CC BY
5УДК 3179.55 Nguyen Khac Hoang Giang, Phi Hong Thinh
Nguyen Khac Hoang Giang
Master of Science in Geology Hanoi University of Natural Resources and Environment (Hanoi city, Vietnam)
Phi Hong Thinh
PhD in Engineering Geology University of Transport and Communications (Hanoi city, Vietnam)
ANALYSIS AND ASSESSMENT OF THE PROBABILITY OF RAINFALL THRESHOLD CAUSING LANDSLIDES IN MOUNTAINOUS AREAS OF QUANG NAM PROVINCE, VIETNAM
Abstract: landslides have caused serious damage to infrastructure in Quang Nam province, Vietnam in recent years. Especially in the context of complex climate change in a negative direction, many large landslides have caused serious damage to people and property on key traffic routes in mountainous areas. From this reality, it is necessary and urgent to study, synthesize, and evaluate landslide points in the study area to maintain the stable and long-term operation of these key routes. This paper aims to introduce new research results on determining the rainfall threshold causing landslides in each area province area. Using statistical data on landslide events in the area combined with actual rainfall data measured at rain gauge stations, the research team used methods of increasing probability assessment to determine the average rainfall threshold of 5 days (Rt5). The north of the study area (Prao gauge station) is represented by the equation RT = -139.2ln(R5) + 278.53 with the reliability coefficient R2 = 0.9731, The northeast of the province (Thanh My gauge station) is represented by the equation RT = -118.3ln(R5) + 222.34, The Southwest of the Province (Kham Duc gauge station) is represented by the equation RT = -150.4ln(R5) + 280.78 with the reliability coefficient R2 = 0.9717, The South of the Province (Tra My gauge station) is represented by the equation RT = -113ln(R5) + 246.28 with the reliability coefficient R2 = 0.9473, The average daily cumulative rainfall threshold for the entire study area is given by the equation RT = -129.9ln(R5) + 258.19 with the reliability coefficient R2 = 0.9221.
Keywords: landslide, quang nam province, rainfall gauge station.
1. INTRODUCTION.
Landslide reports obtained from historical records can be used to perform a number of necessary steps in landslide risk assessment, including Estimating the probability of landslide events over time, Estimating the scale of landslide probability, Estimating the spatial probability of landslide occurrence, Determining the vulnerability of landslide risk factors.
Landslides in the study area are mainly caused by rainfall, so a temporal probability signature of landslides can be obtained by assessing the temporal probability of rainfall events combined with an analysis of rainfall thresholds, intensities, or minimum rainfall durations required to trigger a landslide (Crozier et al., 1997). Such an analysis requires information on the actual date of landslide occurrence and corresponding rainfall data, which in this case is derived from the assumption that the rate of landslide triggering events and landslide occurrence will remain constant in the future under certain geo-environmental conditions, which may be questionable when taking into account the combined effects of global changes, climate, and slope structure. The probability of infrequent landslides can also be estimated using the mean rate of landslide occurrence (Guetti et al., 2005). The results of such studies are often only applicable to the modeled area. Physically based threshold models use local topographic features (e.g., slope gradient, soil depth) in a dynamic hydrological model in which rainfall is the most important variable (Wieczorek et al., 2005). These models are less suitable for larger areas because they require detailed information on parameters (e.g., soil properties, water table changes, discharge conditions), which are difficult to extrapolate outside the measuring equipment (with piezometers, tensiometers, etc.) Experimental methods based on rainfall threshold estimates are obtained by estimating the processing rainfall conditions that lead to landslides. They are typically contained in envelope curves based on variables such as cumulative rainfall, antecedent rainfall, rainfall intensity, and rainfall duration (Wieczorek et al., 1987, Glade et al., 2005, Chen C. W. et al., 2016). The most commonly used empirical model is based on rainfall. This threshold model requires data with high quality and temporal resolution (at least hourly rainfall data), which are often not available. Other
models based on antecedent rainfall operate with daily rainfall data, which are relatively simple and inexpensive to measure over large areas, suitable for the study area throughout the Province. In this paper, the authors propose a method for determining landslide probability based on the rainfall threshold for landslide events using the empirically derived probability of exceeding the rainfall threshold and the probability of landslide occurrence related to the rainfall threshold.
2. RESEARCH METHODS.
2.1. Theoretical basis.
Rainfall data for analysis were collected at 4 out of 7 monitoring stations in the entire study area. The distribution of rainfall stations is shown in Figure 1.
Daily rainfall data from rain gauge stations in the province have been compiled from 2004 to the present. Although there is not much change in the total annual rainfall from the stations, there is a large difference in rainfall during landslide-triggering events. Table 1 describes the heavy rainfall triggering landslides, typical in each area as follows:
Fig. 1. Location of rainfall gauge stations.
