Landslide detection based on contour-based deep learning framework in case of national scale of Nepal in 2015

Yu, Bo; Chen, Fang; Xu, Chong

doi:10.1016/j.cageo.2019.104388

Cited by 85 publications

(39 citation statements)

References 29 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Glacial lakes are formed and develop temporally with the retreat or thinning of glaciers and are directly or indirectly fed by glacier meltwater. They are located within 10 km of the nearest glacier terminus (Wang et al, 2013;Zhang et al, 2015).…”

Section: Study Areamentioning

confidence: 99%

“…They set the minimum mapping unit to 0.0081 km 2 . Zhang et al (2015) presented a database of glacial lakes larger than 0.003 km 2 in the Third Pole for the years 1990, 2000, and 2010. All of these studies significantly help to fill the data gap relating to information about glacial lakes in the HMA region. At the global scale, Pekel et al (2016) used millions of Landsat satellite images to record global surface water over the past 32 years at 30 m resolution, and many large and visible glacial lakes were also included.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Annual 30 m dataset for glacial lakes in High Mountain Asia from 2008 to 2017

Chen

Zhang

Guo

et al. 2021

Earth Syst. Sci. Data

Self Cite

111

137

View full text Add to dashboard Cite

Abstract. Atmospheric warming is intensifying glacier melting and glacial-lake development in High Mountain Asia (HMA), and this could increase glacial-lake outburst flood (GLOF) hazards and impact water resources and hydroelectric-power management. There is therefore a pressing need to obtain comprehensive knowledge of the distribution and area of glacial lakes and also to quantify the variability in their sizes and types at high resolution in HMA. In this work, we developed an HMA glacial-lake inventory (Hi-MAG) database to characterize the annual coverage of glacial lakes from 2008 to 2017 at 30 m resolution using Landsat satellite imagery. Our data show that glacial lakes exhibited a total area increase of 90.14 km2 in the period 2008–2017, a +6.90 % change relative to 2008 (1305.59±213.99 km2). The annual increases in the number and area of lakes were 306 and 12 km2, respectively, and the greatest increase in the number of lakes occurred at 5400 m elevation, which increased by 249. Proglacial-lake-dominated areas, such as the Nyainqêntanglha and central Himalaya, where more than half of the glacial-lake area (summed over a 1∘ × 1∘ grid) consisted of proglacial lakes, showed obvious lake-area expansion. Conversely, some regions of eastern Tibetan mountains and Hengduan Shan, where unconnected glacial lakes occupied over half of the total lake area in each grid, exhibited stability or a slight reduction in lake area. Our results demonstrate that proglacial lakes are a main contributor to recent lake evolution in HMA, accounting for 62.87 % (56.67 km2) of the total area increase. Proglacial lakes in the Himalaya ranges alone accounted for 36.27 % (32.70 km2) of the total area increase. Regional geographic variability in debris cover, together with trends in warming and precipitation over the past few decades, largely explains the current distribution of supraglacial- and proglacial-lake area across HMA. The Hi-MAG database is available at https://doi.org/10.5281/zenodo.4275164 (Chen et al., 2020), and it can be used for studies of the complex interactions between glaciers, climate and glacial lakes, studies of GLOFs, and water resources.

show abstract

Section: Study Areamentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Annual 30 m dataset for glacial lakes in High Mountain Asia from 2008 to 2017

Chen

Zhang

Guo

et al. 2021

Earth Syst. Sci. Data

Self Cite

111

137

View full text Add to dashboard Cite

show abstract

“…Artificial neural networks (ANNs) have proven to be successful in dealing with complicated problems, such as function approximation and pattern recognition [ 51 ]. Due to their powerful capability and functionality, ANNs provide an alternative approach for many assessment problems that are difficult to solve by conventional approaches [ 52 ]. The Backpropagation (BP) neural network is currently the most widely used ANN [ 6 ].…”

