Identification and Prospecting Prediction of Marine Geological

Anomalies Based on Deep Learning

Jian Wang

1,2

Yongfeng Wang

1,2*

, Jianxin Zhang

, Guotao Pang

1.2

, Yinji Ba

1,2

and Guiheng Wang

1,2

Yantai Center of Coastal Zone Geological Survey, China Geological Survey, Yantai Shandong 264000, China

Ministry of Natural Resources Observation and Research Station of Land-Sea,

Interaction Field in the Yellow River Estuary, Yantai Shandong 264000, China

Northwest bureau of China Metallurgical Geology Bureau, Xian Shanxi 710119, China

Keywords: Deep Learning, Identification of Marine Geological Anomalies, Prospecting Prediction, Method.

Abstract: The need for deep-sea mineral exploration has become more urgent as marine resources become increasingly

scarce. In order to effectively identify marine geological anomalies and improve the accuracy of prospecting

prediction, this study proposes a multi-modal data fusion method based on deep learning to achieve anomaly

identification and prospecting prediction. Based on the fusion of multi-source data such as ocean seismic

waves, magnetism, gravity, etc., the method adopts adaptive feature extraction technology, and uses a double-

branch prediction network to perform anomaly identification and mineral enrichment prediction. Finally, the

results of this paper show that the system performs well in multi-regional seabed geological data, among

which the enrichment of copper ore is 3.5%, and the enrichment of nickel ore is 1.2%. The comprehensive

analysis shows that the model and its integrated platform have strong robustness in the complex marine

environment, which can effectively improve the efficiency of mineral exploration.

1 INTRODUCTION

The exploitation of marine mineral resources has

become a key component of the global economy, but

due to the complexity of the deep-sea environment,

traditional geological exploration methods have

many challenges in terms of cost and efficiency. In

order to solve the above problems, some researchers

have proposed that deep-sea geological anomalies

can be identified based on the joint analysis of

magnetic and gravity data (Chao, Wang, et al. 2023),

but this method cannot effectively deal with the

complexity of multimodal data. Some researchers

have also proposed that a simple machine learning

system based on seismic wave data can be used to

accomplish related tasks. However, due to the

ignorance of the importance of spatial features and

multimodal fusion, the results are not accurate

enough. In addition, some researchers have also tried

to estimate mineral enrichment based on geochemical

data analysis (De, Cocchi et al. 2024), but due to the

scarcity of sampling sites, it is not possible to predict

the full range of mineral distribution (Kim, Golynsky,

et al. 2022). In order to solve these limitations, this

paper uses deep learning algorithms and integrates

multi-modal data, and at the same time, based on

adaptive feature extraction and dual-branch network

prediction, in order to significantly improve the

accuracy of marine geological anomaly identification

and mineral prediction. This method can cope with

the complexity of marine data and provide new ideas

for future deep-sea mineral exploration. This chapter

analyzes the complexity of the deep-sea environment

based on the current situation of marine geological

resources development, studies the shortcomings of

various marine geological anomaly identification and

prospecting prediction algorithms, and puts forward

the development and application advantages of deep

learning in the current environment, and helps people

realize that it is of practical significance to apply deep

learning algorithms to marine geological anomaly

identification and prospecting prediction.

Wang, J., Wang, Y., Zhang, J., Pang, G., Ba, Y. and Wang, G.

Identiﬁcation and Prospecting Prediction of Marine Geological Anomalies Based on Deep Learning.

DOI: 10.5220/0013535200004664

Paper published under CC license (CC BY-NC-ND 4.0)

In Proceedings of the 3rd International Conference on Futuristic Technology (INCOFT 2025) - Volume 1, pages 51-58

ISBN: 978-989-758-763-4

2 RELATED WORKS

2.1 Deep Learning and Multimodal

Data Fusion will Better Reflect the

Characteristics of Marine

Geological Anomalies

When dealing with complex geological data, a single

data source generally cannot fully reflect the

characteristics of marine geological anomalies, so

multimodal data fusion technology is necessary

(Kochetov, Shepelev, et al. 2023). The theory of

multimodal data fusion is based on the use of

heterogeneous data from different sources, such as

seismic waves, magnetism, gravity, etc., based on a

unified framework (Kusnida, Albab, et al. 2023), to

perform processing, so that the model can extract

information from multiple dimensions and capture

the internal correlation between different modalities.

