Enhancing Landslide Risk Management in India: A Research Perspective
Introduction
Landslides represent a major global threat, endangering infrastructure, communities, and the environment through loss of life, property damage, and degradation. Landslides can be triggered by various factors like rainfall, seismic activity, and human actions. Key factors influencing landslides include slope steepness, geological and hydrological conditions, soil properties, and land use practices.
Effective landslide hazard management requires a multidisciplinary approach and technological advancements. Remote sensing and numerical modeling have significantly improved our ability to assess landslide susceptibility, predict events, and develop early warning systems (EWSs). Rainfall, a primary trigger for landslides, has been extensively studied, with critical rainfall thresholds for debris flows and shallow landslides established decades ago. Earthquakes also contribute to slope instability, necessitating integrated risk assessment strategies.
Landslide susceptibility mapping (LSM) and EWSs are vital tools in hazard management, identifying vulnerable areas by analyzing topography, geology, seismic activity, and rainfall patterns. Geographic Information Systems (GIS) and statistical modeling play key roles in these assessments, with machine learning techniques increasingly being used to predict landslides.
Rainfall thresholds are crucial in landslide prediction, helping identify precipitation levels that could trigger landslides. By analyzing historical events and rainfall data, researchers can develop predictive models and EWSs, integrating this information with GIS and remote sensing to account for spatial and temporal variations, thereby enhancing system effectiveness. MEMS sensor-based monitoring systems further contribute by providing real-time data that improve landslide hazard assessment and early warning capabilities.
Despite significant global research on rainfall and earthquake-induced LSM and EWS development, critical gaps persist in landslide hazard management. In regions like the Indian Himalayas, Western Ghats, and Darjeeling Himalayas, there is a lack of a comprehensive framework that integrates susceptibility mapping with landslide monitoring for local and regional EWS development. Furthermore, there is a significant gap in research combining landslide susceptibility maps with rainfall thresholds to develop regional-level EWSs. Despite the pressing need, efficient, operational, and cost-effective landslide early warning systems remain largely absent in these regions. Addressing these research gaps is crucial to improving landslide hazard management and mitigating the devastating impacts of landslides in vulnerable areas.
The research at IIT Indore is focused on addressing these critical gaps, particularly in the context of the Indian Himalayas, Western Ghats, and Darjeeling Himalayas. The team is working on developing integrated models that combine landslide susceptibility mapping with rainfall thresholds and real-time monitoring systems. This includes studying the complex interactions between geological, hydrological, and climatic factors that contribute to landslide risk. Additionally, IIT Indore's research aims to enhance the accuracy of machine learning models used for landslide prediction by incorporating more diverse and precise datasets. The ultimate goal is to create a robust and cost-effective early warning system that can be deployed in landslide-prone regions, significantly improving hazard management and reducing the impact of landslides in India. Here is an overview of IIT Indore's research on landslides across various regions of India:
Darjeeling Himalayas
Our research in the Darjeeling Himalayas has aimed to address the severe issue of rainfall-induced landslides, which pose significant risks to life, property, and infrastructure in this region. The Indian Himalayan region, known for its complex geological features, steep slopes, and intense monsoonal rainfall, presents ideal conditions for landslides. The damage caused by these landslides necessitates effective strategies to minimize their impact and enhance disaster preparedness.
To tackle this issue, we have focused on developing and refining early warning systems and rainfall thresholds specific to the Kalimpong area within the Darjeeling Himalayas (Dikshit et al. 2018b; Abraham et al. 2020b, 2022a). Our research involved applying and adapting several models and methodologies to better predict and manage landslide risks. One of the core components of our study was the application of the SIGMA (Sistema Integrato Gestione Monitoraggio Allerta) model, which was originally developed for Emilia Romagna in Italy. This model uses statistical distributions of cumulative rainfall values to predict landslide occurrences.
We tailored this model to the Kalimpong context by utilizing a calibration dataset of daily rainfall and landslide occurrences from 2010 to 2015. The results were validated with independent data from 2016 and 2017. The SIGMA model demonstrated high efficiency in predicting landslides, achieving 92% accuracy with a likelihood ratio of 11.28. This performance highlights the model's potential for integration with rainfall forecasting systems, offering a robust tool for early warning in Landslide- prone areas (Abraham et al. 2020a).
