Improved landslide susceptibility mapping using unsupervised and supervised collaborative machine learning models

Su, Chenxu; Wang, Bijiao; Lv, Yunhong; Zhang, Mingpeng; Peng, Dalei; Bate, Bate; Zhang, Shuai

doi:10.1080/17499518.2022.2088802

Cited by 18 publications

(10 citation statements)

References 37 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The KM clustering method is an unsupervised classification algorithm that is applicable to the classification of unlabeled sample data 36 . The KM clustering method does not need to know the label (landslide or non-landslide) of each sample when training the model.…”

Section: Methodsmentioning

confidence: 99%

Effects of non-landslide sampling strategies on machine learning models in landslide susceptibility mapping

Gu,

Duan,

Wang

et al. 2024

Sci Rep

View full text Add to dashboard Cite

This study aims to explore the effects of different non-landslide sampling strategies on machine learning models in landslide susceptibility mapping. Non-landslide samples are inherently uncertain, and the selection of non-landslide samples may suffer from issues such as noisy or insufficient regional representations, which can affect the accuracy of the results. In this study, a positive-unlabeled (PU) bagging semi-supervised learning method was introduced for non-landslide sample selection. In addition, buffer control sampling (BCS) and K-means (KM) clustering were applied for comparative analysis. Based on landslide data from Qiaojia County, Yunnan Province, China, collected in 2014, three machine learning models, namely, random forest, support vector machine, and CatBoost, were used for landslide susceptibility mapping. The results show that the quality of samples selected using different non-landslide sampling strategies varies significantly. Overall, the quality of non-landslide samples selected using the PU bagging method is superior, and this method performs best when combined with CatBoost for predicting (AUC = 0.897) landslides in very high and high susceptibility zones (82.14%). Additionally, the KM results indicated overfitting, displaying high accuracy for validation but poor statistical outcomes for zoning. The BCS results were the worst.

show abstract

Section: Methodsmentioning

confidence: 99%

Effects of non-landslide sampling strategies on machine learning models in landslide susceptibility mapping

Gu,

Duan,

Wang

et al. 2024

Sci Rep

View full text Add to dashboard Cite

show abstract

“…Kayastha et al (2013) stated that the stream network is crucial for controlling landslides. This occurrence is because streams contribute to the erosion of the bases of slopes and the saturation of the underlying slopeforming material, increasing the probability of landslides (Saha et al, 2002;Chen et al, 2018;Su et al, 2022;Youssef et al, 2022). Several tributaries flow into the Meze and Mazole Rivers in the study area, and landslides frequently occur near these rivers.…”

Section: Curvaturementioning

confidence: 97%

“…Considering precipitation and soil composition factors, higher inclinations tend to be more prone to landslides. Furthermore, the slope aspect can also affect the susceptibility to landslides (Su et al, 2022). Slopes facing a specific direction may occasionally receive more heat and sunlight exposure, leading to the soil drying up and becoming less stable (Haque et al, 2019).…”

Section: Introductionmentioning

confidence: 99%

Assessing landslide susceptibility in Lake Abya catchment, Rift Valley, Ethiopia: A GIS-based frequency ratio analysis

Oyda,

Jothimani,

Regasa

2024

J. Degrade. Min. Land Manage.

View full text Add to dashboard Cite

Ethiopia's varied landscape, significant rainfall, and diverse geological characteristics pose risks of landslides. The specific research area spans 40 km2 within the Lake Abaya catchment area in the Rift Valley of Ethiopia. This investigation aimed to map landslide susceptibility using remote sensing information, GIS technology, and frequency ratio analysis. It evaluated multiple factors influencing landslide susceptibility. The process involved meticulous mapping of thematic layers, utilizing GIS techniques and diverse data sources, including primary data, satellite imagery, and secondary sources. A combination of Google Earth image analysis and field surveys was used to map landslide susceptibility in inaccessible areas. It was determined that 138 landslide sites existed. Of these, 30% (41 points) were assigned to the test of the model and another 30% to the training of the model, for a total of 97 points. The landslide susceptibility was classified into five categories based on frequency ratio analysis of the landslide susceptibility index (LSI): very low, low, moderate, high, and very high. The northeastern sector of the study area demonstrated a comparatively diminished susceptibility to landslides, ranging from low to moderate, whereas the central and southern regions showcased markedly elevated vulnerability. An evaluation of the model's accuracy using the area under the curve (AUC) method based on test inventory landslide data produced encouraging results: 84.8% accuracy on the success rate curve and 78.8% accuracy on the prediction rate curve. Based on the frequency ratio model, a susceptibility map is derived to represent susceptibility levels accurately.