The attention mechanism and convolutional neural

network provided by deep learning will provide a

strong theoretical basis for this kind of fusion (Liu,

Wu, et al. 2023), which can automatically identify

the importance of each modal data and dynamically

adjust the weight of each data in the overall features.

This method can not only improve the efficiency of

data utilization, but also effectively enhance the

system's ability to identify anomalies (Ma, Chao, et

al. 2024), especially to adapt to the complexity of

marine geological data.

2.2 The Research Role of Adaptive

Feature Extraction and

Two-Branch Prediction in the Field

of Marine Geology

The theory of adaptive feature extraction refers to the

dynamic adjustment of the convolution kernel and

feature extraction process to adapt to the changes of

geological features in different regions (Ma, Liu, et

al. 2023), so as to facilitate the efficient extraction of

useful information in complex and changeable

environments. This is of great significance for the

spatial heterogeneity of marine geology, as the

geological structure of different seabed areas varies

significantly. At the same time, the two-branch

prediction theory combined with the idea of multi-

task learning (Sang, Long, et al. 2023) will complete

the task of identifying marine geological anomalies

and predicting mineral enrichment based on the

shared underlying characteristics. This approach can

effectively improve the efficiency of the model and

achieve the prediction of different goals under the

same framework, and it will also be optimized based

on the joint loss function, which will help the system

to have high accuracy and robustness when handling

complex tasks (Zhang, Liu, et al. 2023).

3 ABNORMAL MARINE

GEOLOGICAL STRUCTURES,

YOUR PRESET COMPARISON

FOR MINERAL EXPLORATION

3.1 Construction of a Comprehensive

Platform for Anomaly

Identification and Prospecting

Prediction Based on Deep Learning

A complete and integrated platform requires a

multifaceted composition. In this study, the

comprehensive platform has several functionally

important components, each of which is very

important and has its own function. In this process,

the data acquisition and preprocessing component is

a complex process that collects raw data from a

variety of ocean data sources and cleanses,

normalizes, and normalizes it. This component needs

to process a large amount of multi-modal data, such

as seismic waves, magnetics, gravity, geochemical

data, etc., and after processing, the data format can be

unified and high-quality to adapt to the input of deep

learning models. The multimodal data fusion

component needs to perform weighted fusion of

geological data from different sources based on the

attention mechanism, and further generate a unified

representation of marine geological features. Based

on the dynamic evaluation and fusion of the

importance of each modal data feature, the

component will effectively ensure that the model can

effectively apply the information of each data type, so

as to improve the accuracy of anomaly recognition.

The adaptive feature extraction component is

responsible for the use of adaptive convolution

networks to extract spatial features from marine

geological data. The component can dynamically

adjust the convolution kernel according to the

complexity of the marine geological environment to

capture geological changes at different scales. The

focus of this component is to accurately identify the

spatial location and morphology of anomalies, and to

provide key features for subsequent predictions. What

the Dual Branch Prediction Component does is to

perform geological anomaly identification and

mineral enrichment prediction at the same time. The

INCOFT 2025 - International Conference on Futuristic Technology

component is based on a two-branch architecture, one

of which ensures that the location and scope of the

anomaly can be controlled, and the other part ensures

that the mineral content and distribution of the

anomaly can be effectively assessed. These two

branches share the results of the underlying feature

extraction to ensure the collaborative execution of the

identification and prediction tasks. The system

training and evaluation component is designed to

manage the training process of the system and

continuously evaluate and adjust the performance of

the model. This component is based on techniques

such as backpropagation to effectively optimize the

parameters of the model, and uses the joint loss

function to perform multi-task training. In addition, it

needs to be responsible for ensuring the accuracy and

generalization ability of the validation set evaluation

system during the training process, and if necessary,

the component also needs to adjust the

hyperparameters to improve the prediction effect of

the system. The purpose of the adaptive optimization

component is to dynamically adjust the training

hyperparameters of the model, such as the

improvement rate, to ensure that the system can

converge stably in the complex marine environment.

