 m. This is clear that river widths increased by more than a factor of 2 to those statistical estimate values before the hit of the typhoon. Results show that Typhoon Morakot caused severe riverbank erosion for the

Landslides Triggered by Typhoon Morakot in Taiwan http://dx.doi.org/10.5772/intechopen.76930 35

Comparing these two river channel inventories (in 2008 and in 2009) can show changes in river widths caused by Typhoon Morakot. Subtraction between the pretyphoon and posttyphoon river widths in each river cross section indicates that only about 22 cross sections display contracting adjustment, i.e., the river widths in 2009 smaller than that in 2008, only accounting for

**Figure 18.** Box and whisker plot of river widths along the Cishan River estimated from pretyphoon and posttyphoon

mapping based on a pair of Formosat-II images in 2008 and 2009, respectively.

**Figure 16.** Longitudinal profiles of river reach for the Cishan River from Jiasiang weir to Erren-Yuemei weir obtained from the elevation data of pretyphoon and posttyphoon measurement.

**Figure 17.** River gradients of four reaches along the Cishan River from Jiasiang weir to Erren-Yuemei weir before and after the disturbance of Typhoon Morakot.

channel was elevated by TypvMorakot, leading to reduction in its riverbed gradient and lowered about 10, 7, 29 and 7% of unit stream power for those four river reaches from the upstream to the downstream. These patterns suggest that significant sedimentation processes can be observed in this reach for the Cishan River, because the lowering of river transport capacity and abundant sediment supplied from the hillslopes in the CRW after Typhoon Morakot. Hence, the increased likelihood of flood inundation in the reach of the Cishan River (**Figure 14**) are expected due to its elevated riverbed.

#### *3.4.3. Widening of the Cishan River*

channel was elevated by TypvMorakot, leading to reduction in its riverbed gradient and lowered about 10, 7, 29 and 7% of unit stream power for those four river reaches from the upstream to the downstream. These patterns suggest that significant sedimentation processes can be observed in this reach for the Cishan River, because the lowering of river transport capacity and abundant sediment supplied from the hillslopes in the CRW after Typhoon Morakot. Hence, the increased likelihood of flood inundation in the reach of the Cishan River

**Figure 17.** River gradients of four reaches along the Cishan River from Jiasiang weir to Erren-Yuemei weir before and

**Figure 16.** Longitudinal profiles of river reach for the Cishan River from Jiasiang weir to Erren-Yuemei weir obtained

(**Figure 14**) are expected due to its elevated riverbed.

from the elevation data of pretyphoon and posttyphoon measurement.

34 Environmental Risks

after the disturbance of Typhoon Morakot.

Sediment sources produced by Typhoon Morakot was richly supplied from hillslopes to the fluvial system, elevating the riverbed of the Cishan River. This elevated riverbed lowered the unit stream power of the river channel and could lead to water flow centrally erode the riverbanks. To investigate changes in the riverbanks for the Cishan River, we used a pair of Formosat-II images to digitalize the pretyphoon (in 2008) and posttyphoon (in 2009) edges of the riverbanks within GIS. Then, HEC-RAS was used to extract the river widths per 100 m along the river channel before and after the disturbance of Typhoon Morakot. **Figure 18** shows the box-whisker plot of the river widths estimated from pretyphoon and posttyphoon data for the Cishan River. The second and third quartiles of the pretyphoon river widths are ~110 and 190 m, respectively, with a maximum of 728 m, a minimum of 16 m, and an average of 144 m. However, after the typhoon disturbance, the Cishan River widths were evidently widened and had the second and third quartiles of 260 and 471 m. Its maximum and minimum river widths were shifted into 2090 and 17 m, with an average of 342 m. This is clear that river widths increased by more than a factor of 2 to those statistical estimate values before the hit of the typhoon. Results show that Typhoon Morakot caused severe riverbank erosion for the Cishan River, leading to its consequent river width widening.

Comparing these two river channel inventories (in 2008 and in 2009) can show changes in river widths caused by Typhoon Morakot. Subtraction between the pretyphoon and posttyphoon river widths in each river cross section indicates that only about 22 cross sections display contracting adjustment, i.e., the river widths in 2009 smaller than that in 2008, only accounting for

**Figure 18.** Box and whisker plot of river widths along the Cishan River estimated from pretyphoon and posttyphoon mapping based on a pair of Formosat-II images in 2008 and 2009, respectively.