Table 1. Heavy rainfall events causing landslides.
No. Time of occurrence Landslide event Station Cumulative average rainfall (mm/day)
5 4 3 2 1
1 1/10/2006 Prao 297.6 175.4 118.7 89.1 71.4
2 16/10/2007 311.9 177.6 125.9 94.7 82.5
3 11/11/2007 266.9 140.5 93.7 70.3 56.2
4 18/9/2013 274.1 163.1 115.9 89.9 75.7
5 1/10/2006 Tra My 258.1 139.9 107.8 80.9 64.7
6 15/10/2013 263.4 183.8 122.5 93.0 75.4
7 15/11/2013 265.3 140.7 96.4 72.3 57.8
8 30/10/2007 Kham Duc 305.0 153.9 103.9 80.5 72.9
9 2/10/2013 289.0 146.2 97.8 73.9 62.0
10 15/10/2013 280.7 147.1 98.0 75.5 60.4
No. Time of occurrence Landslide event Station Cumulative average rainfall (mm/day)
5 4 3 2 1
11 17/10/2008 Thanh My 225.1 151.0 100.7 75.5 60.5
12 18/9/2013 231.1 133.0 89.9 70.5 67.1
13 15/11/2013 236.5 121.3 80.9 60.7 48.6
Figure 2 shows the rainfall variation for representative landslide events measured at Prao, Kham Duc, Thanh My, and Tra My stations. The events considered for this analysis resulted in numerous large landslides in different parts of the major traffic routes in the study area. The daily rainfall per event varied considerably and Figure 2 shows no clear distinction across the entire study area. Most of the landslides occurred in areas with relatively high rainfall.
Fig. 2. Daily rainfall during some major landslide events at measuring stations.
2.2. Method for assessing the probability of an increase.
The input of the rainfall threshold analysis is the time series of the average daily rainfall Rd(t) in mm day 1-, where t is the time. For a landslide (L) to occur, the average daily cumulative rainfall must exceed the threshold, which is a relationship R(t)
between the average daily cumulative rainfall over some time (t) and the rainfall of the - Rad premise (t), that is, the average cumulative rainfall (mm/day) that has occurred up to the date of the landslide.
R(t) = /[Rd(t), Rad(t)]. where: Rad (t) is the average cumulative rainfall premise in mm. This function of R determines the probability of a landslide L: P(L). If RT is the threshold value of R then,
P[L|(R > RT)] = 1 and P[L|(R < RT)] = 0. Therefore, in this simplified model, landslides always occur when R exceeds RT and do not occur when the value of R is lower than or equal to RT. In the former case, the probability of landslide occurrence P(L) depends on the probability of exceeding P(R> RT), i.e. P(L) = P(R >RT). However, in practice, the rainfall threshold can be exceeded without leading to any landslide hazard, which can be attributed to several other factors. This discrepancy can be reduced when the latter probability is considered as the conditional probability of exceeding the threshold P(R > RT) and the probability of landslide occurrence P(L) given the exceedance. Thus, the probability of landslide occurrence can be given by the intersection of the two probabilities,
P[(R > RT) |L] = P(R > RT) x P[L|(R > RT)]. This means that the probability of occurrence of both (-R > RT) and (L) is equal to the probability of (R > RT) multiplied by the probability of occurrence of (L). Landslides must occur whenever a certain rainfall threshold is exceeded, which may not be true in all cases. However, to rule out the possibility that landslides will not occur below this assumed rainfall threshold, it is possible to use a minimum acceptable rainfall threshold to work with and estimate the frequency of landslide events by establishing a relationship between the triggering rainfall threshold, the magnitude and the occurrence of landslides.
2.3. Method of determining rainfall thresholds.
To determine the rainfall thresholds, the authors statistically recorded landslide events along the traffic routes in the study area from 2005 to 2017 in the period from October to December. The reasons for this were that the majority of landslide events occurred between October and December, The dates of all landslide events were statistically known for the period from 2005 to 2017, The current status of all landslide events occurring in the study area in which the events from 2005 to 2017 were a threshold model based on previous rainfall was selected due to the availability of rainfall data. In practice, the model is easy to implement and effective because the variation in daily rainfall associated with landslide hazards (Figure 2) and the presence of 7 rainfall gauges in the study area (Figure 1) are accurate and timely. It is important to select rain gauges that represent the relationship between rainfall and landslides along different routes in the study area. The selection of 4 rain gauges Prao, Thanh My, Kham Duc, and Tra My is considered to be representative of the mountainous terrain of the province and is taken to represent different parts spread from the North to the South of the area.