According to the gradient change and the fluctuation

of the loss function, the component adaptively adjusts

the improvement rate to prevent the model from

falling into the local optimal solution and overfitting

phenomenon, so as to improve the overall robustness

and adaptability of the system.

3.2 Fusion of Multi-Source Data Such

as Marine Seismic Waves,

Magnetism, Gravity, etc., Adaptive

Feature Extraction, and Algorithm

Calculation

The process of identifying marine geological

anomalies involves a variety of geological and

physical data, such as seismic waves and magnetic

data, which reveal different aspects of the seabed

environment. Therefore, the model must be based on

multi-modal data fusion to jointly use the information

from these different data sources to obtain more

specific and comprehensive geological structure

characteristics. In this paper, we design an attention-

driven feature fusion method, which is mainly based

on the importance of weighted data of different

modalities to generate a comprehensive feature

representation, so as to facilitate more accurate

identification of the complex structure of anomalies.

See Eq. (1) for this.

fused

=α



(1)

In this formula,

used

it is the integrated feature

representation after fusion, which mainly refers to the

overall characteristics of the final geological

environment obtained by the system when processing

multimodal data. It can ensure that the model will

consider the characteristics of different data sources

when making decisions, so as to enhance the

prediction ability of the system. ℎ



It refers to the

features extracted from different data modes, such as

the waveform features in seismic waves and the

intensity characteristics of magnetic anomalies. The

purpose of these features is to provide the model with

different perspectives of marine geological anomalies

and lay the groundwork for subsequent fusion.

𝛼



Represents weights, which are based on attention

mechanism learning and can dynamically allocate the

contribution of each data source based on the validity

of different modal data. For example, in a certain

region, if the magnetic data is more revealing about

the anomaly, then the model will automatically

increase the weight of the modality to enhance the

influence of the feature.

Marine geological data have significant spatial

heterogeneity, which means that the geological

characteristics of different regions may be

significantly different. In order to accurately capture

these differences, an adaptive convolutional kernel is

designed in the construction of the system, so that the

size and weight of the convolution kernel can be

flexibly adjusted according to the geological

complexity of different regions. This adaptive

convolution enables effective modeling of complex

geological structures, especially when capturing the

spatial distribution of marine anomalies. See Eq. (2)

for this.

ij k i k, j k

fW(x)



(2)

In this formula, it 𝑓



represents the output feature

after adaptive convolution, which refers to the

system's understanding of the anomaly at a certain

spatial location. This output will be used to determine

whether the area contains geological anomalies or

not. (𝑥

,

) Representing the input features, which

specifically refer to the geological data of adjacent

areas, based on capturing these neighborhood

features, the convolution operation will further assist

the model, allowing the model to understand the local

Identiﬁcation and Prospecting Prediction of Marine Geological Anomalies Based on Deep Learning

geological environment more comprehensively.

𝑊



Represents the weight of the convolution kernel,

which can be dynamically adjusted according to the

geological characteristics of different regions. When

the geological information of a certain area has a large

change, then the convolution kernel will fully capture

the details of the anomaly based on weight

adjustment.

In addition to identifying the spatial location of

marine geological anomalies, it is also necessary to

assess the possible mineral enrichment within them.

To this end, the system adopts a double-branch

network structure to perform anomaly identification

and mineral enrichment prediction respectively. This

type of architecture allows the system to complete

both tasks in parallel based on a shared feature

representation to improve the overall efficiency and

prediction accuracy of the model. For this, see Eq. (3).

y f (h ), y f (h )

anomaly anomaly shared mineral mineral shared

(3)

In this formula, 𝑦

anomaly

it represents the output of

the anomaly identification branch, which is used to

predict whether there are geological anomalies in the

seabed area, and their specific location and

morphology. The main purpose of this branch is to

quickly locate anomalous objects. 𝑦

mineral

Represents

the output of the Mineral Enrichment Prediction

Branch, which is used to estimate the mineral content

of the anomaly. Based on this prediction, it can

provide a basis for subsequent prospecting decisions.

