Available via license: CC BY 4.0
Content may be subject to copyright.
Date of publication xxxx 00, 0000, date of current version xxxx 00, 0000.
Digital Object Identifier 10.1109/ACCESS.2024.0429000
Simplified Physical Stability Assessment of
Chilean Mine Waste Storage Facilities Using GIS
and AI: Application in the Antofagasta Region
GABRIEL HERMOSILLA1, GABRIEL VILLAVICENCIO2, GIOVANNI COCCA-GUARDIA1(Member,
IEEE), VICENTE APRIGLIANO2, MANUEL SILVA1, (Member, IEEE), JUAN CARLOS QUEZADA3,
PIERRE BREUL4, VINICIUS MINATOGAWA2, JAIME MORALES5
1Escuela de Ingeniería Eléctrica, Pontificia Universidad Católica de Valparaíso, Valparaíso 2374631, Chile
2Escuela de Ingeniería de Construcción y Transporte, Pontificia Universidad Católica de Valparaíso, Valparaíso 2362804, Chile
3ICUBE, UMR 7357, CNRS, INSA de Strasbourg, Strasbourg, France
4Département Génie Civil, Polytech Clermont, Institut Pascal UMR CNRS 6602, Université Clermont Auvergne, Av. Blaise Pascal SA 60206-63178 Aubière,
CEDEX, 63000 Clermont Ferrand, France
5Escuela de Ingeniería Química. Pontificia Universidad Católica de Valparaíso, Avenida Brasil 2180, Valparaíso 2362854, Chile
Corresponding author: Gabriel Hermosilla (e-mail: gabriel.hermosilla@pucv.cl).
This research was funded and supported by the Vice-Rectorate for Research, Creation, and Innovation (VINCI) at Pontificia Universidad
Católica de Valparaíso (Chile), through the Associative Research Project DI (under grant: 039.301/2024), Centennial Project 2024 (under
grant: 039.310/2024), and FONDECYT under Grant 1240573, and ANID Doctorado Nacional 2023-21232328.
ABSTRACT Chile’s mining industry, a global leader in copper production, faces challenges due to increasing
volumes of mining waste, particularly Waste Rock Dumps (WRD) and Leaching Waste Dumps (LWD). The
National Service of Geology and Mining (SERNAGEOMIN) requires assessment of the physical stability
(PS) of these facilities, but current methods are hindered by data scarcity and resource constraints. This
study proposes a simplified evaluation methodology using first-order parameters from open-access data. By
integrating Geographic Information Systems (GIS) and Artificial Intelligence (AI)—utilizing models like
YOLOv11 and convolutional neural networks—we automate the detection and characterization of WRD
and LWD from satellite imagery, extracting critical parameters for PS assessment. This approach reduces
analysis time and minimizes human error. Validated in the Antofagasta Region, Chile’s primary mining area,
we identified and evaluated 70 WRD and 54 LWD. The results demonstrate the effectiveness of prioritizing
deposits based on potential risk, enhancing SERNAGEOMIN’s capacity for supervision. The successful
application suggests scalability to other mining regions and adaptability to different facility types, including
tailings storage facilities. This work offers a practical tool to improve safety and risk management in the
mining industry, addressing critical challenges in PS evaluation under current regulatory constraints.
INDEX TERMS Artificial intelligence, closure plan, geographical information systems, mine waste storage
facilities, physical stability assessment, Sentinel-2 satellite imagery, YOLOv11.
I. INTRODUCTION
THE mining industry in Chile, a world leader in copper
production, generates a significant amount of mining
waste, which is expected to progressively increase in the
future. By 2026, Chile is projected to generate approximately
1,000 million tons of mining waste annually [1]. These mate-
rials are disposed of and stored in large structures known as
Mine Waste Storage Facilities (MWSF).
Among these waste types are Waste Rock Dumps (WRD)
and Leaching Waste Dumps (LWD), both of which are ex-
pected to increase in size and quantity in the near future.
To date, more than 792 WRDs and LWDs zones have been
identified [2], distributed across mining operations in regions
I, II, III, IV, V, MR, VI, VII, and XI of Chile, reaching
the following sizes: i) WRDs vary between 50 and 500 m
in height, with projections up to 1,000 m in large mining
projects, occupying thousands of hectares (Figure 1.a); ii)
LWDs have heights ranging from 10 to 120 m and occupy
areas covering hundreds of hectares (Figure 1.b).
To date, these types of MWSF have demonstrated adequate
mechanical behavior under static and seismic conditions.
However, in Chile, mining companies are legally obligated
to assess and ensure their physical stability (PS), even after
closure, to safeguard human life and health and to protect the
VOLUME 11, 2023 1
This article has been accepted for publication in IEEE Access. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2025.3530856
This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
Author et al.: Preparation of Papers for IEEE TRANSACTIONS and JOURNALS
FIGURE 1: MWSF under construction, located in the northern regions of Chile. a) Waste rock dumps (WRD), El Soldado,
Nogales, Chile [3]. b) Leaching waste dumps (LWD), BioCobre, Copiapó, Chile [4].
environment, in accordance with national legislation [5], [6].
A potential failure in the PS of these facilities could result in
to environmental disasters, significant material damage, and
even loss of human life, as has been reported in other mining
countries [7], [8].
The National Service of Geology and Mining (SERNA-
GEOMIN), a government entity responsible for the super-
vision and oversight of MWSF, faces a growing challenge
due to the need to ensure the long-term PS of these mining
facilities. To guide companies and inspectors in assessing
PS, SERNAGEOMIN issued the "Methodological Guide for
the Evaluation of the Physical Stability of Residual Mining
Facilities" [9], hereinafter referred to as the PS Guide. This
tool standardizes and regulates the procedures for evaluating
the PS of MWSF. The evaluation process outlined int the PS
Guide requires a range of physical parameters and environ-
mental data, including both the design project and the closure
plan for these facilities [5], [10]. Additionally, information
from periodic monitoring conducted during their operation
and construction phase is essential. Using this data, the PS
Guide employs a matrix analysis to assess stability condition
for various potential failure mechanisms. However, since its
official adoption in 2018, the effective implementation of
this evaluation and oversight tool has faced several technical
barriers. These include the lack of a national MWSF registry,
insufficient number of inspectors, difficulties in obtaining
critical geotechnical parameters (e.g. foundation soils for
MWSF), and stringent administrative requirements imposed
by national regulations.
In this context, to address the current challenges associated
with applying the PS Guide, this work proposes a simplified
evaluation methodology. This methodology uses first-order
parameters to estimate the global PS condition of WRD and
LWD, using state-of-the-art technologies such as geographic
information systems (GIS) and artificial intelligence (AI).
GIS enables the integration and analysis of multiple layers of
geospatial data, such as geology, topography, seismicity, and
precipitation, complemented by specific geotechnical data on
the waste [11], [12]. AI facilitates the automatic detection
and geometric characterization of these deposits, reducing
analysis times and minimizing human errors. This methodol-
ogy, including proposed matrices and operational approach,
will be validated through a case study in the Antofagasta
Region, a key mining area in Chile with major operations
(e.g., CODELCO and Antofagasta Minerals), to demonstrate
its effectiveness. By integrating geospatial data layers and
AI-based detectors, the methodology characterizes WRD and
LWD and assesses critical stability parameters. This enables
the development of a prioritization ranking for MWSF based
on PS levels, facilitating effective supervision and oversight
by SERNAGEOMIN. Furthermore, the methodology has po-
tential applications in other mining regions and for different
types of MWSF, such as tailings storage facilities (TSF).