In addition to the SIGMA model, we focused on establishing specific rainfall intensity–duration thresholds for landslide occurrences in Kalimpong. By using a power law equation, we determined that rainfall events with an intensity of 0.95 mm/h over a duration of 24 hours are associated with a high risk of landslides. Furthermore, our study identified that antecedent rainfall periods of 10 and 20 days require intensities of 88.37 mm and 133.5 mm, respectively, to trigger landslides. These findings are crucial for refining early warning systems and improving predictive accuracy for landslide events (Dikshit and Satyam 2018; Dikshit et al. 2020).
We also explored the application of the hydrological FLaIR (Forecasting of Landslides Induced by Rainfall) model, which incorporates two modules: Rainfall-Landslide (RL) and Rainfall- Forecasting (RF). The RL module correlates rainfall amounts with landslide occurrences, while the RF module simulates future rainfall events. By analyzing the mobility function Y(.) with data from the July 1, 2015 landslide event and validating it with 2016 monsoon data, we demonstrated the model's potential for predicting landslides. This analysis underscores the need for more detailed data to further refine the thresholds and enhance prediction accuracy.
Additionally, we installed an early warning and monitoring system in the Chibo Pashyor region of West Bengal, incorporating MicroElectroMechanical Systems tilt sensors and volumetric water content sensors. The slopes monitored are located on the banks of mountain rivulets (jhoras), known as sinking zones in Kalimpong, which are highly affected by surface displacements during the monsoon season due to heavy rainfall and poor drainage. This study represents the first real- time field monitoring effort in this area, evaluating the applicability of tilt sensors. The sensors, embedded within the soil, measured tilting angles and moisture content at shallow depths.
Data collected during the 2017–2019 monsoon seasons were compared with field observations and rainfall data, revealing that considering long-term rainfall conditions is more effective than focusing solely on immediate rainfall events when developing rainfall thresholds. This research contributes to refining early warning systems by incorporating detailed, real-time monitoring data to better predict and mitigate landslide risks. This approach supports the development of a cost-effective early warning system for slope instability, addressing the challenge of identifying which slope sections are at risk of failure during heavy rainfall (Dikshit et al. 2018b; Abraham et al. 2020b).
Moreover, we explored the integration of real-time monitoring data from MicroElectroMechanical Systems (MEMS) tilt sensors with the SIGMA model to improve its predictive capabilities. The SIGMA model, which uses statistical distributions of rainfall for forecasting landslide occurrences, was enhanced by combining it with tilt meter readings through a decisional algorithm. We compared three approaches: the SIGMA model alone, tilt meter readings alone, and the combination of both, using precipitation and landslide data from Kalimpong between July 2017 and September 2020. The integration of tilt meter readings reduced the number of false alarms issued by the SIGMA model from 70 to 38 and increased the likelihood ratio from 18.10 to 20.23.
The Receiver Operating Characteristic (ROC) curves indicated that the combined approach provided the best performance, with an area under the curve of 0.976. This proposed method outperformed previous rainfall thresholds for the Kalimpong region and holds promise for further refinement into an operational Landslide Early Warning System (LEWS) for the area (Abraham et al. 2022a).
Further enhancing our research, we employed Bayesian analysis to assess the probability of landslide occurrences based on rainfall severity and antecedent soil moisture content. Using the SHETRAN (Système Hydrologique Européen Transport) model for simulating soil moisture and applying event rainfall-duration (ERD) thresholds, we derived two-dimensional Bayesian probabilities for landslides in Kalimpong. This approach highlighted the applicability of the SHETRAN model in improving the prediction capabilities of empirical thresholds and provided a more nuanced understanding of landslide risk (Abraham et al. 2020c).
