Professional Documents
Culture Documents
2016; 8:360–370
- 10.1515/geo-2016-0032
Downloaded from PubFactory at 08/02/2016 04:37:28PM
via free access
362 | A. Amirahmadi et al.
existing landslides and also to calculate depth of slides. Various experimental power relations have been em-
Figure 2 illustrates distribution of landslides in Baqi Basin. ployed for landslide volume calculations by researchers at
locations in different countries. To evaluate these relations
in the study area, the numbers of landslides and the as-
sociated area with calculated volume are given in Table 1.
These relations were applied for 44 observed landslides ar-
eas in the basin. The Volume values calculated by these
relations were compared with the volume of observational
landslide of the basin.
At last, a statistical improved model for Baqi Basin
was also compared with relations of Table 1. Comparison
was made through coefficient of determination (R2 ), per-
centiles statistics values, maximum, minimum and aver-
age and the Root Mean Square Errors (RMSE).
- 10.1515/geo-2016-0032
Downloaded from PubFactory at 08/02/2016 04:37:28PM
via free access
Modeling of landslide volume estimation | 363
experimental relation, the slide volume was calculated; 3.4.2 Estimating the Mean Depth of Landslides in the
and then, according to Equation 2, the model with the least area for evaluating the model
RMSE value was chosen as the best one.
⎯
⎸ n After evaluating the given model, this model was used for
⎸ (o i − p i )2 estimating the mean size of landslides. Then the mean
⎸ ∑︀
depth of landslides in the area were calculated and com-
⎷ i=1
RMSE = (2)
n pared with mean depth of observational data and depths
Where n is number of variables, o i is the observed land- obtained from different models.
slide volume, p i is the estimated landslide volume by each
relation [49].
4 Results
3.4 Estimating the Depth of Landslide 4.1 Statistical Specifications of Geometrical
3.4.1 Estimating the depth of landslides observed in the
Data in Landslides
area for estimating the model
After collecting the data required from surveys, first their
accuracy and quality were controlled and then their statis-
Field and surveying methods have been used to estimate
tical properties were calculated and provided in Table 2.
the depth observed in the landslides of the region, because
According to the results of this table, landslides in the
some regions are impassable, so it was impossible con-
Baghi area are in a relatively wide range of area, size and
ducting field studies. Therefore there has been used of re-
depth, such that their area (A L ) was in the range between
mote sensing and ASTER satellite images as well as the
1.5 × 104 m2 ≤ A L ≤ 2 × 106 m2 . The observational vol-
pair of left and right satellite images in these images and
ume or size of landslides (V L ) was in the range between
Digital Elevation Model, 2003 and digital elevation model
1 × 103 m3 ≤ V L ≤ 8.4 × 106 m3 and their depth (D L ) was
from topographical maps with scale of 1:50000 (prepared
in the range of 0.1 m ≤ D L ≤ 13.85 m. This wide range may
from aerial photographs, 1956). Because landslides have
increase the standard deviation (SD) followed by variation
been occurred in this area during these years and at 70s,
coefficient for each parameter.
the observational depth was calculated by reducing both
layers of 2003 and 1956 DEM from each other.
- 10.1515/geo-2016-0032
Downloaded from PubFactory at 08/02/2016 04:37:28PM
via free access
364 | A. Amirahmadi et al.
Table 2: Descriptive statistics related to range of geometric parameters, volume and depth of landslides in the Baghi Area.
Table 3: The estimative and field depth. Because all calculations and results of this study are
based on data observed in the region, therefore one can
Slide No. Depth estimated (m) Field depth (m) see the importance of observational data more. This can
2 1.92 1.863 be clear more for a part of data with lower frequency or
4 1.4 0.998 not being seen in the observational data; because using a
13 0.52 0.435 part of with frequent data – in this study areas between 1 ×
19 2 1.811 105 m2 < A L < 5×106 m2 , one can extract the relationships
20 5.2 6.534 between volume and area for a part with lower data.
