Snow avalanche hazard prediction using machine learning methods (original) (raw)
Related papers
Remote Sensing, 2019
Although snow avalanches are among the most destructive natural disasters, and result in losses of life and economic damages in mountainous regions, far too little attention has been paid to the prediction of the snow avalanche hazard using advanced machine learning (ML) models. In this study, the applicability and efficiency of four ML models: support vector machine (SVM), random forest (RF), naïve Bayes (NB) and generalized additive model (GAM), for snow avalanche hazard mapping, were evaluated. Fourteen geomorphometric, topographic and hydrologic factors were selected as predictor variables in the modeling. This study was conducted in the Darvan and Zarrinehroud watersheds of Iran. The goodness-of-fit and predictive performance of the models was evaluated using two statistical measures: the area under the receiver operating characteristic curve (AUROC) and the true skill statistic (TSS). Finally, an ensemble model was developed based upon the results of the individual models. Results show that, among individual models, RF was best, performing well in both the Darvan (AUROC = 0.964, TSS = 0.862) and Zarrinehroud (AUROC = 0.956, TSS = 0.881) watersheds. The accuracy of the ensemble model was slightly better than all individual models for generating the snow avalanche hazard map, as validation analyses showed an AUROC = 0.966 and a TSS = 0.865 in the Darvan watershed, and an AUROC value of 0.958 and a TSS value of 0.877 for the Zarrinehroud watershed. The results indicate that slope length, lithology and relative slope position (RSP) are the most important factors controlling snow avalanche distribution. The methodology developed in this study can improve risk-based decision making, increases the credibility and reliability of snow avalanche hazard predictions and can provide critical information for hazard managers.
A machine learning framework for multi-hazards modeling and mapping in a mountainous area
Scientific Reports
This study sought to produce an accurate multi-hazard risk map for a mountainous region of Iran. The study area is in southwestern Iran. The region has experienced numerous extreme natural events in recent decades. This study models the probabilities of snow avalanches, landslides, wildfires, land subsidence, and floods using machine learning models that include support vector machine (SVM), boosted regression tree (BRT), and generalized linear model (GLM). Climatic, topographic, geological, social, and morphological factors were the main input variables used. The data were obtained from several sources. The accuracies of GLM, SVM, and functional discriminant analysis (FDA) models indicate that SVM is the most accurate for predicting landslides, land subsidence, and flood hazards in the study area. GLM is the best algorithm for wildfire mapping, and FDA is the most accurate model for predicting snow avalanche risk. The values of AUC (area under curve) for all five hazards using the ...
Mass wasting susceptibility assessment of snow avalanches using machine learning models
Scientifc Reports, 2020
Snow avalanche is among the most harmful natural hazards with major socioeconomic and environmental destruction in the cold and mountainous regions. The devastating propagation and accumulation of the snow avalanche debris and mass wasting of surface rocks and vegetation particles threaten human life, transportation networks, built environments, ecosystems, and water resources. Susceptibility assessment of snow avalanche hazardous areas is of utmost importance for mitigation and development of land-use policies. This research evaluates the performance of the well-known machine learning methods, i.e., generalized additive model (GAM), multivariate adaptive regression spline (MARS), boosted regression trees (BRT), and support vector machine (SVM), in modeling the mass wasting hazard induced by snow avalanches. The key features are identified by the recursive feature elimination (RFE) method and used for the model calibration. The results indicated a good performance of the modeling process (Accuracy > 0.88, Kappa > 0.76, Precision > 0.84, Recall > 0.86, and AUC > 0.89), which the SVM model highlighted superior performance than others. Sensitivity analysis demonstrated that the topographic position index (TPI) and distance to stream (DTS) were the most important variables which had more contribution in producing the susceptibility maps. Snow avalanche is among the most destructive natural hazards in the cold and mountainous regions with devastating socioeconomic and environmental impacts 1-4. The mobility, transportation, tourism, and the leisure industries of the snowy mountain regions are under the avalanche's uncertain threats. Damaging infrastructures, roads and railways obstruction, threatening human life and the built environments and settlements, and harming the water resources, ecosystems, and vegetations are associated with the propagation and deposition of snow avalanche debris 5-11. The devastating propagation of the snow avalanches may also contribute to the mass wasting of surface rocks and vegetation particles transported along the way and accumulated together with the snow avalanche debris 12. The mass wasting induced by snow avalanche and the deposition of such snow-rock-debris poses longer-lasting damages with more destructive effects 4,13,14. The global snow avalanche regime is increasingly reported to be changing within the past decade 15-18. The climate change is introduced as a significant contributor in raising the occurrence rate and irregularity and increasing the risk and devastation 19-22. Therefore, more than ever, the accurate spatial hazard modeling and susceptibility mapping of the avalanche slopes and hazardous locations are seen crucial for risk management, planning efficient mitigation and adaptation practices, and territorial land-use policies. Modeling the avalanche triggering mechanisms is complicated 23-26. The complexity of avalanche models has been discussed in many studies 27-31. The snowpack, meteorology, terrain, and slope characteristics are the predominant contributing factors initiating the avalanche movement and propagation, and the debris deposition 2,32,33. Based on the interaction of these factors, the motion and run out of snow and eventually, the avalanche formation and propagation can be modeled 2,32,34-36. Various numerical methods for modeling the avalanche flow dynamics 37-41 , as well as statistical approaches for processing historical database information and climatological data sets 15,42-45 , have been proposed to predict the hazard susceptibility mapping. The models have been enhanced with the involvement of recent advanced OPEN