Lastly, we addressed the limitations of deterministic rainfall thresholds by evaluating probabilistic thresholds using statistical methods. By analyzing single and multiple rainfall parameters through variants of Bayes theorem, we calculated the probabilities of landslide occurrences for Kalimpong. Our study found that the probability of landslide initiation increases with higher rainfall intensity, underscoring the importance of considering rainfall event parameters, particularly intensity, in landslide prediction (Dikshit et al. 2018a).
Western Ghats
In the Western Ghats, our research was focused on forecasting the occurrence of rainfall-induced landslides, a critical issue due to the extensive damage landslides cause worldwide. Our research aimed to contribute to the development of operational landslide early warning systems (LEWS) by providing detailed spatial and temporal forecasts. The forecasting method was tailored to the unique topographical, hydrological, and meteorological conditions of our study areas, with rainfall identified as the primary trigger for landslides.
To achieve temporal forecasting, we derived rainfall thresholds using multiple approaches. These thresholds were based on historical data that linked rainfall to landslide occurrences, with the aim of predicting future landslides. Initially, we derived Intensity-Duration thresholds using historical relationships between landslides and rainfall. An algorithm-based approach, the Calculation of Thresholds for Rainfall-induced Landslides Tool (CTRL-T), was employed to define event- duration (ED) thresholds (Abraham et al. 2019).
Additionally, we developed probabilistic thresholds using a Bayesian approach, which involved calculating the posterior probability of landslide occurrence based on marginal and conditional probabilities of control parameters, along with the prior probability of occurrence. We also applied the SIGMA model to create thresholds that considered both long-term and short-term rainfall, tailored for subdivisions within the study area (Abraham et al. 2021e).
After comparing these thresholds quantitatively, we found that probabilistic thresholds incorporating both rainfall severity and antecedent soil wetness performed better than other models across both districts in the study area (Abraham et al. 2021d). These findings open new avenues for developing an operational LEWS in the region, particularly by integrating rainfall and soil moisture data. Moreover, this study provides valuable insights from a monsoon region, highlighting the effectiveness of hydro-meteorological thresholds based on soil moisture—a relatively unexplored area in LEWS development for the study area (Abraham et al. 2020d).
The next stage of our research involved identifying locations susceptible to landslides. We developed Landslide Susceptibility Maps (LSMs) using both data-driven and process-based approaches. Advanced machine learning algorithms played a crucial role, with five different algorithms employed to create LSMs for our study areas. Physically based models were also developed to understand the hydrological mechanisms and assess terrain stability, primarily through factor of safety values (Abraham et al. 2021b). These models are valuable for both spatial and temporal forecasting, as they use precipitation and spatial properties as inputs.
However, unlike physically-based models, data-driven approaches only provide the spatial probability of landslide occurrence, which requires integrating LSMs with rainfall forecasting models for dynamic forecasting—a complex process. For spatial forecasting, the digital elevation model (DEM) of the study area served as the primary input for both data-driven and physically based approaches. We used two DEMs: one from Alos Palsar with a 12.5 m resolution and another from the National Remote Sensing Centre with a 1 arc second resolution (30.4 m). The DEM quality critically affected the output in both methods. Our results indicated that the Random Forest (RF) model performed better for both districts when using the 12.5 m DEM, with effectiveness values of 0.81 and 0.83 for Idukki and Wayanad, respectively. The resolution of the DEM was found to be crucial in the RF model's performance, with finer resolution data leading to better results. The study also demonstrated that process-based approaches' performances were comparable to data- driven models and could be effectively applied for regional-scale forecasting with more precise data collection and fine property zoning.
When comparing the mutual agreement between the models, we observed that the data-driven model satisfactorily classified the green and red alert classes, which were similarly categorized by process-based models. However, there was more disagreement in the orange and yellow alerts, which is significant as orange alerts are considered positive predictions and yellow alerts negative (Abraham et al. 2023b).
The study proposes a spatio-temporal framework for landslide forecasting in the study areas, which can be a crucial input for developing an LEWS in the region. The developed maps can be refined using local administrative boundaries, such as panchayats and municipalities. Idukki district, with 54 local self-governments, and Wayanad district, with 26 local self-governments, can issue alerts based on the pixel-wise distribution of each alert within a local body.