1.6% of total 1357 cross sections. Most of the river widths show widening adjustment after the typhoon disturbance. Here we focused on the widening adjustment of the 1335 cross sections for the Cishan River in 2009. Also, we characterized the noncumulative frequency-widening relations to clarify whether the river widening has a SOC phenomenon. The noncumulative frequency distribution of river widening can be mathematically expressed as

$$p(\mathbb{R}\_{\nu}) = \alpha \mathbb{R}\_{\nu}^{-\rho} \tag{10}$$

from the many regions of the globe. Hence, more data of river channel inventories and its corresponding field checking should be necessary to examine whether the *β* values in (10) still

Landslides Triggered by Typhoon Morakot in Taiwan http://dx.doi.org/10.5772/intechopen.76930 37

In this chapter, we characterize landslides triggered by Typhoon Morakot in 2009, and its corresponding frequency-area distribution. Results show that the exponent value of a noncumulative relation for these landslides approximates the lowest limitation of worldwide observation. This infers that the hillslopes of the CRW has high potentials on landslide triggering. Meanwhile, ambient sediment materials produced by landslides could deposit on hillslopes and river channels and cause the adjustments of hillslope and fluvial systems, which can be observed from raised river-borne suspended-sediment concentration in the Cishan River (i.e., rich-supply hillslopes) and its decreased stream power (severe sedimentation in river channels). These patterns indicate that landslides not only pose threats to people's life and properties, but also have significant influence on the downstream. Hence, long-term and short-term strategies for landslide countermeasures are both necessary. The long-term strategies are the comprehensive management and regulation of basins and watersheds. The shortterm strategies are the development of real-time warning systems for landslide triggering on hillslopes. In Taiwan, the present real-time warning system developed for landslide hazards

Before hit by this typhoon storms, the Central Geological survey, MOEA (2009) in Taiwan has used logistic equation to estimate landslide ratios via the potential values obtained from the combination of 100-year return period hourly rainfall depth and cumulative daily rainfall to map the landslide susceptibility. On the basis of this susceptibility, Taiwan's hillslopes were categorized in three regions of high risk, medium risk and low risk. Comparing with the location and initiation time in situ of landslides or rock avalanching (total of 909; provided by Soil and Water Conservation Bureau and Central Geologic Survey) show that these geomorphic erosion processes crop out the regions of high risk ~43% of totals. 90% of total can be observed when we consider both of high risk and medium risk regions. However, although the construction of landslide susceptibility can provide some useful information on mapping landslide-prone areas, the effect of real-time rainfall during typhoon storms should be necessary for landslide warning, still. Considering only landslide-prone area could also lead to the over-issued orders of hazard mitigation from landslide warning and also the wasting of

Rainfall brought by typhoon storms plays a majorly important role in triggering landslides on hillslopes. Typically, some topographic and geologic regimes could provide suitable conditions for landslide triggering but landslides are still needed to be initiated by external triggers such as rain infiltration and its consequent saturation. The evolution of soil pore pressure can

holds on true in other environments, or these values have a variation.

**4.1. Landslide warning system adopted by the Taiwan's government**

**4. Discussion**

are described as follow.

governmental administrative resources.

where *Rw* is the magnitude of river widening, the *p*(*Rw*) is a PDF (probability density function) equivalent to \_\_\_1 *Nw* <sup>Δ</sup> *<sup>N</sup>* \_\_\_\_*<sup>w</sup>* Δ *Rw* , and the △*Nw* is the number of cross sections with widening between *Rw* and *Rw* + △*Rw* and the *RwT* is the total number of river cross section widening. The constants *α* and *β* are obtained from fitting medium and large river widening in order to detect the right heavy tailed decay of PDF through a power-law. We increase our bin width △*Rw* with increasing area *Rw*, so that bin widths are approximately equal in logarithmic coordinates. **Figure 19** shows that river widening having *Rw* larger than 64 m could be well interpreted by a power-law statistic with *α* = 25.6 and *β* = 1.93 with *r*<sup>2</sup> = 0.98.

Of particular interest, the *β* value for river widening is greater than that for landslides driven by the same external trigger, i.e., Typhoon Morakot, suggesting that the occurrence likelihood of large magnitude of river widening is smaller than that of large landslides in the CRW for the perspective of environmental risks. Also, typhoon-induced river widening could have self-organized criticality. However, the *β* values in (10) are very limited on the basis of worldwide observations and are not like the *β* values in (2) for landslides that can be obtained

**Figure 19.** Noncumulative frequency distribution of river width widening after Typhoon Morakot for the Cishan River. Above the cutoff of 64 m, the river width widening satisfies a power-law relation with exponent β = 1.93.

from the many regions of the globe. Hence, more data of river channel inventories and its corresponding field checking should be necessary to examine whether the *β* values in (10) still holds on true in other environments, or these values have a variation.