II. RELATED WORK
The evaluation of the PS of MWSF has been a fundamental
topic in research, due to the significant risks these facilities
represent for human safety and the environment. Tradition-
2VOLUME 11, 2023
This article has been accepted for publication in IEEE Access. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2025.3530856
This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
Author et al.: Preparation of Papers for IEEE TRANSACTIONS and JOURNALS
ally, this evaluation relies on a series of factors and variables
considered in geotechnical modeling to analyze possible fail-
ure modes (e.g., slope instability, static liquefaction, among
others) during the design phase, as well as requiring peri-
odic in-situ monitoring of each variable. However, economic
resource constraints and the large scale of modern mining
operations demand more efficient and scalable approaches.
The following are some relevant studies in this field.
A. EVALUATION OF MWSF PS
In recent years, remote sensing techniques have transformed
the evaluation of the PS of MWSF in engineering. For in-
stance, satellites such as Sentinel-2 and Landsat have played a
crucial role in providing high-resolution images and enabling
continuous monitoring of the large areas occupied by MWSF.
A recent study utilized the Google Earth Engine platform to
process satellite images, optimizing information extraction
from the deposits through image processing techniques and
cloud cover reduction [13]. Another study demonstrated how
machine learning algorithms can estimate parameters such as
moisture content, facilitating operational periodic control in
thickened tailings deposits [14].
In addition, remote monitoring technologies such as syn-
thetic aperture radar (SAR) and interferometric radar (In-
SAR) have gained prominence in recent years for evaluating
the PS of MWSF. These techniques enable the monitoring
of surface deformation, which can serve as an early warning
indicator for potential physical instability scenarios. Another
example is the use of the Small Baseline Subset (SBAS)
technique to monitor surface deformations in mining areas
[15] [16]. These techniques have proven to be useful tools
to complement classic geotechnical evaluations, offeringcon-
tinuous and real-time information.
B. ARTIFICIAL INTELLIGENCE IN THE DETECTION AND
ANALYSIS OF MWSF
The integration of AI in the detection and analysis of MWSF
has achieved remarkable progress in recent years, particularly
for TSF. Object detection and segmentation models such as
YOLO and Faster R-CNN have significantly improved the
automatic identification of these mining facilities through
satellite imagery [2], [17]. These models not only identify
TSF with high precision but also enable the extraction of geo-
metric and geotechnical parameters related to PS evaluation.
Advances in deep learning have automated processes that
previously required manual inspections, thereby increasing
the efficiency of periodic monitoring.
Moreover, techniques for generating synthetic data have
been developed to complement real datasets and improve the
ability of AI models to learn and generalize across different
scenarios. A recent example includes the use of generative
adversarial networks (GANs) for creating synthetic data and
parameters used in training regression models to estimate PS
[18]. These innovations are facilitating a more automated and
continuous analysis of TSF, a critical step toward improving
risk management and mitigation for such mining facilities.
C. APPLICATIONS OF GIS IN MINING
The use of GIS has been fundamental for management and
planning in the mining sector. GIS facilitates the integration
and analysis of essential geospatial data, such as topography,
geology, and hydrology, enabling informed decision-making
regarding the location of new facilities and the environmental
management of mining operations. In the context of PS eval-
uation, GIS provides a platform for overlaying multiple data
layers, helping to identify critical spatial patterns influencing
the PS of MWSF.
Several studies have implemented GIS methods to monitor,
evaluate, or predict environmental or physical risks associated
with TSF. For instance, [19], [20] applied GIS to identify
historical and spatial patterns of incidents at such mining
facilities and their relationship with climate variables, such
as rainfall. In this regard, [20] used satellite images and GIS
techniques to assess the risk level of TSF based on factors
such as proximity to urban areas and the type of ore ex-
tracted. With an environmental focus, [21] applied a GIS-
based method to assess the level of sediment storage and
transport in rivers impacted by mining activity in Bolivia. [19]
developed a GIS methodology to predict failures, using the
Brumadinho case in Brazil as an example.
The combination of AI and GIS is proposed as an inno-
vative strategy to improve the detection and evaluation of
MWSF. GIS integrates multiple geographical data layers,
while AI models analyze large volumes of satellite data,
automating the identification of potential risks, obtaining
critical parameters for PS evaluation, and optimizing resource
management. This integrated approach prioritizes deposits
with higher instability risks, offering a scalable solution to
enhance safety in remote or hard-to-access areas.
III. PHYSICAL STABILITY OF WRD AND LWD
For the evaluation of the PS of WRD and LWD, SERNA-
GEOMIN as the supervisory body of the State of Chile,
establishes that mining companies must use the classification
system outlined in the PS Guide [9]. The objective is to facil-
itate progressive and safe closure processes for these types of
MWSF. In general terms, the PS Guide was developed based
on the framework of the systems initially proposed by [7]
and later modified by [8] but including technical adaptations
and criteria applicable to the context of the Chilean mining
industry.
Currently, the PS Guide requires the evaluation of 20 fac-
tors or parameters that define PS, organized into 8 evaluation
matrices (Table 1). To reflect the degree of importance on the
PS of WRD and LWD, each factor is assigned a condition and
a numerical rating, defined based on the analysis of the me-
chanical behavior observed in these types of MWSF [9]. The
sum of these numerical ratings allows WRD and LWD to be
evaluated and classified in terms of their PS level for the fail-
ure mechanisms being assessed: slope instability/foundation
instability and static liquefaction. A higher numerical rating
corresponds to a lower PS level and a greater associated
failure occurrence potential (FOP) (Figure 2). It is important
VOLUME 11, 2023 3
This article has been accepted for publication in IEEE Access. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2025.3530856
This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
Author et al.: Preparation of Papers for IEEE TRANSACTIONS and JOURNALS
TABLE 1: Evaluation Matrix for PS Assessment
ID Matrix Factor or attribute
M1 Foundation conditions Geomechanical characteristics of foundation soil, shape foundation, and MWSF ty-
pology.
M2 Geometric configuration Overall height, overall fill slope angle, maximum thickness lift, and maximum overall
thickness.
M3 Constructive background Construction method and loading rate.
M4 Material quality and in-situ state Particle size distribution, Atterberg limits, and in-situ state condition.
M5 Instrumentation and monitoring Piezometrics levels, drainage systems, accelerometers, and inclinometers.
M6 Regional setting for the closure condition Seismicity and precipitation (rain and/or snow).
M7 Stability performance PS observed during the operational phase.
M8 Degree of implementation of measures to ensure
PS in the operation and closure stages
Verification of works and actions implemented as indicated in the approved closure
plan.
FIGURE 2: Evaluation scheme of WRD and LWD according to the PS Guide [9].
to note that this evaluation system was defined by a committee
of national and international experts and subsequently applied
in various national mining operations for calibration.
Although this classification system uses various param-
eters from technical reports that mining companies are re-
quired to submit to SERNAGEOMIN (such as the design
project and closure plan), and basic operational control of
these MWSF (geometric configuration, construction back-
ground, instrumentation and monitoring, stability perfor-
mance, among others), it has been evidenced that this infor-
mation is limited or, in many cases, inaccessible due to poor
management and lack of updates of the required information.
This has hindered the effective application of the PS Guide
for both mining companies and SERNAGEOMIN.
