Page 2251

www.rsisinternational.org

INTERNATIONAL JOURNAL OF LATEST TECHNOLOGY IN ENGINEERING,

MANAGEMENT & APPLIED SCIENCE (IJLTEMAS)

ISSN 2278-2540 | DOI: 10.51583/IJLTEMAS | Volume XV, Issue VI, June 2026

Geological Hazard Prediction and Prevention: A Review of

Mechanisms, Monitoring, and Mitigation

Jeevana Sasindu Wickramaarachchige

, Daim Safeer Mughal

School of Environment and Civil Engineering, Chengdu University of Technology, Chengdu, China

School of Management and Economics, Chongqing University of Post and Telecommunications,

Chongqing, China

*Corresponding Author

DOI:

https://doi.org/10.51583/IJLTEMAS.2026.150600164

Received: 28 June 2026; Accepted: 03 July 2026; Published: 18 July 2026

ABSTRACT

As industrialization expands, modern infrastructure increasingly collides with volatile geological environments,

driving a sharp escalation of complex, cascading geohazards. These destructive hazard chains represent a critical

global threat to structural and economic resilience. Despite this escalating risk, current early warning systems

predominantly rely on fragmented, static methodologies that fail to capture the dynamic reality of temporal

hazard evolution. Furthermore, contemporary predictive models are bifurcated between deterministic physical

models demanding exhaustive geotechnical parameters and data-driven artificial intelligence algorithms

functioning as opaque black boxes devoid of physical interpretability. To resolve these limitations, this review

systematically evaluates the modern geohazard landscape through a coupled active-passive conceptual model.

This framework systematically evaluates how primary high-energy failures trigger subsequent instability in

surrounding geomaterials. The synthesis reveals that integrating space-air-ground multi-scale monitoring

including orbital InSAR, UAV photogrammetry, and distributed ground sensors establishes a vital surveillance

continuum for early hazard identification. Because these heterogeneous data streams possess significant

environmental noise, rigorous multi-source data fusion remains essential to minimize false alarms and resolve

spatial discontinuities. Analytically, physically based models provide indispensable mechanical transparency by

explicitly simulating material deformation, whereas data-driven ensemble algorithms excel at processing high-

dimensional, nonlinear geospatial inputs. Ultimately, transitioning toward proactive, real-time risk reduction

demands the integration of physics-informed neural networks (PINNs) that embed geomechanical constraints

into computational pipelines. Coupling these hybrid architectures with interactive digital twin technologies will

transform static hazard mapping into dynamic virtual environments, empowering engineers to successfully

mitigate evolving disaster chains globally.

Keywords: Geological hazards; Disaster risk reduction; Real-time monitoring; Data-driven forecasting; Early

warning systems; Physics-informed neural networks (PINNs).

INTRODUCTION

As industrialization and urban expansion push deeper into challenging terrains, the global footprint of

infrastructure is colliding head-on with increasingly volatile geological environments. We are witnessing a sharp

escalation in complex geohazards, ranging from earthquake-induced rockfalls along mountain corridors to

catastrophic rockbursts in deep underground excavations. Consider the tragic events at the phosphate mine in

Fuquan City, China, back in 2014. A massive rockfall and landslide plummeted into a deep pond, triggering a

displacement wave of 50,000 cubic meters that resulted in 23 fatalities and extensive environmental devastation.

Such disasters rarely operate in isolation. They frequently evolve into intricate geological-environmental hazard

chains, where the initial mechanical failure triggers cascading secondary impacts like flooding or widespread

soil and water contamination

[1]

. Confronting these cascading risks is no longer just a localized engineering

Page 2252

www.rsisinternational.org

INTERNATIONAL JOURNAL OF LATEST TECHNOLOGY IN ENGINEERING,

MANAGEMENT & APPLIED SCIENCE (IJLTEMAS)

ISSN 2278-2540 | DOI: 10.51583/IJLTEMAS | Volume XV, Issue VI, June 2026

challenge; it is a critical global imperative for safeguarding human life, ecological stability, and economic

resilience.

In an ideal scenario, hazard management would rely on dynamic, real-time early warning systems capable of

seamlessly integrating subsurface mechanics with continuous monitoring data to predict the precise time,

location, and magnitude of impending failures. Infrastructure planners would have access to high-fidelity,

spatiotemporal forecasting that proactively identifies risk well before a steep slope gives way or a tunnel

collapses. Yet, the reality on the ground falls drastically short of this vision. Current early warning systems and

risk assessments predominantly rely on fragmented, reactive, or highly static methodologies. For example,

traditional regional susceptibility maps typically reflect the spatial probability of failure under a specific seismic

or rainfall scenario, but they utterly fail to account for temporal thresholds, progressive rock weakening, or site-

specific energy accumulation

[2]

. We are mapping where disasters might happen based on static variables, but we

are consistently missing the dynamic reality of exactly when and how they will unfold.

Previous attempts to solve this forecasting dilemma have largely bifurcated into two distinct camps: physical

process-based models and data-driven algorithms. Physical approaches, such as the widely utilized Newmark

displacement model, offer excellent mechanical interpretability by calculating critical acceleration and

permanent displacement to assess slope stability

[3,4]

. Unfortunately, these deterministic models demand precise,

comprehensive geotechnical parameters such as transient pore water pressure and shear strength, which are

notoriously difficult and expensive to acquire across broad, heterogeneous regional scales

[3,5]

. Furthermore, they

often rely on simplified assumptions, treating sliding masses as rigid bodies while neglecting complex internal

deformations and dynamic topographic amplification effects

[6]

. Conversely, the recent surge in deep learning and

machine learning models provides a highly scalable alternative capable of processing vast amounts of remote

sensing data

[3]

. Even so, these artificial intelligence approaches function predominantly as black boxes. They

learn statistical correlations rather than physical mechanisms, which makes them highly susceptible to domain

shifts, meaning a model trained on one geological setting frequently fails when applied to another with different

lithology or climate conditions

[7]

. Moreover, optimizing these algorithms solely to maximize statistical hit rates

often generates unacceptable levels of false alarms, ultimately undermining the social credibility of the early

warning systems they are meant to support

[6]

The direct and indirect consequences of relying on these inadequate, decoupled systems are severe. Directly,

unforeseen landslides and deep-tunnel rockbursts lead to catastrophic loss of life and the sudden destruction of

critical transportation networks and energy grids

[8,9]

. Indirectly, the fallout is even more insidious. When a major

transportation artery is severed by a rockfall, the ensuing highway blockages paralyze regional trade, disrupt

emergency response efforts, and trigger prolonged economic stagnation in communities that lack viable alternate

routes

[10]

. Beyond the economic toll, misclassifying or entirely missing the precursory signals of a major slope

failure can lead to environmental disasters that persist for decades, as seen when mining slopes collapse and

release toxic materials into local watersheds

[1,11]

What remains glaringly absent from the current literature is a cohesive framework that bridges the gap between

macroscopic data-driven forecasting and microscopic, physics-based failure mechanisms. For instance, while

recent work has successfully integrated multi-geometry InSAR with explainable machine learning to map

potential landslide areas in reservoir regions, these frameworks still treat susceptibility as a relatively static

baseline rather than a time-resolved, dynamic state

[12]

. Similarly, advancements in microseismic monitoring have

revolutionized our ability to track deep rock microfractures in real-time

[6]

, yet integrating these high-frequency,

localized signals into broader regional vulnerability models remains a significant hurdle

[13]

. This study seeks to

fill that exact void by examining the hazard landscape through the lens of a coupled active-passive conceptual

model. In this framework, primary high-energy brittle failures (the active source) trigger the subsequent

instability of surrounding fractured rock masses (the passive response), a dynamic particularly evident in

compound rockburst-collapse events

[14]

. By adopting this theoretical lens, we can better decode how localized

stress concentrations dynamically interact with external triggers like extreme rainfall or seismic shocks.

Accordingly, the primary objective of this review is to systematically evaluate the underlying mechanisms of

complex geological hazards, critically assess the efficacy and limitations of emerging multi-source monitoring

Page 2253

www.rsisinternational.org

INTERNATIONAL JOURNAL OF LATEST TECHNOLOGY IN ENGINEERING,

MANAGEMENT & APPLIED SCIENCE (IJLTEMAS)

ISSN 2278-2540 | DOI: 10.51583/IJLTEMAS | Volume XV, Issue VI, June 2026

technologies, and test the theoretical boundaries of both physical and AI-driven forecasting models. Specifically,

we aim to investigate how the integration of ground-based sensor networks, such as distributed fiber optic

sensing, with intelligent predictive algorithms can successfully transition hazard management from reactive post-

disaster analysis to proactive, real-time early warning

[15]

. In practical terms, this research is vital for engineers

and policymakers who require robust, uncertainty-aware decision-support tools that function reliably under the

chaotic, non-stationary conditions of real-world environments

[6]

. Academically, it establishes a necessary

dialogue between geomechanics and artificial intelligence, charting a realistic course toward physically-

informed foundation models.

Ultimately, managing geological hazards is paramount for ensuring the resilience and sustainability of our

expanding global infrastructure. Despite an abundance of sophisticated monitoring tools and advanced

computational algorithms, current research remains highly fragmented, leaving a critical gap in our ability to

synthesize real-time dynamic monitoring with transparent, physically grounded predictive models. This paper

occupies that exact niche by offering a comprehensive review. The subsequent sections are logically organized

to unravel this complexity: first, we dissect the intricate failure mechanisms driving modern geohazards; next,

evaluate the state-of-the-art in remote and ground-based monitoring systems; finally, critically analyze current

forecasting models, outlining a strategic roadmap for the future of intelligent, multi-hazard disaster risk reduction.

Formation Mechanisms of Geological Hazards

Internal Geological Factors

The geometric boundary conditions of a landscape establish the fundamental gravitational framework for slope

instability. Terrain steepness directly dictates the magnitude of shear stress acting parallel to potential failure

surfaces. This means steeper gradients inherently elevate the probability of mass displacement

[8]

. This geometric

influence is further modulated by slope aspect. Aspect controls local microclimates by determining solar

radiation and moisture retention

[16,17]

. Such variations alter vegetation density and chemical weathering rates,

which indirectly shape the mechanical resistance of surficial materials

[18]

. Beyond simple inclination and

orientation, hillside curvature determines the pathways of surface runoff and the accumulation of loose

colluvium

[8,17]

. Concave topographic hollows serve as natural convergence zones for groundwater. They

facilitate rapid increases in pore pressure during precipitation events that frequently initiate shallow

landsliding

[16,19]

. Beneath this topographic surface, regional tectonic activity deeply compromises terrain

stability through the formation of complex fault zones

[20,21]

. Major geological faults and their associated damage

corridors shatter intact bedrock. They produce thick bands of low-strength fault gouge and highly permeable

fragmented materials

[12,14,18]

. These structural discontinuities not only reduce the overall compressive and shear

capabilities of the rock mass, but they also create preferential infiltration channels that accelerate deep fluid

migration

[22,23]

. When slopes are situated within these tectonically active corridors, the combination of

mechanically degraded lithology and heightened groundwater accumulation dramatically increases the

propensity for large-scale failure

[18,20]

. Operating at a more localized scale, intrinsic rock mass properties and

secondary discontinuities define the specific kinematic mechanisms of failure

[18]

. The stability of a bedrock slope

is largely governed by its lithological composition alongside the orientation of its internal structural planes, such

as bedding, foliation, and joints

[24,25]

. Intact, massive rocks like granodiorite typically resist weathering and

exhibit high geomechanical strength

[25]

. In contrast, heterogeneous formations containing alternating hard and

soft strata weather unevenly. Interbedded sandstone and mudstone, for instance, rapidly form discrete, low-

strength slip surfaces

[5,24,26]

. The spatial alignment of these planes relative to the slope face determines the

physical mode of collapse

[18]

. Planar sliding readily occurs when discontinuities strike parallel to the slope and

dip at angles exceeding the internal friction angle but less than the slope face. Alternatively, intersecting joint

sets can isolate individual tetrahedral wedges that detach under gravitational pull. Steeply dipping vertical

fractures often promote toppling mechanisms instead

[25]

. Over geological time, intensive weathering degrades

these rock masses into compressible residual soils. This shifts the governing failure mechanics from structural

detachment to hydro-mechanical soil yielding. Within these soil mantles, stability depends entirely on a delicate

physical balance. The shear strength, derived from cohesion and the effective angle of internal friction, must

adequately resist the downslope driving forces

[2,27]

. As meteoric water infiltrates the porous matrix, it steadily

Page 2254

www.rsisinternational.org

INTERNATIONAL JOURNAL OF LATEST TECHNOLOGY IN ENGINEERING,

MANAGEMENT & APPLIED SCIENCE (IJLTEMAS)

ISSN 2278-2540 | DOI: 10.51583/IJLTEMAS | Volume XV, Issue VI, June 2026

dissipates matric suction and elevates positive pore water pressure. This fundamentally diminishes the effective

stress holding the soil particles together

[19,28,29]

. For highly compressible residual soils, rapid shearing can induce

localized pore pressure spikes and severe strain softening, ultimately driving the material into a state of residual

strength

[30]

. Once this localized yielding initiates near the over-steepened toe of a slope, the loss of lateral support

transfers stress upslope. A retrogressive cascade of progressive soil failure then inevitably follows

[31,32]

External Triggering Factors

Geohazards rarely initiate without external forcing mechanisms that disrupt the mechanical equilibrium of slopes.