(ℎ

shared

) Represents a shared feature that is used to

support common feature extraction for both tasks. Its

function is to integrate the features extracted by

multi-modal data and spatial convolution to ensure

the synergy between anomaly identification and

mineral prediction.

3.3 Two-branch Prediction Network

Application

In this link, it is necessary to optimize the

identification of anomalies and the prediction of

mineral enrichment at the same time based on the

joint loss function. Based on multi-task learning, the

model can complete two tasks in continuous

reinforcement learning to improve the efficiency and

accuracy of the system. Depending on the needs of

the task, the system can adjust the weights during the

exercise to ensure that more attention is paid to a

particular task. For example, the value that can be

increased in scenarios where the anomaly

identification task is more important 𝜆

anomaly

In the optimization process of marine geological

anomaly identification and prospecting detection

model based on deep learning, it is necessary to

introduce geological prior knowledge to effectively

guide gradient update. This approach allows the

model to perform faster convergence in targeted

regions, especially in regions where anomalies are

known or present, and based on this, the accuracy of

the system can be realistically improved. Gradient

enhancement means that the gradient update

amplitude of a specific area should be increased when

the parameters are systematically updated based on

prior knowledge in the geological field. Based on this,

it will effectively ensure that the model receives more

attention in key geological anomaly areas and

captures potential geological anomalies more quickly.

The system then has to deal with the multi-scale

characteristics of marine geology, that is, the seabed

topography and geological structure in different

regions may span multiple scales. To this end, a

multi-scale regularization strategy is designed to

prevent the system from overfitting information of a

specific scale based on capturing local and global

geological features at the same time. For this, see Eq.

(4).

()

iij

reg



=λ θ + θ −θ







(4)

In this formula, 𝐿

reg

this is the regularized loss

term to help the model maintain stability and

accuracy. In the identification of marine geological

anomalies, the geological structure is complex and

diverse, so the model has to deal with different types

of anomalies, so if the system is too sensitive to some

specific data, it is likely to make its performance on

new data too poor. In this way, the excessive

fluctuations of system parameters can be effectively

reduced, and the model can be prevented from

"remembering" the characteristics of specific regions,

so as to enhance the generalization ability of the

model in different ocean regions and better identify

unknown geological anomalies.

is the

regularization coefficient, which can control the

intensity of regularization. In the "prospecting

forecast", if the

value is large, it means that the

integrated platform places more emphasis on the

stability of the system to prevent the system from

performing poorly in areas with large data

fluctuations, such as complex seabed geological

structures. However, if the value is too large, it means

that the model may ignore subtle geological changes,

such as the characteristics of small-scale mineral-rich

INCOFT 2025 - International Conference on Futuristic Technology

areas, so the integrated platform will adjust the value

according to the specific geological characteristics

to make the model have a balanced performance in

different scenarios.

θ represents the first parameter

of the model

. For marine geological data, the

operating system parameters

refer to the ability of

the model to deal with specific geological features,

such as seismic wave characteristics and magnetic

anomalies in a certain area. The purpose of the

regularization term

is to constrain the parameters

of the operating system and ensure that the

parameters are not too large, so that the model will

not rely too much on a certain data feature when

dealing with different geological structures, but can

judge the enrichment of minerals in the anomaly

according to a variety of characteristics. In this way,

the overfitting of the operating system to certain

extreme geological features will be avoided and its

adaptability will be enhanced.

∑

𝜃



−𝜃









It is used

to constrain the differences between different model

parameters, especially the correlation between

geological features in adjacent areas. In the marine

geological environment, the geological

characteristics of adjacent areas generally have some

spatial continuity, such as the thickness of

sedimentary layers and the distribution of faults,

which will not change drastically suddenly.

Therefore, the purpose of this project is to ensure that

the operating system can capture the continuity of

geological features when dealing with these areas,

that is, the model parameters will not fluctuate

drastically between adjacent areas. For example,

when an operating system predicts the mineral

enrichment of an area, if the geological conditions of

the area are similar to those of adjacent areas, the

differences in parameters should also be consistent,

which can help the model to more accurately identify

the distribution of potential mineral resources.