This context has affected SERNAGEOMIN’s supervisory
role, particularly in the case of WRD and LWD, facing var-
ious obstacles: i) the absence of a national registry for these
types of MWSF; ii) a shortage of inspectors nationalwide; iii)
the large number of variables required to evaluate PS, many
of which lack quantifiable information; iv) technical and
economic difficulties in obtaining critical parameters, such
as soil conditions, geometric configuration, and geotechnical
characteristics; v) strict administrative requirements imposed
by current national regulations. These challenges highlight
the need for a more efficient and updated system that enhaces
both supervision and compliance with safety standards in the
management of MWSF.
Given the limitations of the PS Guide for evaluating the PS
of WRD and LWD, we propose a simplified system to over-
come data scarcity by using open-source information. This
proposal relies on geospatial data such as surface geology,
active faults, seismicity, topography, and precipitation. In ad-
dition, it considers general knowledge of the geomechanical
properties of WRD and LWD [11], [12] , complemented by
advanced GIS and AI technologies for automated data pro-
cessing, while generally maintaining the evaluation criteria
defined in the PS Guide [9].
4VOLUME 11, 2023
This article has been accepted for publication in IEEE Access. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2025.3530856
This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
Author et al.: Preparation of Papers for IEEE TRANSACTIONS and JOURNALS
IV. PROPOSED PS EVALUATION USING SIMPLIFIED
MATRICES
The proposed simplified system, developed with the partic-
ipation of expert engineers from SERNAGEOMIN, reorga-
nizes the eight original matrices from the PS Guide into four
key matrices. These matrices focus on foundation conditions,
geometric configuration, geomechanical quality of the ma-
terial, and the regional setting of the deposits (Figure 3).
The excluded matrices (M3: constructive background; M5:
instrumentation and monitoring; M7: stability performance;
and M8: degree of implementation of measures to ensure
PS in the operation and closure stages) rely on data that,
to date, are not regularly recorded or monitored during the
construction and operation phases of these types of mining
facilities. Thus, simplifying and adapting the PS Guide is
essential to enable more frequent and accessible evaluations,
especially for MWSF that lack records and are not regularly
monitored.
Each matrix uses open-access data, enabling scalable and
practical implementation. The sum of the scores from each
matrix results in the Physical Global Stability Index (PSGI),
a qualitative measure that facilitates the prioritization of de-
posits for SERNAGEOMIN supervision, focusing on those
with the highest potential risk. The simplified matrices are
presented below.
A. MATRIX M1’: FOUNDATION CONDITIONS
Foundation conditions include key factors for the PS of
WRD and LWD, especially for instabilities associated with
foundation failures, such as rotational failure, non-circular
rotation, wedge failure, base translation, liquefaction, and toe
failure [7]–[9]. In this context, the parameters included in this
matrix are related to the characteristics of the foundation soil:
overburden type and topography.
The foundation soil types of WRD and LWD, in a first ap-
proximation, can be defined based on their geological origin
(rock and residual or sedimentary soils). A complementary
parameter, widely used to define seismic site characterization,
is the stiffness upper 30 m of the soil profile through an
equivalent shear wave velocity Vs30 [22]. The foundation soil
topography has been linked to the average overall foundation
slope angle (α) and the foundation shape of the area where
the WRD and LWD are located, as shown in Figure 4.2a and
Figure 4.2b, respectively. The parameter αcorresponds to the
angle between the toe of the WRD and LWD slopes and their
upper intersection with the natural terrain.
FIGURE 3: Simplified evaluation system of global PS and inspection prioritization of WRD and LWD.
FIGURE 4: Geometric parameters and typology of WRDs and LWDs. a) Geometric parameters of WRDs and LWDs. b) Basic
MWFs types. Adapted from [8].
VOLUME 11, 2023 5
This article has been accepted for publication in IEEE Access. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2025.3530856
This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
Author et al.: Preparation of Papers for IEEE TRANSACTIONS and JOURNALS
TABLE 2: Evaluation matrix M1’: foundation conditions factors
Factor Parameter and condition Category Ratings
Geomechanical characteristics
of foundation soil
Type I: rock and Vs30 >900 (m/s) Very competent 0
Type II: soft or fractured rock, very dense or hard soils
and Vs30 ≥500 (m/s)
Competent 100
Type III: dense or very stiff soils and Vs30 ≥350 (m/s) Intermediate 200
Type III: medium-dense or firm to stiff soils and Vs30
≥180 (m/s)
Weak 300
Type IV: loose to medium-dense or soft to firm soils and
Vs30 <180 (m/s)
Very weak 400
Average overall foundation
slope angle (α)
5°Flat 0
5 – 15°Gentle 50
15 – 25°Moderate 100
25 – 32°Steep 150
> 32°Very steep 200
No information – 200
Shape foundation/MWSF ty-
pology
Valley fill or cross valley fill Confined 0
Sidehill fill Moderately confined 100
Heaped fill or ridge crest fill Unconfined 200
No information – 200
TABLE 3: Evaluation matrix M2’: geometric configuration factors
Factor Parameter and condition Ratings
Overall height (m)
< 100 0
100 – 250 50
250 – 500 100
> 500 200
No information 200
Volume (m3)
Small < 10,000,000 0
Medium 10,000,000 – 100,000,000 50
Large 100,000,000 – 1,000,000,000 100
Very large > 1,000,000,000 200
No information 200
Overall fill slope angle (β°)
Gentle < 26 0
Moderate 26 – 34 50
Steep 34 – 39 100
Very steep > 39 200
No information 200
Lifts
With lifts 0
Without lifts 200
No information 200
The structure of matrix M1’ is presented in Table 2. To
obtain the evaluation parameters contained in matrix M1’, the
following open-source layers were used: i) Geomechanical
characteristics of foundation soil: geological map of Chile,
scale 1:1,000,000 [23] and Vs30: Global Vs30 mosaic map
viewer [24]; ii) foundation soil slope and shape founda-
tion/MWSF typology: ground slope (slope geoprocessing of
digital elevation models [25], and open-access satellite im-
ages (Sentinel-2).
B. MATRIX M2’: GEOMETRIC CONFIGURATION
The geometric configuration of WRD and LWD defines the
PSGI level they present, associated with the generation of the
following failure mechanisms: edge slumping, plane failure,
and rotational failure [7]–[9]. The geometric parameters (Fig-
ure 4) are directly related to the shape and size of these types
of MWSF. For simplified evaluation purposes, the proposed
matrix M2’ includes the following geometric parameters:
overall height, volume, overall fill slope angle, and stepped
construction (single or multiples lifts), as presented in Table 3
These geometric parameters are obtained from the processing
and analysis of satellite images (Sentinel-2). For the single
lift slope angle, based on topographic surveys from WRD
and LWD design projects, [12] propose a range of 34-39ºfor
WRD. Meanwhile, [11] suggest a range of 33-35ºfor LWD.