Precipitation acts as the most ubiquitous catalyst for mass wasting globally

[31]

. When meteoric water infiltrates

the subsurface, it progressively fills pore spaces, which simultaneously dissipates matric suction and elevates

positive pore water pressure

[33–35]

. This hydrodynamic alteration fundamentally diminishes the effective stress

and shear resistance of the geomaterials while adding significant gravitational mass to the potential sliding

body

[18,33,36]

. The temporal distribution of this precipitation dictates the specific failure mechanics. Short-

duration, high-intensity rainstorms typically induce rapid shallow landsliding, whereas prolonged antecedent

moisture accumulation governs the activation of deep-seated structural failures

[18]

. In high-altitude or frigid

mountain catchments, these hydrological effects are further compounded by seasonal snowmelt and thermal

fluctuations

[18,37]

. Rapid spring warming accelerates cryospheric melt, delivering substantial fluid pulses that

infiltrate fractured bedrock and residual soils

[20,37]

. Antecedent freeze-thaw cycles physically degrade the rock

mass by expanding existing fissures during freezing periods

[20,38]

. This repeated expansion and contraction

ultimately leaves the slope highly susceptible to hydro-mechanical yielding once thawing occurs

[18,36]

. Beyond

hydrological drivers, tectonic disturbances represent a dominant dynamic trigger, particularly in active orogenic

belts. Earthquakes subject slope materials to sudden, intense inertial and shear forces that frequently exceed the

available frictional resistance of internal discontinuities

[3,10,39]

. Seismic shaking causes instantaneous co-seismic

fracturing and potential soil liquefaction

[18,40,41]

. This shaking is also subject to significant topographic

amplification along steep ridges and cliff edges

[10,41,42]

. Amplified kinetic energy drives the rapid detachment of

unstable rock blocks and can generate catastrophic, long-runout mass movements

[10]

. Furthermore, the structural

degradation inflicted by seismic wave propagation leaves the landscape preconditioned for subsequent rainfall-

induced secondary hazards, initiating complex disaster chains

[41,43]

. Human engineering interventions have

increasingly become primary instigators of slope instability, mirroring the destructive potential of natural triggers.

Subsurface mining operations fundamentally disrupt the primordial geostress field

[15,44]

. The extraction of

underground orebodies or coal seams induces a massive redistribution of stress, leading to the progressive

deformation, tension cracking, and eventual subsidence or catastrophic collapse of the overlying strata

[44–46]

. To

visualize this mechanical disruption, Figure 1 illustrates the severe concentration and redistribution of principal

stress localized around an advancing underground excavation face compared to the undisturbed geostress field.

Figure 1: Principal stress distribution before and after underground excavation. The initial geostatic equilibrium

(a) is fundamentally disrupted by the advancing face (b), triggering massive, localized stress redistribution

[14]

Page 2255

www.rsisinternational.org

INTERNATIONAL JOURNAL OF LATEST TECHNOLOGY IN ENGINEERING,

MANAGEMENT & APPLIED SCIENCE (IJLTEMAS)

ISSN 2278-2540 | DOI: 10.51583/IJLTEMAS | Volume XV, Issue VI, June 2026

Conversely, open-pit mining, road construction, and agricultural terracing directly modify the terrain

geometry

[15,47]

. These activities often involve aggressive slope cutting that removes essential basal support and

creates over-steepened, artificial free faces. Similarly, the construction and operation of major hydropower

reservoirs impose severe hydro-mechanical stressors on valley flanks

[34,48]

. The periodic fluctuation of reservoir

water levels subjects the submerged rock and soil to cyclic wetting and drying, which progressively softens the

materials and degrades their overall shear strength

[5,12]

. During phases of rapid water drawdown, the delayed

dissipation of internal groundwater creates steep hydraulic gradients

[18]

. This phenomenon generates powerful

outward-directed seepage forces that compromise the stability of the reservoir banks, frequently reactivating

dormant ancient landslides or triggering sudden rockfalls

[5,18,24]

Coupling Effects

Geological hazards rarely emerge from isolated triggers; rather, they manifest through the complex, non-linear

interaction of multiple physical fields. The transition from a stable state to catastrophic failure is fundamentally

governed by hydro-mechanical coupling, particularly during severe precipitation events

[3,49]

. As rainwater

permeates the subsurface, it initiates a sequence of interacting physical alterations within the rock and soil mass.

Initially, infiltration diminishes matric suction and elevates pore water pressure, which directly reduces the

effective stress and frictional resistance along potential slip surfaces

[33,50,51]

. Concurrently, the accumulated fluid

adds significant gravitational mass to the slope, increasing the driving downward forces

[33,40]

. This hydrodynamic

alteration degrades the structural integrity of the geomaterials, often softening clay-rich or heavily weathered

strata

[36]

. Beyond uniform saturation, transient groundwater flow generates powerful seepage forces that can

erode fine particles within the matrix, thereby creating localized preferential flow networks

[31,34]

. In highly

permeable residual soils, this process can trigger a distinct physical cascade where rapid internal erosion and

localized pore pressure spikes overcome the remaining shear strength, driving the material into an unstable

state

[18,34]

This hydrological mechanism is frequently exacerbated by static and dynamic stresses in anthropogenically

modified terrains, initiating a synergistic tri-field coupling effect of cracks, external loads, and rainfall.

Excavations for transportation networks or residential infrastructure impose static stress redistributions that often

generate initial tension cracks along the slope crest or toe. These structural discontinuities dramatically alter the

infiltration geometry by providing direct, low-resistance channels for preferential water flow. When subjected to

cyclic dynamic loading from passing traffic or heavy machinery, the rock-soil mass experiences progressive

material fatigue, causing these initial mechanical fractures to propagate further into the subsurface. The

synergistic amplification between mechanical crack growth and hydraulic fracturing dictates that even short-

duration rainstorms can induce instantaneous pore pressure surges deep within the slope, bypassing the slower

matrix infiltration process entirely and leading to sudden block detachment

[51]

Similar coupled dynamics govern slope behavior during seismic events, where transient ground shaking interacts

violently with subsurface hydrology. Earthquakes subject slopes to intense inertial forces while simultaneously

altering internal fluid pressures. In fractured or partially saturated geological formations, seismic waves can

compress pore spaces, leading to sudden, localized spikes in pore water pressure that drastically reduce inter-

particle cohesion

[10]

. Even if a hillside survives the immediate shaking, co-seismic fracturing permanently lowers

the overall geomechanical stability threshold

[23]

. These newly formed structural weaknesses precondition the

terrain, allowing subsequent rainfall to infiltrate deeply and trigger delayed, post-seismic mass movements

[52]

Ultimately, the synchronized action of these coupled forces dictates not only the initiation of a hazard but also

its evolutionary trajectory. Once a sliding mass detaches, the combined effects of high moisture content and

accumulated kinetic energy can induce rapid fluidization

[1]

. The mobilized debris then aggressively scours the

channel bed and banks, incorporating additional loose alluvium and amplifying its destructive volume as it

accelerates downslope

[1,53]

. The synchronized action of these coupled forces dictates not only the initiation of a

hazard but also its evolutionary trajectory. To visually synthesize these interacting physical fields, Figure 2

illustrates a conceptual cross-section of hydro-mechanical coupling, demonstrating how surface rainfall

infiltration, internal pore pressure accumulation, and anthropogenic toe excavation synergistically drive slope

instability.

Page 2256

www.rsisinternational.org

INTERNATIONAL JOURNAL OF LATEST TECHNOLOGY IN ENGINEERING,

MANAGEMENT & APPLIED SCIENCE (IJLTEMAS)

ISSN 2278-2540 | DOI: 10.51583/IJLTEMAS | Volume XV, Issue VI, June 2026

Figure 2: Conceptual model of hydro-mechanical coupling in slope instability

[34]

Early Identification of Hidden Hazards

Satellite Remote Sensing

Interferometric synthetic aperture radar fundamentally operates by calculating phase shifts between successive

microwave pulses to map surface displacement at the millimeter scale

[7,54]

. Within this technological domain,

the Small Baseline Subset algorithm specifically minimizes both atmospheric phase delays and spatial-temporal

decorrelation by establishing strict thresholds for interferometric pairing

[12,55,56]

. Consequently, this approach

maintains phase stability across highly vegetated slopes and topographically varied environments, thereby

outperforming permanent scatterer techniques that demand dense, stable targets

[56]

. Single-orbit radar

acquisitions inherently constrain deformation data to a one-dimensional line-of-sight vector, frequently

underrepresenting the magnitude of actual ground movement and struggling to detect north-south kinematic

shifts

[12,55]

. Fusing independent ascending and descending flight paths resolves this geometric ambiguity.

Decomposing these intersecting lines of sight allows geoscientists to reconstruct complete two-dimensional and

three-dimensional deformation velocity fields, which more accurately reflect the true translational or rotational

failure mechanisms of a given slope

[12,52]

. Radar monitoring still encounters significant environmental

constraints. Steep terrain produces complex geometric distortions, primarily layover and shadowing effects, that

physically obscure the radar signal in deeply incised valleys

[7,55,57]

. Moreover, sudden displacement accelerations

immediately preceding a collapse, alongside dense forest canopies induce abrupt phase decorrelation, causing

temporary observation gaps during critical pre-failure windows

[7,57,58]

. High-resolution optical sensors

complement these radar deficiencies by capturing precise surface morphology and multi-spectral

reflectance

[18,59,60]

. Data from platforms like Sentinel-2 and the Gaofen constellation provide the sub-meter

optical clarity necessary to visually trace tension cracks, map active scarp retreat, and verify land-cover

alterations following a displacement event

[37,59,60]

. Passive optical imaging is nonetheless severely limited by its

dependence on clear atmospheric conditions. Heavy cloud cover during intense monsoon rainfall, which is the

exact periods when landslide probability maximizes, renders optical satellites essentially blind

[6,15]

. Fusing these

contrasting modalities establishes a comprehensive monitoring continuum

[60]

. Integrating the continuous, cloud-

penetrating kinematic measurements of radar with the discrete, high-fidelity spatial boundaries of optical

Page 2257

www.rsisinternational.org

INTERNATIONAL JOURNAL OF LATEST TECHNOLOGY IN ENGINEERING,

MANAGEMENT & APPLIED SCIENCE (IJLTEMAS)

ISSN 2278-2540 | DOI: 10.51583/IJLTEMAS | Volume XV, Issue VI, June 2026

imagery allows analysts to confirm the geological origin of a moving mass

[12]

. This integrated methodology

successfully differentiates deep-seated slope instability from shallow anthropogenic ground disturbances,

substantially minimizing the occurrence of false-positive hazard classifications

[12,24]

UAV and Ground-Based Surveys

Unmanned aerial vehicles routinely acquire overlapping high-resolution imagery along pre-programmed flight

trajectories, relying on structure-from-motion and aerial triangulation algorithms to reconstruct complex terrain

geometries into dense three-dimensional point clouds, digital surface models, and orthomosaics

[37,61]

. To ensure

precise georeferencing in environments where deploying ground control points is physically hazardous, modern

aerial platforms integrate real-time or post-processed kinematic global navigation satellite systems

[19,40]

. This

positional accuracy enables analysts to apply digital image correlation and optical flow techniques directly to

the sequential aerial datasets, extracting sub-pixel displacement vectors and tracking the progressive deformation

of open-pit slopes or landslide scars

[36,39]

. The acquisition of multi-angle oblique photographs further facilitates

the extraction of structural data and the construction of high-fidelity digital twins for immersive virtual reality

assessments of active mining operations

[36,62]

. Aerial photogrammetry nonetheless suffers from inherent

limitations in forested topographies, as passive optical sensors cannot resolve bare-earth features beneath dense

vegetative canopies

[9,39,63]

. Light detection and ranging systems overcome this observational barrier by emitting

high-frequency laser pulses and measuring the time of flight of the returning backscatter

[9]

. Because multiple

return signals can be isolated from a single emitted pulse, LiDAR effectively penetrates foliage gaps to generate

accurate digital terrain models of the underlying bedrock and geomorphic depressions

[9,40]

. When mounted on

automated aerial platforms, these active sensors rapidly survey vast, inaccessible orogenic belts and fault zones,

capturing the structural architecture of the landscape without exposing field personnel to imminent collapse

risks

[9]

. Operating at the ground level, terrestrial and dynamic laser scanning systems supplement these airborne

campaigns by delivering millimeter-scale resolution across near-vertical rock faces and deeply incised gorges

[15]

Terrestrial scanners project discrete laser beams against exposed outcrops, extracting the orientation, persistence,

and spacing of rock mass discontinuities that dictate the kinematic feasibility of wedge failures or toppling events.

Fusing the multi-spectral textural fidelity of photogrammetry with the structural penetration of LiDAR

establishes a comprehensive surveying continuum

[40]

. By co-registering these multi-source point clouds over

successive monitoring campaigns, geoscientists compute high-resolution digital elevation models of difference

[8]

This subtraction process directly quantifies volumetric material loss from detachment zones, measures the

accumulation of debris along runout pathways, and isolates subtle micro-topographic anomalies such as incipient

tension cracks or localized subsidence that frequently precede catastrophic mass movements

[8,15,37]

Machine Learning for Hazard Detection

Convolutional neural networks automatically extract hierarchical spatial features directly from multi-spectral

and radar imagery, circumventing the need for manual feature engineering

[3,6,64]

. By applying successive

convolutional and pooling layers, these architectures identify low-level edges and high-level morphological

patterns indicative of slope failure

[6]

. Within this domain, computer vision applications generally bifurcate into

object detection and semantic segmentation

[24]

. Object detection algorithms, including Faster R-CNN and YOLO

variants, localize active deformation zones and historical scars by generating bounding boxes around suspected

geohazards

[24,65]

. While these models offer rapid localization, bounding boxes inherently fail to capture the

irregular physical boundaries of mass movements. Semantic segmentation resolves this geometric limitation by

classifying input imagery strictly at the pixel level

[24,64]

. Architectures such as U-Net employ a symmetric

encoder-decoder structure where skip connections transfer low-level spatial details directly across the network.