The proposed framework involves integrating rainfall forecasts, antecedent soil moisture data, and the LSM for landslide forecasting, requiring a clear understanding of threshold conditions, expert decision-making teams, and detailed communication strategies and response plans. These steps are vital for developing an early warning system that can significantly aid regional planning and development activities to minimize the risk of landslides (Abraham et al. 2021a).
Our study in the Idukki district tackled the challenge of predicting landslides, particularly in ecologically sensitive areas of the Western Ghats. We developed a forecasting model using mobility functions to assess the likelihood of landslides based on critical values. The model proposes two warning levels tailored for different areas in the district. It proved highly effective, with a 97% accuracy in smaller, uniform areas, although performance decreases in larger, more diverse regions. This model can be integrated with rainfall forecasting systems to enhance early warning capabilities for landslide risks (Abraham et al. 2024).
In addition to identifying critical locations, we conducted detailed studies on debris flows, calibrating rheological parameters for the region and assessing the potential for future failures. At a debris flow location in Idukki, data collection was limited to lower elevation zones due to site constraints. Similar analysis was carried out for four debris flow sites in Wayanad district (Abraham et al. 2023a). The subsurface was divided into three distinct layers: a topsoil layer of loose debris with boulders and fragmented rocks, a sand-silt-clay matrix layer, and the bedrock beneath the overburden. Soil samples were tested for engineering properties, revealing that coarse-grained particles were more prone to erosion, while fine-grain content increased with depth. The debris volume at the sites varied, with significant amounts at sites 1 and 2, indicating a higher risk of future debris flows.
We employed confidence ellipse analysis to quantify the similarity between sites, finding an overall similarity greater than 0.5. The calibrated friction parameters from one site satisfactorily predicted the shape of other debris flows, indicating that these parameters can be used for regional debris flow simulation (Abraham et al. 2021c).
Given the complexities associated with existing numerical models, we developed an easy-to-use simulation tool, Debris Flow Simulation 2D (DFS 2D), which considers multiple rheologies and bed entrainment. DFS 2D simplifies the modeling process by using shallow water equations and provides information on each input parameter through a user-friendly interface. The model was evaluated using a case study from Yindongzi gully in Sichuan, China, where it provided satisfactory results in simulating debris flows, despite the challenges posed by coarse-resolution DEMs. The study concludes by highlighting the multiple aspects of landslide forecasting addressed for Wayanad and Idukki districts in the Western Ghats of India.
Our research developed a spatio-temporal landslide forecasting framework, evaluated rheological parameters for debris flow simulation, and created a user-friendly numerical model for debris flow modeling, all of which can significantly aid in hazard assessment and regional planning (Abraham et al. 2022c).
Our research was focused on developing advanced data-driven approaches to assess landslide hazards in response to the increasing risks posed by urbanization in hilly regions. To address this, we employed a random forest algorithm to estimate critical landslide parameters—such as projected area, length, travel distance, and width—using elevation and slope information. This approach was tested in Idukki and Wayanad, two Landslide-prone areas, with three different combinations of input features, including elevation, tangential slope, drop height, angle of reach, and profile curvature.
Out of 144 models evaluated, our method showed significant improvements in accuracy. For instance, in Wayanad, the RMSE value for estimating the length of flow-like landslides was reduced from 472.74 m to 204.64 m when using elevation and tangential slope as inputs. Only a few cases showed minor increases in error values, highlighting the model's robustness. To make our findings accessible, we developed an easy-to-use tool based on pre-trained models, allowing untrained personnel to perform preliminary hazard assessments. This tool simplifies complex statistical methods, offering a practical solution for disaster management and risk assessment in landslide-prone areas.
Our work represents a significant step forward in applying machine learning to natural hazard management, providing effective tools for enhancing landslide risk mitigation strategies in rapidly urbanizing hilly regions (Abraham et al. 2022b).
Our research also extends to the critical analysis of land use and land cover (LULC) changes, which are essential for regional planning and disaster risk reduction, particularly in rapidly urbanizing hilly regions. Recognizing the challenges posed by population growth and urbanization in landslide-prone areas, we developed a novel tool using a random forest classifier that automatically generates LULC classification maps from natural color satellite imagery without requiring any training input from the end user.