Depending on the type of landslide and the slope structure, the number of days before can vary from 3 days for shallow landslides to 30 days for deep landslides. To determine the appropriate number of days required in the study area, 71 landslide events were statistically recorded that occurred from 2005 to 2017 in the study area. After analyzing the historical rainfall for 3, 5, 15, and 30 days, according to the method proposed by Zezere et al. (2005), the average cumulative rainfall for the previous 5 days was considered suitable for analysis. To determine the RT, a scatter plot was prepared showing the daily rainfall compared to the average cumulative rainfall for the previous 5 days, respectively, for each day with one or more landslide events recorded. This curve can be represented by a linear mathematical equation (Crozier, 1999). To calculate the thresholds, the study area was divided into four parts according to the rainfall gauge stations, based on the topography, geological conditions, and terrain slope. Rainfall conditions in each section were determined from rainfall gauges in the area. For the four individual sections, a common threshold was established for all heavy
rainfall events resulting in 15 or more landslides. During the period 2003-2017, the study area experienced 26 heavy rainfall events resulting in multiple landslides per day. Determining a threshold for an individual section was not possible, due to the lack of historical data.
3. RESULTS.
The northern part of the study area (Prao gauge station) represents Dong Giang and Tay Giang districts, where 18 heavy rainfall events causing landslides were recorded. Here, the RT (daily cumulative rainfall intensity threshold) for a landslide event to occur with the cumulative rainfall intensity in the 5 days before the event (R5) is expressed by the equation RT = -139.2ln(R5) + 278.53, with a confidence coefficient R2 = 0.9731 (Figure 3). The illustration shows that on 5 consecutive rainy days with an average intensity greater than 63.2mm/day, there is a risk of landslides occurring on the 5th day, in addition, for days with rainfall greater than 266.9mm, landslides can occur on the same day.
Prao
* y = -47,15ln[x) + 93,449
R2 = 0,9583
___
• ..............
0 2 4 6
R5 Cumulative average rainfall days (days)
Fig. 3. Daily cumulative average rainfall limit curve for landslides in the Northern region (Prao gauge station).
The Northeast of the Province, representing Dai Loc, Nong Son districts, and a part of the east of Nam Giang (Thanh My gauge station), recorded 13 heavy rain events causing landslides from 2003 to 2017. The cumulative average rainfall threshold is expressed by the equation RT = -31.13ln(R5) + 89.966 (Figure 4) with a reliability coefficient R2 = 0.8315. The illustration shows that on 5 consecutive rainy days with an average intensity greater than 40mm/day, there will be a risk of landslides on the 5th day. In addition, for days with rainfall greater than 92mm, landslides can occur on the same day.
Thanh My
ns
4—
С 'ГС
w
<u >
£0 _
!|
(J
Ш
ьо
re ._
0) >
nj
120,0 100,0 80,0 60,0 40,0 20,0 0,0
• y = -31,13ln(x) + 89,966 R2 = 0,8135
" » _1
t.......i .........
R5 Cumulative average rainfall days (days)
Fig. 4. Daily cumulative average rainfall limit curve for landslides in the Northeast region (Thanh My gauge station).
The Southwest of the Province, representing Phuoc Son, Nam Giang, and Hiep Duc districts (Kham Duc gauge station), recorded 23 heavy rain events causing landslides from 2003 to 2017. The cumulative average rainfall threshold is expressed by the equation RT = - 62.24ln(R5) + 141.26 (Figure 5) with a reliability coefficient R2 = 0.9497. The illustration shows that on 5 consecutive rainy days with an average intensity greater than 40mm/day, there will be a risk of landslides on the 5th day. In addition, on days with rainfall greater than 140mm, landslides can occur on the same day.
Kham Duc
200
П5
4—
С 'ns 150
<u
> 100
£0
£
E jz 50
и
Ш 0
щ,
re
0)
>
nj
4->
DC
I • Rz = 0 i,9497
1 ' ............ .
»
R5 Cumulative average rainfall days (days)
Fig. 5. Daily cumulative average rainfall limit curve for landslides in the Southwest region (Kham Duc gauge station).
The southern part of the province, representing Tien Phuoc, Bac Tra My and Nam Tra My districts (Tra My gauge station), recorded 18 heavy rain events causing landslides from 2003 to 2017. The cumulative average rainfall threshold is expressed by the equation RT = - 48.85ln(R5) + 114.79 (Figure 6) with a reliability coefficient R2 = 0.9112. The illustration shows that in 5 consecutive rainy days with an average intensity greater than 38mm/day, there will be a risk of landslides on the 5th day. In addition, for days with rainfall greater than 116mm, landslides can occur on the same day.