This mechanism preserves the fine-grained edge resolution necessary to delineate complex landslide scarps from

stable terrain

[57]

. Other specialized segmentation frameworks, such as fully convolutional networks and image

cascade networks, process varying resolutions to capture both broad spatial patterns and localized geomorphic

anomalies simultaneously. To further improve discrimination in visually complex environments, multi-modal

architectures like the RGB-D fusion network integrate optical spectral reflectance with digital elevation depth

data. Processing these distinct data streams through recursive dilated convolutions allows the network to isolate

specific topographic cues, such as steep failure flanks or accumulated debris flows, that purely spectral models

Page 2258

www.rsisinternational.org

INTERNATIONAL JOURNAL OF LATEST TECHNOLOGY IN ENGINEERING,

MANAGEMENT & APPLIED SCIENCE (IJLTEMAS)

ISSN 2278-2540 | DOI: 10.51583/IJLTEMAS | Volume XV, Issue VI, June 2026

frequently overlook

[64]

. Identifying historical hazard scars triggered by specific extreme events relies heavily on

multi-feature change detection algorithms applied to pre- and post-event satellite optical imagery. These methods

often utilize principal component analysis and independent component analysis to highlight structural deviations

while suppressing background environmental noise

[6]

. Recent structural modifications to these visual processing

models increasingly incorporate attention mechanisms and Vision Transformers

[6,7]

. These transformer blocks

capture long-range spatial dependencies across an image that traditional, highly localized convolution kernels

might miss

[7]

. Hybrid models subsequently fuse these transformer outputs with morphological edge detection

techniques to smooth the jagged discretization frequently observed during automated boundary extraction. A

persistent constraint in training these deep learning models is the severe scarcity of high-quality, manually

annotated hazard inventories. To mitigate class imbalance and limited sample sizes, generative adversarial

networks are deployed to synthesize realistic image patches, effectively expanding the available training data

without requiring additional physical field surveys. Transfer learning strategies address this exact computational

bottleneck by initializing networks with weights pre-trained on massive, generalized global datasets before fine-

tuning them on localized target imagery. This targeted knowledge transfer significantly reduces the requirement

for extensive local sampling while maintaining high detection accuracy across highly diverse geological

domains

[6]

Monitoring And Early Warning Systems

Sensor Networks and IoT Integration

Modern geological monitoring systems increasingly rely on the Internet of Things to connect distributed sensor

arrays directly to cloud-based analytical platforms, thereby eliminating the physical vulnerabilities and high

deployment costs associated with traditional wired instrumentation

[66,67]

. Within this architecture, wireless sensor

networks transmit high-frequency environmental and mechanical data across challenging topographies using

low-power, long-range communication protocols such as LoRaWAN or 4G cellular networks

[50,67]

. To capture

macroscopic surface displacements, global navigation satellite system receivers continuously track three-

dimensional coordinate variations.

By employing real-time kinematic positioning, these receivers achieve millimeter-level accuracy and bypass

cumulative errors, although their operational efficacy remains sensitive to signal obstruction in deep canyons or

beneath dense forest canopies

[15]

. Complementing this regional spatial data, localized surface fracturing is

monitored using high-sensitivity crack meters and wireline extensometers. These instruments translate the

minute opening of tension fissures into digital electrical or optical signals, delivering instantaneous alerts when

surface displacement rates suddenly accelerate

[15,36]

Because surface sensors cannot map the internal structural degradation of a sliding mass, subsurface deformation

must be tracked using embedded inclinometers and tiltmeters. Deep borehole inclinometers capture lateral

subsurface movement along the vertical profile to accurately locate deep-seated shear zones

[15]

. Similarly,

advanced micro-electro-mechanical systems measure the gravitational tilt of specific strata, continuously

recording the angular variations that indicate progressive slope yielding

[25]

. Because slope instability is

fundamentally governed by subsurface hydrology, mechanical tracking is systematically coupled with

groundwater instrumentation

[28]

. Piezometers and tensiometers are deployed at varying depths to measure

transient pore water pressures and soil matric suction

[33,68]

When integrated within a unified wireless network, these distributed piezometric readings allow geoscientists to

observe exactly how rainfall infiltration progressively dissipates effective stress within the soil matrix

[18,25]

. The

real-time synchronization of these diverse field instruments successfully transitions hazard assessment from a

sequence of periodic manual surveys into an automated, data-driven surveillance continuum

[50,69]

. To

contextualize the physical architecture supporting this network, Table 1 synthesizes the primary operational

capabilities and environmental limitations of the remote sensing and ground-based monitoring technologies

discussed thus far.

Page 2259

www.rsisinternational.org

INTERNATIONAL JOURNAL OF LATEST TECHNOLOGY IN ENGINEERING,

MANAGEMENT & APPLIED SCIENCE (IJLTEMAS)

ISSN 2278-2540 | DOI: 10.51583/IJLTEMAS | Volume XV, Issue VI, June 2026

Table 1: Operational comparison of remote sensing and ground-based monitoring technologies for geological

hazard assessment

Technology

Category

Specific Methods

Primary Application

Key Strengths

Fundamental

Limitations

Source

Satellite

Radar

SBAS-InSAR, PS-

InSAR, D-InSAR

Regional tracking of

subsidence basins and

slow-moving

landslides.

All-weather

operational

capacity, extensive

spatial coverage,

and millimeter-

level kinematic

precision.

Susceptible to

vegetation

decorrelation,

atmospheric phase

delays, and restricted by

fixed satellite revisit

cycles.

[12,15,60]

Satellite

Optical

Multispectral,

hyperspectral, and

Thermal Infrared

(TIR).

Macroscopic land

cover mapping,

ecological assessment,

and surface thermal

anomaly detection.

Cost-effective

regional

surveillance with

rich spectral

discrimination and

nocturnal

observation

capabilities.

Highly vulnerable to

cloud cover, fog, and

dust; incapable of

penetrating dense

vegetation canopies.

[15,59]

Aerial

Surveying

UAV

photogrammetry,

Airborne LiDAR,

and Structure-

from-Motion

(SfM).

Generating high-

resolution 3D digital

elevation models and

localized

morphological crack

mapping.

Highly flexible

rapid deployment,

sub-centimeter

spatial resolution,

and effective

vegetation

penetration via

LiDAR.

Flight operations are

restricted by adverse

meteorological

conditions, limited

battery endurance, and

heavy data processing

demands.

[15,40]

Surface

Sensors

GNSS, tiltmeters,

and fissure/crack

meters.

Monitoring localized

fracture propagation

and instantaneous

surface coordinate

variations.

Delivers

continuous, real-

time data

transmission with

high sensitivity for

immediate

macroscopic failure

warning.

Sparse point-based

spatial density,

susceptibility to signal

multipath interference,

and physical

vulnerability to

environmental damage.

[15,25]

Subsurface

Sensors

Deep borehole

inclinometers,

piezometers, and

microseismic

arrays.

Isolating deep shear

zones, tracking

transient pore

pressures, and

localizing micro-

fracture acoustic

emissions.

Directly captures

internal hydro-

mechanical yielding

and precursory

stress changes prior

to surface

deformation.

Requires highly

invasive borehole

excavation, prone to

sensor shearing from

structural collapse, and

experiences signal

attenuation.

[15,25,70]

Data Fusion and Processing

Raw monitoring streams inherently contain environmental interference and measurement anomalies that

necessitate aggressive preprocessing before analytical integration. Statistical methods like the Pauta criterion

establish dynamic confidence intervals based on standard deviations to isolate and discard transient outliers

triggered by extreme weather or localized construction vibrations

[62]

. For high-frequency acoustic and

microseismic data, spectral subtraction algorithms decompose original signals into amplitude and phase spectra,

subtracting stationary background noise before reconstructing the clean waveforms

[70]

. Similarly, adaptive time-

frequency transformations, such as db4 wavelet denoising or orthogonal master-slave adaptive notch filters,

systematically eliminate periodic noise components while preserving the physical characteristics of early fracture

signatures

[13,71]

. Once individual sensor streams are purified, the analytical focus shifts to resolving the spatial

and semantic discontinuities between heterogeneous observation platforms. Integrating meteorological radar,

Page 2260

www.rsisinternational.org

INTERNATIONAL JOURNAL OF LATEST TECHNOLOGY IN ENGINEERING,

MANAGEMENT & APPLIED SCIENCE (IJLTEMAS)

ISSN 2278-2540 | DOI: 10.51583/IJLTEMAS | Volume XV, Issue VI, June 2026

ground-based piezometers, and unmanned aerial vehicle topography establishes the necessary boundary

conditions for dynamic hydro-mechanical modeling

[37]

. Figure 3 visualizes this integrated space-air-ground

monitoring architecture, illustrating how these diverse sensor streams converge into a centralized data fusion

center.

Figure 3: Architecture of an integrated space-air-ground multi-scale geological hazard monitoring system

Aligning these ground-level metrics with orbital remote sensing demands specialized cross-modal architectures.

To harmonize optical imagery, which excels at delineating surface morphology but is easily obscured by cloud

cover, with all-weather synthetic aperture radar, deep learning frameworks frequently utilize domain-sharing

encoders. These architectures project disparate spectral and microwave signals into a unified feature space,

applying pointwise convolutions and attention mechanisms to suppress modality-specific biases

[60]

. The

subsequent fusion of these aligned spatial features with temporal meteorological inputs introduces significant

predictive uncertainty. Researchers increasingly apply Dempster-Shafer evidence theory to calculate the joint

probability of multi-source warning indicators to mitigate this ambiguity. This probabilistic formulation treats

conflicting sensor readings as probability masses, dynamically reallocating support to resolve contradictions

between localized crack gauge accelerations and stable regional radar interferograms

[62,72]

. Complementing this

probabilistic reasoning, multi-model ensemble strategies concurrently process the fused data arrays. Hybrid

temporal networks deploy parallel architectures such as Long Short-Term Memory units, Gated Recurrent Units,

and frequency-aware Transformers to extract long-range dependencies from the homogenized time-series data.

By optimizing the predictive weights of each base learner through genetic algorithms or empirical risk functions,

these ensemble matrices effectively stabilize the variance inherent in multi-sensor environments and generate

continuous, physically coherent risk trajectories

[73]

Early Warning Thresholds

Establishing actionable warning triggers requires translating raw environmental and kinematic data into defined

boundaries that signify impending slope failure. In meteorological monitoring, empirical power-law

relationships frequently define the minimum forcing required to initiate instability by plotting accumulated

rainfall against event duration. To optimize these boundaries for public alert systems, analysts employ frequentist

approaches that set specific non-exceedance probabilities, deliberately balancing true positive predictions while

minimizing false alarm rates

[74]

. Beyond single-variable duration limits, contemporary hazard management

utilizes double-index models to capture complex hydrological interactions. These models evaluate short-term

storm intensity, such as daily precipitation, alongside antecedent moisture accumulation spanning five to fifteen

days

[47,63]

. By cross-referencing immediate rainfall with prior effective infiltration, these systems establish

Page 2261

www.rsisinternational.org

INTERNATIONAL JOURNAL OF LATEST TECHNOLOGY IN ENGINEERING,

MANAGEMENT & APPLIED SCIENCE (IJLTEMAS)

ISSN 2278-2540 | DOI: 10.51583/IJLTEMAS | Volume XV, Issue VI, June 2026

graded alert tiers ranging from low advisories to very high emergency evacuation levels

[47]

. Physics-based

models further modulate these meteorological limits by incorporating the initial saturation state of the soil profile.

Field tests demonstrate that during dry seasonal conditions, the required cumulative rainfall to breach safety

thresholds is significantly higher than during wet periods, reflecting the initial buffering capacity of unsaturated

pore spaces

[35]

When environmental forcing translates into physical ground movement, kinematic thresholds provide a direct

measurement of the progression toward catastrophic collapse. A widely adopted standard for defining these

critical failure points is the improved tangent angle method, which mathematically transforms displacement

time-series curves into angular metrics. Within this evaluation system, angles below 45 degrees indicate primary,

stable creep. Once the tangent angle reaches 45 degrees, the geomaterial enters a constant-rate secondary

deformation stage, whereas angles exceeding 85 degrees typically trigger terminal red alerts denoting imminent

tertiary failure. This angular metric is frequently combined with hard limits on cumulative displacement and

daily deformation velocity to formulate a comprehensive instability index

[62]

. To isolate the exact temporal

transition into the dangerous acceleration phase, statistical techniques compute the normal fluctuation range of

a stable slope and identify the onset of acceleration point when deformation velocity permanently deviates from

this baseline

[75]

. For macroscopic regional surveillance utilizing satellite interferometry, distinct kinematic

thresholds isolate hazardous slopes from stable terrain. Researchers have proposed that slopes exhibiting

settlement rates exceeding 30 millimeters per year, or cumulative displacements beyond 80 millimeters, should

be automatically flagged as potential landslide points demanding immediate intervention

[57]

. To visualize this

kinematic progression and the corresponding stability thresholds, Figure 4 illustrates the standard three-stage

creep deformation curve, highlighting the critical temporal transitions from steady-state creep to terminal

acceleration.