show abstract

“…Because of these limitations, SL methods may not be applicable where there are a limited number of labeled samples as they are not always easy to obtain and may be expensive to acquire in abundance through image interpretation and site surveying, especially in a large study area. USL-based approaches are applied and have contributed to improving the implementation and the accuracy of LSM in such situations (Lei et al, 2018;Hu et al, 2021;Yimin et al, 2021;Mao et al, 2022;Su et al, 2022;Liu et al, 2023). USL-based methods such as clustering can be used to map the susceptibility areas, as they can identify the underlying structures in unlabeled datasets, hence, do not require data with predefined labels, and do not involve a training process during their implementation.…”

Section: Introductionmentioning

confidence: 99%

“…Clustering is a common USL method that assigns a set of samples (mapping units) into some subclasses or clusters based on certain similarities so that samples in a certain subclass have a maximum similarity degree to those in other subclasses (Wang et al, 2017). Over decades, these methods have been widely used in other fields such as marketing research, pattern recognition, and image processing, but very rarely explored in LSM studies (Huang et al, 2020;Su et al, 2022). In recent years, making use of the advantages of these methods, some landslide researchers have also shown interest and conducted LSM studies using these methods (Wan et al, 2015;Wang et al, 2017;Hu et al, 2019;Mao et al, 2021a;Mao et al, 2021b;Hu et al, 2021;Pokharel et al, 2021;Yimin et al, 2021;Mao et al, 2022).…”

Section: Introductionmentioning

confidence: 99%

Landslide susceptibility mapping using O-CURE and PAM clustering algorithms

Mwakapesa

Lan²,

Nanehkaran³

et al. 2023

Front. Environ. Sci.

View full text Add to dashboard Cite

Landslide susceptibility mapping (LSM) is a crucial step during landslide assessment and environmental management. Clustering algorithms can construct effective models for LSM. However, a random selection of important parameters, inconsideration of uncertain data, noise data, and large datasets can limit the implementation of clustering in LSM, resulting in low and unreliable performance results. Thus, to address these problems, this study proposed an optimized clustering algorithm named O-CURE, which combines: the traditional Clustering Using REpresentatives algorithm (CURE), that is, efficient for large datasets and noise data, the partition influence weight (PIW)-based method to enhance the selection of sample sets and the city block distance (CIBD) for processing of the uncertain data in CURE clustering during LSM modeling. A database containing 293 landslide location samples, 213 non-landslide samples, and 7 landslide conditioning factors was prepared for the implementation and evaluation of the method. Also, a Multicollinearity analysis was conducted to select the most appropriate factors, and all the factors were acceptable for modeling. Based on O-CURE, landslide density, and the partitioning around medoids (PAM) algorithm a susceptibility map was constructed and classified into very high (33%), high (18%), moderate (24%), low (13%), and very low (12%) landslide susceptible levels. To evaluate the performance of the O-CURE model, five statistic metrics including accuracy, sensitivity, specificity, kappa, and AUC were applied. The analysis shows that O-CURE obtained accuracy = .9368, sensitivity = .9215, specificity = .9577, kappa = .8496, and AUC = .896 is an indication of high-performance capability. Also, the proposed method was compared with the CURE algorithm, three existing clustering methods, and popular supervised learning methods. From this assessment, O-CURE outperformed the other clustering methods while showing significant and more consistent performance than the supervised learning methods. Therefore, we recommend that the O-CURE model and the constructed map can be useful in assessing landslides and contribute to sustainable land-use planning and environmental management in light of future disasters.

show abstract

Improved landslide susceptibility mapping using unsupervised and supervised collaborative machine learning models

Cited by 18 publications

References 37 publications

Effects of non-landslide sampling strategies on machine learning models in landslide susceptibility mapping

Effects of non-landslide sampling strategies on machine learning models in landslide susceptibility mapping

Assessing landslide susceptibility in Lake Abya catchment, Rift Valley, Ethiopia: A GIS-based frequency ratio analysis

Landslide susceptibility mapping using O-CURE and PAM clustering algorithms

Contact Info

Product

Resources

About