Limits...
Optimized volume models of earthquake-triggered landslides.

Xu C, Xu X, Shen L, Yao Q, Tan X, Kang W, Ma S, Wu X, Cai J, Gao M, Li K - Sci Rep (2016)

Bottom Line: The samples were used to fit the conventional landslide "volume-area" power law relationship and the 3 optimized models we proposed, respectively.Two data fitting methods, i.e. log-transformed-based linear and original data-based nonlinear least square, were employed to the 4 models.Results show that original data-based nonlinear least square combining with an optimized model considering length, width, height, lithology, slope, peak ground acceleration, and slope aspect shows the best performance.

View Article: PubMed Central - PubMed

Affiliation: Key Laboratory of Active Tectonics and Volcano, Institute of Geology, China Earthquake Administration, Beijing, 100029, China.

ABSTRACT
In this study, we proposed three optimized models for calculating the total volume of landslides triggered by the 2008 Wenchuan, China Mw 7.9 earthquake. First, we calculated the volume of each deposit of 1,415 landslides triggered by the quake based on pre- and post-quake DEMs in 20 m resolution. The samples were used to fit the conventional landslide "volume-area" power law relationship and the 3 optimized models we proposed, respectively. Two data fitting methods, i.e. log-transformed-based linear and original data-based nonlinear least square, were employed to the 4 models. Results show that original data-based nonlinear least square combining with an optimized model considering length, width, height, lithology, slope, peak ground acceleration, and slope aspect shows the best performance. This model was subsequently applied to the database of landslides triggered by the quake except for two largest ones with known volumes. It indicates that the total volume of the 196,007 landslides is about 1.2 × 10(10) m(3) in deposit materials and 1 × 10(10) m(3) in source areas, respectively. The result from the relationship of quake magnitude and entire landslide volume related to individual earthquake is much less than that from this study, which reminds us the necessity to update the power-law relationship.

No MeSH data available.


Related in: MedlinePlus

Performances and uncertainties of each model in the Tables 1 and 2.The performance means the percentage of calculated total volume to real total volume of the 1,415 samples. Squares denote the regressed values. Upper and lower vertical lines denote the uncertainties of volume% considering of ±1σ on γ1.
© Copyright Policy - open-access
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC4940731&req=5

f3: Performances and uncertainties of each model in the Tables 1 and 2.The performance means the percentage of calculated total volume to real total volume of the 1,415 samples. Squares denote the regressed values. Upper and lower vertical lines denote the uncertainties of volume% considering of ±1σ on γ1.

Mentions: Next, the nonlinear equations were regressed by the nonlinear least square method through iteration. Original data were directly used in modeling without any data transformation but the initial values of every coefficient were set for the iteration. In order to avoid iteration failure caused by big gaps between the initial values and actual values, the coefficients in Table 1 were employed as the initial values for equations (6–9). Results in Table 2 show significant differences to the linear method in Table 1. Similar to Table 1, the orders of R2 generally increase from 0.777 to 0.787. However, standard deviations of the landslide volume generally decrease from 365,614 m2 to 358,79 m2, which is different from the cases in Table 1. Therefore, it shows the increasing reliability of the four models (Table 2) on the basis of the original data-based nonlinear method. The percentage of regressed total volume of the samples to the real one does not show similar tendency, which is probably because the volumes of most landslides are less than real values. The optimal model in this study aims at the closest fitting between the regression value and true value for every landslide, rather than only the accuracy of the total landslide volume. Furthermore, the total landslide volumes of the four models are very close to the true values, with a percentage range 97.98–98.91% (Fig. 3). Therefore, the type 8 (equation 9) is the best model by considering values close enough between regression and reality for individual landslides.


Optimized volume models of earthquake-triggered landslides.

Xu C, Xu X, Shen L, Yao Q, Tan X, Kang W, Ma S, Wu X, Cai J, Gao M, Li K - Sci Rep (2016)

Performances and uncertainties of each model in the Tables 1 and 2.The performance means the percentage of calculated total volume to real total volume of the 1,415 samples. Squares denote the regressed values. Upper and lower vertical lines denote the uncertainties of volume% considering of ±1σ on γ1.
© Copyright Policy - open-access
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC4940731&req=5

f3: Performances and uncertainties of each model in the Tables 1 and 2.The performance means the percentage of calculated total volume to real total volume of the 1,415 samples. Squares denote the regressed values. Upper and lower vertical lines denote the uncertainties of volume% considering of ±1σ on γ1.
Mentions: Next, the nonlinear equations were regressed by the nonlinear least square method through iteration. Original data were directly used in modeling without any data transformation but the initial values of every coefficient were set for the iteration. In order to avoid iteration failure caused by big gaps between the initial values and actual values, the coefficients in Table 1 were employed as the initial values for equations (6–9). Results in Table 2 show significant differences to the linear method in Table 1. Similar to Table 1, the orders of R2 generally increase from 0.777 to 0.787. However, standard deviations of the landslide volume generally decrease from 365,614 m2 to 358,79 m2, which is different from the cases in Table 1. Therefore, it shows the increasing reliability of the four models (Table 2) on the basis of the original data-based nonlinear method. The percentage of regressed total volume of the samples to the real one does not show similar tendency, which is probably because the volumes of most landslides are less than real values. The optimal model in this study aims at the closest fitting between the regression value and true value for every landslide, rather than only the accuracy of the total landslide volume. Furthermore, the total landslide volumes of the four models are very close to the true values, with a percentage range 97.98–98.91% (Fig. 3). Therefore, the type 8 (equation 9) is the best model by considering values close enough between regression and reality for individual landslides.

Bottom Line: The samples were used to fit the conventional landslide "volume-area" power law relationship and the 3 optimized models we proposed, respectively.Two data fitting methods, i.e. log-transformed-based linear and original data-based nonlinear least square, were employed to the 4 models.Results show that original data-based nonlinear least square combining with an optimized model considering length, width, height, lithology, slope, peak ground acceleration, and slope aspect shows the best performance.

View Article: PubMed Central - PubMed

Affiliation: Key Laboratory of Active Tectonics and Volcano, Institute of Geology, China Earthquake Administration, Beijing, 100029, China.

ABSTRACT
In this study, we proposed three optimized models for calculating the total volume of landslides triggered by the 2008 Wenchuan, China Mw 7.9 earthquake. First, we calculated the volume of each deposit of 1,415 landslides triggered by the quake based on pre- and post-quake DEMs in 20 m resolution. The samples were used to fit the conventional landslide "volume-area" power law relationship and the 3 optimized models we proposed, respectively. Two data fitting methods, i.e. log-transformed-based linear and original data-based nonlinear least square, were employed to the 4 models. Results show that original data-based nonlinear least square combining with an optimized model considering length, width, height, lithology, slope, peak ground acceleration, and slope aspect shows the best performance. This model was subsequently applied to the database of landslides triggered by the quake except for two largest ones with known volumes. It indicates that the total volume of the 196,007 landslides is about 1.2 × 10(10) m(3) in deposit materials and 1 × 10(10) m(3) in source areas, respectively. The result from the relationship of quake magnitude and entire landslide volume related to individual earthquake is much less than that from this study, which reminds us the necessity to update the power-law relationship.

No MeSH data available.


Related in: MedlinePlus