Figure 4: Kinematic stages of slope creep deformation

[40]

In geological settings subjected to dynamic disturbances, early warning criteria depend heavily on the

accumulation and dissipation of mechanical energy. For earthquake-induced hazards, stability thresholds are

dictated by a critical acceleration limit, which is derived directly from the inherent shear strength and frictional

resistance of the slope material

[2]

. When transient peak ground acceleration surpasses this specific threshold,

slopes experience rapid non-linear displacement escalation, signaling the onset of co-seismic mass movement

[76]

Similarly, in deep underground excavations, warning parameters are established by monitoring the intensity and

Page 2262

www.rsisinternational.org

INTERNATIONAL JOURNAL OF LATEST TECHNOLOGY IN ENGINEERING,

MANAGEMENT & APPLIED SCIENCE (IJLTEMAS)

ISSN 2278-2540 | DOI: 10.51583/IJLTEMAS | Volume XV, Issue VI, June 2026

spatial concentration of microseismic events. Sudden spikes in the logarithmic energy density or abrupt

variations in dominant frequency amplitudes serve as reliable indicators that the rock mass is transitioning from

localized stress adjustment to macroscopic shear fracturing

[77,78]

. Emerging field investigations even suggest that

thermodynamic responses can define instability limits. Internal slope temperature variations, specifically

anomalous sharp peak and gentle slope fluctuation patterns, strongly correlate with high-intensity rainfall

infiltration. These thermal signatures provide an independent physical metric to establish joint thresholds

alongside conventional precipitation limits, ultimately enhancing the reliability of predictive alarms before

physical detachment occurs

[38]

Prediction And Forecasting Models

Statistical and Heuristic Methods

Heuristic techniques, such as the Analytic Hierarchy Process and fuzzy logic, formulate landslide susceptibility

by relying entirely on expert judgment to assign relative weights to conditioning parameters

[79]

. While these

knowledge-driven strategies successfully integrate qualitative geomorphological experience, their inherent

subjectivity introduces notable inconsistencies and uncertainties into spatial hazard assessments

[64,80]

Quantitative statistical models bypass this subjective reliance. They directly analyze the mathematical

correlations between historical landslide inventories and underlying topographic or environmental datasets

[2,61]

Among bivariate approaches, the Information Value model measures the propensity for slope failure by

calculating the logarithmic ratio of landslide density within a specific factor class against the regional

average

[56,57]

. This formulation yields a transparent, physically interpretable index where positive values denote

a direct contribution to instability

[5,56,81]

. The computational simplicity of IV mapping efficiently isolates the

individual influence of geo-environmental parameters. It is, however, fundamentally restricted by linear

assumptions

[5,81]

. Because it evaluates each variable independently, the traditional IV approach fails to capture

complex, nonlinear interactions among triggering mechanisms, frequently resulting in factor suppression or

misrepresentation in highly heterogeneous terrains

[80,81]

. To resolve missing data anomalies in categories lacking

historical failures, researchers occasionally apply a Modified Information Value technique. This mathematically

adjusts the discrete calculations to ensure all parameter classes contribute meaningfully to the statistical

analysis

[80,82]

. Extending beyond bivariate limitations, multivariate techniques like logistic regression compute

the combined influence of independent continuous and categorical variables on a binary stability outcome

[23,83]

LR models estimate the probability of slope failure through a sigmoidal function. They utilize partial regression

coefficients to quantify the log-odds ratio of an event occurring under specific site conditions

[23,82]

. This

multivariate capacity allows analysts to identify dominant causative factors without assuming a normal

distribution within the input data

[23]

. To ensure the statistical independence of selected factors before executing

logistic regression, researchers must rigorously apply multicollinearity diagnostics, typically identifying and

discarding redundant variables through Variance Inflation Factor and tolerance thresholds

[23,80,84]

. Despite this

analytical rigor, logistic regression remains constrained by its foundational assumption of linearity between

predictors and the target logit

[85]

. Consequently, LR formulations often exhibit diminished predictive accuracy

when processing highly dimensional, non-linear geospatial relationships

[71,85]

. Empirical models therefore

routinely yield moderate discrimination capacities

[63]

. To circumvent these mathematical boundaries while

preserving interpretability, modern assessments increasingly deploy these traditional statistical outputs not as

final susceptibility indices, but as weighted spatial priors that feed directly into advanced ensemble learning

architectures

[56,81]

. Furthermore, the statistical boundaries generated by IV assessments are increasingly

repurposed to optimize non-landslide sampling; extracting negative training points exclusively from IV-

designated low-susceptibility zones systematically reduces the spatial bias associated with purely random

sampling protocols

[80,84]

Physically Based Models

Limit equilibrium methods establish the fundamental physical basis for slope stability forecasting by defining

the factor of safety as the mathematical ratio of the available shear strength to the downslope driving shear

stress

[86]

. For shallow failures where the rupture depth is significantly smaller than the overall slope length,

analysts routinely apply the infinite slope model

[19,27]

. This formulation assumes that the failure plane develops

Page 2263

www.rsisinternational.org

INTERNATIONAL JOURNAL OF LATEST TECHNOLOGY IN ENGINEERING,

MANAGEMENT & APPLIED SCIENCE (IJLTEMAS)

ISSN 2278-2540 | DOI: 10.51583/IJLTEMAS | Volume XV, Issue VI, June 2026

strictly parallel to the topographic surface, effectively allowing stability to be evaluated within a rigid, individual

slice of soil

[27]

. Under this geometry, the driving forces are governed by the gravitational weight of the soil mass,

which depends directly on the slope angle and the saturated or unsaturated unit weight of the material

[27,86]

Conversely, the resisting shear strength is quantified using the Mohr-Coulomb failure criterion, relying on

fundamental geotechnical properties including effective cohesion, the angle of internal friction, and prevailing

pore water pressures

[27]

. When evaluating more complex, deep-seated, or non-circular sliding geometries, single-

body infinite slope assumptions become inadequate. Researchers instead utilize method of slices techniques,

such as the Bishop, Janbu, Spencer, Morgenstern-Price, and Sarma formulations. These approaches partition the

potential sliding mass into multiple discrete vertical blocks to satisfy varied combinations of horizontal, vertical,

and rotational equilibrium

[2]

. Operating solely on mechanical equilibrium, these rigid-body analyses are

frequently coupled with transient hydrological equations to model rainfall-induced instability. Models including

TRIGRS, SHALSTAB, and SINMAP integrate the infinite slope stability equation with physical infiltration

models, such as the Green-Ampt approximation or Richards' equation, to capture the subsurface progression of

wetting fronts

[87]

. As meteoric water percolates into the soil matrix, the resultant increase in pore-water pressure

and concurrent dissipation of matric suction fundamentally reduce the effective stress acting on the failure

plane

[27]

. This hydrodynamic alteration degrades the available shear resistance, progressively lowering the factor

of safety until it drops below unity, marking the theoretical onset of physical failure

[86–88]

. Despite their

computational efficiency, traditional limit equilibrium methods remain constrained by their foundational

mathematical assumptions. They treat the geomaterial as a completely rigid body, ignore internal stress-strain

constitutive relationships, and compute only an average safety factor across the entire predefined slip surface.

These macroscopic averaging obscures localized stress concentrations and prevents the simulation of progressive

mechanical yielding. To transcend these analytical limitations, contemporary physical forecasting increasingly

relies on advanced stress-strain numerical modeling utilizing finite element or finite difference techniques. These

computational environments dispense with predefined slip geometries and explicitly simulate material

deformation, strain localization, and stiffness degradation under dynamic loading

[2,5]

. Within these numerical

frameworks, the factor of safety is typically extracted using the shear strength reduction technique

[5,89]

. This

procedure iteratively divides the input cohesion and internal friction angle by a scalar reduction factor until the

numerical model fails to reach equilibrium

[90]

. The threshold reduction value triggers macroscopic slope failure

often identified through abrupt nodal displacement or the coalescence of continuous plastic shear zones, is then

designated as the system's actual safety factor. By directly linking localized mechanical responses to global

stability limits, this reduction methodology bypasses the rigid-body constraints of limit equilibrium analysis and

precisely localizes irregular failure surfaces without prior geometric estimation

[89,90]

Data-Driven Models

Data-driven prediction techniques bypass the requirement for explicit physical parameters by extracting

mathematical correlations directly from historical observations and environmental datasets

[3]

. Ensemble machine

learning algorithms currently dominate regional susceptibility mapping due to their capacity to process high-

dimensional, nonlinear geospatial inputs

[84,91,92]

. Random Forest aggregates multiple decision trees through

bootstrap sampling and random feature selection, a mechanism that effectively minimizes variance and controls

overfitting

[54,80,84]

. It readily handles multicollinearity and automatically quantifies variable importance through

metrics like Gini impurity

[54,84]

. Alternatively, gradient boosting frameworks such as XGBoost construct decision

trees sequentially to correct the prediction residuals of preceding iterations

[54,80]

. By incorporating second-order

Taylor expansion and explicit regularization terms, XGBoost actively penalizes model complexity

[54,56,84]

. This

ensures high predictive precision even when dealing with severely imbalanced hazard inventories

[54,84]

. While

ensemble techniques require predefined input variables, deep learning architectures automatically extract

hierarchical representations directly from raw multi-source data

[64,93]

. Convolutional neural networks utilize

successive convolutional and pooling operations to capture complex spatial dependencies across topographic

and geological layers

[5,64,86]

. Because they reduce dimensionality without significant loss of information, these

networks translate pixel-level variations into spatial probabilities of slope failure

[6,64,86]

. Predicting the precise

temporal occurrence of a failure demands distinct sequence modeling capabilities to process non-stationary

monitoring streams, such as cumulative displacement or groundwater fluctuations

[34,94]

. Long short-term

memory networks and gated recurrent units overcome the gradient vanishing problems inherent in standard

Page 2264

www.rsisinternational.org

INTERNATIONAL JOURNAL OF LATEST TECHNOLOGY IN ENGINEERING,

MANAGEMENT & APPLIED SCIENCE (IJLTEMAS)

ISSN 2278-2540 | DOI: 10.51583/IJLTEMAS | Volume XV, Issue VI, June 2026

recurrent neural networks by employing specialized memory blocks and gating mechanisms

[34,94]

. These

structures selectively retain informative historical trends while filtering high-frequency noise, making them

highly effective for forecasting progressive step-like displacements and identifying short-term acceleration

precursors

[5,62,94]

. To capture both geometric morphology and dynamic kinematics, contemporary frameworks

frequently fuse these distinct neural architectures. Dual-autoencoder systems couple multilayer perceptrons with

sequence networks to simultaneously compress static environmental parameters and extract deep evolutionary

features from temporal windows

[3,93]

. More recently, attention-based Transformer models have been deployed to

bypass the step-wise information propagation limits of traditional sequential networks. By allocating attention

weights globally across an entire time series, Transformers better capture abrupt, impulse-driven responses and

long-range dependencies, such as the delayed slope degradation triggered by seismic tail waves

[3]

. Furthermore,

graph neural networks address the spatial autocorrelation assumptions of standard grid-based models by treating

monitoring locations as irregular nodes. This approach preserves the non-Euclidean topological relationships of

natural terrain, aggregating feature information through spatial connectivity to improve boundary detection and

hazard localization in complex orogenic environments

[17]

. While these advanced neural architectures offer

unprecedented capabilities in processing complex spatial topologies, selecting the appropriate predictive

framework ultimately requires balancing computational cost, data availability, and the necessity for physical

interpretability. To visually summarize these distinct operational pipelines, Figure 5 contrasts the deterministic

workflow of physically based models against the multi-source architecture of data-driven frameworks.

Furthermore, to synthesize the broader critical trade-offs across all methodologies, Table 2 provides a

comprehensive comparison of the statistical, physically based, and data-driven forecasting models currently

utilized in geohazard assessments.

Figure 5: Comparative workflows of physically based vs. data-driven forecasting models

Table 2: Summary of core mechanisms, primary advantages, and major limitations of geological hazard

forecasting models

Model Category

Specific

Algorithms /

Frameworks

Core

Mechanism

Primary

Advantages

Major

Limitations

Source

Heuristic & Statistical

AHP,

Information

Value (IV),

Logistic

Regression

Weighs

causative

conditioning

factors via

pairwise

comparison

Offers high

structural

transparency,

simple

mathematical

computation,

Relies heavily on

subjective expert

judgment and

lacks the capacity

to simulate

[8,80,95]

Page 2265

www.rsisinternational.org

INTERNATIONAL JOURNAL OF LATEST TECHNOLOGY IN ENGINEERING,

MANAGEMENT & APPLIED SCIENCE (IJLTEMAS)

ISSN 2278-2540 | DOI: 10.51583/IJLTEMAS | Volume XV, Issue VI, June 2026

matrices or

applies relative

statistical

probability

calculations to

historical data.

and the direct

integration of

subjective expert

knowledge.

explicit physical

failure mechanics.

Physically Based

Limit

Equilibrium,

Finite

Element,

TRIGRS

Models

specific

transient

triggering

conditions and

calculates

safety factors

based on

topographic

and

hydrological

inputs.

Grounded in

explicit physical

principles,

providing a

comprehensive

basis for

understanding

the continuous

dynamics of

complex hazard

chains.

Computationally

expensive and

demand precise,

high-resolution

geotechnical and

hydrological

input parameters

to function

accurately.

[88]

Ensemble Machine

Learning

Random

Forest,

XGBoost

Aggregates

numerous

decision trees

utilizing

parallel

bagging or

sequential

gradient

boosting

algorithms.

Excels at

capturing

complex non-

linear

relationships

among

environmental

variables and

quantitatively

ranks feature

importance.

Operates as an

opaque

mathematical

algorithm that

remains highly

susceptible to

sampling biases

and training data

imbalances

[55,84,96]

Deep & Sequence

Learning

CNNs,

LSTMs,

Transformers

Employs gated

memory cells

or multi-head

self-attention

modules to

process

variable-length

sequential

data.

Automatically

extracts long-

term

spatiotemporal

dependencies

and features

from continuous

time-series

monitoring

sequences.

Highly

computationally

intensive,

requiring

extensive training

datasets and

heavy hardware

optimization to

function

effectively.

[62,73,97]

Hazard Prevention and Control

Engineering Interventions

Engineering interventions counteract mass wasting through mechanical resistance and hydrodynamic regulation.

Anti-slide piles and retaining walls deliver immediate lateral support by transferring upper sliding forces into

stable, deeper bedrock strata. When installed at the over-steepened toe of a slope, these rigid boundaries arrest

progressive shear failure and constrain localized bulging

[50]

. To address internal tensile and shear deficiencies,

systematic rock bolting and anchored wire mesh configurations actively bind fractured rock masses together.