22 4 3.914
31 4 3.637
41 5 3.415 4.2 Calculating the Depth of Observational
Landslides (44 samples) by Remote
Table 4: Correlation between field and estimated depth. Sensing and GIS
Estimated Field Because there are no observational data for depth, we de-
.915** 1 Pearson Correlation cided to calculate the observational depth of landslides by
.001 Sig. (2-tailed) Field remote sensing and GIS. In addition, to ensure the relia-
8 8 N bility of data, by selecting 8 sites for landslide out of 44
1 .915** Pearson Correlation landslides occurred in the region and field studies, the ac-
.001 Sig. (2-tailed) Estimated curacy of calculation depth were estimated (Table 3). On
8 8 N the other hand, to determine the accuracy and reliability
** . Correlation is significant at the 0.01 level (2-tailed). of calculation data, correlation between estimated depth
and observational field depth (Table 4 and Figure 3) ob-
According to above models, by estimating the volume tained by confidence of 99%, R2 : 0.84 and correlation rate
based on results of this study for descriptive statistics of of 0.92, we can confide to depth data calculated by remote
range of data, volume and depth of landslides (Table 2), sensing and GIS.
increased variation coefficient, skewness and stretch of
data, it maybe indicate the abnormality of data. It must
be however noted that the size of landslides occurred in 4.3 Developing a Statistic Model for Baghi
the nature are different and a landslide can even cover 1 Area
square meter or a few cubic meters to several km2 or km3
of an area or a volume of a soil in a region, such that the Figure 5 indicates No. 44 of landslide with complete data
range of area as indicated in the study of reference [50] was of area (A L ) and volume (V L ), in a diagram with Log-Log
between 2 m2 to 1 × 109 m2 and or maximum area of land- UTM coordinates. According to Figure 5, the frequency of
slides studied in [43] obtained to 3.9 × 10110 m2 and in such slides is higher in the area range between 100000 m2 to
cases this range of changes may increase the mentioned 1000000 m2 and volumetric range between 500000 m2
coefficients. Therefore higher variation coefficient, skew- to 5000000 m3 . The model exhibits lack of fit for small
ness and stretch may not be a reason for reduced statistical very values of area and volume. Physical review of Fig-
quality of data used in this study. ure 4 indicates an exponential relation in different values
between A L and V L ; with axis indicated in log UTM coor-
- 10.1515/geo-2016-0032
Downloaded from PubFactory at 08/02/2016 04:37:28PM
via free access
Modeling of landslide volume estimation | 365
Figure 3: Diagram of correlation coeflcient between field and Figure 4: Diagram of exponential regression fitness on area and
estimated depth. volume data.
Figure 5: Experimental equation obtained between area and volume for landslides present in the area.
dinates. Exponential regression fitness on the data of ob- log UTM coordinates, there can be more seen the observa-
servational area and volume finally resulted in creation of tional data far from the given fitness line related to regres-
Equation (3). sion equation.
V L = 2.482 × A1.024
L (3)
2
R = 0.99 4.4 Evaluating the Model Provided
Where; A L is the area (m2 ); V L is the volume (m3 ). This To evaluate the calculation values of size of landslide by
exponential model can be used for estimating the size of related equation, these values were compared to volumet-
mass movements of slide type, with known interrupted ric estimations by other equations (Table 1). Figure 6 indi-
range area. cates the results of such comparison in the frame of a cal-
Studying the diagram obtained for calculating the size culating the statistics values of 25, 50 and 75%, min, max
of landslides indicate that despites showing data in log-
- 10.1515/geo-2016-0032
Downloaded from PubFactory at 08/02/2016 04:37:28PM
via free access
366 | A. Amirahmadi et al.