Represents the number of parameters, which refers to

the total number of parameters that must be optimized

on the entire marine geological dataset of the deep

learning system.

Because of the complexity and dynamic changes

of the marine environment, such as violent

fluctuations in seabed topography, ocean currents,

geological movements, etc., the deep learning model

needs to flexibly adjust the learning rate according to

the changes in data characteristics to ensure a stable

and efficient exercise process. The adaptive

improvement rate mechanism can enable the

operating system to balance the learning speed and

training stability in the face of these complex

situations, and at the same time, ensure that the model

will not affect the convergence of the operating

system due to unreasonable improvement rate setting

on the basis of accurately capturing geological

anomalies. See Eq. (5) for this.

𝛼



= 𝛼



⋅

1+𝛽|∇𝐿



(5)

In this formula, represents the

rate of

improvement at the first iteration, which controls the

pace at which the model updates parameters at each

step. Its function is to determine how quickly the

operating system responds to current errors. If the

improvement rate is too large, the deep learning

model may miss the optimal parameters. If it is too

small, the workout time of the operating system will

increase significantly.

L∇ Represents the gradient

of the current loss function, which is the size of the

model's current error on the input data. If the gradient

is large, it means that the characteristics learned by

the operating system at this stage are more complex,

so the improvement rate should be reduced to ensure

steady convergence. If the gradient is small, it means

that the model is already familiar with the current task

and can speed up learning.

is a moderating

parameter that controls the effect of the gradient on

the improvement rate.

The effect is to balance the

magnitude of the gradient change to the improvement

rate. If

the setting is larger, the improvement rate

will be more sensitive to changes, which will help the

model to respond more quickly in the face of complex

marine geological environments. If the setting is

small, it can prevent the operating system from being

overly sensitive to short-term gradient fluctuations to

avoid significant changes in the improvement rate

unnecessarily. Based on the adaptive improvement

rate adjustment strategy, the operation system can

adapt to the dynamically changing marine

environment, and gradually converge in the area with

complex geological structure and unstable

characteristics, so as to better identify anomalies and

improve the accuracy of prospecting prediction.

4 RESULTS AND DISCUSSION

4.1 Marine Geological Area Testing

In order to more efficiently identify marine geological

anomalies and predict the possible enrichment of

mineral resources in a deep-sea mineral exploration

Identiﬁcation and Prospecting Prediction of Marine Geological Anomalies Based on Deep Learning

project, this study introduces this self-designed deep

learning integrated platform, hoping to effectively

cope with the limitations of traditional exploration

methods in complex seabed environments. The area

has a typical multi-layered geological structure, and

after many surveys, it has been found that the area

may contain abundant polymetallic nodules and

natural gas hydrates, which provides an important

strategic reserve potential for the development of

marine resources, the test area results are shown in

Table 1.

Table 1: Distribution of Multimodal Data by Region.

Region Seismic

Wave Data

Volume

(

)

Magnetic

data volume

(GB)

Gravity

data

(GB)

Number of

geochemical

samples

A 120 30 25 500

B 150 40 30 600

C 100 20 15 300

Table I shows the distribution of the amount of

multimodal data collected in each region. Among

them, the seismic wave and magnetic data of area B

are large, reflecting its active geological movements.

This exploration mission covers three areas of sea

area A, B and C, and the geological characteristics of

each area are obviously different. Area A is located in

a sedimentary basin and is mainly composed of fine-

grained sediments; Zone B is an area with a

significant history of volcanic activity with active

fault zones; Area C is located in a deep-sea basin and

initial exploration indicates possible enrichment of

gas hydrates, the structure of the testing area is shown

in Figure 1.

Figure 1: Model construction of marine geological

anomalies.

This exploration mission covers three areas of sea

area A, B and C, and the geological characteristics of

each area are obviously different. Area A is located in

a sedimentary basin and is mainly composed of fine-

grained sediments, with 120 GB of seismic wave data

and 500 geochemical samples. Region B is an area of

significant volcanic activity, with 150 GB of seismic

wave data and 600 geochemical samples. In order to

perform more accurate anomaly identification and

mineral prediction, the design of a comprehensive

platform in this study integrates multi-modal data

fusion, adaptive feature extraction, and dual-branch

prediction network.