These reinforcement anchors perform reliably in high-relief seismic zones, although their holding capacity can

degrade if prolonged dynamic shaking induces anchor pullout

[10]

. To overcome the limitations of strictly rigid

supports, stabilization design increasingly incorporates yielding or composite structures. Deep underground

excavations, for example, utilize energy-absorbing mechanisms such as concrete-filled steel pipes coupled with

high-pressure advanced grouting to accommodate substantial geostatic deformations while sealing surrounding

fissures against fluid ingress

[98]

. Similarly, inclined steel grouting pipes injected into loess embankments

establish an interconnected cementitious network that densifies the soil matrix and physically impedes capillary

moisture migration

[49]

. Hydrological control remains equally necessary to prevent the accumulation of

Page 2266

www.rsisinternational.org

INTERNATIONAL JOURNAL OF LATEST TECHNOLOGY IN ENGINEERING,

MANAGEMENT & APPLIED SCIENCE (IJLTEMAS)

ISSN 2278-2540 | DOI: 10.51583/IJLTEMAS | Volume XV, Issue VI, June 2026

destabilizing pore water pressures

[10]

. Surface interceptor drains and cut-off ditches rapidly channel meteoric

runoff away from vulnerable failure scarps, effectively minimizing infiltration and subsequent matric suction

loss

[50]

. Beneath the topographic surface, perforated subsurface pipes and deep drainage galleries systematically

relieve elevated groundwater levels, directly mitigating hydraulic fracturing and base heave in soft-ground

environments

[10,99]

. In steep catchments prone to high-velocity debris flows, physical mitigation shifts toward

kinetic energy dissipation. Closed-type check dams are strategically positioned along channel convergence

points to intercept fluidized solid volumes. Deploying sequential barriers such as an upper dam to immediately

attenuate initial flow momentum and a lower terminal dam to trap residual materials, dramatically reduces the

destructive velocity reaching downstream infrastructure

[29]

. For isolated kinematic rockfall events, passive

interceptors like flexible catchment nets, earthen bunds, and diversion trenches capture descending blocks

[10]

Stepped buffering platforms filled with composite materials can also smoothly dissipate terminal kinetic energy

before rocks strike transportation corridors

[42]

. In shallower unstable terrains, bioengineering interventions

frequently supplement structural mechanics. Deep-rooted vegetation physically binds topsoil to resist fluvial

scour

[10]

. Planting live cuttings alongside driven steel nails and natural fiber mats yields a hybrid reinforcement

that limits frost heaving and shallow rainfall-induced sliding

[100]

Ecological Mitigation Methods

Plant root systems mechanically reinforce unstable slopes by physically binding loose soil particles and

increasing the apparent cohesion of the localized terrain

[18,19]

. This bio-mechanical interaction elevates the

overall shear strength of the soil matrix without significantly altering its internal angle of friction. Deep-rooted

woody species naturally offer superior anchoring capacity compared to shallow-rooted grasses, as they penetrate

deeper soil horizons to interlock with underlying stable strata

[10,23]

. Beyond physical anchoring, vegetation exerts

profound hydrological control over slope dynamics. Canopy structures intercept incoming precipitation and

shield the vulnerable surface from direct erosive runoff

[18,100]

. Below the surface, active root water uptake and

evapotranspiration consistently reduce volumetric soil moisture, thereby preserving matric suction and aiding in

the rapid desaturation of the soil profile

[18,100]

. This hydrological reinforcement becomes particularly pronounced

during the spring and summer vegetative growing seasons

[100]

. Root systems nonetheless exhibit a complex dual

behavior; while they actively draw down moisture, their physical penetration can simultaneously create

preferential flow channels that, under extreme rainfall, might locally accelerate rapid water infiltration and

momentarily elevate pore water pressures

[19,23]

. To systematically harness these biological benefits, practitioners

increasingly deploy soil and water bioengineering techniques across degraded or artificially cut terrains.

Hydroseeding is frequently applied to steep, bare embankments by spraying a customized slurry containing seeds,

nutrients, and sediments, which accelerates the establishment of a stabilizing surface root network. For somewhat

deeper structural stabilization, brush layering embeds live cut branches such as salix cuttings, directly into

excavated slope terraces to provide immediate mechanical resistance. Because young vegetative cuttings and

seeds remain highly vulnerable to rain washout before establishing a mature root architecture, biodegradable

geotextiles, primarily coconut fiber mats, are typically anchored over the treated areas to prevent immediate

topsoil loss. These biological components are routinely integrated into hybrid stabilization frameworks, such as

live crib walls or combined soil-nailing arrays, where the maturing vegetation progressively assumes the primary

load-bearing function as the temporary wooden or steel structural elements degrade over subsequent decades.

Optimal shear resistance within these systems is achieved by planting a diverse composite of grasses, shrubs,

and trees, which generates an overlapping, multi-tiered root matrix

[100]

. Despite these distinct mechanical and

hydrological advantages, the absolute efficacy of bioengineering remains strictly bounded by the physical limits

of the rhizosphere. Root reinforcement rarely extends beyond a vertical depth of 0.5 to 2.0 meters, rendering

purely vegetative mitigation insufficient against deep-seated kinematic failures dictated by underlying geological

structural discontinuities

[84]

. Because full biological interventions are constrained by the physical limits of the

rhizosphere, strong disaster risk reduction frequently demands a hybrid approach. To visually contextualize this

integration, Figure 6 illustrates a stabilized slope cross-section, contrasting the shallow soil reinforcement

provided by ecological bioengineering against the deep, lateral support required from structural retaining walls.

Page 2267

www.rsisinternational.org

INTERNATIONAL JOURNAL OF LATEST TECHNOLOGY IN ENGINEERING,

MANAGEMENT & APPLIED SCIENCE (IJLTEMAS)

ISSN 2278-2540 | DOI: 10.51583/IJLTEMAS | Volume XV, Issue VI, June 2026

Figure 6: Cross-section of a hybrid mitigation framework illustrating the integration of structural toe support and

shallow ecological root reinforcement

Non-Structural Risk Management

Translating spatially explicit susceptibility models into binding land-use regulations represents the primary

mechanism for decoupling human exposure from geohazard threats. By integrating predictive hazard maps into

regional master plans, municipal authorities can systematically restrict residential and commercial development

within high-probability failure zones, thereby preventing the artificial amplification of risk associated with

unregulated peri-urban expansion

[92,101]

. This regulatory approach frequently involves delineating strict no-

construction buffer zones along steep, unstable topography or river valleys prone to cascading debris flows

[92,102]

In regions already burdened by extensive infrastructure, strategic land-use planning shifts toward managed

retreat, subsidizing the proactive relocation of exposed settlements and critical utilities to geologically stable

lowlands

[85]

. However, spatial zoning inherently requires a complementary evaluation of community

vulnerability to accurately prioritize these interventions. Quantitative vulnerability assessments achieve this by

intersecting physical exposure metrics such as the distribution of building footprints, gross domestic product

density, and agricultural acreage with socioeconomic indicators that dictate local coping capacities

[95]

. Utilizing

composite metrics like the Social Development Index or Human Development Index alongside demographic

data allows analysts to isolate specific social fragilities

[101,102]

. These indices frequently reveal that migrant

populations and economically marginalized groups, who often reside in informal settlements with inadequate

drainage and weak social networks, experience disproportionately high vulnerability to mass movements

[102]

Mapping these socioeconomic disparities ensures that hazard mitigation resources are directed toward

communities lacking intrinsic adaptive capacity rather than relying strictly on geographic proximity to a

hazard

[79,101]

. Once high-vulnerability clusters are identified, risk managers implement targeted evacuation

strategies guided by predictive runout simulations. Delineating the precise spatial trajectory and terminal extent

of potential debris flows or rockfalls provides the physical basis for designing safe, accessible evacuation routes

and strategically positioning emergency shelters away from anticipated impact corridors

[4,92,102]

. The

effectiveness of these logistical protocols depends entirely on localized preparedness and the integration of

community-based early warning systems. Disseminating real-time hazard alerts via telecommunication networks

facilitates rapid vertical or horizontal evacuation before extreme rainfall thresholds are breached

[25,27,47]

. To

ensure these alerts translate into immediate action, authorities must sustain ongoing public awareness campaigns,

conduct routine community drills, and foster local participation in monitoring efforts, which collectively

empower residents to execute orderly evacuations during sudden-onset environmental crises

[25,85,102]

. While these

community-driven protocols represent the final layer of defense, effective disaster risk reduction ultimately relies

on the strategic integration of all available countermeasures.

To synthesize this multi-tiered approach, Table 3 categorizes the structural, ecological, and non-structural

mitigation strategies discussed throughout this section, highlighting their specific target mechanisms and

operational lifespans.

Page 2268

www.rsisinternational.org

INTERNATIONAL JOURNAL OF LATEST TECHNOLOGY IN ENGINEERING,

MANAGEMENT & APPLIED SCIENCE (IJLTEMAS)

ISSN 2278-2540 | DOI: 10.51583/IJLTEMAS | Volume XV, Issue VI, June 2026

Table 3:Classification of structural, ecological, and non-structural geological hazard mitigation strategies

Mitigation

Category

Specific

Interventions

Target Mechanism

Operational

Lifespan & Function

Source

Structural / Rigid

Anti-slide piles,

retaining walls, rigid

linings.

Resists sliding forces;

transfers localized stress to

stable deep bedrock. Directly

opposes shear failure.

Immediate lateral

support; blocks mass

movement. Static,

long-term structural

stabilization.

[50,99]

Structural /

Dynamic

Flexible rock

barriers, yielding

supports, rock bolts,

energy-dissipation

platforms.

Absorbs kinetic energy via

shape-changing properties.

Accommodates large

deformations; enhances rock

mass cohesion.

Dynamic shock

absorption; prevents

sudden brittle failure;

actively adapts to

progressive stress.

[10,42,98]

Hydrological

Control

Deep well

dewatering, surface

drainage ditches,

siphon pipelines.

Lowers groundwater table;

mitigates pore water pressure

surges. Reduces rainfall

infiltration.

Continuous seepage

control; prevents

static liquefaction and

hydraulic fracturing.

[50,103,104]

Ecological

(Bioengineering)

Live crib walls,

hydroseeding, deep-

rooted vegetation

buffers.

Roots mechanically bind soil

particles; increases shear

strength. Intercepts rainfall

runoff; regulates soil

moisture via

evapotranspiration.

Self-strengthening

long-term

stabilization as roots

mature; continuous

erosion control;

ecological restoration.

[18,100]

Non-Structural

Land-use planning,

community

relocation, early

warning systems

(EWS), risk

zonation.

Reduces human and

infrastructure exposure to

hazards. Delivers preemptive

alerts before catastrophic

failure.

Proactive risk

avoidance;

continuous life-cycle

situational awareness;

zero physical terrain

footprint.

[102,105]

DISCUSSION AND FUTURE PERSPECTIVES

Current Technical Limitations

Ground-based sensor networks frequently experience data loss, signal discontinuities, and physical damage when

deployed in harsh operating environments. Extreme weather events, mechanical vibrations, and heavy dust

physically degrade instrumentation, while signal obstruction in deep topographical incisions disrupts wireless

telemetry

[15,62]

. These transmission anomalies create missing values that alter time-series structures, directly

causing analytical failures in predictive algorithms. While statistical filters are routinely applied to remove such

ambient noise, they introduce a secondary operational risk. Rigid statistical boundaries often fail to differentiate

between common measurement noise and the abrupt, high-frequency anomaly signals that characterize actual

slope destabilization

[62]

. This means significant precursory indicators are inadvertently discarded before they

reach the analytical stage. Macro-scale observation platforms face equally significant environmental barriers that

complicate continuous spatial monitoring. Passive optical remote sensing is inherently constrained by its

dependence on clear atmospheric conditions, rendering it largely ineffective during monsoon seasons or

prolonged cloud cover when hydro-meteorological hazards are most likely to initiate

[6,15]

. Conversely, active

microwave systems bypass cloud interference but remain highly susceptible to geometric and atmospheric

distortions. In steep alpine topographies, inherent side-looking radar geometries induce severe foreshortening,

layover, and shadowing effects, which physically obscure deformation signals. Even in observable areas, dense

vegetation canopies and fluctuating soil moisture cause rapid temporal decorrelation. Atmospheric phase screens

generated by tropospheric and ionospheric variations produce spatially coherent phase shifts that artificially

mimic true ground displacement

[7,57]

. The presence of these pervasive environmental artifacts fundamentally

Page 2269

www.rsisinternational.org

INTERNATIONAL JOURNAL OF LATEST TECHNOLOGY IN ENGINEERING,

MANAGEMENT & APPLIED SCIENCE (IJLTEMAS)

ISSN 2278-2540 | DOI: 10.51583/IJLTEMAS | Volume XV, Issue VI, June 2026

undermines the reliability of contemporary machine learning and deep learning forecasting architectures.

Because deep neural networks are highly data-driven, they frequently overfit to site-specific atmospheric or

geometric noise, inadvertently learning to classify these artifacts as actual geodetic deformation

[7]

. This noise

sensitivity drives a persistent operational bottleneck: algorithms optimized to maximize detection hit rates

frequently generate unacceptable volumes of false alarms. Such false positive alerts overwhelm manual

verification capacities and ultimately erode the social credibility of early warning systems during long-term

operation

[6]

. This issue is compounded by severe domain shift, where predictive models trained on specific

lithologies, sensor wavelengths, or temporal sampling intervals experience drastic performance degradation

when transferred to novel geographic regions or different satellite platforms

[7]

. Resolving these false alarm rates

necessitates the fusion of multi-source heterogeneous datasets, yet integrating disparate geological,

meteorological, and kinematic streams introduces profound scaling challenges. Different sensor platforms

operate across incompatible spatial resolutions, temporal frequencies, and data formats, leading to semantic

inconsistency and information loss during analytical integration. Predictive models typically discretize

continuous geological bodies into rigid grid cells, which mathematically homogenizes localized subsurface

properties. This scale-induced simplification obscures the spatial continuity of deep structural planes and weak

interlayers that actually govern failure mechanics. Consequently, the resulting algorithms function

predominantly as opaque black boxes that map statistical correlations rather than underlying physical processes.

The absence of transparent, physically grounded reasoning in these models severely limits their engineering

applicability, as decision-makers cannot readily interpret the causal mechanisms driving the automated hazard

alerts

[6]

Future Research Directions

Future algorithmic architectures are progressively integrating physical laws directly into machine learning

pipelines to resolve the inherent opacity and data dependency of pure data-driven forecasting. Physics-informed

neural networks embed partial differential equations and geomechanical constraints such as elastic dislocation

or poroelastic adjustment, directly into the loss function of the model

[7,106]

. By penalizing predictions that violate

fundamental principles like mass conservation or kinematic consistency, these hybrid approaches force the neural

network to output physically plausible deformation trajectories, even when trained on sparse or noisy datasets

[6]

This regularization strategy is highly effective at mitigating the influence of atmospheric phase screens and

temporal decorrelation in satellite interferometry, which frequently cause standard deep learning models to learn

environmental artifacts as actual ground displacement

[7]

. While the theoretical advantages of physics-informed

artificial intelligence are substantial, applying these coupled models across highly heterogeneous lithologies

remains computationally expensive and mathematically challenging because idealized governing equations

rarely match the discontinuous reality of natural rock masses

[6]

. To overcome these computational barriers,

researchers propose integrating adaptive sampling, active learning, and physics-guided surrogate models to

maintain high predictive accuracy while minimizing sample demand and runtime

[6,107]

. Parallel to these

algorithmic developments, the physical monitoring landscape is transitioning toward the implementation of

digital twin technologies. Virtual replicas of physical infrastructure and geological environments fuse real-time

Internet of Things sensor streams—including groundwater piezometers, microseismic geophones, and

distributed fiber optics—with three-dimensional geological models and Building Information Modeling systems.

