Application of Analytical Hierarchy Process (AHP) and Frequency Ratio (FR) Model in Assessing Landslide Susceptibility and Risk - Semi-quantitative Approaches for Landslide Assessment and Prediction - page 199

indicates, strong and
positive relationship between landslide occurrences in each class of the data layers
and high landslide susceptibility where

The derived frequency ratio (FR) value of more than ' 1 '

' FR ' value of less than ' 1 ' depicts the

negative and low landslide susceptibility. In this study, values for each class

were accepted as prioritized class rating value (PCRV) or prioritized class weight
(PCW).

' FR '

7.2.1.6 Linear Combination Model and Landslide Susceptibility Map

Avinash and Ashamanjari ( 2010 ) and Intarawichian and Dasananda ( 2011 ) used a
landslide susceptibility index value (LSIV) which is the summation of class-and
factor-weighted values.
'

values for each class (PCRV) or prioritized class rating value, (Table G.1,
Appendix G) as well as prioritized factor

FR '

s weighted values (PFRV) for each factor
map was taken into account in calculating the landslide susceptibility index value
(LSIV) with the following linear combination model:

'

X

n
i ¼ 1 ð

LSIV ¼ W i FR i Þ FV ð 7 : 5 Þ

where, n: total number of factors included in the study (n = 10); Wi: i: Factor s weight

(PFRV), FV IS factor value, and FRi: i : Class Frequency Ratio/class weight.
The

'

, greater was
the propensity of landslide phenomena and vice versa. The LPIV based frequency
curve showed many oscillations. To classify the watershed into 5 susceptibility
zones moving averages with averaging window lengths of 3, 5, 7, and 9 were
considered for smoothing the frequency distribution curve (Fig. 7.2 ). After analyzing
four new curves, the Shivkhola Watershed was classi

' LSIV ' varied from ' 4.81 ' to ' 16.00 ' . Higher the value of ' LSIV '

ed into 5 landslide suscep-
tibility zones i.e. very low, low, moderate, high, and very high with class boundaries
were demarcated at the signi cant changes of gradient of the curves. The abrupt

change points on frequency curve (landslide threshold boundaries) were 7.05, 9.29,
11.5, and 13.8 which were recognized as class boundaries to classify the
map. A 3 × 3 ' majority lter ' technique was applied to the map as a post-classifi- -

cation filter to reduce the high frequency variation.

To verify the landslide susceptibility map, landslide density under each
susceptibility class was computed. The landslide inventory map was crossed with
prepared landslide susceptibility map to derive landslide affected pixels for each
susceptibility classes (zones). Research by Sarkar and Kanungo ( 2004 ) indicates
that the higher the landslide density, greater is the probability and larger the area is
affected by landslide in each landslide susceptibility class.