Figure 6: Comparing the statistics of 25%, 50%, 75%, min, max, mea of volumes estimated by different equations and equations provided
in Baghi area.
and total mean of data estimated by different experimental 4.5 Statistical comparing the Developed
equations as well as equation provided for Baghi area. Model with Other Models
After calculating the size of landslides in the Baghi
area, the diagram for range of area determined (Table 1) Each different model as introduced in Table 1 were used
were drawn by different equations (Figure 7). for 44 landslides present in the Baghi area and data pre-
According to Figures 6 and 7, it can confirm that equa- dicted by each equation were comparing to observational
tions like references [30, 33, 39] were relatively in good data by standard coefficient and Chi-Square Errors (RMSE)
conformity to diagram of this study (equation 3); however, (Table 6).
models developed by [35, 41, 42] , have relatively lower esti- Results for standard coefficient indicated that there is
mation than developed equation. In the above area range, a significant correlation in the confidence level of 99% be-
of course, there were also developed some models like [31, tween predicted values by all these equations with obser-
36, 37, 50] with relatively higher estimations than equa- vational volume. Because standard coefficient can accu-
tions provided here. As indicated in Figure 4, for equa- rately indicate the accuracy of different equations, there-
tion developed by [37, 43] for very great landslides (Ta- fore RMSE was also used. As indicated in Table 4, be-
ble 1), the minimum predicted volume by such equations sides equation developed by this study, according to [38]
was 91.289279 m3 and 1157256.40 m3 ; while minimum ob- has a lower RMSE value than other equations followed
served volume in the province was 10000 m3 and mini- by [30, 34, 39], having predictions relatively close to the
mum volume calculated by related equation was 15000 m3 observational data.
and was almost reliable estimation. As mentioned above,
such limitation for maximum data can be also seen in the
equations like [35, 41, 42].
- 10.1515/geo-2016-0032
Downloaded from PubFactory at 08/02/2016 04:37:28PM
via free access
Modeling of landslide volume estimation | 367
Table 6: Results for studying the values of standard coeflcient and RMSE between observed and predicted data by different relations.
Experimental This Guz Sim Inn Guth Kor Ima. Guz. Ima
equations study (2009) (1967) (1983) (2004) (2005) (2007) (2008) (2008)
Standard 0.646 0.574 0.59 0.587 0.638 0.476 0.633 0.577 0.623
coeflcient
(sig.) 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000
RMSE 1.245 15.757 9.165 1.785 2.097 10.730 1.500 13.794 1.443
× 106 × 106 × 106 × 106 × 106 × 106 × 106 × 106 × 106
Experimental Ric. Abe Whi. Lar. Mar Haf. Ric. Ten. Omi.
equations (1971) (1974). (1983) (1998) (2002). (2005) (1969) (2006) (2011)
Standard 0.637 0.602 0.613 0.655 0.656 0.644 0.636 0.605 0626
coeflcient
(sig.) 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000
RMSE 1.805 5.798 8.942 2.129 2.221 8.241 1.910 116.148 3.118
× 106 × 106 × 106 × 106 × 106 × 106 × 106 × 106 × 106
Figure 7: Diagram of different experimental equations as well as diagram obtained by this study.
4.6 Calculating the mean depth of tions obtained, it seems that this developed model include
landslides in the area an acceptable model for predicting the volume of land-
slides. Therefore, the volumetric mean for 44 landslides
According to the similarity between equation obtained for with given data have been estimated by using the equa-
determining the size of landslides in the Baghi area (Equa- tion developed by this study equal to 922658.42 m3 . Ac-
tion 3) with equations developed internationally (Table 1) cording to mean area for these landslides, the mean depth
as well as relative conformity of these equations with equa- for landslides in Baghi area was estimated about 3.314 m.
- 10.1515/geo-2016-0032
Downloaded from PubFactory at 08/02/2016 04:37:28PM
via free access
368 | A. Amirahmadi et al.