This continuous data assimilation allows analysts to simulate subsurface stress redistributions and fluid

migrations dynamically, effectively replacing static regional susceptibility maps with interactive, time-resolved

virtual testing environments. Integrating these digital replicas with edge computing architectures further reduces

transmission latency by enabling localized anomaly detection and immediate decision-making at the sensor node

before data is relayed to centralized cloud servers. Beyond predicting isolated slope failures, future hazard

assessments must explicitly quantify the complex dynamics of cascading multi-hazard disaster chains

[27,95]

Extreme meteorological forcing or seismic shocks frequently initiate primary instabilities, such as massive

rockfalls, which rapidly fluidize into debris flows, dam adjacent river channels, and ultimately trigger

catastrophic outburst floods

[53,108]

. Because traditional models generally evaluate these phenomena

independently, they severely underestimate the compounded spatial exposure, kinetic energy amplification, and

secondary environmental contamination generated during a continuous disaster sequence

[1,95]

. To capture these

cascading interactions, emerging theoretical frameworks propose generalized geological-environmental hazard

Page 2270

www.rsisinternational.org

INTERNATIONAL JOURNAL OF LATEST TECHNOLOGY IN ENGINEERING,

MANAGEMENT & APPLIED SCIENCE (IJLTEMAS)

ISSN 2278-2540 | DOI: 10.51583/IJLTEMAS | Volume XV, Issue VI, June 2026

chain effect matrices that iteratively calculate the overlapping spatial and temporal probabilities of primary mass

movements and their secondary impacts

[1]

. Translating these conceptual matrices into operational tools requires

coupled numerical simulations, such as combined discrete element and shallow-water equation solvers, that can

explicitly track the transfer of momentum, progressive material entrainment, and rheological transformations

across the entire hazard cascade. To ensure these advanced modeling chains can be deployed across diverse

geographic domains without suffering severe performance degradation, ongoing research is increasingly

experimenting with self-supervised foundation models. By pre-training neural architectures on massive,

unlabeled, multi-modal Earth observation datasets, these foundation models extract highly generalized

spatiotemporal representations that can be transferred to new topographies or climatic zones with minimal local

fine-tuning

[6,7]

CONCLUSION

The management of geological hazards currently occupies a transitional space between reactive post-disaster

evaluation and proactive, real-time assessment. As infrastructure increasingly permeates volatile terrain, the

necessity of understanding the intricate physical fields driving these failures has never been more urgent. Hazards

do not emerge from isolated triggers. They evolve through complex hydro-mechanical and tectonic interactions

that dynamically degrade material stability. Technological interventions have significantly enhanced our ability

to track these destructive processes. We now integrate orbital synthetic aperture radar, aerial point clouds, and

distributed wireless sensor networks to construct continuous, high-resolution surveillance systems. Concurrently,

predictive analytics has transitioned from rigid heuristic equations to advanced data-driven algorithms. Deep

learning architectures successfully digest vast, high-dimensional datasets. They allow researchers to isolate

spatial anomalies and morphological patterns with unprecedented speed. Despite these observational and

computational achievements, significant bottlenecks obstruct the realization of intelligent forecasting systems.

Environmental artifacts profoundly limit data acquisition. Atmospheric noise and severe terrain distortions

regularly compromise remote sensing fidelity. More fundamentally, prevailing artificial intelligence models

operate predominantly as opaque systems. They map mathematical correlations instead of underlying physical

mechanisms. This structural opacity makes them highly vulnerable to domain shifts and environmental

interference, frequently resulting in unacceptable false alarm rates that undermine early warning networks.

Furthermore, existing evaluations often treat susceptibility as a static baseline, failing to accommodate the

cascading nature of complex disaster sequences. To transcend these limitations, the next generation of hazard

management must actively bridge the persistent divide between macroscopic statistics and localized

geomechanics. Incorporating partial differential equations directly into machine learning pipelines offers a

promising pathway. This forces computational outputs into physically plausible trajectories. Scaling these hybrid

systems alongside interactive digital twins will effectively transition regional static maps into dynamic virtual

testing environments. Ultimately, coupling physically grounded predictive architectures with self-supervised

foundation models will empower engineers to simulate and mitigate evolving multi-hazard disaster chains

globally.

REFERENCES

1. Zhang, S., Song, J., Zhan, L., Xu, Q., Zhang, L., & Chen, Y. (2025). Geological-environmental

hazard chain: Phenomena, process, and principle. In Science China Earth Sciences (Vol. 68,

Number 10, pp. 3085–3107). Science China Press. https://doi.org/10.1007/s11430-024-1602-x

2. Gupta, K., Satyam, N., Falak, & Rahim, S. (2026). Mapping Co-seismic Landslide Susceptibility:

An Overview of Current and Emerging Methods. Indian Geotechnical Journal.

https://doi.org/10.1007/s40098-026-01495-5

3. Zhang, X., Huang, S., & Gao, B. (2026). Full-waveform CNN–transformer neural network for

regional coseismic landslide susceptibility modeling: A case study of the 2022 Luding earthquake,

China. Engineering Geology, 362. https://doi.org/10.1016/j.enggeo.2025.108520

4. Wang, H., Sun, P., Mao, J., Wang, T., & Li, K. (2026). Dynamic-process-based quantitative hazard

assessment for earthquake-induced landslide clusters in the Northern Mountain of Tianshui.

Landslides. https://doi.org/10.1007/s10346-026-02776-z

Page 2271

www.rsisinternational.org

INTERNATIONAL JOURNAL OF LATEST TECHNOLOGY IN ENGINEERING,

MANAGEMENT & APPLIED SCIENCE (IJLTEMAS)

ISSN 2278-2540 | DOI: 10.51583/IJLTEMAS | Volume XV, Issue VI, June 2026

5. Huang, Y., Xu, C., Shao, X., He, X., Xiao, Z., Xu, X., Xie, Y., Nie, X., & Li, X. (2026). Landslide

assessment research in the three gorges reservoir area: A review of methodological advances and

future directions. In Bulletin of Engineering Geology and the Environment (Vol. 85, Number 2).

Springer Science and Business Media Deutschland GmbH. https://doi.org/10.1007/s10064-025-

04733-x

6. Zhou, M., Mei, G., Ma, Z., Xu, N., & Peng, J. (2026). AI for geosafety: motivations, advances,

challenges, and opportunities. In Bulletin of Engineering Geology and the Environment (Vol. 85,

Number 5). Springer Science and Business Media Deutschland GmbH.

https://doi.org/10.1007/s10064-026-04961-9

7. Alonso-Díaz, A., Fontes, M., Teixeira, A. C., Wdowinski, S., & Sousa, J. J. (2026). Multi-

Temporal InSAR and Machine Learning for Geohazard Monitoring: A Systematic Review with

Emphasis on Noise Mitigation and Model Transferability. In Remote Sensing (Vol. 18, Number

9). Multidisciplinary Digital Publishing Institute (MDPI). https://doi.org/10.3390/rs18091356

8. Okegbola, M. O., Owolabi, O., Sheng, Z., & Liu, Y. (2026). Integrating Multi-Temporal LiDAR

and AHP for Landslide Detection and Susceptibility Mapping in Prince George’s County,

Maryland. Journal of Safety Science and Resilience, 100311.

https://doi.org/10.1016/j.jnlssr.2026.100311

9. Naskar, D. C. (2026). Advances in Engineering Geophysics: Concepts and Applications (Vol. 9).

Springer Nature Switzerland. https://doi.org/10.1007/978-3-032-28000-8

10. Eze, K. N., Ilesanmi, O. O., Igah, G. C., Abidola, A. Q., Ojefia, F. E., & Adekoya, A. M. (2025).

Seismic rockfall risk assessments and mitigation strategies for transportation infrastructure in

high-risk regions. Discover Geoscience, 3(1). https://doi.org/10.1007/s44288-025-00182-x

11. Yao, X., Huang, M., Shi, F., & Yu, L. (2025). Intelligent and Sustainable Classification of Tunnel

Water and Mud Inrush Hazards with Zero Misjudgment of Major Hazards: Integrating Large-Scale

Models and Multi-Strategy Data Enhancement. Sustainability (Switzerland), 17(24).

https://doi.org/10.3390/su172411286

12. Xing, K., Li, S., Dou, J., Chen, F., Wang, L., Dong, A., Daud, H., & Zhang, L. (2026). Integrating

multi-geometry InSAR and explainable machine learning for mapping potential landslide areas in

reservoir regions. Remote Sensing of Environment, 341.

https://doi.org/10.1016/j.rse.2026.115442

13. Ma, T., Duan, Y., Duan, W., Wang, H., Tang, C., Wang, K., & Cheng, G. (2025). Research on

Intelligent Early Warning System and Cloud Platform for Rockburst Monitoring. Applied Sciences

(Switzerland), 15(20). https://doi.org/10.3390/app152011098

14. Kuang, Z., Li, S., Qiu, S., Huang, Y., & Chang, S. (2026). Identifying zones of rockburst-collapse

compound hazards in deep tunnels. Journal of Rock Mechanics and Geotechnical Engineering.

https://doi.org/10.1016/j.jrmge.2026.03.019

15. Zhang, Y., Sun, Y., Yan, Y., Wang, S., & Ge, L. (2026). Research Status, Challenges and Future

Perspectives of Geological Hazard Monitoring Methods in Mining Areas. In Remote Sensing (Vol.

18, Number 9). Multidisciplinary Digital Publishing Institute (MDPI).

https://doi.org/10.3390/rs18091333

16. Zhang, X., Duan, X., Khalil, U., Muhammadi, M. A., Elmannai, H., Algarni, A. D., & Kucher, D.

E. (2026). Multi-sensor deep learning approach for landslide susceptibility mapping under climate

and anthropogenic pressures. Physics and Chemistry of the Earth, 143.

https://doi.org/10.1016/j.pce.2026.104305

17. Goli Mokhtari, L. (2026). Optimizing graph neural networks for rockfall susceptibility mapping:

a feature selection and hazard prediction approach. Natural Hazards, 122(1).

https://doi.org/10.1007/s11069-025-07855-3

18. Ehsan, M., Anees, M. T., Bakar, A. F. B. A., & Ahmed, A. (2025). A review of geological and

triggering factors influencing landslide susceptibility: artificial intelligence-based trends in

mapping and prediction. In International Journal of Environmental Science and Technology (Vol.

22, Number 16, pp. 17347–17382). Springer Nature. https://doi.org/10.1007/s13762-025-06741-

Page 2272

www.rsisinternational.org

INTERNATIONAL JOURNAL OF LATEST TECHNOLOGY IN ENGINEERING,

MANAGEMENT & APPLIED SCIENCE (IJLTEMAS)

ISSN 2278-2540 | DOI: 10.51583/IJLTEMAS | Volume XV, Issue VI, June 2026

19. Liao, Z., Wu, J., Deng, Z., Lü, Q., & Yu, Y. (2026). Tectonic and microtopographic controls on

aspect-dependent distribution of rainfall-induced shallow landslides. Geomorphology, 511,

110416. https://doi.org/10.1016/j.geomorph.2026.110416

20. Li, Y., Yang, H., Zheng, L., Qi, W., Chongyi, E., & Gao, T. (2026). Formation mechanism and

stability assessment of the Attabad landslide-dammed lake along the China-Pakistan Karakoram

Highway using remote sensing technology. Landslides. https://doi.org/10.1007/s10346-026-

02697-x

21. Zhang, J. W., Xu, W. Z., Song, Z. X., Zhang, Y., Yang, J. N., Dong, X. K., Wu, S. K., Bai, X. Y.,

Zhang, S. L., & Li, X. (2026). An overview of coal-rock-gas compound dynamic disasters in China

and the structural classification of risk-prone areas. In Journal of Central South University (Vol.

33, Number 3, pp. 997–1027). Central South University. https://doi.org/10.1007/s11771-026-

6221-6

22. Shi, W., Chen, G., Meng, X., Bian, S., Jin, J., Wu, J., Huang, F., & Chong, Y. (2023). Formation

and Hazard Analysis of Landslide Damming Based on Multi-Source Remote Sensing Data.

Remote Sensing, 15(19). https://doi.org/10.3390/rs15194691

23. Bukhari, S. N. U. H., Ahmed, K. S., Ikram, N., Usmani, N. A., Sadiq, S., & Gardezi, S. A. H.

(2026). Post-earthquake landslide hazard evolution: Spatio-temporal analysis of active fault zone

in Western Himalayas. Journal of Mountain Science, 23(5), 2088–2110.

https://doi.org/10.1007/s11629-025-0204-1

24. Ran, S., Zhou, W., Weng, Y., Chi, F., Wang, D., Cao, X., & Ma, G. (2026). Causative environment-

informed active landslide detection with InSAR monitoring: Empowering reservoir slope hazard

management. International Journal of Applied Earth Observation and Geoinformation, 146.

https://doi.org/10.1016/j.jag.2025.105079

25. Malik, B. A., & Koner, R. (2024). Comprehensive review of the monitoring and sensing system

in slopes with a special focus on the mining sector. In Environmental Science and Pollution

Research (Vol. 31, Number 59, pp. 66588–66614). Springer. https://doi.org/10.1007/s11356-024-

35693-6

26. Nan, K., Luo, Y., Xu, Q., Zhao, B., & Song, H. (2025). Reactivation mechanisms of the ancient

Dahekou landslide in Hanzhong City, Shaanxi Province, China. Journal of Mountain Science,

22(4), 1245–1260. https://doi.org/10.1007/s11629-024-9130-x

27. Dolojan, N. L. J., Takahashi, T., Hashimoto, M., Shibayama, A., Nomura, R., Terada, K., &

Moriguchi, S. (2025). Integrated multihazard study combining qualitative and quantitative

analyses of floods, landslides, and debris flows: A case study on the impacts of Typhoon Yun-

Yeung on Iwaki City, Fukushima. International Journal of Disaster Risk Reduction, 127.

https://doi.org/10.1016/j.ijdrr.2025.105647

28. Moradi, S., Heinze, T., Budler, J., Gunatilake, T., Kemna, A., & Huisman, J. A. (2021). Combining

site characterization, monitoring and hydromechanical modeling for assessing slope stability. Land,

10(4). https://doi.org/10.3390/land10040423

29. Lee, D. H., Lee, S. R., & Park, J. Y. (2025). Coupled modeling framework for proactive design of

debris-flow barrier placements. Scientific Reports, 15(1). https://doi.org/10.1038/s41598-025-

15290-4

30. You, Q., Wang, F., Ma, H., Fu, Z., & Feng, Y. (2026). Lithological controls on clustered landslides:

a case study of landslides triggered by Typhoon Gaemi (2024) in Zixing, Hunan Province, China.