This innovative approach demonstrated an overall accuracy of 0.75 and an AUC score of 0.95, making it a reliable tool for mapping built-up area expansion in regions susceptible to rainfall-induced landslides. We applied this framework to the Idukki block panchayat in Kerala, India, comparing LULC data from 2012 and 2022.
Our findings revealed a significant increase in built-up areas, from 12.76% of the total area in 2012 to 26.48% in 2022. This rapid expansion occurred predominantly in zones classified as 'very high' in landslide susceptibility. These results highlight the urgent need for hazard-inclusive planning and continuous monitoring of LULC changes to effectively mitigate disaster risks in such vulnerable regions (Sunil et al. 2024).
Western Himalayas
In the Western Himalayas, we focused on addressing significant gaps in landslide hazard management, particularly within the highly vulnerable Uttarakhand state. Landslides are a critical global hazard, with immense potential to damage infrastructure, disrupt communities, and cause environmental degradation. They can be triggered by various factors, including rainfall, seismic activity, and human intervention, necessitating a deep understanding of the underlying influences such as slope steepness, geological and hydrological characteristics, soil properties, and land use practices. These factors are essential in assessing and mitigating landslide risks.
Our research began with an extensive data collection phase, focused on Uttarakhand and the Chamoli district, utilizing field visits, laboratory analyses, subsurface explorations, and collaborations with governmental agencies. This comprehensive approach aimed to capture the region's geological, geomorphological, hydrological, and seismotectonic characteristics, which are crucial for understanding landslide susceptibility and hazards in Chamoli. We conducted a thorough examination of the district's geological formations, geomorphological features, hydrological components, seismotectonic activity, historical landslide data, and local meteorological parameters.
This detailed regional analysis provided nuanced insights into the specific geological characteristics of Chamoli and identified the rock types and structures influencing stability and susceptibility to landslides.
In the subsequent phase, we conducted a detailed seismic landslide hazard assessment across Uttarakhand, employing both Probabilistic Seismic Hazard Assessment (PSHA) and Scenario- Based Seismic Hazard Assessment methodologies (Gupta and Satyam 2022a, 2023). These approaches provided valuable insights into the spatial distribution of seismic hazards, highlighting the areas which are most susceptible to landslides.
We further optimized our methodology by evaluating the performance of various models, including the Conventional and Modified Newmark's Models, using metrics like the Area Under the Receiver Operating Characteristic Curve (AUC-ROC). Our findings revealed that the Modified Newmark's model, demonstrated superior predictive capability for seismic-induced landslides in the region (Gupta and Satyam 2022b; Gupta et al. 2023c, b).
Building on this, we refined the model through parametric optimization, incorporating Monte Carlo simulations to address uncertainties in geotechnical and geological parameters. We also considered multiple seismic inputs, including those derived from PSHA, to ensure a comprehensive seismic hazard characterization. The optimized methodology demonstrated high predictive accuracy, particularly with PSHA-based inputs, making it a reliable framework for assessing seismic landslide hazards in Uttarakhand (Gupta and Satyam 2024a).
Recognizing that Chamoli district would be one of the most severely affected areas, we extended our study to focus on landslide susceptibility mapping (LSM) and the development of a dynamic early warning system (EWS) for the region. We employed five distinct machine learning models, rigorously evaluating their performance using AUC-ROC, Accuracy, Precision, Recall, and F1- Score. The Random Forest (RF) model emerged as the most effective, achieving an AUC of 0.9704 and an accuracy of 92.86%.
Further analysis using classified data from the National Landslide Susceptibility Mapping (NLSM) database highlighted the importance of leveraging comprehensive datasets for more accurate susceptibility assessments (Mittal et al. 2024).