Section: Discussionmentioning

confidence: 99%

Vulnerability Analysis to Drought Based on Remote Sensing Indexes

Jia

Chen

Zhang

et al. 2020

IJERPH

Self Cite

View full text Add to dashboard Cite

A vulnerability curve is an important tool for the rapid assessment of drought losses, and it can provide a scientific basis for drought risk prevention and post-disaster relief. Those populations with difficulty in accessing drinking water because of drought (hereon “drought at risk populations”, abbreviated as DRP) were selected as the target of the analysis, which examined factors contributing to their risk status. Here, after the standardization of disaster data from the middle and lower reaches of the Yangtze River in 2013, the parameter estimation method was used to determine the probability distribution of drought perturbations data. The results showed that, at the significant level of α = 0.05, the DRP followed the Weibull distribution, whose parameters were optimal. According to the statistical characteristics of the probability density function and cumulative distribution function, the bulk of the standardized DRP is concentrated in the range of 0 to 0.2, with a cumulative probability of about 75%, of which 17% is the cumulative probability from 0.2 to 0.4, and that greater than 0.4 amounts to only 8%. From the perspective of the vulnerability curve, when the variance ratio of the normalized vegetation index (NDVI) is between 0.65 and 0.85, the DRP will increase at a faster rate; when it is greater than 0.85, the growth rate of DRP will be relatively slow, and the disaster losses will stabilize. When the variance ratio of the enhanced vegetation index (EVI) is between 0.5 and 0.85, the growth rate of DRP accelerates, but when it is greater than 0.85, the disaster losses tend to stabilize. By comparing the coefficient of determination (R2) values fitted for the vulnerability curve, in the same situation, EVI is more suitable to indicate drought vulnerability than NDVI for estimating the DRP.

show abstract

“…In mountainous areas, altitude affects vegetation distribution, land use and slope, which can also be observed from the correlation analysis mentioned in Section Ⅳ-A, and altitude further affects the stress value of the slope body, thereby affecting the occurrence of landslides. Vegetation can play an active role in the stability of shallow soil on the slope through the roots, and NDVI can reflect the distribution of vegetation, so NDVI has a strong correlation with the occurrence of shallow landslides [60]. The importance of land use factors illustrates the huge impact of human activities on landslides, which reminds people to carry out more detailed geological survey before land development and utilization in this area.…”

Section: B Importance Of Landslide Causative Factorsmentioning

confidence: 99%

Potential of Ensemble Learning to Improve Tree-Based Classifiers for Landslide Susceptibility Mapping

Song

Wang

Fang

et al. 2020

IEEE J. Sel. Top. Appl. Earth Observations Remote Sensing

View full text Add to dashboard Cite

Ensemble learning methods have been widely used due to their remarkable generalized performance, but their potential in landslide spatial prediction application is not fully studied. To take full advantage of ensemble learning techniques, the classification and regression tree classifier and four treebased ensemble classifiers of random forest, extremely randomized tree, gradient boosting decision trees and extreme gradient boosting decision trees are used in this study for landslide susceptibility assessment. Specifically, a stacking ensemble learning framework coupled with embedded feature selection is presented, consisting of multiple tree-based classifiers mentioned previously as base learners and logistic regression as a meta-learner in a two-layer structure. In the study area of Yongxin, China, 364 historical landslide locations were first randomly partitioned into a ratio of 7/3 for training and testing the model. Then, a spatial database of sixteen landslide causative factors were constructed for landslide prediction. Meanwhile, the relative importance of these factors were quantified by using the total number of feature splits and the average Gini index during the training process, and a novel embedded feature selection method was used in the base learner of the proposed framework to further improve computational efficiency and predictive performance by allowing each base learner to obtain its own optimal sub-feature space. Finally, different methods were assessed by using several evaluation criteria. Experimental results demonstrated that the proposed ensemble learning framework had the highest area under the curve value of 0.864, and it is more effective than the conventional tree-based classifiers and other ensemble learning methods.

show abstract

Landslide detection based on contour-based deep learning framework in case of national scale of Nepal in 2015

Cited by 85 publications

References 29 publications

Annual 30 m dataset for glacial lakes in High Mountain Asia from 2008 to 2017

Annual 30 m dataset for glacial lakes in High Mountain Asia from 2008 to 2017

Vulnerability Analysis to Drought Based on Remote Sensing Indexes

Potential of Ensemble Learning to Improve Tree-Based Classifiers for Landslide Susceptibility Mapping

Contact Info

Product

Resources

About