Landslides. https://doi.org/10.1007/s10346-026-02762-5

31. Wang, J. H., & Xu, W. J. (2025). Slope stability and failure dynamics of rainfall-induced landslide:

Algorithm and applications. Computers and Geotechnics, 177.

https://doi.org/10.1016/j.compgeo.2024.106919

32. Zhang, L., Wang, Y., & Liu, J. (2024). Mechanical interpretation of retrogressive failure in mild

dip rock slopes induced by slope cutting and rainfall: a case study in Yunnan, China. Bulletin of

Engineering Geology and the Environment, 83(1). https://doi.org/10.1007/s10064-023-03503-x

33. Li, K., Sun, P., Wang, H., & Ren, J. (2024). Insight into failure mechanisms of rainfall induced

mudstone landslide controlled by structural planes: From laboratory experiments. Engineering

Geology, 343. https://doi.org/10.1016/j.enggeo.2024.107774

Page 2273

www.rsisinternational.org

INTERNATIONAL JOURNAL OF LATEST TECHNOLOGY IN ENGINEERING,

MANAGEMENT & APPLIED SCIENCE (IJLTEMAS)

ISSN 2278-2540 | DOI: 10.51583/IJLTEMAS | Volume XV, Issue VI, June 2026

34. Li, Y., Hu, X., Zhang, H., Zheng, H., & Li, N. (2025). Displacement prediction and failure

mechanism analysis of rainfall-induced colluvial landslides. Journal of Hydrology, 660.

https://doi.org/10.1016/j.jhydrol.2025.133361

35. Fusco, F., Bordoni, M., Tufano, R., Vivaldi, V., Meisina, C., Valentino, R., Bittelli, M., & De Vita,

P. (2022). Hydrological regimes in different slope environments and implications on rainfall

thresholds triggering shallow landslides. Natural Hazards, 114(1), 907–939.

https://doi.org/10.1007/s11069-022-05417-5

36. Du, H., Wan, Y., Wang, D., Cao, B., & Liu, G. (2025). Research on deformation characteristics

and mechanisms of an open pit coal mine landslide event in extremely cold region. Scientific

Reports, 15(1). https://doi.org/10.1038/s41598-025-27509-5

37. An, H., & Ouyang, C. (2026). From sky to ground: A multi-source observations-physical model

cascade for debris flow forecasting in inaccessible alpine regions. Engineering Geology, 369.

https://doi.org/10.1016/j.enggeo.2026.108810

38. Li, Z., Chen, J. P., Cao, C., Zhang, W., Xu, P., Zheng, L. J., Jing, Y., & Hu, J. Y. (2025). Temperature

and deformation response under the influence of continuous typhoons in seasonal permafrost

rainfall-induced landslide evolution. Natural Hazards, 121(11), 13093–13116.

https://doi.org/10.1007/s11069-025-07312-1

39. Weng, M. C., Chao, W. A., Yang, C. M., Teo, T. A., Li, K. W., & Hsiao, Y. T. (2026). Assessing

earthquake-induced displacements of large-scale landslide by in-situ monitoring and discrete

element analysis. Engineering Geology, 368. https://doi.org/10.1016/j.enggeo.2026.108775

40. Minghui, M., Hongsheng, M., Ningfei, L., Dong, S., Liang, Q., Mingli, L., Xu, Z., & Wulin, M.

(2025). The research on the evolution characteristics of multi-geneses Dashui gully high-locality

landslide in the upper reaches of Minjiang river. Bulletin of Engineering Geology and the

Environment, 84(7). https://doi.org/10.1007/s10064-025-04389-7

41. He, N., Qu, X., Yang, Z., Xu, L., & Gurkalo, F. (2023). Disaster Mechanism and Evolution

Characteristics of Landslide–Debris-Flow Geohazard Chain Due to Strong Earthquake—A Case

Study of Niumian Gully. Water (Switzerland), 15(6). https://doi.org/10.3390/w15061218

42. Zhao, X., Fen, W., Dai, Z., Jiao, W., Geng, J., Xiong, Q., & Zhang, N. (2026). Dynamic rockfall

risk assessment using multi-source data fusion and 3D simulation: a case study of Jiaohua rock.

Scientific Reports, 16(1). https://doi.org/10.1038/s41598-026-36769-8

43. Zhang, M., Xing, A., Li, K., Zhuang, Y., Chang, W., & Liu, Y. (2023). Debris flows in Lebai gully

along the Yarlung Tsangpo River in Tibet: characterization, causes, and dynamic prediction of

potential debris flows. Environmental Earth Sciences, 82(1). https://doi.org/10.1007/s12665-022-

10694-1

44. Wang, F., Ren, Q., Wu, K., & Zheng, C. (2025). Formation and Evolutionary Mechanisms of Slope

Instability Disasters at Exposed Surfaces of Deep-Large Collapse Pit Induced by Underground

Mining Activities. Indian Geotechnical Journal. https://doi.org/10.1007/s40098-025-01319-y

45. Xia, K., Chen, C., Zheng, Y., Zhang, H., Liu, X., Deng, Y., & Yang, K. (2019). Engineering

geology and ground collapse mechanism in the Chengchao Iron-ore Mine in China. Engineering

Geology, 249, 129–147. https://doi.org/10.1016/j.enggeo.2018.12.028

46. Yang, Y., Li, J., Shi, W., Yang, C., & Yan, L. (2024). Numerical Investigation of Failure

Characteristics and Debris Movement in the Madaling Landslide Using Coupled FDM–DEM.

Geotechnical and Geological Engineering, 42(4), 2745–2765. https://doi.org/10.1007/s10706-

023-02703-2

47. Zeng, T., Jin, B., Liu, Y., Glade, T., Wang, F., Yin, K., & Peduto, D. (2024). Cut slope hazard

analysis and management based on a double-index precipitation threshold: a case study in the

Miaoyuan area (Eastern China). Environmental Earth Sciences, 83(24).

https://doi.org/10.1007/s12665-024-11987-3

48. Yang, F., & Hasan, M. (2026). Comprehensive risk mapping and management of rainfall-induced

slope failures in urbanizing subtropical regions. Scientific Reports, 16(1).

https://doi.org/10.1038/s41598-026-49114-w

Page 2274

www.rsisinternational.org

INTERNATIONAL JOURNAL OF LATEST TECHNOLOGY IN ENGINEERING,

MANAGEMENT & APPLIED SCIENCE (IJLTEMAS)

ISSN 2278-2540 | DOI: 10.51583/IJLTEMAS | Volume XV, Issue VI, June 2026

49. Fan, J., Zhang, Y., Peng, Y., Xing, Z., Yuan, K., Cui, J., Liu, B., & Zhou, W. (2025). Study on the

disaster mechanism and prevention technology of embankment slip-collapse after extreme rainfall

in the loess area. Scientific Reports, 15(1). https://doi.org/10.1038/s41598-025-04920-6

50. Guo, J., Meng, F., & Guo, J. (2024). Gradual Failure of a Rainfall-Induced Creep-Type Landslide

and an Application of Improved Integrated Monitoring System: A Case Study. Sensors, 24(22).

https://doi.org/10.3390/s24227409

51. Fei, P., Yi, Q., Deng, M., Wang, B., Song, Y., & Liu, L. (2025). Study on the Deformation

Mechanism of Shallow Soil Landslides Under the Coupled Effects of Crack Development, Road

Loading, and Rainfall. Water (Switzerland), 17(8). https://doi.org/10.3390/w17081196

52. Zhao, W., Cong, P., Ma, X., Yi, M., Liu, C., Gao, J., & Zhang, Y. (2026). Landslide Hazard

Identification and Prediction in Complex Mountainous Areas Using Ascending and Descending

Orbits InSAR Technology. Sensors, 26(8). https://doi.org/10.3390/s26082455

53. Yu, Z., Zhan, J., Yao, Z., & Peng, J. (2024). Characteristics and mechanism of a catastrophic

landslide-debris flow disaster chain triggered by extreme rainfall in Shaanxi, China. Natural

Hazards, 120(8), 7597–7626. https://doi.org/10.1007/s11069-024-06518-z

54. Zhao, F., Hu, B., Wu, H., Li, W., Zhang, C., Hei, L., Wang, L., Wu, W., & Yang, L. (2026).

Landslide susceptibility assessment based on a frequency ratio-ensemble learning framework and

InSAR constraints. Advances in Space Research. https://doi.org/10.1016/j.asr.2026.05.039

55. Guan, X., Chen, P., Liu, H., Li, Z., Qiu, L., Zhu, C., Hong, Y., Yao, Y., & Li, Z. (2026). Dynamic

landslide susceptibility assessment in Zhouqu County, China, based on SBAS-InSAR and a

Stacking ensemble learning model. Advances in Space Research.

https://doi.org/10.1016/j.asr.2026.04.081

56. Zhou, H., Lu, M., Liu, F., Jiang, K., & Wang, Y. (2026). Landslide Hazard InSAR Monitoring and

Stability Evaluation Method for Large Open-Pit Mines: A Case Study of the Baiyinhua Open-Pit

Mine, China. Processes, 14(12), 1844. https://doi.org/10.3390/pr14121844

57. Cheng, X., Zeng, B., Tang, L., Yuan, J., Ai, D., & Huang, W. (2026). Multi-model landslide

integrated hazard detection using SBAS-InSAR and U-Net segmentation. Bulletin of Engineering

Geology and the Environment, 85(3). https://doi.org/10.1007/s10064-026-04780-y

58. Yang, A., Wang, Y., Yin, K., Gui, L., Liu, X., Liu, Y., Li, X., & Zhao, B. (2026). Progressive

deformation and post-failure residual displacement of the 16 October 2025 Tongren loess landslide,

Qinghai, China: insights for long-term extra-high voltage transmission line monitoring. Landslides.

https://doi.org/10.1007/s10346-026-02786-x

59. Guo, Z., Zeng, T., Zhang, Y., Yu, W., Wang, L., Guo, Z., & Glade, T. (2025). A novel hybrid model

integrating high resolution remote sensing and stacking ensemble techniques for landslide

susceptibility mapping: Application to event-based landslide inventory. Geomorphology, 486.

https://doi.org/10.1016/j.geomorph.2025.109886

60. Ahmadi, P., Valadan Zoej, M. J., Mokhtarzade, M., Kardan, N., Ahmadi, P., & Ghaderpour, E.

(2026). TLE-FEDformer: A Frequency-Domain Transformer Framework for Multi-Sensor Multi-

Temporal Flood Inundation Mapping. Remote Sensing, 18(6). https://doi.org/10.3390/rs18060895

61. Gurung, B., Chen, N., Hu, G., Khadka, N., Sapkota, L., Gouli, M. R., & Tian, S. (2026). Integrating

machine learning and numerical methods for enhanced landslide susceptibility and hazard

mapping in the Bhotekoshi watershed, central Nepal. Journal of Mountain Science.

https://doi.org/10.1007/s11629-025-9951-2

62. Dong, X., Li, S., Ma, R., Tian, W., Zhao, K., Xiang, H., Zhu, J., & Qiu, Y. (2025). Cloud-based

slope risk monitoring and early warning system for open-pit coal mines: a case study of Zhonglian

Runshi. Scientific Reports, 15(1). https://doi.org/10.1038/s41598-025-28190-4

63. Pornbunyanon, T., Jitpat, P., & Suwanno, P. (2026). Rainfall-conditioned landslide susceptibility

mapping using an FR–CART hybrid framework in southern Thailand. Progress in Disaster Science,

29. https://doi.org/10.1016/j.pdisas.2026.100530

64. Rezaie, F., Eghbali, M., Panahi, M., Shafapourtehrany, M., Batur, M., Moeini, H., Özener, H., &

Kalantari, Z. (2025). Advanced deep learning–based approaches for semantic segmentation in

precise landslide detection and susceptibility assessment. Ecological Informatics, 92.

https://doi.org/10.1016/j.ecoinf.2025.103447

Page 2275

www.rsisinternational.org

INTERNATIONAL JOURNAL OF LATEST TECHNOLOGY IN ENGINEERING,

MANAGEMENT & APPLIED SCIENCE (IJLTEMAS)

ISSN 2278-2540 | DOI: 10.51583/IJLTEMAS | Volume XV, Issue VI, June 2026

65. Kulkarni, S., Jasani, S., Kulkarni, S., & Wadmare, J. (2025). Datasets, Features, and Advanced

Techniques in Landslide Susceptibility Prediction: a Review. In SN Computer Science (Vol. 6,

Number 7). Springer. https://doi.org/10.1007/s42979-025-04395-2

66. Luo, W., Zheng, J., Miao, Y., & Gao, L. (2024). Raspberry Pi-Based IoT System for Grouting Void

Detection in Tunnel Construction. Buildings, 14(11). https://doi.org/10.3390/buildings14113349

67. Cacciuttolo, C., Atencio, E., Komarizadehasl, S., & Lozano-Galant, J. A. (2024). Internet of

Things Long-Range-Wide-Area-Network-Based Wireless Sensors Network for Underground

Mine Monitoring: Planning an Efficient, Safe, and Sustainable Labor Environment. In Sensors

(Vol. 24, Number 21). Multidisciplinary Digital Publishing Institute (MDPI).

https://doi.org/10.3390/s24216971

68. Kondo, A., & Matsushi, Y. (2026). Mechanisms of rainfall-induced shallow landslides regulated

by hydrological subsurface structures: Cases in granite and granodiorite areas in Northern

Abukuma Mountains, Japan. Engineering Geology, 366.

https://doi.org/10.1016/j.enggeo.2026.108692

69. Pei, X. (2022). Design of a Wireless Sensor Network-Based Risk Assessment Algorithm for Cave

Collapse. Mobile Information Systems, 2022. https://doi.org/10.1155/2022/8568181

70. Su, G., Huang, J., Jiang, J., Niu, W., & Liu, Z. (2026). Real-time microseismic monitoring and

intelligent early warning of collapse in the shallow-buried tunnel in relatively fractured hard rocks.