Because mean depth for these 44 landslides was also esti- and other morphometric parameters for landslide. By de-
mated about 4.06 m, therefore it can be seen a slight dif- veloping this model, when calculating the area of a land-
ference between observed depth and depth calculated by slide, one can find its volume and depth. By providing this
this equation. model indeed, one can save the cost and time for calculat-
These calculations were also conducted for different ing the volume and depth of landslide. Because equations
relations and mean depth was calculated based on mean with higher priority have some predictions with higher
volume predicted by any equation (Figure 8). As indicated and lower mean than observational data and because even
in Figure 8, the equations with higher predictions for mean lower difference between means when a study area is wide
total volume of slides, they seemed to have higher depth with higher landslides, may cause high changes in the to-
as well. For equations with lower predictions however, the tal volume of region, therefore it can be concluded that it is
calculated depth is lower than observed depth. necessary to develop a unique equation for each region, if
there are data for volume and area of landslides. Therefore
it is recommended to design and develop such models for
regions susceptible to landslide through the country hav-
5 Conclusions ing such data.
- 10.1515/geo-2016-0032
Downloaded from PubFactory at 08/02/2016 04:37:28PM
via free access
Modeling of landslide volume estimation | 369
[5] Souri S., Lashkari Pour G.R., Ghafoori M., Farhadinejad Taher., [20] Karimi H., Naderi F., Morshedi E., Nikseresht M., 2011. Zonation
2011. Zoning of landslide risk using Artificial Neural Network, a of landslide hazard in the Charavel Watershed, Ilam, using GIS
Case Study: National Area (Nojian), Journal of Engineering Ge- System; Quarterly of Applied Geology, 7th year, No. 4, 319–332,
ology, 5th Vol. No. 2, Fall and Winter, 2011, 1269–1286, (In Per- (In Persian).
sian). [21] Soeters R., van Westen C.J., 1996. Slope instability recog-
[6] Roering J.J., Kirchner J.W., Dietrich W.E., 2005. Characterizing nition, analysis and zonation. In:Turner, A.K., Schuster, R.L.
structural and lithologic controls on deep-seated landsliding: (Eds.), Landslide Investigation and Mitigation. Transportation
Implications for topographic relief and landscape evolution in Research Board Special Report, vol. 247. National Research
the Oregon Coast Range, USA. Geological Society of America Council, pp. 129–177.
Bulletin 117, 654–668. [22] Guzzetti F., Carrara A., Cardinali M., Reichenbach P., 1999. Land-
[7] Hattanji T., Moriwaki H., 2009. Morphometric analysis of relic slide hazard evaluation: a review of current techniques and their
landslides using detailed landslide distribution maps: Implica- application in a multi-scale study. Geomorphology 31, 181–216.
tions for forecasting travel distance of future landslides, Geo- [23] Malamud B.D., Turcotte D.L., Guzzetti F., Reichenbach P., 2004.
morphology 103, 447–454. Landslide inventories and their statistical properties. Earth Sur-
[8] George D.B., Kalliopi G.P., Hariklia D.S., Dimitrios P., Kon- face Processes and Landforms 29, 687–711.
stantinos G.Ch., 2012. Potential suitability for urban plan- [24] Cardinali M., Reichenbach P., Guzzetti F., Ardizzone F., Antonini
ning and industry development using natural hazard maps G., Galli M., Cacciano M., Castellani M., Salvati P., 2002. A geo-
and geological–geomorphological parameters. Environ Earth morphologic approach to estimate landslide hazard and risk in
Sci (2012) 66:537–548, DOI 10.1007/s12665-011-1263-x. urban and rural areas in Umbria, central Italy. Natural Hazards
[9] George D.B., Kalliopi G.P., Hariklia D.S., Georgios A.S., Kon- and Earth System Sciences 2 (1–2), 57–72.
stantinos G.Ch., 2013. Assessment of rural community and agri- [25] Reichenbach P., Galli M., Cardinali M., Guzzetti F., Ardizzone F.,
cultural development using geomorphological–geological fac- 2005. Geomorphologic mapping to assess landslide risk: con-
tors and GIS in the Trikala prefecture (Central Greece). Stoch En- cepts, methods and applications in the Umbria Region of cen-
viron Res Risk Assess (2013) 27:573–588, DOI 10.1007/s00477- tral Italy. In: Glade, T., Anderson, M.G., Crozier, M.J. (Eds.), Land-
012-0602-0 slide Risk Assessment. John Wiley, Chichester, pp. 429–468.