Toward the development of deep learning analyses for snow avalanche releases in mountain regions
Geocarto International, 2021
Snow avalanches impose a considerable threat to infrastructure and human safety in snow bound mountain areas. Nevertheless, the spatial prediction of snow avalanches has received little research attention in many vulnerable parts of the world, particularly in developing countries. The present study investigates the applicability of a stand-alone convolutional neural network (CNN) model, as a deep-2 learning approach, along with two metaheuristic algorithms including grey wolf optimization (CNN-GWO) and imperialist competitive algorithm (CNN-ICA) in snow avalanche modeling in the Darvan watershed, Iran. The analysis was based on thirteen potential drivers of avalanche occurrence and an inventory map of previously documented avalanche occurrences. The efficiency of models' performance was evaluated by Area Under the Receiver Operating Characteristic curve (AUC) and the Root Mean Square Error (RMSE). The CNN-ICA model yielded the highest accuracy in both training (AUC= 0.982, RMSE=0.067) and validation (AUC= 0.972, RMSE=0.125) steps, followed by the CNN-GWO model (AUC of 0.975 for training, RMSE of 0.18 for training, AUC of 0.968 for validation, RMSE of 0.157 for validation). However, the standalone CNN model showed lower goodness-of-fit (AUC= 0.864, RMSE=0.22) and predictive performance (AUC= 0.811, RMSE=0.330). The approach utilized in this study is broadly applicable for identifying areas where avalanche hazard is likely to be high and where mitigation measures or corresponding land use planning should be prioritized.
IEEE Access
Snow avalanche as a natural disaster severely affects socio-economic and geomorphic processes through damaging ecosystems, vegetation, landscape, infrastructures, transportation networks, and human life. Modeling the snow avalanche has been seen as an essential approach for understanding the mountainous landscape dynamics to assess hazard susceptibility leading to effective mitigation and resilience. Therefore, the main aim of this study is to introduce and implement an ensemble machine learning model of random subspace (RS) based on a classifier, functional tree (FT), named RSFT model for snow avalanche susceptibility mapping at Karaj Watershed, Iran. According to the best knowledge of literature, the proposed model, RSFT, has not earlier been introduced and implemented for snow avalanche modeling and mapping over the world. Four benchmark models, including logistic regression (LR), logistic model tree (LMT), alternating decision tree (ADT), and functional trees (FT) models were used to check the goodnessof-fit and prediction accuracy of the proposed model. To achieve this objective, the most important factors among many climatic, topographic, lithologic, and hydrologic factors, which affect the snow accumulation and snow avalanche occurrence, were determined by the information gain ratio (IGR) technique. The goodness-of-fit and prediction accuracy of the models were evaluated by some statistical-based indexes including, sensitivity, specificity, accuracy, kappa, and area under the ROC curve, Friedman and Wilcoxon sign rank tests. Results indicated that the ensemble proposed model (RSFT), had the highest performance (Sensitivity = 94.1%, Specificity = 92.4%, Accuracy = 93.3%, and Kappa = 0.782) rather than the other soft-computing benchmark models. The snow avalanche susceptibility maps indicated that the high and very high susceptibility avalanche areas are located in the north and northeast parts of the study area, which have a higher elevation with more precipitation and lower temperatures.
Journal of the Indian Society of Remote Sensing, 2019
The aim of this study is to generate a reliable dynamic snow avalanche hazard map using analytical hierarchy process method based on multisource geo-spatial data for the Chowkibal-Tangdhar (C-T) road axis in Jammu and Kashmir (J&K), India. Avalanche-prone areas of C-T axis have been demarcated using three causative parameters, i.e., terrain, ground cover and meteorological parameters. Terrain parameters, e.g., elevation, slope, aspect and curvature, have been estimated from Advanced Spaceborne Thermal Emission and Reflection Radiometer, Global Digital Elevation Model Version 2. Ground cover information has been extracted from Landsat-8 data. Meteorological parameters maps, i.e., snow depth, relative humidity and air temperature, have been generated using geo-spatial interpolation techniques of in situ data. All the parameters have been incorporated in Geographic Information System environment to generate the hazard map. Validation of hazard map was accomplished with the area under the curve method. The prediction rate was observed to be 93.2%. Further, 20% of the study area was estimated having no hazard, 55% as low hazard, 12% as moderate hazard and 13% as high hazard on April 13, 2015. Dynamic hazard map thus generated using remote sensing and in situ data will be useful for mitigation of snow avalanche hazard, regional planning for development of infrastructure, transportation, troops movement, army deployments and communication network.