Engineering Failure Analysis, 185. https://doi.org/10.1016/j.engfailanal.2025.110364

71. Ramya, H. N., & Nagesh, M. A. (2026). Advanced generative adversarial network for rainfall-

induced landslide hazard mapping. Journal of the Brazilian Society of Mechanical Sciences and

Engineering, 48(4). https://doi.org/10.1007/s40430-025-06288-0

72. Wu, P., Yang, L., Li, W., Huang, J., & Xu, Y. (2023). Construction Safety Risk Assessment and

Early Warning of Nearshore Tunnel Based on BIM Technology. Journal of Marine Science and

Engineering, 11(10). https://doi.org/10.3390/jmse11101996

73. Li, W., Li, Y., Zhao, Y., & Xu, D. (2025). Optimization of monitoring and early warning technology

for mine water disasters using microservices and long short-term memory algorithm. Journal of

Supercomputing, 81(4). https://doi.org/10.1007/s11227-025-07033-z

74. Satyaningsih, R., Jetten, V., Ettema, J., Sopaheluwakan, A., Lombardo, L., & Nuryanto, D. E.

(2023). Dynamic rainfall thresholds for landslide early warning in Progo Catchment, Java,

Indonesia. Natural Hazards, 119(3), 2133–2158. https://doi.org/10.1007/s11069-023-06208-2

75. Feng, B., Zeng, P., Li, T., Sun, X., Zhu, X., Fan, X., & Xu, Q. (2025). Acceleration stage detection

and dynamic model selection for real-time landslide time-of-failure predictions. Acta Geotechnica.

https://doi.org/10.1007/s11440-025-02878-3

76. Pitilakis, K., Fotopoulou, S., Manakou, M., Karafagka, S., Petridis, C., Pitilakis, D., & Raptakis,

D. (2024). Towards seismic risk reduction of critical facilities combining earthquake early warning

and structural monitoring: a demonstration study. Bulletin of Earthquake Engineering, 22(14),

6893–6927. https://doi.org/10.1007/s10518-024-02046-0

77. Wang, C., Guo, M., Ji, Z., Gong, Y., Lu, C., Qi, P., Zhu, K., Zhu, C., Gao, M., & Chen, X. (2026).

Degradation-buckling-shear mechanism of earthquake-induced bedding rock landslides in the

Three-rivers basin: Insights from field investigation and shaking table test. Engineering Geology,

367. https://doi.org/10.1016/j.enggeo.2026.108722

78. Kuang, Z., Qiu, S., Li, S., Xiao, Y., Feng, G., Huang, Y., & Chang, S. (2025). A new early warning

method for rockbursts and compound rockburst–collapse hazards in deeply-buried tunnels based

on energy density. Tunnelling and Underground Space Technology, 164.

https://doi.org/10.1016/j.tust.2025.106828

79. Sharma, A., Sajjad, H., Roshani, & Rahaman, M. H. (2024). A systematic review for assessing the

impact of climate change on landslides: research gaps and directions for future research. In Spatial

Information Research (Vol. 32, Number 2, pp. 165–185). Springer Science and Business Media

B.V. https://doi.org/10.1007/s41324-023-00551-z

80. Abdelkader, M. M., Daoud, A. M. A., Abdel-Gawad, A. G., & Csámer, Á. (2025). Improving

landslide susceptibility mapping: The role of non-landslide sample selection in machine learning

Page 2276

www.rsisinternational.org

INTERNATIONAL JOURNAL OF LATEST TECHNOLOGY IN ENGINEERING,

MANAGEMENT & APPLIED SCIENCE (IJLTEMAS)

ISSN 2278-2540 | DOI: 10.51583/IJLTEMAS | Volume XV, Issue VI, June 2026

models. Remote Sensing Applications: Society and Environment, 40.

https://doi.org/10.1016/j.rsase.2025.101759

81. Wang, L., Wang, X., Zhang, H., Xu, Y., Chen, Q., Wu, J., Wu, W., & Ma, D. (2026). Regional

landslide susceptibility assessment via coupled modeling and sample optimization: A case study

of Dujiangyan, China. Journal of Hydrology: Regional Studies, 66.

https://doi.org/10.1016/j.ejrh.2026.103570

82. Zhang, Y., Yang, Z., Liu, J., Zeng, Y., Sun, Y., & Tan, J. (2026). Vulnerability of mountain road

networks to rainfall-induced landslide hazards. Journal of Mountain Science, 23(1), 188–202.

https://doi.org/10.1007/s11629-025-9673-5

83. Singh, G., Kumar, S., Karmakar, R., & Mishra, A. K. (2025). Rapid assessment of landslide

exposure to elements at risk for decision support in regional landslide forecasting. Discover

Geoscience, 3(1). https://doi.org/10.1007/s44288-025-00299-z

84. Joshi, B. R., Bhandary, N. P., Acharya, I. P., Niraj, K. C., & Bhandari, C. (2026). Integration of

information value with machine learning method for an enhanced predictive performance in

landslide susceptibility mapping. Discover Geoscience, 4(1), 152. https://doi.org/10.1007/s44288-

026-00517-2

85. Shen, L., Zhao, X., Liu, H., Sheng, H., Chi, Z., & Zhang, C. (2026). Enhancing landslide

susceptibility mapping analysis through neighborhood feature aggregation and interpretable

machine learning: Implications for disaster prevention in valley-type urban planning. Journal of

Environmental Management, 402. https://doi.org/10.1016/j.jenvman.2026.129117

86. Nguyen, H. H. D., Song, C. H., & Kim, Y. T. (2025). Semi-quantitative risk assessment: From

rainfall-induced landslides to the risk of persons in buildings. Bulletin of Engineering Geology

and the Environment, 84(9). https://doi.org/10.1007/s10064-025-04420-x

87. Dolojan, N. L. J., Moriguchi, S., Hashimoto, M., & Terada, K. (2021). Mapping method of rainfall-

induced landslide hazards by infiltration and slope stability analysis: A case study in Marumori,

Miyagi, Japan, during the October 2019 Typhoon Hagibis. Landslides, 18(6), 2039–2057.

https://doi.org/10.1007/s10346-020-01617-x

88. Musaib, A., Aparna, V., & Divya, P. V. (2026). Integrating TRIGRS and RAMMS for the

spatiotemporal prediction of rainfall induced landslides and landslide trajectory: a case study.

Natural Hazards, 122(3). https://doi.org/10.1007/s11069-025-07779-y

89. Chavali, R. V. P., Bayati, Z., Tremblay, A., Saeidi, A., Lévesque, Y., & Lambert, M. (2026).

Integrated Remote Sensing and Slope Stability Modeling for Back Analysis of Sensitive Clay

Landslides. Geotechnical and Geological Engineering, 44(2). https://doi.org/10.1007/s10706-026-

03628-2

90. Meng, G., Li, H., Wu, B., Liu, G., Ye, H., & Zuo, Y. (2023). Prediction of the Tunnel Collapse

Probability Using SVR-Based Monte Carlo Simulation: A Case Study. Sustainability

(Switzerland), 15(9). https://doi.org/10.3390/su15097098

91. Sahin, E. K., Demir, S., Ozturk, M., & Duzce, M. S. (2025). Geoscience in the era of generative

artificial intelligence (Geo[AI]-LSM): understanding the potential benefits of Google Gemini in

producing landslide susceptibility mapping. Advances in Space Research.

https://doi.org/10.1016/j.asr.2025.11.048

92. Utthasini, M., Ilampooranan, I., Singh, S. K., Kanga, S., Kumar, P., Halder, K., Pradhan, B.,

Srivastava, A. K., Chatterjee, R. S., Chakrabortty, R., Ali, T., & Meraj, G. (2026). Enhancing

landslide susceptibility mapping in the Himalayas: geospatial and machine learning with

explainable AI (XAI). Gondwana Research, 149, 262–290.

https://doi.org/10.1016/j.gr.2025.08.003

93. Ou, L., Zhang, Y., & Chen, Y. (2026). Intelligent Triggering of Safety Risk Warning in Metro

Tunnel Construction: A Two-Stage Framework Integrating Static and Dynamic Data. Buildings,

16(8). https://doi.org/10.3390/buildings16081550

94. Peng, Y., Li, Q., Wang, H., Liu, Z., & Xia, Y. (2026). Automated TBM tunnel-face collapse

detection system based on spatiotemporal fusion and attention mechanism with muck

characteristics and operational data. Tunnelling and Underground Space Technology, 174.

https://doi.org/10.1016/j.tust.2026.107688

Page 2277

www.rsisinternational.org

INTERNATIONAL JOURNAL OF LATEST TECHNOLOGY IN ENGINEERING,

MANAGEMENT & APPLIED SCIENCE (IJLTEMAS)

ISSN 2278-2540 | DOI: 10.51583/IJLTEMAS | Volume XV, Issue VI, June 2026

95. Yang, X., Yan, Y., Zhou, X., Zhu, L., Ma, M., Zhang, J., Chen, Y., & Gao, L. (2025). Risk of

Compound Typhoon Disaster Chains: Insights from Southeastern China. International Journal of

Disaster Risk Science, 16(5), 870–887. https://doi.org/10.1007/s13753-025-00674-x

96. Sarna, S., Gutierrez, M., Mooney, M., & Zhu, M. (2022). Predicting Upcoming Collapse Incidents

During Tunneling in Rocks with Continuation Length Based on Influence Zone. Rock Mechanics

and Rock Engineering, 55(10), 5905–5931. https://doi.org/10.1007/s00603-022-02971-z

97. Liu, H., Ma, T., Lin, Y., Peng, K., Hu, X., Xie, S., & Luo, K. (2024). Deep Learning in Rockburst

Intensity Level Prediction: Performance Evaluation and Comparison of the NGO-CNN-BiGRU-

Attention Model. Applied Sciences (Switzerland), 14(13). https://doi.org/10.3390/app14135719

98. Yu, X., Zhang, X., Feng, X. T., Wang, F., Shi, M., Yang, C., Luo, H., Yang, Y., & Wang, Y. (2026).

Fracture evolution and differential mechanical response of surrounding rock in deep tunnel

excavation: A case study under complex geological conditions. Engineering Geology, 362.

https://doi.org/10.1016/j.enggeo.2025.108534

99. Calvi, A. (n.d.). Tunnels Collapses: Case Studies and Key Failure Mechanisms.

100. Capobianco, V., Palau, R. M., Solheim, A., Gisnås, K., Gilbert, G., Danielsson, P., & van der Keur,

P. (2024). The potential use of nature-based solutions as natural hazard mitigation measure for

linear infrastructure in the Nordic Countries. In Geoenvironmental Disasters (Vol. 11, Number 1).

Springer Science and Business Media Deutschland GmbH. https://doi.org/10.1186/s40677-024-

00287-4

101. Mondragón-Rodríguez, A., & Quesada-Román, A. (2026). Integrating geomorphological and

socio-spatial factors for landslide risk zonation in Atenas, Costa Rica. Discover Hazards, 2(1), 8.

https://doi.org/10.1007/s44475-026-00012-9

102. Jain, S. K., Jain, S. K., & Ahmad, R. (2025). Glacial lake outburst floods (GLOFs): causes,

modeling, and mitigation in Indian Himalayan region. GeoJournal , 90(2).

https://doi.org/10.1007/s10708-025-11326-4

103. Gong, Z., Fu, X., & Cheng, X. (2026). Excavation Stability of Tunnel Structure in Water-Rich

Areas. Applied Sciences, 16(2), 902. https://doi.org/10.3390/app16020902

104. Nguyen, H. S., Khau, T. L., & Huynh, T. T. (2025). Investigation of Natural and Human-Induced

Landslides in Red Basaltic Soils. Water (Switzerland), 17(9). https://doi.org/10.3390/w17091320

105. He, X., Xu, C., Huang, Y., Yang, Q., Wang, W., & Xiao, Z. (2026). Evolution of Geological Hazard

Prevention in the Three Gorges Reservoir Area: Insights from Planning and Implementation.

Earthquake Research Advances, 100464. https://doi.org/10.1016/j.eqrea.2026.100464

106. Ling, Q., Chen, W., Zhang, Q., Wang, L., Qu, W., Shu, B., Ren, J., Guo, Q., Qu, J., Li, W., Yang,

C., Chen, Y., & Sha, J. (2026). A PINN-based hybrid model with APIDr for interpretable and cross-

scenario landslide displacement prediction. Journal of Hydrology, 674.

https://doi.org/10.1016/j.jhydrol.2026.135493

107. Xu, L., Zhanping, S., Xiaole, S., Lianbaichao, L., Tong, W., & Shengyuan, F. (2026). Damage

quantification and response surface prediction of blasting-induced slope stability based on DFN.

Bulletin of Engineering Geology and the Environment, 85(1). https://doi.org/10.1007/s10064-

025-04723-z

108. Li, Y., Zhang, C., & Zhang, G. (2025). The development characteristics and formation modes of

rainstorm-triggered flash flood disasters in the Hengduan Mountains. Journal of Geographical

Sciences, 35(3), 619–640. https://doi.org/10.1007/s11442-025-2337-0