[10] Dai F.C., Lee C.F., Zhang X.H., 2001. GIS-basedgeo- [26] Hovius N., Stark C.P., Allen P.A., 1997. Sediment flux from a
environmental evaluation for urban land use planning: a mountain belt derived by landslide mapping. Geology 25, 231–
case study. Eng Geol 61:257–271. 234.
[11] Apostolidis E., Koukis G., 2013. Engineering-geological condi- [27] Harmon R.S., Doe III W.W., 2001. Landscape Erosion and Evolu-
tions of the formations in the Western Thessaly basin, Greece. tion Modeling. Springer–Verlag. 535 p.
Cent. Eur. J. Geosci. 5(3), 2013, 407–422, DOI: 10.2478/s13533- [28] Lavé J., Burbank D., 2004. Denudation processes and rates
012-0200-1 in the transverse ranges,southern California: erosional re-
[12] Papadopoulou-Vrynioti K., Bathrellos G., Skilodimou H., Kaviris sponse of a transitional landscape to external and anthro-
G., Makropoulos K., 2013. Karst collapse susceptibility mapping pogenic forcing. Journal of Geophysical Research 109 (F01006).
using seismic hazard in a rapid urban growing area. Eng. Geol., doi:10.1029.2003JF000023, 2004.
2013 158, 77–88, doi:10.1016/j.enggeo. 2013.02.009 [29] Korup O., 2005a. Geomorphic imprint of landslides on alpine
[13] Smith K., 2003, Environmental Risks, translated by Goodarzine- river systems, southwest New Zealand. Earth Surface Processes
jad Shapour and Moghimi Ebrahim, 1st edition, Samt Press. and Landforms 30, 783–800.
[14] Kanungo D.P., Arora M.K., Sarcar S., Gupta R.P., 2006. A com- [30] Imaizumi F., Sidle R.C., 2007. Linkage of sediment sup-
parative study of conventional, ANN black box, fuzzy and com- ply and transport processes in Miyagawa Dam catch-
bined neural and fuzzy weighting procedures for landslide sus- ment, Japan. Journal Geophysical Research 112 (F03012).
ceptibility zonation ln Darjeeling Himalayas. Engineering Geol- doi:10.1029.2006JF000495.
ogy, 85: 347–366. [31] Guzzetti F., Ardizzone F., Cardinali M., Galli M., Reichenbach P.,
[15] Crozier M.J., 1986. Landslides: Causes, Consequences & Envi- Rossi M., 2008. Distribution of landslides in the Upper Tiber
ronment. Croom Helm Pub.,London. 245 p. River basin, central Italy. Geomorphology 96, 105–122.
[16] Turner A.K., Schuster R.L. (Eds.), 1996. Landslides: Investiga- [32] Simonett D.S., 1967. Landslide distribution and earthquakes in
tion and Mitigation. the Bewani and Torricelli Mountains, New Guinea. In: Jennings
[17] Omidvar E., Kavian A, 2011. Estimating the size of landslide J.N., Mabbutt J.A. (Eds.), Landform Studies from Australia and
based on area in the regional scale (A Case Study: Mazandaran NewGuinea. Cambridge University Press, Cambridge, 64–84.
province); Journal of Range and Watershed, Iranian Journal of [33] Rice R.M., Corbett E.S., Bailey R.G., 1969. Soil slips related to
Natural Resources, period 63, No. 4, Winter 2010, 439–455, (In vegetation, topography, and soil in Southern California. Water
Persian). Resources Research 5 (3), 647–659.
[18] Rezaei Moghadam M.H., Feizallah Pour M., Asghari Sayad., [34] Innes J.N., 1983. Lichenometric dating of debris-flow deposits in
2011. Estimating the mathematics between factors of volume, the Scottish Highlands. Earth Surface Processes and Landforms
area of mass slide in the Saein neck (Nir town), Journal of Sci- 8, 579–588.
ence and Research of Iranian Geographical Club, 9th year. No. [35] Guthrie R.H., Evans S.G., 2004. Analysis of landslide frequen-
28, Spring 203, 2011–218, (In Persian). cies and characteristics ina natural system, coastal British
[19] Crosta B.G. 2009. Dating, triggering, modeling and hazard as- Columbia. Earth Surface Processes and Landforms 29, 1321–
sessment of large landslides, Geomorphology, 103, 1–4. 1339.