Tra My
_ 140 Е
120 100
с
tu
>
+J
_п) 3
Е
и 01 ьо
та
,
01
>
та
80 60
40 4 20 О
_•_
• 4 V = -4 8,85ln(x) + 114,79
» у ^ R2 = 0,9112
w •••m
О
1 2 3 4 5
R5 Cumulative average rainfall days (days)
Fig. 6. Daily cumulative average rainfall limit curve for landslides in the Southern region (Tra My gauge station).
The average daily cumulative rainfall threshold for the entire study area is given by the equation RT = -40.51ln(R5) + 84.988 (Figure 7) with a reliability coefficient R2 = 0.9112, which shows that the average daily cumulative rainfall threshold is quite low, landslides can occur in any area when the daily rainfall reaches more than 80mm, and the average cumulative rainfall of 20mm/day for 5 consecutive days will also cause landslides.
All Province
_юо
I 90
= 80 га
1 70
га
5 60
В 50 -G!
I 40
g 30
6 20 га
Si ю
™ о
D£
i » 1
я 1 ^ y = -40,51ln(x} + 84,988
< 9 \ R 2 = 0,9582
1_
Г"""--....1!
•
0
2 3 4 5 6
R5 Cumulative average rainfall days (days)
Fig. 7. Daily cumulative mean rainfall limit curve for landslides over the entire study area.
4. CONCLUSION.
The analysis of the average rainfall threshold causing landslides was conducted based on the statistical ratio of the number of landslide events in the study area combined with the analysis of the average cumulative rainfall data in 5 actual days collected in the area where landslide events occurred in the mountainous areas of Quang Nam province.
Figure 3. shows that for the northern areas of the province (Prao gauge station), on 5 consecutive rainy days with an average intensity greater than 63.2mm/day, there will be a risk of landslides on the 5th day. In addition, for days with rainfall greater than 266.9mm, landslides can occur on the same day. Observing figures 4, and 5, it can be seen that for the areas in the northeast and west of the province, the average rainfall accumulated over 5 consecutive days is relatively low (greater than 40.1mm/day in the northeast areas, greater than 50.62mm/day in the southwest areas) which can cause landslides, for the areas in the northeast of the province, the daily rainfall only reaches greater than 225.1mm to cause landslides, for the western areas, the daily rainfall needs to reach a threshold greater than 280.7mm to cause landslides. Figure 6 represents the southern areas of the province, showing that on 5 consecutive rainy days with an
average intensity greater than 73.08mm/day, there will be a risk of landslides on the 5th day, in addition, for days with rainfall greater than 258.1mm, landslides can occur on the same day. The average daily cumulative rainfall threshold for the entire study area is given by the equation RT = -129.9(R5) + 258.19 (Figure 7) with a reliability coefficient R2 = 0.9221, which shows that the average daily cumulative rainfall threshold is quite low, landslides can occur in any area when the daily rainfall reaches more than 225.1mm, and the average cumulative rainfall of 56.2mm/day for 5 consecutive days will also cause landslides.
REFERENCE:
1. Chi Wen Chen, Hongey Chen, Takashi Oguchi. Distributions of landslides, vegetation, and related sediment yields during typhoon events in northwestern Taiwan // Geomorphology. 2016. Vol. 273. P. 51-63;
2. Crozier M.J. Prediction of rainfall-triggered landslides: A test of the Antecedent Water Status model // Earth Surface Processes and Landforms 24. 1999. P. 825-833;
3. Glade T. Establishing the frequency and magnitude of landslide-triggering rainstorm events in New Zealand // Environmental Geology. 1998. Vol. 35. P. 160-174;
4. Glade T., Anderson M.G., Crozier M.J., (eds.). Vulnerability to landslides // Landslide Hazard and Risk, Wiley, London. 2005. P. 175-198;
5. Guetti F., Peruccacci S., Rossi M., Stark C.P., 2007. Rainfall thresholds for the initiation of landslides in Central and Southern Europe // Meteorology and Atmospheric Physics. 2007. Vol. 98. P. 239-267;
6. Wieczorek G.F. Effect of rainfall intensity and duration on debris flows in Central Santa Cruz Mountains, California. Geological Society of America // Reviews in Engineering Geology. 1987. Vol. 7. P. 93-104;
7. Zezere J.L., Trigo R.M., Trig I.F. Shallow and deep landslides induced by rainfall in the Lisbon region (Portugal): assessment of relationships with the North Atlantic Oscillation // Nat. Hazards Earth Syst. 2005. No. 5. P. 331-344
