Spatial clustering and modelling for landslide susceptibility mapping in the north of the Kathmandu Valley, Nepal
- PDF / 1,403,546 Bytes
- 17 Pages / 595.276 x 790.866 pts Page_size
- 16 Downloads / 227 Views
Badal Pokharel I Omar F. Althuwaynee I Ali Aydda I Sang-Wan Kim I Samsung Lim I Hyuck-Jin Park
Spatial clustering and modelling for landslide susceptibility mapping in the north of the Kathmandu Valley, Nepal
Abstract In this article, we propose and test alternative sampling strategies based on clustering distribution concepts to increase the efficiency of the landslide susceptibility model outcomes, instead of common random selection method for training and testing samples. To that end, we prepared a comprehensive landslide inventory and used six unsupervised clustering algorithms (Kmeans, K-medoids, hierarchical cluster (HC) analysis, expectation–maximization using Gaussian mixture models (EM/ GMM), affinity propagation, and mini batch K-means) to generate six different training datasets. After getting the cluster pattern in each technique, we classified it into 70% and 30% for training and testing samples, respectively. We generated an additional training dataset using random selection procedure to test the hypothesis. The EM/GMM model exhibited the highest accuracy than the other methods. The findings confirm the hypothesis and recommend investing in natural distribution of landslides incident, as training concepts, instead of random sampling. Keywords Landslides . Unsupervised clustering . Spatial patterns . Logistic regression . Parametric analysis Introduction Landslides are one of the most significant secondary hazards associated with the earthquakes (Jibson et al. 2000). An earthquake with a strong magnitude has the potential to triggered thousands of landslides in a large spatial range (Jelének et al. 2018). Earthquake-induced landslides are frequent in mountain areas; they contribute to erosional budgets and causing loss of life (Keefer 2002; Dadson et al. 2004; Hovius et al. 2011; Li et al. 2014). A recent example of such disastrous earthquake event is the 7.8 magnitude Gorkha Earthquake 2015 which occurred at Barpark, Gorkha, Nepal, on 25 April 2015. In central Nepal, about 25,000 landslides covering an area of approximately 87 km2 of sloping land were triggered by the main shock and aftershock sequences (Roback et al. 2017). Nearly 9000 people lost their lives, and 14 districts in the central Nepal were severely affected due to the earthquake event and associated hazards. The earthquakeinduced landslides, particularly in the highways, caused most of the damages (Collins and Jibson 2015). Landslide susceptibility mapping defined as the likelihood of landslides in a geographical location for a given set of environmental factors (Guzzetti et al. 2005) and as a primary step of mitigating the disaster and extracting information of the potentially susceptible regions. Over the several past decades, in literature, several heuristic, physically or statistically based modelling methods were considered to generate the landslide susceptibility maps (Reichenbach et al. 2018). In addition, new techniques, in the domain of machine learning (ML) and data mining, have been introduced to analyse the landslide suscept
Data Loading...