4.2 Abnormal Ocean Address, Testing

the Mining Area

Based on the automatic extraction and fusion of key

features from multi-modal data such as seismic

waves, magnetism, and gravity, the integrated

platform can accurately identify the spatial

distribution of geological anomalies and predict the

mineral enrichment in them, the test results are shown

in Table 2.

Table 2: Feature extraction results after data

processing.

Regio

Seismic

Wave

Extractio

Features

(

Wei

hts

)

Magnetic

Characteristics

(Weights)

Gravity

Feature

(Weights)

Geochemical

characteristics

(weights)

A 0.45 0.30 0.15 0.10

B 0.50 0.35 0.10 0.05

C 0.40 0.25 0.20 0.15

Table 2 shows the features extracted from the

different modal data and their importance weights. It

can be seen that seismic waves and magnetic

characteristics dominate the identification of

anomalies, Determine the range of regional structure,

as shown in Figure 2.

Figure 2: Identification of marine geological anomalies.

INCOFT 2025 - International Conference on Futuristic Technology

From the data analysis in Figure 2, it can be found

that there are certain anomalous points in the process

of determining the range of anomalous structures in

the ocean. Mainly distributed in the left and right

parts of the graph.

4.3 Abnormal Ocean Address, Test

Results for Predicting Mining

Areas

Based on the research and analysis of the above three

table data, it can be seen that the seismic wave

reflection layer in region B has obvious anomalies,

corresponding to the traces of volcanic activity in this

area. In the magnetic data, the magnetic anomalies in

this area are strong and show potential mineralisation.

In contrast, the seismic wave reflection in region C is

relatively uniform, but the local anomalies in the

magnetic data indicate that it may have a small-scale

gas hydrate enrichment zone. Combined with these

features, the integrated platform successfully

identified anomalies in the volcanic region and

speculated on their possible mineral distribution. The

identification areas of different outlier points are

shown in Table 3.

Table 3: Predictions of mineral enrichment by region.

Region Copper

Ore

Enrichmen

Forecast

(

)

Nickel Ore

Enrichment

Forecast (%)

Gas Hydrate

Enrichment

Forecast (%)

region

A 2.5 1.5 0.8 A

B 3.8 2.0 0.5 B

C 1.2 0.9 4.5 C

Table 3 shows the projections of mineral enrichment

in each region from the integrated platform. Zone B

is prominent in copper forecasts, while Zone C shows

high enrichment in gas hydrates. In the gravity data

of regions A and B, it can be observed that the gravity

anomalies in region A are relatively uniform, while

the gravity anomalies in region B are obvious, which

indicates that the area has a dense rock structure,

which is consistent with the enrichment of copper and

nickel elements in its geochemical data. Geochemical

data indicate that area B has a high content of copper

and nickel ore, which confirms the impact of volcanic

activity on mineral enrichment. In the mineral

enrichment prediction, the copper enrichment in Zone

B is 3.8%, which is much higher than that of other

regions, indicating that this area is a key area for

future exploration. In contrast, Region C has the

highest gas hydrate enrichment of 4.5%, indicating

that this region has some potential for energy

development. Based on the above analysis, the

comprehensive platform designed in this study can

effectively combine multi-modal data to accurately

predict mineral enrichment in complex geological

environment, and provide a reliable basis for actual

exploration.

5 CONCLUSIONS

Based on the designed integrated platform for the

identification and prospecting prediction of marine

geological anomalies based on deep learning, this

paper successfully realizes the efficient identification

and mineral enrichment prediction of marine

geological anomalies. Based on the integration of

multi-source geological data, the integrated platform

significantly improves the accuracy of anomaly

identification in complex seabed environments, and

shows strong prospecting and prediction capabilities

in different regions. In short, the comprehensive

performance of this comprehensive platform can

provide important technical support for future marine

resource exploration, and at the same time, it can also

provide a strong and scientific basis for the

development of marine minerals. This study is fully

reliable, but it has some limitations in terms of data,

and it can be expanded in the future. There are

limitations in this study, mainly the deep learning

dataset and the incomplete collection of marine

geological anomaly identification and selection,

which will be analyzed in the future.