- 10.1515/geo-2016-0032
Downloaded from PubFactory at 08/02/2016 04:37:28PM
via free access
370 | A. Amirahmadi et al.
[36] Korup O., 2005b. Distribution of landslides in southwest New [45] Klar A., Aharonov E., Kalderon-Asael B., Katz O., 2011.Analyti-
Zealand. Landslides 2, 43–51. cal and observational relations between landslide volume and
[37] Tenbrink U.S., Geist E.L., Andrews B.D., 2006. Size distribution surface area, Journal of Geophysical Research, vol. 116, 1–10.
of submarine landslides and its implication to tsunami hazard [46] Hosseini S.A., Lotfi R., 2007. Studying the Landslide Event phys-
in Puerto Rico. Geophysical Research Letters 33, L11307. Trans- iographically (A Case Study of Chaibagh, Wood and Paper Indus-
portation Research Board Special Report, vol. 247. National Re- tries Hall of Mazandaran), 2nd conference for combating with
search Council,Washington, D.C. 67 natural, (In Persian).
[38] Imaizumi F., Sidle R.C., Kamei R., 2008. Effects of forest harvest- [47] Javan Dolouei G., Abbasi M., Jafarian S.M., 2011. Determining
ing on the occurrence of landslides and debris flows in steep the area of slide and volume of overburden volume of falling
terrain of central Japan. Earth Surface Processes and Landforms mass based on geophysical measurements (A Case Study:
33, 827–840. doi:10.1002/esp.1574. Falling mass next to the Khanjele Tunnel, Bulletin of Seismology
[39] Rice R.M., Foggin III G.T., 1971. Effects of high intensity storms and Earthquake Engineering, 14th year, Nos. 3& 4, Fall& winter,
on soil slippage on mountainous watersheds in Southern Cali- 2011, 9–1, (In Persian).
fornia. Water Resources Research 7 (6), 1485–1496. [48] Freund J, 1992. Mathematical statistics. 5th edition. 650 p, (In
[40] Abele G., 1974. Bergsturze in den Alpen – ihre Verbreitung, Mor- Persian).
phologie und Folgeerscheinungen, Wiss. Alpenvereinshefte, [49] Kim S., Kim H.S., 2008. Neural networks and genetic algo-
25, 247 p. rithm approach for nonlinear evaporation and evapotranspira-
[41] Larsen M.C., Torres Sanchez A.J., 1998. The frequency and dis- tion modeling. Journal of Hydrology, vol. 351, pp. 299–317.
tribution of recent landslides in three montane tropical regions [50] Guzzetti F., Ardizzone F., Cardinali M., Rossi M., Valigi D., 2009.
of Puerto Rico: Geomorphology, v. 24, p. 309–331. Landslide volumes and landslide mobilization rates in Umbria,
[42] Martin Y., Rood K., Schwab J.W., Church M., 2002. Sediment central Italy, Earth and Planetary Science Letters, vol. 279, 222–
transfer by shallow landsliding in the Queen Charlotte Islands, 229.
British Columbia. Canadian Journal of Earth Sciences 39 (2), [51] Whitehouse I.E. (1983), Distribution of large rock avalanche de-
189–205. posits in the central Southern Alps, New Zealand, N. Z. J. Geol.
[43] Haflidason H., Lien R., Sejrup H.P., Forsberg C.F., Bryn P., 2005. Geophys., 26, 272–279.
The dating and morphometry of the Storegga Slide. Marine and
Petroleum Geology.
[44] Kalderon-Asael B., Katz O., Aharonov E., Marco Sh., 2008. Mod-
elling the relation between area and volume of landslides. Min-
istry of National Infrastructures, Geological Survey of Israel,
Jerusalem, April 2008, 1–21.
- 10.1515/geo-2016-0032
Downloaded from PubFactory at 08/02/2016 04:37:28PM
via free access