The study also focused on deriving rainfall thresholds tailored specifically for Chamoli, comparing empirical, probabilistic, and SIGMA thresholds to determine the most effective for landslide prediction and early warning. We developed a slope unit-based dynamic early warning system, incorporating the best-performing machine learning model and optimal rainfall thresholds. This system continuously assesses slope unit susceptibility based on static maps and real-time rainfall data, generating warnings that are tailored to prevailing conditions.
Additionally, we developed a site-specific rainfall-triggered landslide EWS for the Joshimath- Badrinath Highway, utilizing an IoT-based framework that accounted for the hydrological dynamics of unsaturated slopes. Our findings underscored the importance of calibrated hydrogeological models, particularly in reflecting the complex interactions between climate, vegetation, and slope stability. The integration of machine learning techniques further enhanced the system's ability to assess slope stability in real-time, contributing to more effective risk mitigation (Gupta and Satyam 2024b).
In addition to this our research also tackled the challenge of accurately determining soil thickness in large, heterogeneous areas—an essential yet difficult parameter in environmental modeling. The study focused on three key roads in the Joshimath region of the Indian Himalayas, employing three different methods to create soil thickness maps: a customized version of the conventional geomorphologically indexed soil thickness (GIST) model, the GIST model enhanced with Monte Carlo simulations (GIST-MCS), and the random forest algorithm integrated with the GIST model (GIST-RF).
By assessing the errors and validating the results with geophysical tests, the study found that while the standard GIST model struggled to account for the area's spatial variations, the GIST-MCS model showed improvement, and the GIST-RF model demonstrated the best performance, yielding the most accurate soil thickness maps. These maps are crucial for future geotechnical assessments and environmental modeling (Gupta et al. 2024).
In the Joshimath region, we also examined the impact of a rockslide-triggered debris flow that struck the Rishiganga-Dhauliganga valley in Chamoli district, Uttarakhand, on February 7, 2021. To investigate this event, we utilized a dense seismic monitoring network in the region and applied advanced signal processing techniques—including band-pass filtering, Ensemble Empirical Mode Decomposition (EEMD), Short-Time Fourier Transform (STFT), and Power Spectrum Density (PSD). By analyzing seismic signals from nine stations and integrating these with satellite imagery, we aimed to improve the understanding of the hazard chain and enhance landslide detection systems (Gupta et al. 2023a).
The research conducted on debris flow dynamics has focused extensively on understanding the entrainment and deposition processes and the rheological behavior of debris materials. Keeping the Western Himalayas as the study area, we have modeled some important debris flows along Himachal Pradesh and Uttarakhand. At the same time, we have experimentally validated our findings using a fabricated flume setup at IIT Indore.
This study provides a comprehensive analysis of debris flow dynamics through a combination of flume-based experiments and advanced numerical modeling, offering valuable insights into how debris flows interact with erodible beds and replicate real-world scenarios. The flume experiments revealed that the entrainment process typically begins with the arrival of the first surge body, which exhibits the highest scouring potential. This initial surge led to significant gully formations, particularly between 0.1 m and 0.3 m from the start of the erodible bed, mimicking natural debris flows observed in the field. The deposition process, especially at the tail end of the flow, was also observed to closely correspond with the behavior of real debris flows, thereby validating the experimental setup and the results obtained.
The entrainment volume varied significantly based on the water content of the debris flow material, with peak volumes recorded at higher water content levels. This variation underscores the complex interactions between flow dynamics and bed composition, providing a deeper understanding of the scouring and deposition mechanisms in debris flows. The experimental findings highlight the critical role of water content in determining the extent of entrainment and its subsequent impact on deposition patterns. These insights are crucial for predicting debris flow behavior in different environmental conditions and for designing effective mitigation strategies.
Building on these experimental and computational findings, the study employed the r.avaflow simulation tool to model the Kotrupi debris flow event in Himachal Pradesh, India. This numerical model accurately reproduced key characteristics of the event, including a peak velocity of 18.7 m/s and a maximum flow height of 7.34 m. The model's ability to closely match observed runout lengths and deposition patterns further validated its use as a reliable tool for simulating real-world debris flows. The integration of experimental data with numerical simulations provided a robust framework for understanding debris flow behavior, highlighting the importance of combining multiple methodologies to capture the complexities of such natural hazards (Pandey et al 2024). The findings provide critical insights that can inform the development of more effective hazard prediction and mitigation strategies in regions prone to debris flows, ensuring better preparedness and response to these destructive natural events.