Natural Hazards, 2020
Artificial neural network (ANN)-based models have been developed for simulation of snowpack parameters-RAM hardness, shear strength, temperature, density, thickness and settlement of snowpack layers using manually observed weather data. The simulated snowpack parameters have been used for development of ANN for avalanche prediction. The complete scheme of simulation of snowpack parameters and avalanche prediction has been named as HIM-STRAT and developed for Chowkibal-Tangdhar region in NorthWest Himalaya using weather and snow stratigraphy data collected at a representative observatory in that region. Weather and snowpack data collected during the period from 1992 to 2016 have been analysed to generate the database of prominent snowpack layers and associated weather variables. Snowpack parameters have been simulated using randomly selected 490 (80%) data points and tested with 123 (20%) data points. Simulated snowpack parameters-RAM hardness, shear strength of the weakest snowpack layer and overburden pressure on the weakest layer-have been used to derive stability index of snowpack (SIS). The SIS and other snowpack parameters such as snowpack height, RAM hardness, shear strength, storm snow and snow temperature have been derived for past 22 winters (1992-2014) and used for prediction of avalanches. The HIM-STRAT has been validated through computation of root-mean-square error of simulated snowpack parameters and accuracy, bias, false alarm rate and Heidke Skill Score of avalanche prediction for five winters from 2014 to 2019. The performance of HIM-STRAT for simulation of snowpack parameters and prediction of avalanches has been found reasonably good and discussed in detail.
On Estimating Avalanche Danger From Simulated Snow Profiles
International Snow Science Workshop Grenoble Chamonix Mont Blanc October 07 11 2013, 2013
Estimating avalanche danger is the primary goal of avalanche warning services. Typically avalanche danger is estimated based on a variety of information such as manual snow profiles, avalanche observations as well as weather data. However, this required information is often not available especially in data sparse areas, which are common in Canada. It has been shown that coupled snow cover and numerical weather prediction models can provide such information on the snow cover. For this study we simulated the snow cover for three elevation bands -alpine, tree-line, below tree-line -at Glacier National Park, B.C., Canada for the winter season 2012-2013 between December and March. Snow cover simulations were performed using the Swiss snow cover model SNOWPACK forced by weather data from the Canadian high-resolution numeric weather prediction model GEM-LAM. Experienced forecasters estimated the regional avalanche danger (Low to Extreme) daily during the same period for the three elevation bands. Multivariate classification trees were used to estimate the avalanche danger from the simulated profiles. Classification trees were built using four parameters derived from the simulated profiles. These four parameters were the new snow amountsmaximum over 24-hours and 3-days -as well as measures for the likelihood of triggering and the expected avalanche size -based on a skier stability index and the depth of a critical layer. A comparison of the avalanche danger estimated from the simulated profiles with the forecasted avalanche danger showed that the avalanche danger was estimated correctly with an accuracy of 77% for the alpine, 76% for tree-line and 70% below tree-line -overall accuracy about 74%. Although the simulated avalanche danger tends to be slightly underestimated, especially for the alpine and treeline, such a model chain can be a valuable tool for avalanche warning services especially for data sparse areas.
An innovative methodology to perform avalanche hazard mapping that combines open source GIS tools, expert rules, computational routines, and statistical analysis is herewith briefly presented. The method provides a "semi-automatic" definition of areas potentially affected by avalanche release, motion, and run-out and is based on a combination of an ad-hoc developed "flowrouting algorithm" and a statistically-based estimation of the run-out distance of avalanches. The proposed methodology allows obtaining a preliminary, cost-efficient assessment of the avalanche hazard zones, especially for those areas where historical information is lacunose or even missing. Furthermore it allows including confidence bonds into the analysis, providing different avalanche outline as a function of the safety requirements.
Computer Assisted Avalanche Forecasting: Skier-Triggered Avalanches
2004
Computer assisted avalanche forecasting has become a valuable tool in some forecasting operations in Canada. Avalanche forecasting models are based on statistical, numerical and/or rule-based methods, but usually predict natural rather than skier-triggered avalanches. A nearest neighbour forecasting model for skier-triggered avalanches on post storm weak layers was improved by adding snowpack properties including a stability index to meteorological variables. However, a skier stability index did not improve the forecasting success on storm snow instabilities. The best forecast could be achieved by using meteorological parameters.