6VOLUME 11, 2023
This article has been accepted for publication in IEEE Access. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2025.3530856
This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
Author et al.: Preparation of Papers for IEEE TRANSACTIONS and JOURNALS
TABLE 4: Evaluation matrix M3’: geomechanical quality factors of waste rock and leached ore material
Parameter and condition Category Ratings
Very coarse grained (WRD):
- % fines (passing # 200 sieve; < 0.075 mm) < 10%
- % great than 3" > 50%
- Plasticity: N/A
- Intact strength: very strong to strong (UCS > 100 MPa)
- Durability: very high to high
High 0
Mixed grained (LWD):
- % fines (passing # 200 sieve; < 0.075 mm), between 10-25%
- % great than 3", between 25-50%
- Low plasticity: LL < 35%; PI < 10%
- Intact strength: medium strong (UCS: 50 - 100 MPa)
- Durability: medium high
Moderate 100
No information – 200
TABLE 5: Evaluation matrix M4’: regional setting for the closure condition of WRD and LWD
Parameter Description Condition Category Ratings
Seismicity
Peak Ground Acceleration (PGA)
or A0, based on seismic zones of
Chile (INN, 2012)
A0≤0.2g Low 0
A0: 0.2g – 0.3g Moderate 200
A0> 0.4g High 400
Distance to active and
potentially-active conti-
nental faults
Distance from the trace of faults
(polygon defined at 300m distance
from the fault trace)
Outside the polygon Low 0
Inside the polygon High 200
Precipitation Average annual rainfall and/or
snowfall
Rainfall < 100 (mm)
Snowfall < 10 (cm) Very low 0
Rainfall: 100 – 350 (mm)
Snowfall: 10 – 35 (cm) Low 100
Rainfall: 350 – 1000 (mm)
Snowfall: 35 – 100 (cm) Moderate 200
Rainfall: 1000 – 2000 (mm)
Snowfall: 100 – 200 (cm) High 300
Rainfall > 2000 (mm)
Snowfall > 200 (cm) Very high 400
TABLE 6: ILP in WRD and LWD, based on PSGI and FOP
PSGI FOP IPL
< 600 Very low I: very reduced
[600 – 1200[ Low II: reduced
[1200 – 1800[ Moderate III: moderate
[1800 – 2200[ High IV: priority
[2200 – 2800] Very high V: very priority
C. MATRIX M3’: GEOMECHANICAL QUALITY OF THE
MATERIAL
The geomechanical properties of WRD and LWD, such as
gradation (particle size distribution, PSD), plasticity of fines
(Plasticity Index, PI, and Liquid Limit, LL), intact strength
(Uniaxial Compressive Strength, UCS), and durability (Slake
Durability Index, SDI), among others, are considered key
factors influencing overall PS [7], [8]. Although these MWSF
are not usually characterized during the construction and
operation phases through in-situ and laboratory tests, recently
[11], [12] proposed databases with variation ranges for PSD,
plasticity, intact strength, and durability of waste rock and
leached ore materials generated by the copper mining in-
dustry in Chile. Based on this information, it is possible to
globally define the geomechanical quality depending on the
type of identified MWSF, which will be used to apply matrix
M3’ (Table 4).
D. MATRIX M4’: REGIONAL SETTING FOR THE CLOSURE
CONDITION
To consider the regional setting where WRD and LWD are
located, in the national context for the closure condition of
these MWSF, the following parameters have been consid-
ered: seismicity, precipitation (rain and/or snow), and active
and potentially active continental faults, which form matrix
M4’ (Table 5). The open-source data layers used for matrix
M4’ are as follows: i) seismicity: seismic zones of Chile
[26] created in the context of this work using open-source
GIS; ii) geological faults: database of active and potentially
active continental faults in Chile at a 1:25,000 scale [27]; iii)
precipitation: layers from the National Information System of
Groundwater of Chile [28].
E. CALCULATION OF THE PSGI
The PSGI is calculated by summing the scores obtained in
each matrix (Eq. 1). The PSGI provides a qualitative approx-
imation of the PS global of the MWSF, linked to FOP and
with an inspection prioritization level (IPL) (Table 6). This
classification enables the prioritization of inspector actions
and resource allocation by SERNAGEOMIN, focusing on
WRD and LWD that present the highest PSGI.
PSGI =X
rating
M1′+X
rating
M2′+X
rating
M3′+X
rating
M4′(1)
VOLUME 11, 2023 7
This article has been accepted for publication in IEEE Access. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2025.3530856
This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
Author et al.: Preparation of Papers for IEEE TRANSACTIONS and JOURNALS
FIGURE 5: Overview of the Proposed Methodology automate the PS evaluation of the global PS for WRD and LWD.
V. METHODOLOGY
To automate the PS evaluation process using the proposed
simplified matrices, remote sensing, GIS, and artificial intelli-
gence techniques will be used. This methodology is organized
into several key stages, including:
•Creation of the WRD and LWD database.
•Automatic detection and segmentation of WRD and
LWD deposits.
•Integration of GIS layers.
•Application of AI models to the proposed matrices.
•Automatic evaluation of the global PS through simpli-
fied matrices.
•Calculation of the PSGI and categorization of WRD and
LWD deposits.
Figure 5 illustrates the overall flow of the proposed
methodology, applied to the Antofagasta region in Chile,
which represents both state-owned and private companies
involved in large- and medium-scale mining. The flow shows
how PS is evaluated through the detection and segmentation
of WRD and LWD using satellite images processed by an
AI model. The detected deposits are georeferenced within
GIS layers to obtain the parameters needed for evaluation
in the simplified matrices, followed by categorization and
prioritization.
A. DATABASE: WRD AND LWD
To date, SERNAGEOMIN does not have an official registry
of WRD and LWD. The Atlas de Faenas Mineras, an official
SERNAGEOMIN databasewith information on mining oper-
ations approved until 2020, was used to define the search area
for WRD and LWD, potentially including nearby MWSF.
Using this approach, images of the mining areas available
in the Atlas for the Antofagasta Region were downloaded
using the Sentinel-2 satellite API, considering 16 instances
between 2023 and 2024 for each mining area at the maximum
available resolution of 10x10 m/px. Labeling of WRD and
LWD was conducted using Label-Studio software [29], with
input from a geotechnical expert.
The images were stored in the following filename format to
maintain the geolocation of the area: "[x_min,y_min,x_max,y_max]
- (date) - band.png", where (x,y) corresponds to (longitude,
latitude), "(date)" refers to the date of the mining area, and
"band" is the name of the band downloaded from Sentinel-
2, which in this case corresponds to RGB. A total of 2,500
images were downloaded, of which 1,733 were labeled after
discarding those with clouds or noise. The labeled dataset
was divided into a training set (75%) and a test set (25%),
ensuring that the first 12 instances of each mining area were
used for training, while the last 4 instances were reserved for
testing. The details of the downloaded and labeled database
are provided in Table 7.
B. AUTOMATIC DETECTION AND SEGMENTATION OF WRD
AND LWD
For detecting and segmenting WRD and LWD, the YOLOv11
model was chosen for its efficiency and accuracy in satellite
8VOLUME 11, 2023
This article has been accepted for publication in IEEE Access. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2025.3530856
This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
Author et al.: Preparation of Papers for IEEE TRANSACTIONS and JOURNALS
TABLE 7: Distribution of the database used for training the
YOLOv11 segmentation model
Total (100%) Train (75%) Test (25%)
Images 1,733 1,268 465
WRD 5,206 3,853 1,353
LWD 2,532 1,873 659
TABLE 8: Evaluation metrics results of the trained YOLOv11
segmentation model
Class Metric Precision Recall mAP50 mAP50-95
All bbox 0.998 0.608 0.804 0.773
segm 0.995 0.606 0.802 0.680
WRD bbox 0.999 0.614 0.807 0.776
segm 0.993 0.610 0.803 0.665
LWD bbox 0.996 0.716 0.857 0.832
segm 0.992 0.713 0.855 0.762
image analysis [30] [2]. YOLOv11 was selected due to its
superior performance in processing high-resolution satellite
images, its capability to handle complex object shapes, and
its balance between speed and precision, making it well-
suited for large-scale mining applications. Fine-tuning was
performed using the “YOLOv11x-seg” weights on the labeled
database, which allowed the detections to be specifically
adapted to the deposits of interest. This fine-tuning was vali-
dated using the test set, employing metrics such as precision,
recall, and mAP to evaluate model performance.
The results of these metrics, presented in Table 8, indicate
that the YOLOv11 model successfully detected all labeled
deposits, achieving a precision of up to 0.998 for bounding
box detection and 0.995 for segmentation. Despite this high
precision, slightly lower recall values indicate some addi-
tional false positives, which may be attributed to the presence
of visually similar features in the satellite images. These
false positives may affect the overall evaluation process by
increasing the need for additional verification steps to ensure
accuracy. Notably, the model performed exceptionally well in
detecting LWD, with mAP50 values of 0.857 for bounding
boxes and 0.855 for segmentation. The high precision of
YOLOv11 in detecting WRD and LWD ensures high con-
fidence in the automatic evaluation process of global PS.
This precision enables the effective integration of segmented
masks with GIS elevation and slope layers for a comprehen-
sive analysis of the geospatial characteristics of the WRD and
LWD.
C. INTEGRATION OF GIS LAYERS
Once the YOLOv11 model was trained, it was used to detect
and segment WRD and LWD in mining areas of interest,
which were then processed alongside GIS layers. The GIS
layers used in the matrix evaluation are described in Annex
A. These layers have a resolution of 12.5x12.5 m/px, which
differs from the satellite image resolution of 10x10 m/px used
in the database. This difference in resolution occurs because
GIS layers are typically optimized for regional-scale analysis,
whereas satellite images are captured at a finer resolution
to provide detailed local information. The detections made
by YOLOv11 are adjusted to match the GIS layer size be-
fore processing the segmented deposits, which might slightly
impact the accuracy of parameter extraction due to scaling
adjustments.
Thus, the bounding boxes obtained from YOLOv11 detec-
tion are used to crop WRD and LWD deposits, preserving
their aspect ratio relative to the original image. Preserving the
aspect ratio helps maintain the accuracy of georeferencing by
avoiding distortions during parameter extraction. The masks
resulting from the segmentations performed by YOLOv11 are
used to georeference the deposits within the GIS layers to
obtain the parameters needed for PS evaluation through the
proposed simplified matrices.
D. APPLICATION OF AI MODELS TO THE PROPOSED
MATRICES
To evaluate the proposed matrices, most parameters are ob-
tained from GIS layers. However, matrices M1’ and M2’
require visual and geometric parameters, such shape founda-
tion/MWSF typology (degree of confinement) and the pres-
ence of lifts in WRD and LWD. For these cases, two mul-
ticlass classification models were trained: one to determine
shape foundation/MWSF typology (degree of confinement)
and another to identify the presence or absence of lifts (as
shown in Figure 4b).
A ResNet18 neural network [31], a variant of the residual
network architecture, was used for these classifiers. ResNet18
was chosen due to its ability to mitigate the vanishing gradi-
ent problem, its efficiency in training on relatively smaller
datasets, and its proven effectiveness in learning complex
patterns, making it well-suited for this specific classification
task. A new database was generated from the segmented
masks of the initial database, manually labeled with an ex-
pert’s assistance, to train these models. A balanced set of
classes was obtained and divided into training (80%), vali-
dation (10%), and test (10%) sets, totaling 5,940 labeled de-
posits for the presence of lifts and 10,800 MWSF with degree
of confinement levels. The class distribution is detailed in
Table 9.
The evaluation metrics for the degree of confinement clas-
sifier showed outstanding results, with precision ranging from
0.92 to 0.99 and an F1-score from 0.96 to 0.99 across all
classes (see Table 10). These metrics indicate the model’s
high reliability in accurately classifying the degree of confine-
ment, which is crucial for understanding the structural stabil-
ity of the deposits. For the lift classifier, results also showed
high precision and recall, reaching values between 0.96 and
1.00 across all classes and datasets. These results, detailed in
Table 11, demonstrate excellent classification capabilities for
distinguishing between MWSF with and without lifts.
VOLUME 11, 2023 9
This article has been accepted for publication in IEEE Access. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2025.3530856
This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
Author et al.: Preparation of Papers for IEEE TRANSACTIONS and JOURNALS
TABLE 9: Details of the database used to train the lifts classification model and confinement classification model
Multiclass classifier model Category Total (100%) Train (80%) Val (10%) Test (10%)
Lifts classifier With lifts 2,970 2,376 297 297
Without lifts 2,970 2,376 297 297
Confinement classifier
Valley fill 2,700 1,889 405 406
Sidehill fill 2,700 1,889 405 406
Heaped fill 2,700 1,889 405 406
TABLE 10: Evaluation metrics results of the trained confine-
ment classifier model
Metric Valley fill Sidehill fill Heaped fill
Precision 0.920 0.953 0.985
Recall 0.931 0.919 0.990
F1-score 0.967 0.945 0.980
TABLE 11: Evaluation metrics results of the trained lifts
classifier model
Metric With lift Without lift
Precision 1.000 0.997
Recall 0.997 1.000
F1-score 0.998 0.998
E. AUTOMATIC EVALUATION OF GLOBAL PS THROUGH
SIMPLIFIED MATRICES
The automatic evaluation of the global PS of WRD and LWD
integrates detection, segmentation, and geospatial analysis,
supported by the trained YOLOv11 model (Table 12). The
resulting masks are georeferenced within GIS layers to extract
parameters for each evaluation matrix. Trained multiclass
classifiers are essential for extracting geometric parameters,
such as the presence of lifts or the degree of confinement. This
automated approach relies on consistent data quality, which
may limit accuracy in areas with incomplete or outdated
information. Once a WRD or LWD is evaluated, the PSGI
is calculated using the following expression:
PSGI =
4
X
i=1
ni
X
j=1
mscore
ij (2)
Where mscore
ij the score obtained after evaluating factor j
within matrix i, and nirepresents the total number of factors
corresponding to matrix ifor each WRD or LWD detected.
VI. EXPERIMENTS AND RESULTS
Two experiments were conducted to validate the proposed
methodology for evaluating the PS WRD and LWD. The
first experiment applied the methodology to a case study at
the Radomiro Tomic Division (RTD), Chuquicamata South
Mine (Antofagasta region, Chile), involving satellite imagery
acquisition using Sentinel-2, segmentation with YOLOv11,
and parameter extraction through GIS georeferencing. Mul-
ticlass classification models assessed structural features like
the degree of confinement and presence of lifts. The second
FIGURE 6: Detection and segmentation results of the case
study. (a) Unprocessed mining area. (b) Mining area seg-
mented into WRDs and LWDs by YOLOv11.
experiment applied the methodology to all available images
from the Antofagasta region to generate a comprehensive
PS assessment of WRD and LWD, with the aim of creating
a proposed official registry to assist in the inspection and
supervision by SERNAGEOMIN.
A. EXPERIMENT 1: AUTOMATIC EVALUATION OF PS AT
THE RTD
Figure 6 shows a satellite image of the mining area under
study, along with the corresponding detection and segmenta-
tion performed by YOLOv11. The model demonstrated effec-
tive detection and segmentation of MWSF in the mining area,
accurately delineating the WRD and LWD boundaries and
minimizing false positives. For RTD Chuquicamata South
Mine, 3 WRD zones and 3 LWD zones were detected, cover-
ing a total area of 1327.03 (ha) and 986.53 (ha), respectively.
Following detection, the segmented masks were georef-
erenced within GIS layers, including geological units, ac-
tive and potentially active continental faults, Vs30 parame-
ter, ground slope, seismic zone, and precipitation layers, to
extract specific parameters needed for evaluating each of the
proposed matrices (Figures A.3 and A.2). By integrating these
GIS data, a detailed PS assessment of the detected WRD and
LWD was performed using simplified evaluation matrices,
ensuring that each aspect of the physical environment was
accurately considered.
Subsequently, multiclass classification models were used
to determine the degree of confinement (shape founda-
tion/MWSF typology) and the presence of lifts in each de-
10 VOLUME 11, 2023
This article has been accepted for publication in IEEE Access. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2025.3530856
This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
Author et al.: Preparation of Papers for IEEE TRANSACTIONS and JOURNALS
TABLE 12: Materials required for automatic evaluation of the global PS for WRD and LWD
Matrix Factors Materials Scores
M’1
Geomechanical characteristics of foundation soil Mask + GIS Geology layer + GIS VS30 layer mscore
11
Average overall foundation slope angle (α) Mask + GIS ground slope layer mscore
12
Shape foundation and MWSF typology Mask + Confinement classifier mscore
13
M’2
Overall height Mask + GIS ground elevation layer mscore
21
Volume Mask + GIS ground elevation layer mscore
22
Overall fill slope angle (β) Mask + Lift classifier mscore
23
Lifts Mask + Lift classifier mscore
24
M’3 Geomechanical quality Mask and material database mscore
31
M’4
Seismicity Mask + GIS seismic zones layer mscore
41
Distance to geological faults Mask + GIS active and potentially-active conti-
nental faults mscore
42
Precipitation Mask + GIS precipitation layer mscore
43
TABLE 13: Results of the evaluation of global PS for each detected WRDs and LWDs in the case study mining area
Rating Mask ID 1 Mask ID 2 Mask ID 3 Mask ID 4 Mask ID 5 Mask ID 6
mscore
11 200 100 100 300 300 300
mscore
12 50 0 100 50 0 50
mscore
13 200 200 200 200 200 200
mscore
21 200 200 200 200 200 200
mscore
22 200 200 200 200 200 200
mscore
23 5000000
mscore
24 0 200 200 200 200 200
mscore
31 200 200 0 0 200 0
mscore
41 200 200 200 200 0 0
mscore
42 200 200 200 200 0 0
mscore
43 000000
PSGI 1300 1300 1200 1350 1300 1150
FOP Moderate Moderate Moderate Moderate Moderate Low
IPL III III III III III II
tected WRD and LWD in the studied mining area. The results
are presented in Figure 9. This workflow ensured a consistent
evaluation of each WRD and LWD using the proposed matri-
ces, with a summary of the scores obtained for each matrix
shown in Table 13. Detailed ratings for each evaluation ma-
trix, along with the resulting PSGI score, FOP, and associated
IPL, are also included.
B. EXPERIMENT 2: APPLICATION OF THE METHODOLOGY
IN THE ANTOFAGASTA REGION, CHILE
In the second experiment, the methodology was expanded
to cover the entire Antofagasta region, using all available
satellite images to conduct a comprehensive assessment of
existing WRDs and LWDs. By processing these images, an
Excel file was generated containing detailed information on
each mining area, including central geolocation and detected
WRD and LWD. This Excel file serves as a valuable tool
for stakeholders, such as SERNAGEOMIN, to improve mon-
itoring and management efforts by providing centralized,
accessible data on deposit locations and characteristics. As
SERNAGEOMIN currently lacks an updated registry, our
work provides a critical foundation for their oversight efforts.
The generated Excel file also formed the basis for project-
ing the detected WRD and LWD on a map of the Antofagasta
TABLE 14: Results of the evaluation of global PS of WRDs
and LWDs in the Antofagasta region
FOP IPL MWFS WRD LWD
Very low I: Very reduced 0 0 0
Low II: Reduced 18 13 5
Moderate III: Moderate 105 57 48
High IV: Priority 1 0 1
Very high V: Very priority 0 0 0
region, offering a clear visual representation of the deposits
and their corresponding locations (Figure 10a). Table 14
provides details on the number of detected mining facilities
and MWSF, including 70 WRD and 54 LWD deposits, along
with their categorization based on FOP and IPL (Figure 10b).
This categorization is crucial for prioritizing inspections or in-
terventions, allowing SERNAGEOMIN to allocate resources
efficiently and focus on areas of higher risk. By providing
an updated registry, our work supports SERNAGEOMIN’s
efforts to improve their regulatory oversight and management
capabilities across the region.
VOLUME 11, 2023 11
This article has been accepted for publication in IEEE Access. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2025.3530856
This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
Author et al.: Preparation of Papers for IEEE TRANSACTIONS and JOURNALS
(a) Geological units. (b) Global Vs30 values.
(c) Ground elevation. (d) Ground slope.
FIGURE 7: Detection and segmentation of WRDs and LWDs in Chuquicamata, South Mine (CODELCO), georeferenced within
GIS layers.
12 VOLUME 11, 2023
This article has been accepted for publication in IEEE Access. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2025.3530856
This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
Author et al.: Preparation of Papers for IEEE TRANSACTIONS and JOURNALS
FIGURE 8: Detection and segmentation of WRDs and LWDs in Chuquicamata, South Mine (CODELCO), georeferenced within
GIS layers: a) Seismic zone. b) Precipitations.
FIGURE 9: Detection, lift classifier and confinement classifier results shown as: Class, lift presence and typology. (a) LWD,
Without lift, Heaped fill. (b) LWD, With lift, Heaped fill. (c) WRD, With lift, Heaped fill. (d) WRD, With lift, Heaped fill. (e)
LWD, With lift, Heaped fill. (f) WRD, With lift, Heaped fill
VII. CONCLUSION
The present work introduces an innovative solution to the
challenges in evaluating the physical stability of WRD and
LWD in Chile. The proposal of simplified matrices repre-
sents a significant advancement, enabling SERNAGEOMIN
to overcome current data and resource limitations. By fo-
cusing on first-order parameters obtainable through open-
access information, these matrices simplify and streamline
the evaluation process without compromising accuracy.
The integration of GIS technologies and AI techniques
has proven highly effective in the automatic detection and
characterization of WRD and LWD. The developed method-
ology utilizes detection and classification models, such as
YOLOv11 and convolutional neural networks, to analyze
satellite images and extract critical parameters for PS evalua-
tion. This approach not only reduces analysis times and mini-
mizes human errors but also allows for significant scalability
and adaptability.
The validation of the methodology in the Antofagasta Re-
gion, a key mining area in Chile, enabled the identification
and evaluation of 70 WRD and 54 LWD. The results ob-
tained demonstrate the capability of the proposed approach
VOLUME 11, 2023 13
This article has been accepted for publication in IEEE Access. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2025.3530856
This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
Author et al.: Preparation of Papers for IEEE TRANSACTIONS and JOURNALS
FIGURE 10: Georeferencing of LWDs and WRDs detected and evaluated in the Antofagasta region. IPL: a) LWD, b) WRD.
to prioritize deposits according to their potential risk, thereby
facilitating the supervision and oversight work of SERNA-
GEOMIN. The successful application in this region suggests
that the methodology can be extended to other mining areas
and adapted to different types of facilities, including tailings
storage facilities.
VIII. ACKNOWLEDGEMENTS
The authors gratefully acknowledge the valuable contribution
of SERNAGEOMIN in validating the information used and
providing feedback for the evaluation methodology proposed
in this work. We also extend our sincere thanks to under-
graduate students from the School of Electrical Engineering
at Pontificia Universidad Católica de Valparaíso—Rodrigo
Pereira, Gonzalo Caballero, and Felipe Aravena—and to ge-
ographer Emilio Bustos, who contributed to the downloading
and labeling of the database.
IX. FUNDING
This research was funded and supported by the Vice-
Rectorate for Research, Creation, and Innovation (VINCI)
at Pontificia Universidad Católica de Valparaíso (Chile),
through the Associative Research Project DI (under grant:
039.301/2024), Centennial Project 2024 (under grant:
039.310/2024), FONDECYT under Grant 1240573, and
ANID Doctorado Nacional 2023-21232328.
REFERENCES
[1] COCHILCO, ‘‘Monitoreo del estado de los relaves mineros en chile.’’
[2] M. Silva, G. Hermosilla, G. Villavicencio, and P. Breul, ‘‘Automated
detection and analysis of massive mining waste deposits using sentinel-2
satellite imagery and artificial intelligence,’’ vol. 15, no. 20.
[3] G. Earth, ‘‘El soldado, nogales, chile.’’ Version Number: 7.3.6.9796
Published: Accessed: November 18, 2024.
[4] G. Earth, ‘‘BioCobre, copiapó, chile.’’ Version Number: 7.3.6.9796 Pub-
lished: Accessed: November 18, 2024.
[5] Ministerio de Minería, Chile, ‘‘Decreto n°41: Aprueba el reglamento de la
ley de cierre de faenas e instalaciones mineras [in spanish].’’
[6] Ministerio Secretaría General de la Presidencia, Chile, ‘‘Ley 19.300:
Aprueba ley sobre bases generales del medio ambiente [in spanish].’’
[7] British Columbia Ministry of Energy and Mines, ‘‘Investigation and design
manual interim guidelines, mine rock and overburden piles.’’ Place: British
Columbia, Canada.
[8] M. Hawley and J. Cunning, Guidelines for Mine Waste Dump and Stockpile
Design. CSIRO Press/Balkema.
[9] SERNAGEOMIN, ‘‘Guía metodológica para evaluación de la estabilidad
física de instalaciones mineras remanentes.’’
[10] Ministerio de Minería, Chile, ‘‘Ley n°20.551: Regula el cierre de faenas e
instalaciones mineras [in spanish].’’
14 VOLUME 11, 2023
This article has been accepted for publication in IEEE Access. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2025.3530856
This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
Author et al.: Preparation of Papers for IEEE TRANSACTIONS and JOURNALS
[11] A. Fourie, G. Villavicencio, J. Palma, P. Valenzuela, and P. Breul, ‘‘Eval-
uation of the physical stability of leaching waste deposits for the closure
stage,’’ in Proceedings of the 20th International Conference on Soil Me-
chanics and Geotechnical Engineering (ICSMGE 2022).
[12] J. C. Quezada and G. Villavicencio, ‘‘Characterisation of the internal
friction angle of waste rock material from large triaxial tests using the
contact dynamics method,’’ vol. 0, no. 0, pp. 1–20. Publisher: Taylor &
Francis _eprint: https://doi.org/10.1080/17480930.2024.2367977.
[13] E. Ferreira, M. Brito, R. Balaniuk, M. S. Alvim, and J. A. d. Santos,
‘‘Brazildam: A benchmark dataset for tailings dam detection,’’ in 2020
IEEE Latin American GRSS & ISPRS Remote Sensing Conference (LA-
GIRS), pp. 339–344.
[14] G. V. Arancibia, O. P. Bustamante, G. H. Vigneau, H. Allende-Cid, G. S.
Fuentelaba, and V. A. Nieto, ‘‘Estimation of moisture content in thickened
tailings dams: Machine learning techniques applied to remote sensing
images,’’ vol. 9, pp. 16988–16998.
[15] A. Pepe and F. Calò, ‘‘A review of interferometric synthetic aperture
RADAR (InSAR) multi-track approaches for the retrievalof earth’s surface
displacements,’’ vol. 7, no. 12.
[16] X. Wu, X. Qi, B. Peng, and J. Wang, ‘‘Optimized landslide susceptibility
mapping and modelling using the SBAS-InSAR coupling model,’’ vol. 16,
no. 16.
[17] D. Yan, G. Li, X. Li, H. Zhang, H. Lei, K. Lu, M. Cheng, and F. Zhu,
‘‘An improved faster r-CNN method to detect tailings ponds from high-
resolution remote sensing images,’’ vol. 13, no. 11.
[18] F. Pacheco, G. Hermosilla, O. Piña, G. Villavicencio, H. Allende-Cid,
J. Palma, P. Valenzuela, J. García, A. Carpanetti, V. Minatogawa, G. Suazo,
A. León, R. López, and G. Novoa, ‘‘Generation of synthetic data for
the analysis of the physical stability of tailing dams through artificial
intelligence,’’ vol. 10, no. 23.
[19] I. Atif, F. T. Cawood, and M. A. Mahboob, ‘‘Modelling and analysis
of the brumadinho tailings disaster using advanced geospatial analytics,’’
vol. 120, no. 7, pp. 405–414.
[20] S. K. Sarker, N. Haque, W. Bruckard, M. Bhuiyan, and B. K. Pramanik,
‘‘Development of a geospatial database of tailing storage facilities in
australia using satellite images,’’ vol. 303, p. 135139.
[21] S.-I. Balaban, K. A. Hudson-Edwards, and J. R. Miller, ‘‘A GIS-based
method for evaluating sediment storage and transport in large mining-
affected river systems,’’ vol. 74, no. 6, pp. 4685–4698.
[22] R. Verdugo, ‘‘Seismic site classification,’’ vol. 124, pp. 317–329.
[23] ‘‘Mapa geológico de chile,’’ Publisher: SERNAGEOMIN.
[24] D. C. Heath, D. J. Wald, C. B. Worden, E. M. Thompson, and G. M.
Smoczyk, ‘‘A global hybrid VS30 map with a topographic slope–based
default and regional map insets,’’ vol. 36, no. 3, pp. 1570–1584. _eprint:
https://doi.org/10.1177/8755293020911137.
[25] IDE, ‘‘DEM alos palsar región de antofagasta.’’
[26] Instituto Nacional de Normalización (INN), Earthquake Resistant Design
of Buildings, NCh 433.Of 96_Mod 2012 [in Spanish]. Instituto Nacional
de Normalización.
[27] V. Maldonado, M. Contreras, and D. Melnick, ‘‘A comprehensive database
of active and potentially-active continental faults in chile at 1:25,000
scale,’’ vol. 8, no. 1, p. 20.
[28] Sistema Nacional de Información de Aguas Terrestres (SNIAT), ‘‘Sistema
nacional de información de aguas terrestres.’’
[29] M. Tkachenko, M. Malyuk, A. Holmanyuk, and N. Liubimov, ‘‘Label
studio: Data labeling software.’’
[30] G. Jocher, J. Qiu, and A. Chaurasia, ‘‘Ultralytics YOLO.’’ original-date:
2022-09-11T16:39:45Z.
[31] K. He, X. Zhang, S. Ren, and J. Sun, ‘‘Deep residual learning for image
recognition.’’ _eprint: 1512.03385.
GABRIEL HERMOSILLA VIGNEAU was born in
Chillan, Chile, in 1982. He received the degree in
electronic engineering from the University of La
Frontera, Temuco, Chile, in 2007, and the Ph.D.
degree in electric engineering from the University
of Chile, Santiago, Chile, in 2012. Currently, he
is an Associate Professor with the School of Elec-
trical Engineering, Pontificia Universidad Católica
de Valparaiso (PUCV), Valparaiso, Chile. His main
areas of research interest are thermal face recogni-
tion, pattern recognition, computer vision, and deep learning.
GABRIEL VILLAVICENCIO ARANCIBIA was
born in Valparaíso, Chile in 1977. He received
the Ph.D. degree in civil engineering from Ecole
Doctorale des Sciences pour l’Ingénieur. Univer-
sité Blaise Pascal, Clermont II, France in 2009.
From 2004 to 2009, he was a temporary teaching
and research associates with the civil engineering
department, Polytech Clermont-Ferrand, France.
Since 2010 to date, he has been an assistant profes-
sor with the Construction and Transportation En-
gineering of the Pontificia Universidad Católica de Valparaíso, and geotech-
nical engineer with the LEPUCV laboratory. His research interests include
geotechnical engineering applications for tailings storage facilities (TSF),
wasted rocks dumps and leaching waste dumps. In addition, in topics such as
physical stability TSF, evaluation of liquefaction potential, slope stability and
geotechnical modeling of urban sites. He is an active member of the Chilean
Geotechnical Society.
GIOVANNI COCCA GUARDIA (Member, IEEE)
was born in Calama, Chile. He received his B.Sc.
degree in Engineering Sciences and his Civil Elec-
trical Engineering degree from the Pontificia Uni-
versidad Católica de Valparaíso, where he is cur-
rently pursuing an M.Sc. degree. His research in-
terests include Graph Neural Networks, Electrical
Power System, Generative AI and Efficient Learn-
ing AI.
VICENTE APRIGLIANO Professor at the School
of Construction and Transportation Engineering at
PUCV, Valparaíso, Chile. He completed a post-
doctoral fellowship at the Center for Sustainable
Urban Development (PUC-Chile) and holds a PhD
in Human Geography from Universität Tübingen,
Germany. With a Master’s in Transportation Engi-
neering (COPPE/UFRJ) and a Bachelor’s in Ge-
ography (UFRJ), his research focuses on transport
geography, economic geography, urban geography,
and urban mobility planning.
VOLUME 11, 2023 15
This article has been accepted for publication in IEEE Access. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2025.3530856
This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
Author et al.: Preparation of Papers for IEEE TRANSACTIONS and JOURNALS
MANUEL SILVA VEGA (Member, IEEE) was born
in Valparaíso, Chile, in 1990. He received his BS
in Electronics Engineering in 2022 and his MSc in
Electrical Engineering in 2023 from the School of
Electrical Engineering at the Pontificia Universi-
dad Católica de Valparaiso.
He is currently pursuing a Ph.D. in Electrical
Engineering in the Robotics and Vision Lab at
the same university, where his research focuses on
computer vision, remote sensing, machine learn-
ing, and robotics.
JUAN CARLOS QUEZADA was born in Val-
paraíso, Chile, in 1982. He completed his un-
dergraduate studies at the Pontificia Universidad
Católica de Valparaíso, earning a degree in Con-
struction Engineering in 2008. In 2009, he ob-
tained a Master’s degree from the University Blaise
Pascal (Clermont II) in France. From 2009 to 2012,
he pursued a doctoral degree through a collabora-
tive program between the University Montpellier II
in France and the French National Railway Com-
pany (SNCF), culminating in a Ph.D. in Mechanics and Civil Engineering.
Following his doctoral studies, Dr. Quezada worked as an engineer in the
Innovation and Research Department of SNCF in Paris, France, until 2014.
He then joined the LTDS Laboratory at École Centrale de Lyon, France, as
a postdoctoral researcher, where he focused on the numerical modeling of
dry-stone retaining walls. In 2015, he was appointed Associate Professor in
the Civil Engineering Department at INSA Strasbourg, France, a position he
continues to hold. His teaching portfolio encompasses topics in structural
engineering, geotechnics, and numerical methods. Dr. Quezada is an active
researcher affiliated with the ICube Laboratory, where his work centers
on numerical modeling using Discrete Element Method (DEM) and Finite
Element Method (FEM) in geomechanics, with applications to geomaterials.
PIERRE BREUL was born in France, in 1972.
He received the degree in civil engineering from
the University Blaise Pascal, Clermont-Ferrand,
France, in 1995, and the Ph.D. degree in civil
engineering from the University Blaise Pascal,
Clermont-Ferrand, France, in 1999. After having
been engineer for 5 years in a geotechnical en-
gineering company, he is currently Professor at
the National Polytechnic Institute of Clermont Au-
vergne University and head of the engineering
school Polytech Clermont. His research interests include geotechnics, soils
and granular materials mechanics, images and data analysis for soils recog-
nition and identification, infrastructure diagnosis and risk analysis.
VINICIUS MINATOGAWA was born in Ameri-
cana, São Paulo, Brazil, in 1990. He received the
Ph.D. degree in Mechanical Engineering and the
M.Sc. degree from the University of Campinas,
Brazil. Since 2021 to date, he has been an Asso-
ciate Professor with the Construction and Trans-
portation Engineering School of the Pontificia
Universidad Católica de Valparaíso. His research
interests include digital transformation projects,
aiming to integrate emerging technologies, data-
driven decision-making, and organizational innovation in engineering and re-
lated sectors. In addition, he is a Project Management Professional (PMP®)
certified by the Project Management Institute (PMP®Number: 3778535),
with a particular focus on agile projects. His work has been featured in high-
impact journals such as Automation in Construction, the Review of Manage-
rial Science, and the Journal of Manufacturing Technology Management.
JAIME WILSON MORALES SAAVEDRA , born
in the Maria Elena saltpeter office in 1973. Profes-
sional trained in teaching and research, with com-
munication and leadership skills. Experience in
teaching in the area of Thermodynamics and Unit
Operations, applying evaluation concepts through
competencies and strategies of new evaluation al-
ternatives. Development of research in the study of
Thermodynamic, Volumetric and Transport Prop-
erties applying Electrochemistry, Thermodynamic
Modeling in Systems Containing Electrolytes and Electrochemistry of
Lithium Batteries, chemical stability of tailings and slag valorization based
on circular economy.
APPENDIX A GIS CARTOGRAPHY
16 VOLUME 11, 2023
This article has been accepted for publication in IEEE Access. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2025.3530856
This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
Author et al.: Preparation of Papers for IEEE TRANSACTIONS and JOURNALS
FIGURE A.1: Geological map of Chile, scale 1:1,000,000. Antofagasta region, Chile. SERNAGEOMIN (2002).
FIGURE A.2: Precipitation mean 2023. Antofagasta region, Chile. SNIAT (2024)
VOLUME 11, 2023 17
This article has been accepted for publication in IEEE Access. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2025.3530856
This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
Author et al.: Preparation of Papers for IEEE TRANSACTIONS and JOURNALS
(a) (b)
(c) (d)
FIGURE A.3: Antofagasta region, Chile. a) Global Vs30 Mosaic Map Viewer. Heath et al. (2020). b) Digital elevation models.
Chile. IDE (2016). c) Slope geoprocessing of digital elevation models. Antofagasta region, Chile. IDE (2016). d) Seismic zones
of Chile. Antofagasta region. INN (2012).
18 VOLUME 11, 2023
This article has been accepted for publication in IEEE Access. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2025.3530856
This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