References
- Abraham MT, Pothuraju D, Satyam N (2019) Rainfall Thresholds for Prediction of Landslides in Idukki, India: An Empirical Approach. Water 11:2113. https://doi.org/10.3390/w11102113
- Abraham MT, Satyam N, Jain P, et al (2021a) Effect of spatial resolution and data splitting on landslide susceptibility mapping using different machine learning algorithms. Geomatics, Nat Hazards Risk 12:3381–3408. https://doi.org/10.1080/19475705.2021.2011791
- Abraham MT, Satyam N, Kowalski J (2024) Numerical Modelling of Debris Flows for Simulation- Based Decision Support: An Indian Perspective. Indian Geotech J. https://doi.org/10.1007/s40098-024-00988-5
- Abraham MT, Satyam N, Kushal S, et al (2020a) Rainfall Threshold Estimation and Landslide Forecasting for Kalimpong, India Using SIGMA Model. Water 12:1195. https://doi.org/10.3390/w12041195
- Abraham MT, Satyam N, Lokesh R, et al (2021b) Factors Affecting Landslide Susceptibility Mapping: Assessing the Influence of Different Machine Learning Approaches, Sampling Strategies and Data Splitting. Land 10:989. https://doi.org/10.3390/land10090989
- Abraham MT, Satyam N, Pradhan B, et al (2022a) Developing a prototype landslide early warning system for Darjeeling Himalayas using SIGMA model and real-time field monitoring. Geosci
- J 26:289–301. https://doi.org/10.1007/s12303-021-0026-2
- Abraham MT, Satyam N, Pradhan B (2023a) A novel approach for quantifying similarities between different debris flow sites using field investigations and numerical modelling. Terra Nov. https://doi.org/10.1111/ter.12679
- Abraham MT, Satyam N, Pradhan B, Alamri AM (2020b) IoT-Based Geotechnical Monitoring of Unstable Slopes for Landslide Early Warning in the Darjeeling Himalayas. Sensors 20:2611. https://doi.org/10.3390/s20092611
- Abraham MT, Satyam N, Pradhan B, Alamri AM (2020c) Forecasting of Landslides Using Rainfall Severity and Soil Wetness: A Probabilistic Approach for Darjeeling Himalayas. Water 12:804. https://doi.org/10.3390/w12030804
- Abraham MT, Satyam N, Pradhan B, Segoni S (2022b) Proposing an easy-to-use tool for estimating landslide dimensions using a data-driven approach. All Earth 34:243–258. https://doi.org/10.1080/27669645.2022.2127549
- Abraham MT, Satyam N, Pradhan B, Tian H (2022c) Debris flow simulation 2D (DFS 2D): Numerical modelling of debris flows and calibration of friction parameters. J Rock Mech Geotech Eng 14:1747–1760. https://doi.org/10.1016/j.jrmge.2022.01.004
- Abraham MT, Satyam N, Reddy SKP, Pradhan B (2021c) Runout modeling and calibration of friction parameters of Kurichermala debris flow, India. Landslides 18:737–754. https://doi.org/10.1007/s10346-020-01540-1
- Abraham MT, Satyam N, Rosi A, et al (2021d) Usage of antecedent soil moisture for improving the performance of rainfall thresholds for landslide early warning. CATENA 200:105147. https://doi.org/10.1016/j.catena.2021.105147
- Abraham MT, Satyam N, Rosi A, et al (2020d) The Selection of Rain Gauges and Rainfall Parameters in Estimating Intensity-Duration Thresholds for Landslide Occurrence: Case Study from Wayanad (India). Water 12:1000. https://doi.org/10.3390/w12041000
- Abraham MT, Satyam N, Shreyas N, et al (2021e) Forecasting landslides using SIGMA model: a case study from Idukki, India. Geomatics, Nat Hazards Risk 12:540–559. https://doi.org/10.1080/19475705.2021.1884610