ACKNOWLEDGEMENTS

China Geological Survey, Investigation, Monitoring

and Evaluation of Water Sediment Interaction and

Ecological Environment Carrying Capacity in the

Middle and Lower Reaches of the Yellow River

Basin (Henan Section, DD20220885.

REFERENCES

Chao, N.F., Wang, S., Ouyang, G.C., Hwang, C., Jin, T.Y.,

Zhu, C.C., Abulaitijiang, A., Zhang, S.J., Yue, L.Z.,

Chen, G., Zhang, Y.Z. (2023). An improved triple

collocation-based integration of multiple gravity

anomaly grids from satellite altimetry: Contribution of

ICESat-2. Remote Sensing of Environment, 292.

Identiﬁcation and Prospecting Prediction of Marine Geological Anomalies Based on Deep Learning

De Ritis, R., Cocchi, L., Passaro, S., Chiappini, M. (2024).

Giant landslide, hidden caldera structure, magnetic

anomalies and tectonics in southern Tyrrhenian Sea

(Italy). Geomorphology, 466.

Kim, H.R., Golynsky, A.V., Golynsky, D.A., Yu, H., von

Frese, R.R.B., Hong, J.K. (2022). New Magnetic

Anomaly Constraints on the Antarctic Crust. Journal of

Geophysical Research-Solid Earth, 127(3).

Kochetov, M.V., Shepelev, A.A., Chelyshev, S.V. (2023).

The Possibilities of On-Board Gravimetric and

Hydromagnetic Surveys in Solving Engineering-

Geological Problems on the Shelf. Moscow University

Geology Bulletin, 78(4), 531-540.

Kusnida, D., Albab, A., Nainggolan, T.B., Firdaus, Y.

(2023). The Occurrence of Recent-Subrecent Seabed

Acoustic Anomalies and Its Relationship with

Structural Uplift around the Waipoga Trough, West

Papua-Indonesia. Indonesian Journal of Geoscience,

10(1), 109-117.

Liu, Y.T., Wu, Y.Q., Yang, L., Zhou, P.Z., Kuang, J.X., Yu,

W.J., Wang, J.Q., Xu, Z., Li, G. (2023). A Multi-Task

Learning for Submarine Cable Magnetic Anomaly

Recognition. Journal of Marine Science and

Engineering, 11(5).

Ma, A.Y., Chao, N.F., Qin, P.B., Hwang, C., Zhu, C.C.,

Chen, G., Wang, Z.T., Wang, S. (2024). Performance

assessment of sentinel-3/6 altimeter data for marine

gravity recovery. Frontiers in Marine Science, 11.

Ma, L., Liu, C.G., Yang, A., Liu, B.H., Xia, C.L. (2023).

Characteristics of gravity anomalies and tectonic

analysis of Enderby Land in East Antarctica and its

adjacent areas. Acta Oceanologica Sinica, 42(2), 94-

103.

Sang, N.V., Long, K.V., Dung, T.T., Nguyen, L.V., Que,

B.C., Mong, D., Quang, B.D., Andersen, O.B.,

Forsberg, R., Bui, D.T. (2023). Seafloor depth mapping

of central Vietnam's sea area and its surrounding using

gravity anomaly data and gravity geological method.

Advances in Space Research, 72(5), 1721-1738.

Zhang, W., Liu, W.H., Wang, X.F., Huang, Z.L., Kong,

Q.F., Luo, H.Y., Zhang, D.D., Liu, P., Chen, X.Y., Cai,

Z.H. (2023). Formation Mechanism of Heavy

Hydrocarbon Carbon Isotope Anomalies in Natural Gas

from Ordovician Marine Carbonate in the Ordos Basin.

Journal of Marine Science and Engineering, 11(11).

INCOFT 2025 - International Conference on Futuristic Technology