- Abraham MT, Vaddapally M, Satyam N, Pradhan B (2023b) Spatio-temporal landslide forecasting using process-based and data-driven approaches: A case study from Western Ghats, India. CATENA 223:106948. https://doi.org/10.1016/j.catena.2023.106948
- Dikshit A, Sarkar R, Satyam N (2018a) Probabilistic approach toward Darjeeling Himalayas landslides-A case study. Cogent Eng5:1537539. https://doi.org/10.1080/23311916.2018.1537539
- Dikshit A, Satyam DN (2018) Estimation of rainfall thresholds for landslide occurrences in Kalimpong, India. Innov Infrastruct Solut 3:24. https://doi.org/10.1007/s41062-018-0132-9
- Dikshit A, Satyam DN, Towhata I (2018b) Early warning system using tilt sensors in Chibo, Kalimpong, Darjeeling Himalayas, India. Nat Hazards 94:727–741. https://doi.org/10.1007/s11069-018-3417-6
- Dikshit A, Satyam N, Pradhan B, Kushal S (2020) Estimating rainfall threshold and temporal probability for landslide occurrences in Darjeeling Himalayas. Geosci J 24:225–233. https://doi.org/10.1007/s12303-020-0001-3
- Gupta K, Abraham MT, Satyam N (2023a) Characterization and interpretation of seismic signals generated by 7th February 2021, Dhauliganga disaster (India). J Appl Geophys 213:105018. https://doi.org/10.1016/j.jappgeo.2023.105018
- Gupta K, Satyam N (2023) Probabilistic Arias Intensity Maps of Uttarakhand State (India). pp 1– 15
- Gupta K, Satyam N (2022a) Estimation of Arias intensity and peak ground acceleration (PGA) using probabilistic seismic hazard assessment of Uttarakhand state (India). Arab J Geosci 15:437. https://doi.org/10.1007/s12517-022-09733-9
- Gupta K, Satyam N (2022b) Co-seismic landslide hazard assessment of Uttarakhand state (India) based on the modified Newmark model. J Asian Earth Sci X 8:100120. https://doi.org/10.1016/j.jaesx.2022.100120
- Gupta K, Satyam N (2024a) Optimizing seismic hazard inputs for co-seismic landslide susceptibility mapping: a probabilistic analysis. Nat Hazards 120:8459–8481. https://doi.org/10.1007/s11069-024-06517-0
- Gupta K, Satyam N (2024b) Integrating real-time sensor data for improved hydrogeotechnical modelling in landslide early warning in Western Himalaya. Eng Geol 338:107630. https://doi.org/10.1016/j.enggeo.2024.107630
- Gupta K, Satyam N, Gupta V (2023b) Probabilistic assessment of seismically triggered landslide hazard for Uttarakhand (India) in the Western Himalayas. Nat Hazards 118:669–689. https://doi.org/10.1007/s11069-023-06023-9
- Gupta K, Satyam N, Gupta V (2023c) Probabilistic physical modelling and prediction of regional seismic landslide hazard in Uttarakhand state (India). Landslides. https://doi.org/10.1007/s10346-022-02013-3
- Gupta K, Satyam N, Segoni S (2024) A comparative study of empirical and machine learning approaches for soil thickness mapping in the Joshimath region (India). CATENA 241:108024. https://doi.org/10.1016/j.catena.2024.108024
- Mittal A, Gupta K, Satyam N (2024) Landslide Susceptibility Mapping of Chamoli (Uttarakhand) Using Random Forest Machine Learning Method. pp 207–217
- Pandey NK, Satyam N, Gupta K (2024) Landslide-induced debris flows and its investigation using r.avaflow: A case study from Kotrupi, India. J Earth Syst Sci 133:97. https://doi.org/10.1007/s12040-024-02315-1
- Sunil LS, Abraham MT, Satyam N (2024) Mapping built-up area expansion in landslide susceptible zones using automatic land use/land cover classification. J Earth Syst Sci 133:132. https://doi.org/10.1007/s12040-024-02345-9
For more details, please contact at:
