Systematic Review

Artificial Intelligence in Aquatic Biodiversity Research:

A PRISMA-Based Systematic Review
Tymoteusz Miller 1, * , Grzegorz Michoński 1 , Irmina Durlik 2,3 , Polina Kozlovska 4 and Paweł Biczak 2

1 Institute of Marine and Environmental Sciences, University of Szczecin, 71-415 Szczecin, Poland;
[Link]@[Link]
2 Polish Society of Bioinformatics and Data Science, Biodata, 71-214 Szczecin, Poland;
[Link]@[Link] (I.D.); pawel@[Link] (P.B.)
3 Faculty of Navigation, Maritime University of Szczecin, 70-500 Szczecin, Poland
4 Faculty of Economics, Finance and Management, University of Szczecin, 71-412 Szczecin, Poland;
231805@[Link]
* Correspondence: [Link]@[Link]

Simple Summary: This review explores the emerging role of artificial intelligence in
supporting aquatic biodiversity research. We present a structured analysis of how AI
techniques—such as machine learning, deep learning, and transformers—are applied to
key ecological tasks, including species identification, habitat modeling, ecological risk
assessment, remote sensing analysis, and conservation planning. For each domain, we
link biological questions to computational problems and discuss the suitability and limi-
tations of specific AI algorithms. We also highlight real-world case studies and provide
a methodological classification of AI approaches based on the data type and modeling
goal. This work offers both ecologists and data scientists a comprehensive perspective
on how AI can advance biodiversity monitoring and support conservation strategies in
freshwater ecosystems.

Abstract: Freshwater ecosystems are increasingly threatened by climate change and anthro-
pogenic activities, necessitating innovative and scalable monitoring solutions. Artificial
intelligence (AI) has emerged as a transformative tool in aquatic biodiversity research,
enabling automated species identification, predictive habitat modeling, and conservation
planning. This systematic review follows the PRISMA framework to analyze AI applica-
Academic Editor: Jordan G. Okie tions in freshwater biodiversity studies. Using a structured literature search across Scopus,
Received: 10 March 2025 Web of Science, and Google Scholar, we identified 312 relevant studies published between
Revised: 30 April 2025 2010 and 2024. This review categorizes AI applications into species identification, habitat as-
Accepted: 6 May 2025
sessment, ecological risk evaluation, and conservation strategies. A risk of bias assessment
Published: 8 May 2025
was conducted using QUADAS-2 and RoB 2 frameworks, highlighting methodological
Citation: Miller, T.; Michoński, G.;
challenges, such as measurement bias and inconsistencies in the model validation. The
Durlik, I.; Kozlovska, P.; Biczak, P.
citation trends demonstrate exponential growth in AI-driven biodiversity research, with
Artificial Intelligence in Aquatic
Biodiversity Research: A PRISMA- leading contributions from China, the United States, and India. Despite the growing use of
Based Systematic Review. Biology 2025, AI in this field, this review also reveals several persistent challenges, including limited data
14, 520. [Link] availability, regional imbalances, and concerns related to model generalizability and trans-
biology14050520 parency. Our findings underscore AI’s potential in revolutionizing biodiversity monitoring
Copyright: © 2025 by the authors. but also emphasize the need for standardized methodologies, improved data integration,
Licensee MDPI, Basel, Switzerland. and interdisciplinary collaboration to enhance ecological insights and conservation efforts.
This article is an open access article
distributed under the terms and Keywords: artificial intelligence; aquatic biodiversity; machine learning; deep learning; species
conditions of the Creative Commons
identification; habitat modeling; conservation; ecological monitoring; freshwater ecosystems
Attribution (CC BY) license
([Link]
nses/by/4.0/).

Biology 2025, 14, 520 [Link]

Biology 2025, 14, 520 2 of 41

1. Introduction
Freshwater ecosystems, including rivers, lakes, wetlands, and groundwater systems,
are among the most biologically diverse environments on Earth. Despite covering only
about 2.5% of the planet’s total water resources, they support nearly 10% of all known
species, including a vast array of fish [1], amphibians, invertebrates, and aquatic plants.
These ecosystems play a crucial role in maintaining ecological balance, providing essential
services, such as water purification, flood regulation, and carbon sequestration [2].
However, freshwater biodiversity is under severe threat due to climate change, habitat
destruction, pollution [3], invasive species, and overexploitation. According to recent
reports, freshwater species populations have declined by nearly 84% since 1970, making
freshwater ecosystems among the most endangered on the planet. Monitoring and conserv-
ing biodiversity in these habitats is therefore a global priority, requiring innovative and
scalable approaches to assess ecosystem health and implement conservation strategies [4].
Artificial intelligence (AI) has emerged as a powerful tool in biodiversity research,
revolutionizing traditional monitoring methods by enabling automated species identifi-
cation [5,6], habitat mapping, and ecological risk assessment [7]. Machine learning (ML)
and deep learning (DL) techniques, particularly computer vision, bioacoustics, natural
language processing (NLP), and remote sensing AI models, have significantly enhanced
the accuracy and efficiency of biodiversity studies [8].
Key AI applications in aquatic biodiversity research include:
1. Species identification using convolutional neural networks (CNNs) for image recognition.
2. Bioacoustic analysis for detecting species presence through underwater sound recordings.
3. Predictive habitat modeling for assessing ecosystem changes due to climate variability.
4. Ecological risk assessment to evaluate pollution impact and detect environmental
threats [9].
5. Remote sensing AI for large-scale biodiversity monitoring using satellite and drone imagery.
These AI-driven approaches offer unprecedented capabilities for processing vast
amounts of ecological data, enabling researchers to develop more accurate models for
biodiversity conservation, policymaking, and ecosystem management [10]. However,
despite its potential, AI applications in freshwater biodiversity research still face chal-
lenges related to data availability, model generalization, ethical concerns, and the need for
standardized methodologies.
The primary objective of this review is to provide a comprehensive synthesis of the
current state of AI applications in freshwater biodiversity research, following the Pre-
ferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) framework.
Specifically, this study aims to:
1. Systematically analyze AI-based methodologies applied to freshwater biodiversity
monitoring and conservation.
2. Categorize AI applications into key areas, such as species identification, habitat
modeling, ecological risk assessment, and conservation strategies.
3. Evaluate methodological strengths and limitations, including risk of bias, data quality
issues [11], and validation challenges.
4. Identify citation trends and geographical contributions in AI-driven biodiversity research.
5. Highlight knowledge gaps and propose future research directions to improve AI’s
role in freshwater biodiversity conservation.
This systematic review is intended to serve as a reference for researchers, ecologists,
and policymakers, facilitating the development of more robust, scalable, and effective
AI-based conservation strategies to protect and restore freshwater biodiversity.
Biology 2025, 14, 520 3 of 41

2. Methodology (PRISMA Framework)

2.1. Literature Search Strategy
This systematic review was conducted following the Preferred Reporting Items for
Systematic Reviews and Meta-Analyses (PRISMA) guidelines. The literature search was
performed across Scopus, Web of Science, and Google Scholar databases, ensuring compre-
hensive coverage of AI applications in aquatic biodiversity research. The following Boolean
search string was utilized to retrieve relevant publications:
(TITLE-ABS-KEY(“artificial intelligence” OR “machine learning” OR “deep learning”
OR “computer vision” OR “natural language processing” OR “AI”) AND TITLE-ABS-KEY
(“aquatic biodiversity” OR “freshwater biodiversity” OR “water quality” OR “species
identification” OR “habitat modeling” OR “ecological monitoring”) AND TITLE-ABS-KEY
(“conservation” OR “risk assessment” OR “environmental impact” OR “ecological model-
ing”)) AND PUBYEAR > 2010 AND PUBYEAR < 2025 ([Link]
[Link]?sort=plf-f&src=s&sid=ea56e0dd621c4d30fcb500e9e1857c03&sot=a&sdt=a&sl=461&
s=(TITLE-ABS-KEY(%22artificial+intelligence%22+OR+%22machine+learning%22+OR+%22
deep+learning%22+OR+%22computer+vision%22+OR+%22natural+language+processing%
22+OR+%22AI%22)+AND+TITLE-ABS-KEY(%22aquatic+biodiversity%22+OR+%22freshwa
ter+biodiversity%22+OR+%22water+quality%22+OR+%22species+identification%22+OR+%2
2habitat+modeling%22+OR+%22ecological+monitoring%22)+AND+TITLE-ABS-KEY(%22c
onservation%22+OR+%22risk+assessment%22+OR+%22environmental+impact%22+OR+%
22ecological+modeling%22))AND+PUBYEAR+&gt;+2010+AND+PUBYEAR+&lt;+2025&orig
in=searchadvanced&editSaveSearch=&txGid=7494e69bf030c58a1e2071ffa8ca9026&sessionSe
archId=ea56e0dd621c4d30fcb500e9e1857c03&limit=10, accessed on 6 February 2025).
This query was designed to ensure that only studies related to AI applications in fresh-
water biodiversity research were retrieved. The search was limited to studies published in
peer-reviewed journals, conference proceedings, and institutional reports. No language
restrictions were applied, but non-English papers were excluded if no translated version
was available.
In constructing the search strategy, we employed Boolean logic using keywords such
as “artificial intelligence”, “machine learning”, “deep learning”, “computer vision”, and
“aquatic biodiversity”. While this strategy enabled us to capture a broad swath of the
literature, we acknowledge the inherent risk of bias introduced by specific term selection.
For instance, the inclusion of “computer vision” may have skewed the results toward
image-based applications (e.g., visual species recognition) while underrepresenting studies
focused on environmental DNA (eDNA), natural language processing (e.g., ecological
literature mining), or acoustic monitoring. To partially mitigate this, we performed a
sensitivity analysis using additional terms (e.g., “eDNA”, “bioacoustics”, and “NLP”) and
noted that while the overall thematic structure remained consistent, several relevant studies
were indeed uncovered that had been missed by the initial query.
We recommend that future systematic reviews in this domain employ hybrid ap-
proaches that combine keyword-based searches with citation network exploration or
machine-assisted discovery tools to reduce thematic bias and increase comprehensiveness.

2.2. Inclusion and Exclusion Criteria

The selection of studies was based on predefined inclusion and exclusion criteria to
ensure relevance and quality. Studies were included if they applied AI techniques, such
as machine learning, deep learning, natural language processing, or computer vision, to
aquatic biodiversity research. Additionally, research needed to focus specifically on fresh-
water ecosystems, including rivers, lakes, and wetlands. Only studies providing empirical
Biology 2025, 14, 520 4 of 41

evidence or case studies that demonstrated AI applications in species identification, habitat

modeling, or conservation strategies were considered for inclusion.
Several exclusion criteria were applied to refine the dataset. Studies that focused on
marine biodiversity rather than freshwater ecosystems were excluded, as they fell outside
the scope of this review. Papers that lacked a substantial AI methodology or did not clearly
specify how AI was applied in their research were also omitted to maintain methodological
rigor. Review articles that did not contribute original empirical findings were not con-
sidered, ensuring that this review remained focused on primary research. Furthermore,
studies that did not validate AI model performance or lacked sufficient methodological
transparency were excluded due to concerns regarding reliability and reproducibility.
Lastly, non-English studies without an available translated version were not included
to maintain consistency in the data analysis and interpretation. These exclusion criteria
were applied systematically to ensure that only studies meeting high-quality standards
contributed to the final synthesis.

2.3. Study Selection (PRISMA Flow Diagram)

The study selection process followed the PRISMA flow diagram, consisting of four
main phases: identification, screening, eligibility assessment, and inclusion. The initial
search yielded a total of 975 records. After removing duplicates and irrelevant studies based
on title and abstract screening, 615 studies remained. A full-text review was conducted on
these, leading to the final selection of 312 studies for inclusion in the systematic review
(Figure 1).
To enrich the bibliometric analysis and ensure thematic breadth, we generated a
citation network based on core studies identified during initial screening. Using Research
Rabbit, we visualized how foundational and contemporary works are interlinked through
citation patterns. As illustrated in Figure 2, the graph reveals multiple clusters, ranging
from early ecological modeling frameworks to recent advances in AI-based monitoring,
bioacoustics, and eDNA analytics.
This visual exploration confirmed the presence of cross-disciplinary bridges and
highlighted several influential publications that were not highly ranked by a keyword
search alone. Citation mapping thus served as a complementary tool to reduce thematic
bias and identify structurally important but terminologically divergent contributions.

2.4. Assessment of Risk of Bias

To ensure the validity of findings, the assessment of the risk of bias was conducted
using established frameworks. The QUADAS-2 tool was applied for diagnostic studies,
and the RoB 2 tool was used for interventional research. Each study was evaluated across
three key domains:
1. Selection bias: evaluated based on transparency in inclusion criteria and representa-
tiveness of study populations.
2. Information bias: assessed through data source reliability and robustness of
AI methodologies.
3. Measurement bias: analyzed through the consistency and repeatability of AI
model performance.
The findings revealed that while selection and information bias were generally low
across most studies, measurement bias was high in studies that lacked clear validation
methodologies or relied on secondary data sources. A summary of bias distribution is
illustrated in Figure 3.
Figure 3 illustrates the distribution of risk of bias across three domains: selection bias,
information bias, and measurement bias. The majority of studies were rated as having low
Biology 2025, 14, 520 5 of 41

risk for selection bias (60%), indicating that inclusion criteria and population representa-
tiveness were generally well reported. Similarly, 50% of studies had low information bias,
reflecting adequate data source reliability and clear methodological descriptions.
However, measurement bias emerged as a more prevalent issue, with only 30% of
studies rated as low risk in this category. A notable proportion (40%) exhibited moderate
measurement bias, and another 30% were rated as high risk. This trend highlights the
frequent absence of external validation, inconsistent reporting of performance metrics (e.g.,
precision, recall), and a lack of reproducibility in AI model assessments. In some cases,
Biology 2025, 14, x FOR PEER REVIEW 5 of 43
performance evaluations relied solely on internal cross-validation or secondary datasets,
further undermining reliability.

Figure 1. Prisma flow diagram.

Figure 1. Prisma flow diagram.

2.4. Assessment of Risk of Bias

To ensure the validity of findings, the assessment of the risk of bias was conducted
using established frameworks. The QUADAS-2 tool was applied for diagnostic studies,
and the RoB 2 tool was used for interventional research. Each study was evaluated across
three key domains:
 Biology 2025,14,
Biology2025, 14,520
x FOR PEER REVIEW 66 of
of 41
43

Figure2.2.A
Figure A citation
citation network
network visualization
visualizationcentered
centeredon
onkey
keystudies
studiesin
in AI-driven
AI-drivenaquatic
aquaticbiodiversity
biodiversity
monitoring.
Biology 2025, 14, x FOR PEER REVIEWmonitoring. The
The graph
graphwas
wasgenerated
generated using Research
using Rabbit
Research and
Rabbit reveals
and thematic
reveals clusters,
thematic including
clusters, includ- 7 of 43
foundational ecological modeling, machine learning theory and contemporary AI applications
ing foundational ecological modeling, machine learning theory and contemporary AI applications in
species detection
in species andand
detection environmental genomics.
environmental genomics.

The findings revealed that while selection and information bias were generally low
across most studies, measurement bias was high in studies that lacked clear validation
methodologies or relied on secondary data sources. A summary of bias distribution is
illustrated in Figure 3.
Figure 3 illustrates the distribution of risk of bias across three domains: selection bias,
information bias, and measurement bias. The majority of studies were rated as having low
risk for selection bias (60%), indicating that inclusion criteria and population representa-
tiveness were generally well reported. Similarly, 50% of studies had low information bias,
reflecting adequate data source reliability and clear methodological descriptions.
However, measurement bias emerged as a more prevalent issue, with only 30% of
studies rated as low risk in this category. A notable proportion (40%) exhibited moderate
measurement bias, and another 30% were rated as high risk. This trend highlights the
frequent absence of external validation, inconsistent reporting of performance metrics
(e.g., precision, recall), and a lack of reproducibility in AI model assessments. In some
cases, performance evaluations relied solely on internal cross-validation or secondary da-
tasets, further undermining reliability.
[Link]
Figure
Summary of bias distribution.
of bias distribution.

This distribution underscores a key limitation in current AI-driven aquatic biodiver

sity studies: while data collection and algorithm specification are increasingly standard
ized, the evaluation and reporting of model performance remain inconsistent. Future
studies should adopt more transparent and rigorous validation protocols to mitigate
Biology 2025, 14, 520 7 of 41

This distribution underscores a key limitation in current AI-driven aquatic biodiversity

studies: while data collection and algorithm specification are increasingly standardized,
the evaluation and reporting of model performance remain inconsistent. Future studies
should adopt more transparent and rigorous validation protocols to mitigate measurement
bias and improve the reproducibility of AI-based ecological findings.
A significant challenge in systematic reviews is the potential for publication bias, where
studies reporting positive or significant findings are more likely to be published than those
with null or negative results. This can skew the perceived effectiveness of AI applications
in aquatic biodiversity research and limit the reliability of synthesized conclusions.
To mitigate this risk, additional searches were conducted beyond the traditional peer-
reviewed literature. This included screening preprints from repositories, such as arXiv and
bioRxiv, reviewing institutional reports and white papers from research organizations, and
analyzing conference proceedings from AI and ecology-related conferences. These sources
provided valuable insights into ongoing research that may not yet be formally published.
Despite these efforts, the potential for underreporting of negative AI outcomes remains
a limitation of this review. Studies with inconclusive or unfavorable results may not be as
readily accessible, potentially leading to an overestimation of AI’s effectiveness in biodiver-
sity monitoring. To improve transparency and reduce reporting bias, future research should
prioritize the use of comprehensive databases, encourage the publication of negative or
neutral findings, and foster collaborations with research groups that promote open sci-
ence initiatives. Standardized reporting guidelines for AI-driven ecological research could
further support balanced representation of both successful and unsuccessful applications.
For QUADAS-2, we retained the four standard domains—patient selection, index
test, reference standard, and flow and timing—and mapped them to AI-specific elements,
as follows:
1. Patient selection was interpreted as the data sampling strategy, including whether
datasets were balanced, representative, and clearly defined.
2. The index test corresponded to the AI model itself, including algorithm specification,
tuning procedures, and availability of implementation details.
3. The reference standard referred to the ground truth quality—i.e., how labels were
obtained and whether they were verified by domain experts.
4. Flow and timing were aligned with temporal consistency in data collection and model
evaluation procedures.
We applied the following simplified thresholds during assessment (Table 1):

Table 1. Risk levels explanations.

Risk Level Criteria (Example)

Low risk Publicly available dataset, clear model architecture, external validation present
Some concerns Model details incomplete, internal validation only, unclear label derivation
High risk Proprietary dataset, no model transparency, no validation steps reported

For the RoB 2 tool, the primary domains—randomization process, deviations from
intended interventions, missing outcome data, measurement of outcome, and selection of
the reported result—were similarly mapped. Given the non-randomized nature of most AI
studies, we emphasized:
1. Bias due to missing data (e.g., unreported performance metrics);
2. Bias in outcome measurement (e.g., lack of precision/recall/F1-score reporting);
3. Selective reporting (e.g., absence of comparison with baseline models).
Biology 2025, 14, 520 8 of 41

2.5. Data Extraction and Analysis

For each included study, a structured data extraction framework was used to ensure
consistency. The extracted data included research objectives, methodologies, AI tech-
niques, target species, and evaluation metrics. Studies were categorized thematically into
the following:
1. AI advancements in species identification and biodiversity assessment.
2. Habitat modeling and ecological impact prediction.
3. Ethical and regulatory considerations of AI applications in conservation.
A mixed-methods approach combined quantitative metrics (e.g., citation frequency,
model accuracy benchmarks) with qualitative insights from expert evaluations. This
ensured a comprehensive synthesis of findings.
To enhance methodological clarity, we adopted a mixed-methods approach that com-
bined both quantitative and qualitative analyses. This dual strategy enabled us to sys-
tematically assess trends in AI applications in aquatic biodiversity research while also
interpreting the broader ecological and methodological contexts.
The quantitative component included descriptive statistics, such as publication counts
over time, citation frequencies, geographic distribution of studies, model performance
metrics (e.g., accuracy, precision, recall), and frequency of specific AI methods. These
allowed us to establish macro-level patterns in the research landscape.
The qualitative component focused on thematic categorization of studies, evaluation
of methodological rigor, ecological relevance of AI techniques, presence or absence of ex-
plainability tools (e.g., SHAP, LIMEs), and identification of ethical and practical challenges
raised by the authors.
The two components were integrated during the synthesis phase (Sections 3 and 4),
where numerical findings were paired with contextual interpretation. For example, we
combined citation data with model transparency ratings to assess which methods gained
traction while maintaining ethical robustness. Similarly, regional trends were analyzed in
light of biodiversity protection goals.
To facilitate understanding, we created an integration flowchart (Figure 4) that maps
the mixed-methods process, from data extraction to result synthesis. This visual repre-
Biology 2025, 14, x FOR PEER REVIEW 9 of 43
sentation helps clarify how quantitative and qualitative insights were blended to support
evidence-based conclusions.

Figure4.4.
Figure Integration
Integration flowchart.
flowchart.

This approach strengthens the internal coherence of the review and supports a more
holistic understanding of how AI tools are deployed in the service of biodiversity moni-
toring and conservation.
 Biology 2025, 14, 520 9 of 41

This approach strengthens the internal coherence of the review and supports a more
holistic understanding of how AI tools are deployed in the service of biodiversity monitor-
ing and conservation.

2.6. Citation Trends

The citation trends for artificial intelligence applications in aquatic biodiversity re-
search exhibit a significant upward trajectory, particularly in the last five years. The number
of research papers published per year has steadily increased, reflecting the growing integra-
tion of AI in ecological and environmental studies. In 2011, only 22 papers were published,
with a modest rise in subsequent years. By 2015, the count reached 17, and by 2020, the
annual number of publications had grown to 55. The rapid acceleration began in 2021 with
99 papers, followed by 105 in 2022, 187 in 2023, and a substantial surge to 342 publications
Biology 2025, 14, x FOR PEER REVIEW in 2024. This pattern demonstrates a strong academic and research community focus 10on
of 43
applying AI methodologies to address biodiversity challenges in freshwater ecosystems
(Figure 5).

Figure 5. Publication trends.

Figure 5. Publication trends.
The overall trend in citation numbers also shows a remarkable increase. From
7 citations in 2010, the count rose steadily to 45 in 2011, 69 in 2012, 109 in 2013, and
157 in 2014. A sharper rise was observed from 2015 onwards, with citations increasing to
237 in 2015, 280 in 2016, 450 in 2017, and 625 in 2018. The most significant growth occurred
from 2019 to 2024, with citations reaching 1022 in 2019, 1705 in 2020, 2587 in 2021, 3472 in
2022, 5215 in 2023, and a peak of 15,980 in 2024, reflecting the increasing recognition and
influence of AI-driven biodiversity research in the scientific community (Figure 6).
Regional contributions to the field reveal a dominant presence from technologically
advanced and ecologically diverse nations. China leads with 209 publications, followed
closely by the United States with 191, and India with 185. Other significant contributors
include Australia (52), Germany (47), Canada (44), and the United Kingdom (43). Spain,
South Korea, and France have also maintained active research output, each contribut-
ing between 36 and 42 papers. Notably, countries such as Saudi Arabia, Italy, and the
Netherlands exhibit increasing engagement, suggesting a more global adoption of AI in
biodiversity research.
 Biology 2025, 14, 520 10 of 41
Figure 5. Publication trends.

Biology 2025, 14, x FOR PEER REVIEW Figure

Figure 6. Citation trends.
6. Citation trends. 11 of 43

In Asia, Malaysia (24), Bangladesh (23), and Iran (22) show steady research activity.
Regional contributions to the field reveal a dominant presence from technologically
South Africa (23) and Brazil (17) represent the leading contributors from the Global South.
advanced and ecologically diverse nations. China leads with 209 publications, followed
Czech Republic
European (14). Smaller
participation but highlighted
is further notable contributions
by Switzerlandcome from
(21), countries
Belgium likethe
(15), and Taiwan,
closely by the United States with 191, and India with 185. Other significant contributors
Turkey, Portugal,(14).
Czech Republic andSmaller
Denmark, each producing
but notable over
contributions tenfrom
come research papers.
countries like Interestingly,
Taiwan,
include Australia (52), Germany (47), Canada (44), and the United Kingdom (43). Spain,
Turkey,
several Portugal, and Denmark, each producing over ten research papers. Interestingly,
South emerging
Korea, andresearch hubs,
France have including
also Chile,
maintained Ecuador,
active researchHungary,
output, eachandcontributing
Iraq, have begun
several
engaging emerging research hubs, including Chile, Ecuador, Hungary, and Iraq, have begun 7).
between in36 this
and interdisciplinary
42 papers. Notably,research domain,
countries such as albeit at a lower
Saudi Arabia, volume
Italy, and the(Figure
Neth-
engaging in this interdisciplinary research domain, albeit at a lower volume (Figure 7).
erlands exhibit increasing engagement, suggesting a more global adoption of AI in biodi-
versity research.
In Asia, Malaysia (24), Bangladesh (23), and Iran (22) show steady research activity.
South Africa (23) and Brazil (17) represent the leading contributors from the Global South.
European participation is further highlighted by Switzerland (21), Belgium (15), and the

Figure 7. Regional contribution.

Figure 7. Regional contribution.
Overall, the increasing publication rate and the broad geographic distribution of
Overall,
research thesuggest
efforts increasing publication
a widespread rate andofthe
recognition AI’sbroad geographic
potential distribution
in revolutionizing bio- of re-
search efforts
diversity suggest and
monitoring a widespread
conservationrecognition
efforts [12].ofThe
AI’s potential
growing in revolutionizing
impact of these studies biodi-
versity monitoring
underscores and conservation
the necessity efforts [12].
for interdisciplinary The growing
collaboration impact of these
and methodological studies un-
advance-
ments to refine AI applications in aquatic biodiversity studies [13–300].
derscores the necessity for interdisciplinary collaboration and methodological advance-
ments to refine AI applications in aquatic biodiversity studies [13–300].

3. AI Applications in Aquatic Biodiversity Research

Biology 2025, 14, 520 11 of 41

3. AI Applications in Aquatic Biodiversity Research

Artificial intelligence has emerged as a transformative force in biodiversity research,
offering novel solutions for species identification [13], habitat assessment, ecological risk
modeling [14], and conservation planning. By leveraging machine learning and deep
learning techniques [15], researchers can analyze vast datasets, automate labor-intensive
processes, and enhance the precision of biodiversity monitoring efforts [16]. This section
explores the major applications of AI in aquatic biodiversity research, categorizing them
into species identification [17], habitat and risk assessment, remote sensing applications,
and conservation strategies (Table 2).

Table 2. Summary of AI applications in aquatic biodiversity research.

AI Application Methods Used Key Benefits

Computer Vision (CNNs) [158–160], Automated classification [164], high
Species Identification [157]
Bioacoustics [161], eDNA-based AI [162,163] accuracy, scalable monitoring [165,166]
Predictive Habitat Mapping, ML-based
Habitat and Ecological Risk Predicts ecosystem changes, identifies
Ecological Risk Models, AI-driven Water
Assessment [167] threats, supports mitigation [170,171]
Quality Monitoring [168,169]
Large-scale monitoring, real-time
Remote Sensing for Satellites/Drones with AI, Deep Learning for
analysis, improved data
Biodiversity Monitoring Land–Water Interface Analysis [172]
accessibility [173–175]
AI-powered Decision Support Systems, Optimizes conservation decisions,
AI in Conservation and Management
AI-driven Conservation Planning, Citizen enhances public participation, enables
Strategies [176]
Science Integration [177] proactive strategies

In this section, we examine key AI applications by structuring each analysis around the
following elements: (i) the biological research question addressed, (ii) the computational
modeling task derived from the problem (e.g., classification, regression, or clustering),
(iii) the selection and suitability of AI and ML methods, and (iv) the limitations of the
applied techniques. This framework allows for a deeper understanding of how AI method-
ologies align with specific ecological objectives, guiding future developments in aquatic
biodiversity monitoring and conservation.

3.1. AI for Species Identification

Accurate species identification forms the cornerstone of aquatic biodiversity research,
enabling ecological monitoring, population assessment, and conservation planning. The
biological research question centers on whether species can be reliably and automatically
identified based on observable data modalities, such as images, acoustic signals, or ge-
nomic sequences [18–20]. From a computational perspective, these challenges are typically
formalized as supervised classification tasks, where the model must assign an input (e.g.,
an image, audio spectrogram, or DNA sequence) to a predefined taxonomic label.
Advances in machine learning and deep learning have substantially accelerated this
process by minimizing reliance on manual taxonomy and mitigating observer bias [21–25].
Three principal modalities have emerged: image-based recognition using computer vi-
sion, bioacoustic monitoring with deep learning models, and genomic-based classification
leveraging AI.

3.1.1. Image-Based Species Recognition Using Computer Vision

Image-based species recognition addresses the biological question of whether aquatic
organisms can be automatically identified from visual data, such as photographs or under-
water videos. This challenge is typically approached as an image classification problem,
where the goal is to map image data to species labels using machine learning models.
Convolutional neural networks (CNNs) have become the dominant method in this do-
Biology 2025, 14, 520 12 of 41

main [26,27]. Architectures such as ResNet, EfficientNet, and YOLO (You Only Look
Once) are commonly used because of their ability to extract complex hierarchical fea-
tures from visual inputs, achieving high classification accuracy even under challenging
conditions [28–30].
YOLO models, in particular, enable real-time species detection and tracking in dynamic
underwater environments [31–33]. CNNs are highly suitable due to their capacity for end-
to-end feature learning, which eliminates the need for manual feature engineering. Transfer
learning further enhances their utility by allowing models pre-trained on large datasets to
be fine-tuned with limited aquatic data.
Nevertheless, these models require substantial volumes of labeled training data,
which are often scarce in freshwater research [34–36]. Additionally, CNNs are typically
black-box systems with limited interpretability, which may hinder biological validation.
Their computational demands for training and fine-tuning also pose a barrier for smaller
research groups.

3.1.2. Bioacoustic Monitoring with Deep Learning

Bioacoustic monitoring explores the biological question of whether species presence
and behavior can be detected and classified based on underwater sound recordings. In
computational terms, this task is framed as time-series classification, where sequences of
acoustic features—often derived from spectrograms—are mapped to species identities.
Deep learning models, particularly recurrent neural networks (RNNs) such as Long Short-
Term Memory (LSTM) and Gated Recurrent Units (GRUs), are widely used due to their
capacity to capture temporal dependencies within sequential data [37–40]. These models
are especially effective in distinguishing species-specific vocalizations and ecological events
from environmental noise. By analyzing the broader temporal structure of acoustic signals,
RNNs outperform traditional models that rely solely on static features [41–43].
However, the underwater acoustic environment presents numerous challenges, in-
cluding variability caused by anthropogenic interference and overlapping calls between
species [44–46]. Furthermore, RNNs are prone to overfitting when trained on small or
imbalanced datasets, and their optimization requires substantial computational resources.

3.1.3. DNA and eDNA-Based Species Identification

DNA and environmental DNA (eDNA) analysis addresses the biological question
of whether aquatic species can be reliably detected and classified using genetic material
extracted from environmental samples. The computational task involved is typically
sequence classification, where nucleotide patterns are analyzed to infer species identity.
Traditional machine learning algorithms, such as Support Vector Machines (SVMs) and
Random Forests (RFs), have been successfully applied to classify species based on genetic
markers obtained from DNA barcoding and eDNA analyses [47–52]. These models offer
interpretability and robustness, particularly for structured, low-dimensional datasets. More
recently, transformer-based architectures like DNABERT have emerged as powerful tools
for modeling sequential and contextual properties of genetic sequences [53,54]. These
models improve accuracy in distinguishing closely related or cryptic species by capturing
long-range dependencies.
Despite their promise, transformer models require large, high-quality annotated
datasets and are computationally intensive, which limits their adoption in aquatic biodiver-
sity research [55]. Additionally, the lack of comprehensive reference libraries for aquatic
genomes hinders the generalizability of these approaches.
Biology 2025, 14, 520 13 of 41

3.2. AI for Habitat and Ecological Risk Assessment

Beyond species-level identification, aquatic biodiversity research requires the assess-
ment of habitat quality, the prediction of ecological changes, and the evaluation of environ-
mental risks. The biological research questions in this domain focus on whether machine
learning can reliably model habitat suitability, forecast ecosystem changes, and assess risks
posed by pollution, invasive species, and land-use alterations [56–63]. Computationally,
these challenges are formalized primarily as regression and classification problems, often
requiring spatial prediction and risk categorization.
Machine learning models provide flexible, data-driven alternatives to traditional
correlational ecological approaches by capturing nonlinear interactions among multiple
environmental variables.

3.2.1. Predictive Habitat Modeling

Predictive habitat modeling addresses the biological question of whether suitable
aquatic habitats for various species can be reliably predicted based on environmental
parameters. From a computational standpoint, this involves supervised regression and
classification tasks, where models estimate habitat suitability scores or assign habitat types
using predictor variables, such as water temperature, pH, turbidity, or dissolved oxygen. A
wide range of machine learning algorithms has been employed in this context, including
Random Forests, Gradient Boosting Machines (e.g., XGBoost 3.0.0, LightGBM 4.6.0), and
Deep Neural Networks [64,65]. These models are well-suited for capturing nonlinear and
complex interactions among environmental variables, often outperforming traditional
ecological approaches [66–92]. Random Forests are especially robust in the presence of
noisy or high-dimensional data, while Gradient Boosting methods excel in structured and
sparse datasets. Although Deep Neural Networks offer greater modeling flexibility, they
often require large datasets and pose challenges in interpretability.
Limitations in predictive habitat modeling frequently arise from inconsistent spatial
and temporal data coverage, which can lead to biased predictions. Additionally, complex
models may overfit regional data and generalize poorly to new environmental conditions,
while explaining their internal decision-making remains difficult for ecological validation.

3.2.2. Machine Learning in Ecological Risk Assessment

Ecological risk assessment supported by AI investigates the biological question of
whether vulnerable areas or ecosystems can be identified and classified based on exposure
to stressors, such as pollutants or invasive species. This task is typically formalized as
classification or probabilistic modeling, where risk levels are predicted using a combination
of environmental and biological indicators. Ensemble learning methods, notably Extreme
Gradient Boosting (XGBoost) and Bagging algorithms, have proven effective in this do-
main [84,85]. These models integrate heterogeneous data sources—including pollutant
concentrations, land use patterns, and species sensitivity indices—to estimate the likelihood
of ecological degradation [86–90]. By combining predictions from multiple weak learners,
ensemble methods increase robustness, improve generalization, and reduce variance in
high-uncertainty environments.
However, ecological risk models often face challenges, such as imbalanced datasets
(few high-risk events compared to many low-risk ones), missing or noisy input data, and
multicollinearity among predictors, which can obscure causal relationships crucial for
designing mitigation strategies [91,92].
Biology 2025, 14, 520 14 of 41

3.2.3. AI-Based Water Quality Monitoring

AI-based water quality monitoring addresses the biological question of whether critical
ecosystem parameters—such as nutrient levels or algal blooms—can be reliably predicted
or monitored in real time using sensor data. Computationally, this challenge is modeled as
a combination of time-series forecasting and anomaly detection, where algorithms predict
future trends or flag abnormal fluctuations in water quality indicators. Deep learning mod-
els such as Long Short-Term Memory (LSTM) networks and Autoencoders have become
increasingly prominent for these tasks [93–107]. LSTMs are particularly effective in han-
dling temporal dependencies in sequential sensor readings, enabling accurate forecasting
of parameters like turbidity, temperature, and dissolved oxygen [98–103]. Autoencoders,
on the other hand, are often used for detecting anomalies that may signal pollution events
or system failures.
Despite their potential, the success of these models depends on the availability of dense
and high-quality sensor networks, which remain sparse in many freshwater ecosystems.
Moreover, model performance can be sensitive to sensor drift, missing data, and outliers,
while deep learning models’ limited interpretability often hinders their acceptance in
environmental decision-making contexts [104–111].

3.3. AI in Remote Sensing for Freshwater Biodiversity Monitoring

Remote sensing technologies, particularly satellites and unmanned aerial vehicles
(UAVs), have transformed the monitoring of freshwater biodiversity by providing large-
scale, high-frequency environmental data. The central biological questions involve whether
remote sensing can reliably detect ecological patterns, identify habitat changes, and monitor
environmental disturbances [112–125]. Computationally, these challenges are typically
formalized as image classification, change detection, and anomaly detection tasks.
Deep learning methods, particularly convolutional architectures, have been pivotal in
automating the extraction of meaningful ecological information from remotely sensed data.

3.3.1. Satellite- and Drone-Based AI Applications

This application addresses the biological question of whether remotely sensed imagery
from satellites or drones can be used to classify aquatic habitats and detect ecologically rel-
evant environmental changes. From a computational perspective, the challenge is typically
framed as image classification or semantic segmentation, where models assign ecologi-
cal labels to image regions or delineate habitat features. Convolutional neural networks
(CNNs) and, in some cases, recurrent neural networks (RNNs), have been successfully
employed to process and analyze satellite and drone-acquired data [119–121]. CNNs are
particularly effective at extracting spatial features from visual inputs, enabling the detection
of submerged vegetation, shoreline morphology, and water body extent [122–124]. When
analyzing temporal sequences of images, RNNs are useful in capturing dynamic changes
over time. Deep learning approaches enhance the granularity of habitat classification and
allow for detailed ecological analysis even at the microhabitat level. However, several limi-
tations remain. Satellite imagery can be hindered by cloud cover, radiometric inconsistency,
and resolution constraints, especially in freely available datasets [124–132].
Although drones provide higher resolution, they require manual deployment, limited
flight range, and complex post-processing. Moreover, deep models used in these applica-
tions require substantial amounts of labeled data and can overfit when training datasets
are small.
Biology 2025, 14, 520 15 of 41

3.3.2. Pollution Detection via Remote Sensing and AI

Pollution detection through remote sensing and AI addresses the biological ques-
tion of whether aquatic pollution events and degradation processes can be detected and
monitored effectively using image-based data. The computational task involves anomaly
detection and regression modeling, where the goal is to identify pollution signatures—such
as sediment plumes, eutrophication, or oil spills—and quantify their spatial extent or sever-
ity. Supervised machine learning classifiers, including Support Vector Machines (SVMs)
and Random Forests, along with CNN-based anomaly detection methods, are commonly
applied to hyperspectral and multispectral satellite imagery [130–132]. These approaches
allow for the extraction of spectral patterns associated with pollutants and enable monitor-
ing across large spatial scales. Time-series analyses can further aid in identifying seasonal
trends and detecting sudden disturbances [133–147].
However, the success of these methods depends heavily on access to high-resolution
imagery, reliable ground-truth calibration, and the spectral distinctiveness of pollutants.
In highly turbid or complex aquatic environments, misclassification risks increase, and
performance may degrade without proper preprocessing or regional model adaptation.

3.3.3. Land–Water Interface Analysis

The biological question addressed in this context is whether artificial intelligence
methods can reliably detect and analyze changes occurring at the land–water interface,
which are critical for understanding aquatic biodiversity dynamics. Computationally, this
problem is formalized as change detection and segmentation, where models identify shifts
in shoreline position, wetland coverage, or riparian zone configuration. Deep learning
architectures, particularly U-Net and other fully convolutional networks (FCNs), have
shown high effectiveness in extracting morphological features from satellite imagery and
elevation data [148–151]. These models support the delineation of fine-scale aquatic-
terrestrial boundaries and allow for monitoring ecosystem processes like erosion, wetland
degradation, and habitat fragmentation [150–155].
Despite their utility, land–water boundary detection is challenged by seasonal hy-
drological fluctuations, vegetation cover changes, and atmospheric effects on imagery.
Moreover, models trained in specific geographic regions may struggle to generalize else-
where without transfer learning strategies or robust domain adaptation techniques.

3.4. AI in Conservation and Management Strategies

The integration of AI technologies into conservation and management efforts offers a
paradigm shift toward more adaptive, data-driven decision-making in aquatic biodiversity
research. The biological research questions focus on whether AI can optimize conservation
interventions, enhance planning strategies, and broaden public engagement through citizen
science initiatives [154–156]. Computationally, these tasks are typically formalized as
optimization, scenario modeling, spatial prioritization, and classification problems.
By combining ecological, climatic, and socio-economic data, AI-powered systems can
dynamically respond to environmental changes and policy needs.

3.4.1. AI-Powered Decision Support Tools

AI-powered decision support tools address the biological question of whether artificial
intelligence can assist in forecasting species responses and evaluating alternative conserva-
tion strategies. These tasks are typically formalized as predictive modeling, optimization,
and decision analysis, allowing researchers to simulate ecological outcomes under different
intervention scenarios. Machine learning models, such as ensemble predictors (Random
Forests, Gradient Boosting Machines) and simulation-based optimizers, are frequently
Biology 2025, 14, 520 16 of 41

integrated into conservation decision-making pipelines [178,179]. These tools synthesize

multiple data sources—including species distributions, climate projections, land use, and
socio-economic indicators—to support adaptive conservation planning [180–182]. Ensem-
ble methods are especially robust in handling heterogeneous data, while optimization
algorithms like genetic algorithms or reinforcement learning frameworks can identify the
most effective management strategies under uncertainty.
Nonetheless, the effectiveness of these tools depends on the quality and completeness
of input data. Inadequate ecological or socio-economic datasets can introduce bias, while
the complexity and opacity of some AI models may limit stakeholder engagement and
interdisciplinary collaboration [183–185].

3.4.2. AI-Assisted Conservation Planning

AI-assisted conservation planning tackles the biological question of whether artifi-
cial intelligence can optimize the design and spatial prioritization of protected areas to
maximize biodiversity conservation. Computationally, this involves spatial optimization
and network analysis, where algorithms are used to select areas that best meet ecological
objectives under given constraints. Tools like MARXAN and Zonation, originally built
on heuristic approaches, have been enhanced through the integration of machine learn-
ing and predictive habitat modeling techniques [186–189]. These AI-driven systems can
simultaneously evaluate multiple objectives—such as species richness, habitat connectivity,
and climate resilience—enabling more nuanced reserve design. Machine learning algo-
rithms are also used to assess the performance of existing conservation areas and guide the
expansion of protected networks.
However, planning outcomes remain sensitive to subjective inputs, such as target
weights, conservation goals, and cost assumptions. Furthermore, computational complex-
ity can be a barrier when processing large-scale spatial datasets, requiring specialized
infrastructure and optimization expertise [190–192].

3.4.3. Citizen Science and AI Integration

The integration of AI with citizen science initiatives addresses the biological question
of whether non-experts can meaningfully contribute to aquatic biodiversity monitoring
through AI-supported tools. This challenge is formalized as image classification and species
recognition, where AI models embedded in mobile applications assist users in identifying
and reporting aquatic organisms. Lightweight convolutional neural networks (CNNs) are
frequently used in such platforms to enable real-time species identification with minimal
computational demand [193–195].
These models are often pre-trained on extensive biodiversity image datasets and then
fine-tuned for specific regional use cases. When deployed on smartphones, they allow citi-
zen scientists to collect valuable ecological data across large geographic areas, particularly
in regions where professional monitoring is limited. Crowdsourced observations, once
validated, can significantly enhance biodiversity databases and improve spatial coverage.
However, data quality remains a concern. Variability in camera quality, user expertise, and
sampling effort introduces noise and biases, which may reduce classification accuracy and
ecological representativeness.

3.5. Certainty of Evidence

The strength of evidence supporting AI applications in aquatic biodiversity research
varies substantially across studies and application areas. To evaluate the certainty of find-
ings, we considered several technical and methodological criteria, including the validation
strategy, data diversity, model generalizability, and reporting transparency [193–198].
Biology 2025, 14, 520 17 of 41

High confidence was attributed to studies that employed externally validated models,
reported standard performance metrics (e.g., accuracy, precision, recall, and F1-score),
and used diverse, multi-source datasets [199]. These works demonstrated robust model
generalization across spatial, temporal, and taxonomic scales, particularly in domains such
as CNN-based species image classification and LSTM-based water quality forecasting.
Moderate confidence characterized studies relying on internal validation only (e.g.,
k-fold cross-validation without external test sets), limited geographic scope, or imbalanced
datasets that may impair model generalizability [200]. Examples include regional habi-
tat modeling or acoustic classification efforts using restricted sample collections, where
overfitting remains a concern despite promising reported accuracies.
Low confidence was associated with studies that lacked methodological transparency,
did not report benchmarking against alternative methods, or minimally validated model
outputs [201,202]. In several cases, the absence of cross-study comparisons, missing data
handling, or unbalanced label distributions significantly reduced the reliability of con-
clusions, particularly in early-stage applications of deep learning to genomic sequence
classification or unsupervised remote sensing analyses.
To strengthen the certainty of future findings, it is essential to adopt standardized
evaluation protocols, including:
(a) Transparent reporting of model hyperparameters and training procedures,
(b) Consistent use of external and multi-environmental validation,
(c) Open-source publication of datasets and models where feasible,
(d) Adoption of cross-domain benchmarks to enable systematic comparisons.
Advancing the field will also require increased interdisciplinary collaboration between
domain experts in aquatic ecology and technical experts in AI and ML, ensuring that
computational advances are firmly grounded in ecological relevance.

3.6. Classification and Evaluation of AI Methods

To move beyond a simple enumeration of algorithms, we developed a structured
classification framework for evaluating AI methodologies in aquatic biodiversity research.
Each method is assessed based on the type of machine learning task it addresses, the
data modality it operates on (e.g., image, acoustic, genomic, and structured data), and its
strengths and limitations within the ecological application context.
The reviewed studies employed a wide spectrum of AI techniques (Table 3), ranging from
traditional machine learning models to modern deep learning and transformer architectures.

Detailed Evaluation Insights

A. Traditional ML models (RFs, SVMs) remain prevalent for structured ecological
datasets where feature engineering can encapsulate domain knowledge. They offer
transparent decision boundaries but often struggle with complex, high-dimensional,
or noisy inputs without preprocessing [57,58].
B. Convolutional neural networks (CNNs) dominate tasks involving raw visual input
due to their hierarchical feature extraction capabilities. However, they require large,
annotated datasets and are susceptible to adversarial noise or domain shifts [28–30].
C. Recurrent neural networks (RNNs) are well-suited for bioacoustics and time-
dependent environmental monitoring but present optimization challenges, especially
for long sequences, where vanishing gradients and temporal noise may impair learn-
ing [38–40].
D. Transformer models are emerging as promising alternatives for sequence-based eco-
logical data (e.g., DNA, multi-sensor fusion), offering improved context modeling and
Biology 2025, 14, 520 18 of 41

parallelization advantages. Nevertheless, their application is currently limited by com-

putational demands and the scarcity of domain-specific pre-training corpora [53,54].
E. Hybrid and explainable AI (XAI) approaches represent a critical future direction,
aiming to bridge the gap between predictive accuracy and ecological interpretability.
Examples include post hoc explanations using SHAP (Shapley Additive Explanations)
or integrating feature attribution into CNN outputs.

Table 3. Summary of AI classes.

Example
AI Method Type Data Modality ML Task Strengths Limitations
Algorithms
Random Forests Structured tabular Requires feature
Interpretable;
(RFs), Support data (e.g., engineering;
Classification, robust to
Traditional ML Vector Machines environmental sensitive to class
Regression overfitting; low
(SVMs), k-Nearest variables, imbalance;
computational cost
Neighbors (k-NNs) genetic markers) limited flexibility
Data-hungry;
Images (e.g., High accuracy;
Image requires extensive
Deep Learning ResNet, YOLO, species photos, automatic feature
Classification, labeling; low
(CNNs) VGGNet satellite and extraction; scalable
Object Detection interpretability
drone imagery) to large datasets
without XAI tools
Prone to
Sequential/Temporal Captures temporal overfitting;
Time-Series
Deep Learning Data (e.g., acoustic dependencies; training instability;
LSTMs, GRUs Classification,
(RNNs) signals, water effective for high
Forecasting
quality time series) sequential patterns computational
resource demand
High
Context-aware
Sequence computational
BERT, DNABERT, Genomic modeling; superior
Transformer Classification, demand; scarce
Vision sequences, scalability;
Models Multi- applications in
Transformer (ViT) multi-modal data transferable
Modal Fusion aquatic
across domains
field studies
Balances predictive Experimental;
performance with limited
Hybrid and CNNs + SHAP, Classification,
Various interpretability; standardization;
XAI Systems RFs + LIMEs Interpretation
enhances low adoption in op-
model trust erational projects

3.7. Case Studies: Real-World Applications of AI in Aquatic Biodiversity Monitoring

Several real-world implementations demonstrate the practical potential of AI methods
in aquatic biodiversity research. Each case study is analyzed based on the addressed
biological question, the formalized computational problem, the applied AI technique, the
achieved outcomes, and the identified limitations.

3.7.1. Case Study 1: Real-Time Catch Estimation in Demersal Trawl Fisheries

This case study addresses the biological question of whether aquatic species can be
detected and counted in real time during trawling operations to optimize gear deployment
and reduce bycatch. The computational challenge involves object detection and track-
ing, specifically identifying and counting individual specimens in live underwater video
streams. Avşar et al. (2023) [203] implemented a YOLOv4-based system integrated with
the SORT (Simple Online and Realtime Tracking) algorithm to detect and track Nephrops
norvegicus (Norway lobster) in trawl footage [203].
YOLOv4 was selected for its ability to perform high-speed object detection suitable
for real-time applications, while SORT supported continuous tracking and counting under
Biology 2025, 14, 520 19 of 41

dynamic conditions. The system achieved a mean average precision (mAP) of 97.8% and
80.7% accuracy in live species counts.
However, its performance was sensitive to underwater turbidity, camera placement,
and lighting, highlighting the challenges associated with deploying vision-based systems
in variable environmental conditions.

3.7.2. Case Study 2: Lightweight Fish Species Identification on Embedded Systems

This case study explores the biological question of whether fish species can be ac-
curately identified on low-power, portable devices for use in citizen science and de-
centralized biodiversity monitoring. The computational task involves image classifica-
tion, specifically recognizing fish species from images captured on resource-constrained
embedded hardware.
Chan et al. (2022) [204] developed a lightweight YOLOv4-tiny model deployed on a
Raspberry Pi 4 to identify aquarium fish species such as guppies and dwarf gouramis [204].
YOLOv4-tiny was chosen for its compact architecture, making it suitable for edge devices
with limited computational capabilities. The system achieved over 90% classification
accuracy and high mAP scores across tested species classes, demonstrating its feasibility
for real-world, low-cost monitoring.
Nonetheless, its performance depended on well-lit and moderately controlled en-
vironments. Generalizing the system to wild aquatic ecosystems would require further
retraining and dataset expansion to handle environmental variability.

3.7.3. Case Study 3: Biodiversity Text Mining with Domain-Specific Language Models
This case study examines the biological question of whether structured ecological
knowledge—such as species–habitat relationships—can be automatically extracted from
the unstructured biodiversity literature. The corresponding computational task includes
Named Entity Recognition (NER) and Relation Extraction (RE), aimed at identifying species,
habitats, and ecological relationships from scientific text.
Abdelmageed et al. (2023) [205] developed BiodivBERT, a domain-specific transformer
model pre-trained on the biodiversity literature, which significantly outperformed general-
purpose models like BERT and BioBERT in ecological text analysis tasks [205]. By fine-
tuning a transformer on biodiversity-specific corpora, the model achieved high precision
in extracting ecological entities and relations and enabled the construction of biodiversity
knowledge graphs.
However, broader adoption is hindered by the high computational costs associated
with training such models and the limited availability of high-quality, annotated corpora
for aquatic taxa.
Artificial intelligence has introduced powerful tools for advancing aquatic biodiversity
research, enabling the automation, acceleration, and refinement of ecological monitoring
and conservation practices. By formalizing biological questions into computational tasks,
such as classification, regression, and optimization, researchers can leverage a diverse array
of AI methodologies adapted to specific data modalities and ecological contexts. Although
substantial progress has been made across species identification, habitat assessment, re-
mote sensing, and conservation planning, challenges remain related to data availability,
model generalizability, and interpretability. Hybrid models integrating deep learning
with explainable AI techniques, as well as domain-specific adaptations of transformer
architectures, offer promising future directions. Continued interdisciplinary collaboration
and methodological transparency will be critical for ensuring that AI-driven insights con-
tribute effectively to the preservation of aquatic biodiversity in an increasingly complex
environmental landscape.
Biology 2025, 14, 520 20 of 41

4. Challenges and Limitations

Despite the significant advancements in artificial intelligence for aquatic biodiversity
research, several challenges and limitations hinder its full potential. AI applications in bio-
diversity monitoring, species identification, and conservation face obstacles related to data
quality, model transferability, computational constraints, methodological inconsistencies,
and ethical concerns. Addressing these issues is crucial to ensuring that AI-based tools
provide reliable and actionable insights for biodiversity conservation.

4.1. Data Quality, Availability, and Bias in AI Models

One of the fundamental challenges in AI-driven biodiversity research is the quality,
availability [203], and representativeness of training data [204]. AI models, particularly
deep learning systems [205,206], require large, well-labeled datasets to function effectively.
However, in the field of aquatic biodiversity, data collection remains sporadic, fragmented,
and often biased toward certain regions or well-studied species.
The limited access to high-quality biodiversity datasets poses a major obstacle in
AI implementation. Many ecological datasets are either proprietary, inaccessible due to
institutional restrictions or suffer from inconsistencies in data recording methodologies. Fur-
thermore, the use of AI in biodiversity research often relies on publicly available databases,
which may lack sufficient diversity or include outdated and inaccurate records [207].
Bias in AI models is another critical concern. If training datasets overrepresent certain
species or habitats while underrepresenting others, the AI system may produce skewed
predictions. For example, models trained primarily on temperate freshwater ecosystems
may struggle to accurately classify species in tropical environments [208–210]. Addition-
ally, data bias can lead to misclassification errors, disproportionately affecting rare and
endangered species that have limited available data for training.
To mitigate these issues, it is essential to develop comprehensive, standardized, and
publicly accessible biodiversity datasets. Collaboration between researchers, conservation
organizations, and governmental agencies can help improve data sharing and ensure that
AI models are trained on diverse and representative datasets [211].

4.2. Model Transferability and Generalization Issues

A persistent limitation in AI-based biodiversity research is the poor transferability
of trained models across heterogeneous environmental contexts. AI systems designed for
species identification [212,213], habitat assessment, or ecological monitoring often perform
well within the specific dataset or region on which they were trained but generalize poorly
to new ecosystems [214]. This issue, rooted in ecological variability, undermines the
scalability of AI applications in biodiversity science.
One of the main causes of low generalization is the high variability in environmental
conditions across geographic regions [215]. Factors such as seasonal shifts [216], light avail-
ability, water turbidity [217], or background noise can significantly alter the quality of data
inputs, thereby reducing the effectiveness of image recognition and remote sensing models.
Similarly, AI models trained on bioacoustic datasets from one habitat often underperform
when applied to different soundscapes due to changes in species vocalizations and ambient
interference [218].
Furthermore, many machine learning models are strongly dependent on domain-
specific features. Algorithms learn from the data they are exposed to [219], which makes
it challenging to apply them to unseen regions or species without retraining [220]. While
transfer learning techniques have been proposed to improve adaptability, they require
additional computational resources and annotated data, which may not always be available
to conservation practitioners [221].
Biology 2025, 14, 520 21 of 41

A growing body of research, however, demonstrates that these challenges can be

addressed through specific technical adaptations. For instance, Guo (2024) [222] applied
transfer learning by fine-tuning pre-trained CNNs (initially trained on ImageNet) with
locally gathered fish species images from Central Africa. This significantly improved
classification accuracy, despite substantial ecological differences from the original training
data. Transfer learning thus offers a scalable method for regional adaptation when full
model retraining is not feasible.
In addition, federated learning has emerged as a promising approach for decentralized
ecological modeling. Yang (2019) [223] and Saha (2020) [224] implemented federated
frameworks in bioacoustic monitoring networks, where edge devices trained on site-
specific audio data transmitted model updates, rather than raw data, to a central aggregator.
This strategy preserved data privacy while enabling the model to adapt to region-specific
acoustic profiles.
Despite these advances, barriers to widespread adoption remain. Transfer learning
and federated approaches are constrained by computational infrastructure, data avail-
ability, and the lack of standardized protocols across institutions. Nonetheless, such
methods point the way toward more robust, flexible, and ethically aligned AI systems for
biodiversity conservation.
To improve model generalization, future AI frameworks should integrate cross-
domain learning, data augmentation, and adaptive modeling strategies [222]. Where
possible, federated learning architectures should be prioritized to avoid central data depen-
dencies and promote inclusive, multi-regional model development [225].

4.3. Computational and Infrastructure Constraints

The deployment of AI models for biodiversity research requires substantial computa-
tional resources, which can be a limiting factor for many research institutions, particularly
those in developing regions. Deep learning models [226], such as convolutional neural
networks for species identification [227] or recurrent neural networks for bioacoustic analy-
sis [228], demand high-performance computing infrastructure, including powerful GPUs
and cloud computing access.
However, many conservation organizations, especially those operating in biodiversity-
rich but resource-limited areas, lack access to such infrastructure. The cost of AI model
training [229], particularly for deep learning applications, can be prohibitive, restricting
the adoption of AI-based biodiversity monitoring systems [230,231]. Moreover, real-time
AI applications, such as automated species recognition using underwater cameras [232] or
drones, require edge computing capabilities to process data on-site, which may not always
be available [233].
Another challenge is the energy consumption associated with AI computations [234].
Training deep neural networks requires significant computational power, leading to con-
cerns about the environmental impact of large-scale AI applications [235]. The paradox of
using AI for conservation while contributing to carbon emissions from energy-intensive
computations highlights the need for more sustainable AI frameworks [236,237].
To address these computational constraints, researchers should explore lightweight AI
models that can run on lower-resource hardware [238]. Techniques such as model pruning,
quantization, and knowledge distillation can help reduce computational requirements
without compromising accuracy. Additionally, increased investment in decentralized com-
puting infrastructures, such as edge AI and distributed processing, can improve accessibility
for biodiversity researchers in remote areas [239,240].
Biology 2025, 14, 520 22 of 41

4.4. Lack of Standardized AI Methodologies in Biodiversity Research

One of the most pressing challenges in the application of AI in biodiversity research is
the lack of standardized methodologies. AI models used for species classification, habitat
modeling [241,242], and conservation planning often employ different algorithms [243],
data preprocessing techniques, and evaluation metrics, making it difficult to compare
results across studies [244,245].
Currently, there is no universally accepted AI framework tailored for biodiversity
research. Researchers frequently develop custom AI pipelines, leading to inconsistencies in
data collection [246], model validation, and performance reporting. This lack of standard-
ization limits the reproducibility of studies and hinders the broader adoption of AI-based
tools in conservation efforts.
Moreover, many AI-based biodiversity studies do not adequately report model lim-
itations, such as overfitting, data biases, or in predictions. Without transparency in AI
methodologies, conservation practitioners may struggle to assess the reliability of AI-
generated insights for decision-making [247].
To overcome these challenges, the research community should establish best practices
for AI applications in biodiversity science. Developing standardized benchmarks, shared
biodiversity datasets, and open-source AI toolkits can enhance reproducibility and facilitate
cross-study comparisons. Additionally, incorporating explainable AI (XAI) techniques can
improve the interpretability of AI models, ensuring that their predictions are transparent
and actionable for conservation planning.

4.5. Ethical and Policy Challenges in AI-Driven Conservation

The integration of AI in biodiversity research raises ethical and policy-related concerns
that must be carefully addressed. One major ethical challenge is data privacy, particularly
when using AI to monitor biodiversity in protected areas or Indigenous lands. The collec-
tion and processing of ecological data must align with ethical guidelines that respect local
communities and biodiversity conservation laws.
A growing concern is the unintended exposure of sensitive biodiversity data through
AI-assisted monitoring. For example, real-time geotagged images or camera trap footage
could inadvertently reveal the locations of endangered species, potentially facilitating
illegal wildlife trade if accessed by malicious actors [248,249]. Some open biodiversity
platforms have been criticized for lacking adequate spatial data obfuscation, increasing the
risk of exploitation.
The use of bioacoustic monitoring technologies, often deployed in wetlands and
river systems, further raises privacy issues when devices collect ambient human sounds
alongside ecological signals. In some regions, these recordings have sparked community-
level concerns about surveillance, especially where consent protocols are lacking or
poorly communicated.
AI models applied to environmental DNA (eDNA) data introduce additional chal-
lenges. As such, models become more powerful in detecting rare or commercially valuable
species, ethical questions emerge regarding genomic privacy, the potential for bioprospect-
ing, and disputes over data ownership and use rights.
Policy challenges also arise due to the lack of regulatory frameworks governing AI
applications in conservation. Many countries lack specific legislation addressing AI-driven
ecological research, leading to legal ambiguity around data governance, accountability,
and AI model deployment. Moreover, the global disparity in infrastructure and technical
capacity means that low- and middle-income countries may face barriers in adopting
advanced AI tools equitably.
Biology 2025, 14, 520 23 of 41

To mitigate these risks, policymakers and researchers must co-develop transparent and
inclusive AI governance frameworks. Ethical AI principles, such as the CARE Principles
(Collective benefit, Authority to control, Responsibility, and Ethics), should guide data
stewardship, especially in contexts involving Indigenous knowledge and community-based
conservation. Additionally, responsible AI practices—such as fairness-aware learning, ex-
plainability tools, and bias mitigation—should be embedded from the model development
phase onward [250–252].
Collaborative efforts between governments, research institutions, and conservation
organizations are essential to establish internationally recognized standards that ensure re-
sponsible innovation while protecting ecological and human interests. Table 4 summarizes
the main ethical and technical challenges facing AI in aquatic biodiversity and proposes
solutions for each domain.

Table 4. Key challenges and solutions for AI in aquatic biodiversity research.

Challenge Proposed Solution

Data Quality and Bias Develop comprehensive, diverse datasets, improve data-sharing frameworks
Model Transferability Use transfer learning and domain adaptation techniques to enhance generalization
Computational Constraints Implement lightweight AI models, invest in cloud and edge computing
Lack of Standardization Establish standardized benchmarks and protocols for biodiversity AI
Ethical and Policy Concerns Develop transparent governance, apply ethical AI principles, engage stakeholders

While AI has the potential to transform aquatic biodiversity research, its widespread
adoption is hindered by challenges related to data availability, model generalization, compu-
tational demands, methodological inconsistencies, and ethical concerns. Addressing these
limitations requires interdisciplinary collaboration, investment in sustainable AI infrastruc-
ture, and the development of standardized methodologies. By overcoming these challenges,
AI can play a pivotal role in enhancing biodiversity monitoring, conservation planning,
and environmental decision-making for freshwater ecosystems worldwide [251–254].

5. Future Perspectives and Research Directions

The integration of artificial intelligence in aquatic biodiversity research has yielded sig-
nificant advancements, yet several challenges remain that must be addressed to maximize
its potential. Future research should focus on developing standardized AI frameworks,
improving interpretability, integrating citizen science and big data, enhancing AI’s role
in policymaking, and identifying emerging AI trends that could revolutionize freshwater
biodiversity monitoring. These research directions will ensure that AI-driven conservation
approaches are more robust, transparent, and actionable [255,256].

5.1. Need for Standardized AI Frameworks in Aquatic Ecology

The absence of standardized AI frameworks in aquatic ecology poses a major obstacle
to reproducibility, comparability, and large-scale implementation. Current AI applications
in biodiversity research rely on a diverse range of methodologies, including different
machine learning models [257], data preprocessing techniques, and evaluation metrics.
This variability makes it difficult to benchmark AI models and assess their effectiveness
across different ecosystems.
To address this issue, the development of a universal AI framework tailored for
aquatic ecology is crucial. Such a framework should provide standardized protocols for
data collection, model training, validation, and reporting. The establishment of common
performance metrics, similar to those used in other AI-driven domains, would enable
researchers to systematically compare models and refine methodologies.
Biology 2025, 14, 520 24 of 41

Furthermore, standardized datasets [258] and benchmark repositories should be cre-

ated to facilitate AI training and testing across multiple ecosystems. Open-access databases
that integrate water quality parameters [259], species distributions [260], and environmen-
tal stressors will enhance the scalability of AI models [261]. Collaboration among ecologists,
data scientists, and AI researchers is essential to ensure that such frameworks align with
ecological realities while maintaining computational efficiency.

5.2. Enhancing AI Interpretability and Explainability in Conservation

One of the most pressing concerns in AI-driven conservation is the black-box [262]
nature of many machine learning models. Deep learning models, particularly convolutional
neural networks and recurrent neural networks, often provide highly accurate predictions
but lack transparency regarding their decision-making processes [263]. This lack of inter-
pretability limits their adoption in ecological studies, where researchers and policymakers
need to understand how AI-generated insights are derived [264].
Future research should prioritize the integration of explainable AI (XAI) techniques in
aquatic biodiversity monitoring. Methods such as feature attribution analysis, attention
mechanisms, and model-agnostic interpretability tools can help to demystify AI predictions.
For instance, saliency maps can be used in computer vision models to highlight which
features in an image were most influential in species classification [263–269]. Similarly,
SHAP (Shapley Additive Explanations) and LIMEs (Local Interpretable Model-Agnostic
Explanations) can be applied in ecological risk assessment models to provide insight into
which environmental variables contribute most to predictions [267,268].
Moreover, enhancing AI interpretability is crucial for gaining stakeholder trust in
conservation initiatives [269]. By making AI outputs more transparent, conservationists and
policymakers will be better equipped to implement data-driven decisions with confidence.

5.3. Integration of AI with Citizen Science and Big Data

Citizen science initiatives have gained momentum in biodiversity monitoring, enabling
large-scale data collection by engaging non-expert volunteers in species identification,
habitat assessment, and environmental monitoring. However, the integration of AI with
citizen science remains an underutilized avenue for enhancing freshwater biodiversity
research [270].
Future studies should explore how AI models can process and validate citizen science
data in real time. AI-powered mobile applications equipped with image recognition
capabilities could enable volunteers to accurately identify species and report observations
with minimal error [271]. These applications could leverage federated learning approaches,
where AI models continuously improve by learning from decentralized datasets contributed
by citizen scientists worldwide [272].
Additionally, AI should be integrated with big data analytics platforms to enhance
biodiversity monitoring at regional and global scales [273]. By merging remote sensing
data, climate models [274], and citizen science observations, AI-driven systems can detect
long-term ecological trends, forecast biodiversity changes [275], and provide early warning
signals for habitat degradation. Such integrative approaches will make biodiversity research
more inclusive, participatory, and scalable [276].

5.4. AI in Policymaking for Biodiversity and Environmental Protection

AI has the potential to play a transformative role in policymaking for biodiversity
conservation and environmental protection. By analyzing vast amounts of ecological and
socio-economic data, AI models can provide policymakers with actionable insights for
designing effective conservation strategies [277]. However, the adoption of AI in policy
Biology 2025, 14, 520 25 of 41

frameworks remains limited due to regulatory challenges and a lack of interdisciplinary

collaboration between AI experts and environmental policymakers [278].
Future research should focus on developing AI-driven decision support systems
tailored to environmental policymaking [279]. These systems could use predictive modeling
to assess the long-term impact of conservation policies, optimize resource allocation for
protected areas, and evaluate trade-offs between economic development and biodiversity
conservation [280,281].
AI can also enhance environmental impact assessments (EIAs) by automating the
analysis of human-induced pressures on freshwater ecosystems [282]. For example, ma-
chine learning algorithms could assess the ecological consequences of dam construction,
agricultural runoff, or urbanization [283], enabling policymakers to implement mitigation
measures proactively [284].
To ensure ethical AI deployment in conservation policy, there is a need for international
guidelines and governance frameworks. Policymakers should work collaboratively with AI
researchers to establish transparent and accountable AI systems that align with biodiversity
conservation goals.

5.5. Future AI Trends in Freshwater Biodiversity Research

As AI technology continues to evolve, several emerging trends are expected to shape
the future of freshwater biodiversity research. One such trend is the increasing adoption
of self-supervised and unsupervised learning techniques [285]. Unlike traditional super-
vised learning [286], which requires large labeled datasets, self-supervised AI models can
learn meaningful patterns from unstructured biodiversity data [287], making them highly
suitable for analyzing complex ecosystems with limited labeled information [288].
Another promising trend is the use of AI-generated synthetic data for biodiversity
modeling. Synthetic data augmentation techniques, such as generative adversarial net-
works (GANs), can create realistic ecological datasets to train AI models [289], addressing
the issue of data scarcity in freshwater biodiversity studies. This approach can help improve
AI performance in regions where biodiversity data collection is challenging [290].
AI-driven robotics [289–291] and autonomous monitoring systems are also likely to
revolutionize freshwater biodiversity research. Autonomous underwater drones equipped
with AI-based imaging and environmental sensors [292,293] can conduct continuous biodi-
versity assessments in remote and inaccessible aquatic habitats [293–295]. These systems
could operate autonomously, collecting data on species distributions, water quality param-
eters, and habitat changes with minimal human intervention [296].
Additionally, AI-enabled multi-modal analysis is expected to enhance biodiversity
assessments by integrating diverse data sources, including satellite imagery [297], acoustic
recordings, genetic data, and climate models. Multi-modal AI approaches will provide a
more holistic understanding of biodiversity dynamics, allowing for more accurate predic-
tions and conservation planning [298].
Finally, the integration of blockchain technology with AI for biodiversity data manage-
ment and conservation incentives may emerge as a novel trend [299,300]. Blockchain-based
conservation frameworks could ensure data integrity, promote transparent biodiversity
monitoring [301,302], and facilitate the creation of incentive mechanisms for local commu-
nities engaged in conservation efforts [303] (Table 5).
The future of AI in freshwater biodiversity research is filled with opportunities and
challenges. Standardizing AI frameworks, enhancing interpretability, integrating AI with
citizen science and big data, embedding AI in conservation policy, and exploring emerging
AI trends will be crucial for unlocking the full potential of artificial intelligence in biodiver-
sity conservation. Interdisciplinary collaboration between AI researchers, ecologists, and
Biology 2025, 14, 520 26 of 41

policymakers will be essential to ensure that AI-driven approaches are ethical, transpar-
ent, and effective in addressing the global biodiversity crisis. By addressing these future
research directions, AI can become an indispensable tool for monitoring, preserving, and
restoring freshwater ecosystems worldwide.

Table 5. Future research directions in AI for freshwater biodiversity.

Research Direction Expected Impact

Standardized AI Frameworks Ensures comparability and reproducibility of studies across ecosystems
AI Interpretability and Explainability Improves trust and usability of AI-driven insights for conservation
Integration with Citizen Science and Big Data Expands data collection and analysis through crowdsourcing and automation
AI in Policymaking Enhances evidence-based conservation policies and decision-making
Emerging AI Trends Explores novel AI methodologies to advance biodiversity research

6. Conclusions
The application of artificial intelligence in aquatic biodiversity research has demon-
strated its transformative potential in species identification, habitat assessment, ecological
risk modeling, and conservation planning. AI-driven methodologies have significantly en-
hanced biodiversity monitoring by automating complex processes, analyzing vast datasets,
and improving the accuracy of ecological predictions. This systematic review has high-
lighted the growing adoption of AI technologies in freshwater biodiversity research, reveal-
ing both their advantages and limitation.
A key takeaway from this review is the increasing role of AI in species identification,
particularly through deep learning models such as convolutional neural networks (CNNs)
and bioacoustic analysis. AI has enabled the rapid classification of freshwater organisms,
including rare and cryptic species, with unprecedented precision. Additionally, environ-
mental DNA (eDNA)-based AI applications have opened new frontiers in non-invasive
biodiversity monitoring, allowing for the detection of species presence through genetic
material in water samples.
In the realm of habitat assessment and ecological risk modeling, AI has proven valu-
able in predicting habitat changes, assessing pollution risks, and identifying emerging
environmental threats. Machine learning models have been instrumental in analyzing
complex ecological data, offering insights into the effects of climate change, land-use mod-
ifications and anthropogenic activities on freshwater ecosystem. However, challenges
remain in ensuring that AI-generated predictions are robust, transferable across different
ecosystems, and supported by high-quality training data.
Remote sensing and AI integration have facilitated large-scale biodiversity monitor-
ing using satellite imagery, drone-based surveys, and water pollution detection models.
These advancements have enabled researchers to monitor ecosystem health at spatial and
temporal resolutions previously unattainable through conventional fieldwork. Despite
this progress, concerns related to the accessibility of remote sensing data, computational
demands, and model accuracy persist.
One of the most promising applications of AI lies in conservation planning and
decision-making, where AI-powered tools have aided in designing protected areas, opti-
mizing resource allocation, and developing adaptive conservation strategies. AI has also
facilitated the expansion of citizen science initiatives, empowering non-experts to contribute
to biodiversity research through AI-driven mobile applications. However, ethical concerns
regarding data privacy, algorithmic bias, and the potential misuse of AI in biodiversity
exploitation must be carefully managed.
Despite its remarkable potential, AI in biodiversity research faces significant chal-
lenges, including issues related to data quality, model generalization, computational infras-
Biology 2025, 14, 520 27 of 41

tructure, lack of standardization, and ethical considerations. The effectiveness of AI models

is highly dependent on the availability of comprehensive and unbiased datasets, which
remain limited for many freshwater ecosystems. Furthermore, the lack of standardized AI
frameworks has hindered cross-study comparability and reproducibility, emphasizing the
need for universally accepted methodologies in AI-driven biodiversity research.
Looking ahead, future research should focus on developing standardized AI frame-
works, improving model interpretability, integrating AI with citizen science and big data
platforms, and embedding AI in policymaking for biodiversity conservation. Emerging AI
trends, such as self-supervised learning, synthetic data generation, autonomous AI-driven
monitoring systems, and blockchain-integrated conservation strategies, hold promise for
further advancing biodiversity research.
In conclusion, AI has the potential to revolutionize freshwater biodiversity conserva-
tion by providing scalable, efficient, and precise tools for ecological monitoring. However,
to fully harness its benefits, interdisciplinary collaboration among AI researchers, ecologists,
policymakers, and conservation practitioners is essential. Addressing current limitations
and ensuring ethical AI deployment will be crucial in shaping the future of biodiversity re-
search, ultimately contributing to more effective conservation strategies and the protection
of freshwater ecosystems worldwide.

Author Contributions: Conceptualization, T.M., G.M., I.D., P.K. and P.B.; methodology, T.M., G.M.,
I.D., P.K. and P.B.; formal analysis, T.M., G.M., I.D., P.K. and P.B.; investigation, T.M., G.M., I.D., P.K.
and P.B.; resources, T.M., G.M., I.D., P.K. and P.B.; writing—original draft preparation, T.M., G.M.,
I.D., P.K. and P.B.; writing—review and editing, T.M., G.M., I.D., P.K. and P.B.; visualization, T.M.,
G.M., I.D., P.K. and P.B. All authors have read and agreed to the published version of the manuscript.

Funding: This research was co-financed by the Minister of Science (Poland) under the “Regional
Excellence Initiative” Program for 2024–2027 (RID/SP/0045/2024/01) for University of Szczecin,
Institute of Marine and Environmental Sciences.

Institutional Review Board Statement: Not applicable.

Informed Consent Statement: Not applicable.

Data Availability Statement: All data generated and analyzed during this study are available in the
Open Science Framework (OSF) repository and can be accessed at: [Link] (accessed on
11 March 2025).

Conflicts of Interest: The authors declare no conflicts of interest.

Abbreviations
The following abbreviations are used in this manuscript:

AI Artificial intelligence
CNN Convolutional neural network
DL Deep learning
eDNA Environmental DNA
EIA Environmental impact assessment
GAN Generative adversarial network
GPS Global Positioning System
ML Machine learning
NLP Natural language processing
PRISMA Preferred Reporting Items for Systematic Reviews and Meta-Analyses
QUADAS-2 Quality Assessment of Diagnostic Accuracy Studies-2
RNN Recurrent neural network
Biology 2025, 14, 520 28 of 41

RoB 2 Risk of Bias 2 Tool

SHAP Shapley Additive Explanations
SVM Support Vector Machine
UAV Unmanned aerial vehicle
XAI Explainable artificial intelligence
YOLO You Only Look Once (Object Detection Algorithm)

References
1. Rizqi Basuki, N.A. You Only Look Once v8 for Fish Species Identification. IAES Int. J. Artif. Intell. 2024, 13, 3314–3321. [CrossRef]
2. Smith, J.; Wycherley, A.; Mulvaney, J.; Lennane, N.; Reynolds, E.; Monks, C.-A.; Evans, T.; Mooney, T.; Fancourt, B. Man versus
Machine: Cost and Carbon Emission Savings of 4G-Connected Artificial Intelligence Technology for Classifying Species in
Camera Trap Images. Sci. Rep. 2024, 14, 14530. [CrossRef] [PubMed]
3. Podgorski, J.; Wu, R.; Chakravorty, B.; Polya, D.A. Groundwater Arsenic Distribution in India by Machine Learning Geospatial
Modeling. Int. J. Environ. Res. Public Health 2020, 17, 7119. [CrossRef]
4. Song, Y.; Pan, Y.; Xiang, M.; Yang, W.; Zhan, D.; Wang, X.; Lu, M. A WebGIS-Based System for Supporting Saline–Alkali Soil
Ecological Monitoring: A Case Study in Yellow River Delta, China. Remote Sens. 2024, 16, 1948. [CrossRef]
5. Barone, M.; Mollen, F.H.; Giles, J.L.; Marshall, L.J.; Villate-Moreno, M.; Mazzoldi, C.; Pérez-Costas, E.; Heine, J.; Guisande, C.
Performance of iSharkFin in the Identification of Wet Dorsal Fins from Priority Shark Species. Ecol. Inform. 2022, 68, 101514.
[CrossRef]
6. Beloiu, M.; Heinzmann, L.; Rehush, N.; Gessler, A.; Griess, V.C. Individual Tree-Crown Detection and Species Identification in
Heterogeneous Forests Using Aerial RGB Imagery and Deep Learning. Remote Sens. 2023, 15, 1463. [CrossRef]
7. Moe, S.J.; Carriger, J.F.; Glendell, M. Increased Use of Bayesian Network Models Has Improved Environmental Risk Assessments.
Integr. Environ. Assess. Manag. 2021, 17, 53–61. [CrossRef]
8. Imani, M.; Hasan, M.M.; Bittencourt, L.F.; McClymont, K.; Kapelan, Z. A Novel Machine Learning Application: Water Quality
Resilience Prediction Model. Sci. Total Environ. 2021, 768, 144459. [CrossRef]
9. Uddin, M.G.; Imran, M.H.; Sajib, A.M.; Hasan, M.A.; Diganta, M.T.M.; Dabrowski, T.; Olbert, A.I.; Moniruzzaman, M. Assessment
of Human Health Risk from Potentially Toxic Elements and Predicting Groundwater Contamination Using Machine Learning
Approaches. J. Contam. Hydrol. 2024, 261, 104307. [CrossRef]
10. Sörensen, J.; Nilsson, E.; Nilsson, D.; Gröndahl, E.; Rehn, D.; Giertz, T. Evaluation of ANN Model for Pipe Status Assessment in
Drinking Water Management. Water Supply 2024, 24, 1985–1998. [CrossRef]
11. Alfwzan, W.F.; Selim, M.M.; Almalki, A.S.; Alharbi, I.S. Water Quality Assessment Using Bi-LSTM and Computational Fluid
Dynamics (CFD) Techniques. Alex. Eng. J. 2024, 97, 346–359. [CrossRef]
12. Domisch, S.; Kakouei, K.; Martínez-López, J.; Bagstad, K.J.; Magrach, A.; Balbi, S.; Villa, F.; Funk, A.; Hein, T.; Borgwardt, F.; et al.
Social Equity Shapes Zone-Selection: Balancing Aquatic Biodiversity Conservation and Ecosystem Services Delivery in the
transboundary Danube River Basin. Sci. Total Environ. 2019, 656, 797–807. [CrossRef] [PubMed]
13. Hussain, A.; Barua, B.; Osman, A.; Abozariba, R.; Taufiq Asyhari, A. Performance of MobileNetV3 Transfer Learning on Handheld
Device-Based Real-Time Tree Species Identification. In Proceedings of the 2021 26th International Conference on Automation and
Computing (ICAC), Portsmouth, UK, 2–4 September 2021.
14. Di Marco, M. Reptile Research Shows New Avenues and Old Challenges for Extinction Risk Modelling. PLoS Biol. 2022,
20, e3001719. [CrossRef]
15. Alghanmi, A.F.; Aljahdali, B.M.; Sulaimani, H.T.; Turan, O.; Alshareef, M.H. An Innovative Deep-Learning Technique for Fuel
Demand Estimation in Maritime Transportation: A Step Toward Sustainable Development and Environmental Impact Mitigation.
Water 2024, 16, 3325. [CrossRef]
16. Buters, T.M.; Bateman, P.W.; Robinson, T.; Belton, D.; Dixon, K.W.; Cross, A.T. Methodological Ambiguity and Inconsistency
Constrain Unmanned Aerial Vehicles as a Silver Bullet for Monitoring Ecological Restoration. Remote Sens. 2019, 11, 1180.
[CrossRef]
17. Cabrera-Ariza, A.M.; Peralta-Aguilera, M.; Henríquez-Hernández, P.V.; Santelices-Moya, R. Using UAVs and Machine Learning
for Nothofagus Alessandrii Species Identification in Mediterranean Forests. Drones 2023, 7, 668. [CrossRef]
18. Burroughs, R.W.; Parham, J.F.; Stuart, B.L.; Smits, P.D.; Angielczyk, K.D. Morphological Species Delimitation in The Western Pond
Turtle (Actinemys): Can Machine Learning Methods Aid in Cryptic Species Identification? Integr. Org. Biol. 2024, 6, obae010.
[CrossRef]
19. Schneider, S.; Taylor, G.W.; Kremer, S.C.; Fryxell, J.M. Getting the Bugs out of AI: Advancing Ecological Research on Arthropods
through Computer Vision. Ecol. Lett. 2023, 26, 1247–1258. [CrossRef]
Biology 2025, 14, 520 29 of 41

20. Spiesman, B.J.; Gratton, C.; Hatfield, R.G.; Hsu, W.H.; Jepsen, S.; McCornack, B.; Patel, K.; Wang, G. Assessing the Potential for
Deep Learning and Computer Vision to Identify Bumble Bee Species from Images. Sci. Rep. 2021, 11, 7580. [CrossRef]
21. Yuan, F.; Huang, Y.; Chen, X.; Cheng, E. A Biological Sensor System Using Computer Vision for Water Quality Monitoring. IEEE
Access 2018, 6, 61535–61546. [CrossRef]
22. Rocha, P.C.; Romano, P.S.R. The Shape of Sound: A New R Package That Crosses the Bridge between Bioacoustics and Geometric
Morphometrics. Methods Ecol. Evol. 2021, 12, 1115–1121. [CrossRef]
23. Liu, Y.; Che, S.; Ai, L.; Song, C.; Zhang, Z.; Zhou, Y.; Yang, X.; Xian, C. Camouflage Detection: Optimization-Based Computer
Vision for Alligator Sinensis with Low Detectability in Complex Wild Environments. Ecol. Inform. 2024, 83, 102802. [CrossRef]
24. Unnikrishnan, R.; Sumod, M.; Jayaraj, R.; Sujanapal, P.; Dev, S.A. The Efficacy of Machine Learning Algorithm for Raw Drug
Authentication in Coscinium fenestratum (Gaertn.) Colebr. Employing a DNA Barcode Database. Physiol. Mol. Biol. Plants 2021, 27,
605–617. [CrossRef]
25. Rebelo, A.R.; Fagundes, J.M.G.; Digiampietri, L.A.; Francoy, T.M.; Bíscaro, H.H. A Fully Automatic Classification of Bee Species
from Wing Images. Apidologie 2021, 52, 1060–1074. [CrossRef]
26. Brito de Jesus, S.; Vieira, D.; Gheller, P.; Cunha, B.P.; Gallucci, F.; Fonseca, G. Machine Learning Algorithms Accurately Identify
Free-Living Marine Nematode Species. PeerJ 2023, 11, e16216. [CrossRef] [PubMed]
27. Kulits, P.; Wall, J.; Bedetti, A.; Henley, M.; Beery, S. ElephantBook: A Semi-Automated Human-in-the-Loop System for Elephant
Re-Identification. In Proceedings of the 4th ACM SIGCAS Conference on Computing and Sustainable Societies, Virtual, 28 June–2
July 2021; pp. 88–98.
28. Huang, L.; Zhang, Z.; Xing, H.; Luo, Y.; Yang, J.; Sui, X.; Wang, Y. Risk Assessment Based on Dose-Responsive and Time-
Responsive Genes to Build PLS-DA Models for Exogenously Induced Lung Injury. Ecotoxicol. Environ. Saf. 2023, 256, 114891.
[CrossRef]
29. Anderson, C.B. The CCB-ID Approach to Tree Species Mapping with Airborne Imaging Spectroscopy. PeerJ 2018, 2018, e5666.
[CrossRef]
30. Dritsas, E.; Trigka, M. Efficient Data-Driven Machine Learning Models for Water Quality Prediction. Computation 2023, 11, 16.
[CrossRef]
31. Deng, T.; Chau, K.-W.; Duan, H.-F. Machine Learning Based Marine Water Quality Prediction for Coastal Hydro-Environment
Management. J. Environ. Manag. 2021, 284, 112051. [CrossRef]
32. Liu, Z.; Jin, Y.; Yang, L.; Yuan, X.; Yan, L.; Zhang, Y.; Zhang, H.; Xu, M.; Song, X.; Tang, J.; et al. Improving Prediction for Potential
Spawning Areas from a Two-Step Perspective: A Comparison of Multi-Model Approaches for Sparse Egg Distribution. J. Sea Res.
2024, 197, 102460. [CrossRef]
33. Albraikan, A.A.; Aljebreen, M.; Alzahrani, J.S.; Othman, M.; Mohammed, G.P.; Ibrahim Alsaid, M. Modified Barnacles Mating
Optimization with Deep Learning Based Weed Detection Model for Smart Agriculture. Appl. Sci. 2022, 12, 12828. [CrossRef]
34. Chowdury, M.S.U.; Emran, T.B.; Ghosh, S.; Pathak, A.; Alam, M.M.; Absar, N.; Andersson, K.; Hossain, M.S. IoT Based Real-Time
River Water Quality Monitoring System. Procedia Comput. Sci. 2019, 155, 161–168. [CrossRef]
35. Nazarov, D.; Sulimin, V.; Shvedov, V. Advancing Environmental Stewardship: The Role of Automation in Enhanced Environmental
Monitoring. E3S Web Conf. 2024, 542, 05005. [CrossRef]
36. Upton, A.; Jefferson, B.; Moore, G.; Jarvis, P. Rapid Gravity Filtration Operational Performance Assessment and Diagnosis for
Preventative Maintenance from On-Line Data. Chem. Eng. J. 2017, 313, 250–260. [CrossRef]
37. Sinha, K.; Srivastava, D.K.; Bhatnagar, R. Water Quality Management through Data Driven Intelligence System in Barmer Region,
Rajasthan. Procedia Comput. Sci. 2018, 132, 314–322. [CrossRef]
38. Guisasola, A.; Baeza, J.A. Smart-Plant Decision Support System (SP-DSS): Defining a Multi-Criteria Decision-Making Framework
for the Selection of WWTP Configurations with Resource Recovery. J. Clean. Prod. 2022, 367, 132873. [CrossRef]
39. Almodfer, R.; Mudhsh, M.; Zhao, J. Pyramided and Optimized Blurred Shape Model for Plant Leaf Classification. IET Image
Process. 2023, 17, 2838–2854. [CrossRef]
40. Habib, S.; Ahmad, M.; Haq, Y.U.; Sana, R.; Muneer, A.; Waseem, M.; Pathan, M.S.; Dev, S. Advancing Taxonomic Classification
through Deep Learning: A Robust Artificial Intelligence Framework for Species Identification Using Natural Images. IEEE Access
2024, 12, 146718–146732. [CrossRef]
41. Yang, B.; Zhang, Z.; Yang, C.-Q.; Wang, Y.; Orr, M.C.; Wang, H.; Zhang, A.-B. Identification of Species by Combining Molecular
and Morphological Data Using Convolutional Neural Networks. Syst. Biol. 2022, 71, 690–705. [CrossRef]
42. Hou, Y. Implementation of Water Quality Management by Fish School Detection Based on Computer Vision Technology. Adv. J.
Food Sci. Technol. 2015, 9, 422–427. [CrossRef]
43. Piechaud, N.; Hunt, C.; Culverhouse, P.F.; Foster, N.L.; Howell, K.L. Automated Identification of Benthic Epifauna with Computer
Vision. Mar. Ecol. Prog. Ser. 2019, 615, 15–30. [CrossRef]
Biology 2025, 14, 520 30 of 41

44. Woo, D.K.; Ji, J.; Song, H. Subsurface Drainage Pipe Detection Using an Ensemble Learning Approach and Aerial Images. Agric.
Water Manag. 2023, 287, 108455. [CrossRef]
45. Blair, J.D.; Gaynor, K.M.; Palmer, M.S.; Marshall, K.E. A Gentle Introduction to Computer Vision-Based Specimen Classification in
Ecological Datasets. J. Anim. Ecol. 2024, 93, 147–158. [CrossRef]
46. Ficsór, M.; Csabai, Z. Machine Learning Model Ensemble Based on Multi-Scale Predictors Confirms Ecological Segregation
and Accurately Predicts the Occurrence of Net-Spinning Caddisfly Larvae Species Groups (Trichoptera: Hydropsychidae) at
Catchment-Scale. Ecol. Indic. 2023, 146, 109769. [CrossRef]
47. Li, J.; Liu, C.; Wang, L.; Liu, Y.; Li, R.; Lu, X.; Lu, J.; Shen, J. Multi-Species Identification and Number Counting of Fish Passing
through Fishway at Hydropower Stations with LigTraNet. Ecol. Inform. 2024, 82, 102704. [CrossRef]
48. Norman, D.L.; Bischoff, P.H.; Wearn, O.R.; Ewers, R.M.; Rowcliffe, J.M.; Evans, B.; Sethi, S.; Chapman, P.M.; Freeman, R. Can
CNN-Based Species Classification Generalise across Variation in Habitat within a Camera Trap Survey? Methods Ecol. Evol. 2023,
14, 242–251. [CrossRef]
49. Robillard, A.J.; Trizna, M.G.; Ruiz-Tafur, M.; Dávila Panduro, E.L.; de Santana, C.D.; White, A.E.; Dikow, R.B.; Deichmann, J.L.
Application of a Deep Learning Image Classifier for Identification of Amazonian Fishes. Ecol. Evol. 2023, 13, 9987. [CrossRef]
[PubMed]
50. Figueroa-Mata, G.; Mata-Montero, E.; Valverde-Otárola, J.C.; Arias-Aguilar, D.; Zamora-Villalobos, N. Using Deep Learning to
Identify Costa Rican Native Tree Species From Wood Cut Images. Front. Plant Sci. 2022, 13, 789227. [CrossRef]
51. Norouzzadeh, M.S.; Nguyen, A.; Kosmala, M.; Swanson, A.; Palmer, M.S.; Packer, C.; Clune, J. Automatically Identifying,
Counting, and Describing Wild Animals in Camera-Trap Images with Deep Learning. Proc. Natl. Acad. Sci. USA 2018, 115,
E5716–E5725. [CrossRef]
52. Alhashmi, F.; Alhefeiti, M.; Mirza, S.B.; Ridouane, F.L. Efficient Dorsal Fin-Based Classification of Risso’s and Common Bottlenose
Dolphins Using YOLOv7 and YOLOv8 Models for Real-Time Applications. Int. J. Adv. Technol. Eng. Explor. 2024, 11, 875–887.
[CrossRef]
53. Cau, A.; Sbrana, A.; Franceschini, S.; Fiorentino, F.; Follesa, M.C.; Galgani, F.; Garofalo, G.; Gerigny, O.; Profeta, A.; Rinelli, P.; et al.
What, Where, and When: Spatial-Temporal Distribution of Macro-Litter on the Seafloor of the Western and Central Mediterranean
Sea. Environ. Pollut. 2024, 342, 123028. [CrossRef] [PubMed]
54. Lago, A.; Patel, S.; Singh, A. Low-Cost Real-Time Aerial Object Detection and GPS Location Tracking Pipeline. ISPRS Open J.
Photogramm. Remote Sens. 2024, 13, 100069. [CrossRef]
55. Norouzzadeh, M.S.; Morris, D.; Beery, S.; Joshi, N.; Jojic, N.; Clune, J. A Deep Active Learning System for Species Identification
and Counting in Camera Trap Images. Methods Ecol. Evol. 2021, 12, 150–161. [CrossRef]
56. Xhina, E.; Bilo, I.; Ktona, A.; Paparisto, A.; Liçi, O.; Qirinxhi, X.; Haxhiu, D. Machine learning methods for understanding
biodiversity and its conservation. Int. J. Ecosyst. Ecol. Sci. 2023, 13, 39–44. [CrossRef]
57. Yang, S.; Liang, R.; Chen, J.; Wang, Y.; Li, K. Estimating the Water Quality Index Based on Interpretable Machine Learning Models.
Water Sci. Technol. 2024, 89, 1340–1356. [CrossRef] [PubMed]
58. Ntalaperas, D.; Christophoridis, C.; Angelidis, I.; Iossifidis, D.; Touloupi, M.-F.; Vergeti, D.; Politi, E. Intelligent Tools to Monitor,
Control and Predict Wastewater Reclamation and Reuse. Sensors 2022, 22, 3068. [CrossRef]
59. Cha, S.; Lee, J.; Choi, E.; Lim, J. Unveiling the Past: Deep-Learning-Based Estimation of Historical Peatland Distribution. Land
2024, 13, 328. [CrossRef]
60. Dema, T.; Zhang, L.; Towsey, M.; Truskinger, A.; Sherub, S.; Zhang, J.; Brereton, M.; Roe, P. An Investigation into Acoustic
Analysis Methods for Endangered Species Monitoring: A Case of Monitoring the Critically Endangered White-Bellied Heron in
Bhutan. In Proceedings of the 2017 IEEE 13th International Conference on e-Science (e-Science), Auckland, New Zealand, 24–27
October 2017; pp. 177–186.
61. Clements, E.; Thompson, K.A.; Hannoun, D.; Dickenson, E.R.V. Classification Machine Learning to Detect de Facto Reuse and
Cyanobacteria at a Drinking Water Intake. Sci. Total Environ. 2024, 948, 174690. [CrossRef]
62. Alric, B.; Dézerald, O.; Meyer, A.; Billoir, E.; Coulaud, R.; Larras, F.; Mondy, C.P.; Usseglio-Polatera, P. How Diatom-, Invertebrate-
and Fish-Based Diagnostic Tools Can Support the Ecological Assessment of Rivers in a Multi-Pressure Context: Temporal Trends
over the Past Two Decades in France. Sci. Total Environ. 2021, 762, 143915. [CrossRef]
63. Neupane, N.; Goswami, R.; Harrison, K.; Oberhauser, K.; Ries, L.; McCormick, C. Artificial Intelligence Correctly Classifies
Developmental Stages of Monarch Caterpillars Enabling Better Conservation through the Use of Community Science Photographs.
Sci. Rep. 2024, 14, 27039. [CrossRef]
64. Mustafa, H.M.; Hayder, G.; Abba, S.I.; Algarni, A.D.; Mnzool, M.; Nour, A.H. Performance Evaluation of Hydroponic Wastewater
Treatment Plant Integrated with Ensemble Learning Techniques: A Feature Selection Approach. Processes 2023, 11, 478. [CrossRef]
65. Dubinsky, E.A.; Butkus, S.R.; Andersen, G.L. Microbial Source Tracking in Impaired Watersheds Using PhyloChip and Machine-
Learning Classification. Water Res. 2016, 105, 56–64. [CrossRef]
Biology 2025, 14, 520 31 of 41

66. Anees, S.A.; Mehmood, K.; Khan, W.R.; Sajjad, M.; Alahmadi, T.A.; Alharbi, S.A.; Luo, M. Integration of Machine Learning and
Remote Sensing for above Ground Biomass Estimation through Landsat-9 and Field Data in Temperate Forests of the Himalayan
Region. Ecol. Inform. 2024, 82, 102732. [CrossRef]
67. Li, C.; Ma, X.; Teng, Y.; Li, S.; Jin, Y.; Du, J.; Jiang, L. Quantitative Analysis of Forest Water COD Value Based on UV–Vis and FLU
Spectral Information Fusion. Forests 2023, 14, 1361. [CrossRef]
68. Stefanni, S.; Mirimin, L.; Stanković, D.; Chatzievangelou, D.; Bongiorni, L.; Marini, S.; Modica, M.V.; Manea, E.; Bonofiglio, F.;
del Rio Fernandez, J.; et al. Framing Cutting-Edge Integrative Deep-Sea Biodiversity Monitoring via Environmental DNA and
Optoacoustic Augmented Infrastructures. Front. Mar. Sci. 2022, 8, 797140. [CrossRef]
69. Hartvig, I.; Czako, M.; Kjær, E.D.; Nielsen, L.R.; Theilade, I. The Use of DNA Barcoding in Identification and Conservation of
Rosewood (Dalbergia spp.). PLoS ONE 2015, 10, e0138231. [CrossRef]
70. Meher, P.K.; Sahu, T.K.; Gahoi, S.; Tomar, R.; Rao, A.R. FunbarRF: DNA Barcode-Based Fungal Species Prediction Using Multiclass
Random Forest Supervised Learning Model. BMC Genet. 2019, 20, 2. [CrossRef]
71. Paylar, B.; Längkvist, M.; Jass, J.; Olsson, P.-E. Utilization of Computer Classification Methods for Exposure Prediction and Gene
Selection in Daphnia Magna Toxicogenomics. Biology 2023, 12, 692. [CrossRef] [PubMed]
72. Elsayed, A.; Ghaith, M.; Yosri, A.; Li, Z.; El-Dakhakhni, W. Genetic Programming Expressions for Effluent Quality Prediction:
Towards AI-Driven Monitoring and Management of Wastewater Treatment Plants. J. Environ. Manag. 2024, 356, 120510. [CrossRef]
73. Cai, R. Automating Bird Species Classification: A Deep Learning Approach with CNNs. J. Phys. Conf. Ser. 2023, 2664, 012007.
[CrossRef]
74. Duong, T.K.C.; Tran, V.L.; Nguyen, T.B.; Nguyen, T.T.; Ho, N.T.K.; Nguyen, T.Q. Ensemble Learning-Based Approach for
Automatic Classification of Termite Mushrooms. Front. Genet. 2023, 14, 1208695. [CrossRef] [PubMed]
75. Pasanisi, E.; Pace, D.S.; Orasi, A.; Vitale, M.; Arcangeli, A. A Global Systematic Review of Species Distribution Modelling
Approaches for Cetaceans and Sea Turtles. Ecol. Inform. 2024, 82, 102700. [CrossRef]
76. Bernardelli, A.; Marsili-Libelli, S.; Manzini, A.; Stancari, S.; Tardini, G.; Montanari, D.; Anceschi, G.; Gelli, P.; Venier, S. Real-Time
Model Predictive Control of a Wastewater Treatment Plant Based on Machine Learning. Water Sci. Technol. 2020, 81, 2391–2400.
[CrossRef]
77. Meshgi, B.; Hanafi-Bojd, A.A.; Fathi, S.; Modabbernia, G.; Meshgi, K.; Shadman, M. Multi-Scale Habitat Modeling Framework for
Predicting the Potential Distribution of Sheep Gastrointestinal Nematodes across Iran’s Three Distinct Climatic Zones: A MaxEnt
Machine-Learning Algorithm. Sci. Rep. 2024, 14, 2828. [CrossRef] [PubMed]
78. Gebler, D.; Kolada, A.; Pasztaleniec, A.; Szoszkiewicz, K. Modelling of Ecological Status of Polish Lakes Using Deep Learning
Techniques. Environ. Sci. Pollut. Res. 2021, 28, 5383–5397. [CrossRef] [PubMed]
79. Guénard, G.; Morin, J.; Matte, P.; Secretan, Y.; Valiquette, E.; Mingelbier, M. Deep Learning Habitat Modeling for Moving
Organisms in Rapidly Changing Estuarine Environments: A Case of Two Fishes. Estuar. Coast. Shelf Sci. 2020, 238, 106713.
[CrossRef]
80. Abdulkareem, S.A.; Augustijn, E.-W.; Filatova, T.; Musial, K.; Mustafa, Y.T. Risk Perception and Behavioral Change during
Epidemics: Comparing Models of Individual and Collective Learning. PLoS ONE 2020, 15, e0226483. [CrossRef]
81. O’Malley, D.; Delorey, A.A.; Guiltinan, E.J.; Ma, Z.; Kadeethum, T.; Lackey, G.; Lee, J.; Santos, J.E.; Follansbee, E.; Nair, M.C.; et al.
Unlocking Solutions: Innovative Approaches to Identifying and Mitigating the Environmental Impacts of Undocumented Orphan
Wells in the United States. Environ. Sci. Technol. 2024, 58, 19584–19594. [CrossRef]
82. Satish, N.; Anmala, J.; Rajitha, K.; Varma, M.R.R. A Stacking ANN Ensemble Model of ML Models for Stream Water Quality
Prediction of Godavari River Basin, India. Ecol. Inform. 2024, 80, 102500. [CrossRef]
83. Msaddek, M.H.; Moumni, Y.; Zouhri, L.; Chenini, I.; Zghibi, A. Groundwater Quality Evaluation of Fractured Aquifers Using
Machine Learning Models and Hydrogeochemical Approaches to Sustainable Water-Irrigation Security in Arid Climate (Central
Tunisia). Water 2023, 15, 3332. [CrossRef]
84. Araya, D.; Podgorski, J.; Berg, M. Groundwater Salinity in the Horn of Africa: Spatial Prediction Modeling and Estimated People
at Risk. Environ. Int. 2023, 176, 107925. [CrossRef] [PubMed]
85. Maddikunta, P.K.R.; Srivastava, G.; Gadekallu, T.R.; Deepa, N.; Boopathy, P. Predictive Model for Battery Life in IoT Networks.
IET Intell. Transp. Syst. 2020, 14, 1388–1395. [CrossRef]
86. Eze, E.; Halse, S.; Ajmal, T. Developing a Novel Water Quality Prediction Model for a South African Aquaculture Farm. Water
2021, 13, 1782. [CrossRef]
87. Koudenoukpo, Z.C.; Odountan, O.H.; Agboho, P.A.; Dalu, T.; Van Bocxlaer, B.; Janssens de Bistoven, L.; Chikou, A.; Back-
eljau, T. Using Self–Organizing Maps and Machine Learning Models to Assess Mollusc Community Structure in Relation to
Physicochemical Variables in a West Africa River–Estuary System. Ecol. Indic. 2021, 126, 107706. [CrossRef]
88. Kukartsev, V.; Orlov, V.; Semenova, E.; Rozhkova, A. Optimizing Water Quality Classification Using Random Forest and Machine
Learning. BIO Web Conf. 2024, 130, 03007. [CrossRef]
Biology 2025, 14, 520 32 of 41

89. Luzorata, J.G.; Bocobo, A.E.; Detera, L.M.; Pocong, N.J.B.; Sajonia, A.P. Assessment of Land Use Land Cover Classification
Using Support Vector Machine and Random Forest Techniques in the Agusan River Basin through Geospatial Techniques.
In Proceedings of the International Exchange and Innovation Conference on Engineering & Sciences (IEICES), Fukuoka, Japan,
19–20 October 2023; Volume 9, pp. 240–246.
90. Rather, T.A.; Kumar, S.; Khan, J.A. Multi-Scale Habitat Modelling and Predicting Change in the Distribution of Tiger and Leopard
Using Random Forest Algorithm. Sci. Rep. 2020, 10, 11473. [CrossRef] [PubMed]
91. Soumya, S.; Tamatgar, N.; Dilli, R.; Kanthi, M. Deployment of Random Forest Algorithm for Prediction of Ammonia in River
Water. In Proceedings of the 2024 13th International Conference on Software and Computer Applications, Bali Island, Indonesia,
1–3 February 2024; pp. 18–23.
92. Jose, A.; Yasala, S. Machine Learning-Based Ensemble Model for Groundwater Quality Prediction: A Case Study. Water Pract.
Technol. 2024, 19, 2364–2375. [CrossRef]
93. Chung, J.; Yoo, G.; Jo, J.-S.; Choi, J.; Lee, J.-H. Development of a Machine Learning Model to Estimate the Biotic Ligand
Model–Based Predicted No-Effect Concentrations for Copper in Freshwater. Environ. Toxicol. Chem. 2023, 42, 2271–2283.
[CrossRef]
94. Fan, J.; Li, M.; Guo, F.; Yan, Z.; Zheng, X.; Zhang, Y.; Xu, Z.; Wu, F. Priorization of River Restoration by Coupling Soil and Water
Assessment Tool (SWAT) and Support Vector Machine (SVM) Models in the Taizi River Basin, Northern China. Int. J. Environ. Res.
Public Health 2018, 15, 2090. [CrossRef]
95. Morovati, M.; Karami, P.; Amjas, F.B. Accessing Habitat Suitability and Connectivity for the Westernmost Population of Asian
Black Bear (Ursus Thibetanus Gedrosianus, Blanford, 1877) Based on Climate Changes Scenarios in Iran. PLoS ONE 2020,
15, e0242432. [CrossRef]
96. Shi, K.; Han, J.-C.; Wang, P. Near Real-Time Retrieval of Lake Surface Water Temperature Using Himawari-8 Satellite Imagery
and Machine Learning Techniques: A Case Study in the Yangtze River Basin. Front. Environ. Sci. 2023, 11, 1335725. [CrossRef]
97. Kuang, L.; Shi, P.; Hua, C.; Chen, B.; Zhu, H. An Enhanced Extreme Learning Machine for Dissolved Oxygen Prediction in
Wireless Sensor Networks. IEEE Access 2020, 8, 198730–198739. [CrossRef]
98. Tiyasha, T.; Tung, T.M.; Bhagat, S.K.; Tan, M.L.; Jawad, A.H.; Mohtar, W.H.M.W.; Yaseen, Z.M. Functionalization of Remote Sensing
and On-Site Data for Simulating Surface Water Dissolved Oxygen: Development of Hybrid Tree-Based Artificial Intelligence
Models. Mar. Pollut. Bull. 2021, 170, 112639. [CrossRef]
99. Zhao, Y.; Shen, J.; Feng, J.; Sun, Z.; Sun, T.; Liu, D.; Xi, M.; Li, R.; Wang, X. The Estimation of Chemical Oxygen Demand of Erhai
Lake Basin and Its Links with Dom Fluorescent Components Using Machine Learning. Water 2021, 13, 3629. [CrossRef]
100. Knoll, L.; Breuer, L.; Bach, M. Nation-Wide Estimation of Groundwater Redox Conditions and Nitrate Concentrations through
Machine Learning. Environ. Res. Lett. 2020, 15, 064004. [CrossRef]
101. Qiao, Z.; Sun, S.; Jiang, Q.; Xiao, L.; Wang, Y.; Yan, H. Retrieval of Total Phosphorus Concentration in the Surface Water of Miyun
Reservoir Based on Remote Sensing Data and Machine Learning Algorithms. Remote Sens. 2021, 13, 4662. [CrossRef]
102. Ju, F. Mapping the Knowledge Structure of Image Recognition in Cultural Heritage: A Scientometric Analysis Using CiteSpace,
VOSviewer, and Bibliometrix. J. Imaging 2024, 10, 272. [CrossRef] [PubMed]
103. Cuesta, F.; Calderón-Loor, M.; Rosero, P.; Calispa, M.; Zisling, H.; Pérez-Castillo, Y.; Echevarría, G.; Ríos-Touma, B. Seasonally
Flooded Landscape Connectivity and Implications for Fish in the Napo Moist Forest: A High-Resolution Mapping Approach.
Glob. Ecol. Conserv. 2024, 56, e03257. [CrossRef]
104. Cushman, S.A.; Kilshaw, K.; Campbell, R.D.; Kaszta, Z.; Gaywood, M.; Macdonald, D.W. Comparing the Performance of Global,
Geographically Weighted and Ecologically Weighted Species Distribution Models for Scottish Wildcats Using GLM and Random
Forest Predictive Modeling. Ecol. Model. 2024, 492, 110691. [CrossRef]
105. Bartkiewicz, E.; Zimoch, I. Analysis of the Risk of Pipe Breaks Based on Hydraulic Model. In Safety and Reliability—Safe Societies in
a Changing World; CRC Press: Boca Raton, FL, USA, 2018; pp. 1511–1516.
106. Liu, Y.; Engel, B.A.; Flanagan, D.C.; Gitau, M.W.; McMillan, S.K.; Chaubey, I.; Singh, S. Modeling Framework for Representing
Long-Term Effectiveness of Best Management Practices in Addressing Hydrology and Water Quality Problems: Framework
Development and Demonstration Using a Bayesian Method. J. Hydrol. 2018, 560, 530–545. [CrossRef]
107. Sun, X.; Cao, W.; Pan, D.; Li, Y.; Ren, Y.; Nan, T. Assessment of Aquifer Specific Vulnerability to Total Nitrate Contamination
Using Ensemble Learning and Geochemical Evidence. Sci. Total Environ. 2024, 912, 169497. [CrossRef]
108. Aswal, R.S.; Prasad, M.; Singh, J.; Singh, H.; Shrivastava, U.; Wadhwa, M.; Pandey, O.P.; Egbueri, J.C. Spatial Analysis and Soft
Computational Modeling for Hazard Assessment of Potential Toxic Elements in Potable Groundwater. Sci. Rep. 2024, 14, 25473.
[CrossRef] [PubMed]
109. Luo, M.; Liu, M.; Zhang, S.; Gao, J.; Zhang, X.; Li, R.; Lin, X.; Wang, S. Mining Soil Heavy Metal Inversion Based on Levy Flight
Cauchy Gaussian Perturbation Sparrow Search Algorithm Support Vector Regression (LSSA-SVR). Ecotoxicol. Environ. Saf. 2024,
287, 117295. [CrossRef]
Biology 2025, 14, 520 33 of 41

110. Batchelor, J.L.; Hudak, A.T.; Gould, P.; Moskal, L.M. Terrestrial and Airborne Lidar to Quantify Shrub Cover for Canada Lynx
(Lynx canadensis) Habitat Using Machine Learning. Remote Sens. 2023, 15, 4434. [CrossRef]
111. Karimzadeh Motlagh, Z.; Derakhshani, R.; Sayadi, M.H. Groundwater Vulnerability Assessment in Central Iran: Integration of
GIS-Based DRASTIC Model and a Machine Learning Approach. Groundw. Sustain. Dev. 2023, 23, 101037. [CrossRef]
112. Bazzo, C.O.G.; Kamali, B.; dos Santos Vianna, M.; Behrend, D.; Hueging, H.; Schleip, I.; Mosebach, P.; Haub, A.; Behrendt, A.;
Gaiser, T. Integration of UAV-Sensed Features Using Machine Learning Methods to Assess Species Richness in Wet Grassland
Ecosystems. Ecol. Inform. 2024, 83, 102813. [CrossRef]
113. Banerjee, P.; Ghose, M.K.; Pradhan, R. AHP-Based Spatial Analysis of Water Quality Impact Assessment Due to Change in
Vehicular Traffic Caused by Highway Broadening in Sikkim Himalaya. Appl. Water Sci. 2018, 8, 72. [CrossRef]
114. Shamsuddin, I.I.S.; Othman, Z.; Sani, N.S. Water Quality Index Classification Based on Machine Learning: A Case from the
Langat River Basin Model. Water 2022, 14, 2939. [CrossRef]
115. Fuentes-Pérez, J.F.; Sanz-Ronda, F.J. A Custom Sensor Network for Autonomous Water Quality Assessment in Fish Farms.
Electronics 2021, 10, 2192. [CrossRef]
116. Park, J.; Kim, K.T.; Lee, W.H. Recent Advances in Information and Communications Technology (ICT) and Sensor Technology for
Monitoring Water Quality. Water 2020, 12, 510. [CrossRef]
117. Karthick, K.; Krishnan, S.; Manikandan, R. Water Quality Prediction: A Data-Driven Approach Exploiting Advanced Machine
Learning Algorithms with Data Augmentation. J. Water Clim. Change 2024, 15, 431–452. [CrossRef]
118. Stoica, C.; Camejo, J.; Banciu, A.; Nita-Lazar, M.; Paun, I.; Cristofor, S.; Pacheco, O.R.; Guevara, M. Water Quality of Danube
Delta Systems: Ecological Status and Prediction Using Machine-Learning Algorithms. Water Sci. Technol. 2016, 73, 2413–2421.
[CrossRef] [PubMed]
119. Razzano, F.; Di Stasio, P.; Mauro, F.; Meoni, G.; Esposito, M.; Schirinzi, G.; Ullo, S.L. AI Techniques for Near Real-Time Monitoring
of Contaminants in Coastal Waters on Board Future Φsat-2 Mission. IEEE J. Sel. Top. Appl. Earth Obs. Remote Sens. 2024, 17,
16755–16766. [CrossRef]
120. Wang, Y.; Ho, I.W.-H.; Chen, Y.; Wang, Y.; Lin, Y. Real-Time Water Quality Monitoring and Estimation in AIoT for Freshwater
Biodiversity Conservation. IEEE Internet Things J. 2022, 9, 14366–14374. [CrossRef]
121. Ma, Y.; Song, K.; Wen, Z.; Liu, G.; Shang, Y.; Lyu, L.; Du, J.; Yang, Q.; Li, S.; Tao, H.; et al. Remote Sensing of Turbidity for Lakes in
Northeast China Using Sentinel-2 Images with Machine Learning Algorithms. IEEE J. Sel. Top. Appl. Earth Obs. Remote Sens. 2021,
14, 9132–9146. [CrossRef]
122. Chojnacki, A.; Dai, C.; Farahi, A.; Shi, G.; Webb, J.; Zhang, D.T.; Abernethy, J.; Schwartz, E. A Data Science Approach to
Understanding Residential Water Contamination in Flint. In Proceedings of the 23rd ACM SIGKDD International Conference on
Knowledge Discovery and Data Mining, Halifax, NS, Canada, 13–15 August 2017; Volume F129685; pp. 1407–1416.
123. Fan, R.; Deng, Y.; Du, Y.; Xie, X. Predicting Geogenic Groundwater Arsenic Contamination Risk in Floodplains Using Interpretable
Machine-Learning Model. Environ. Pollut. 2024, 340, 122787. [CrossRef]
124. Kumar, S.; Pati, J. Assessment of Groundwater Arsenic Contamination Using Machine Learning in Varanasi, Uttar Pradesh, India.
J. Water Health 2022, 20, 829–848. [CrossRef]
125. Lu, H.; Ding, A.; Zheng, Y.; Jiang, J.; Zhang, J.; Zhang, Z.; Xu, P.; Zhao, X.; Quan, F.; Gao, C.; et al. Securing Drinking Water Supply
in Smart Cities: An Early Warning System Based on Online Sensor Network and Machine Learning. Aqua Water Infrastruct.
Ecosyst. Soc. 2023, 72, 721–738. [CrossRef]
126. Blazejewski, A.; Pecolt, S.; Królikowski, T.; Grunt, M.; Bielicki, F. Project of Innovative Open Multi-Domain Early Warning
Platform Enviwise for Adverse Events in Water Bodies and Streams. Procedia Comput. Sci. 2023, 225, 2743–2753. [CrossRef]
127. Li, X.; Liang, G.; Wang, L.; Yang, Y.; Li, Y.; Li, Z.; He, B.; Wang, G. Identifying the Spatial Pattern and Driving Factors of Nitrate in
Groundwater Using a Novel Framework of Interpretable Stacking Ensemble Learning. Environ. Geochem. Health 2024, 46, 482.
[CrossRef]
128. Jeong, B.; Chapeta, M.R.; Kim, M.; Kim, J.; Shin, J.; Cha, Y.K. Machine Learning-Based Prediction of Harmful Algal Blooms in
Water Supply Reservoirs. Water Qual. Res. J. 2022, 57, 304–318. [CrossRef]
129. Lee, S.; Lee, D. Improved Prediction of Harmful Algal Blooms in Four Major South Korea’s Rivers Using Deep Learning Models.
Int. J. Environ. Res. Public Health 2018, 15, 1322. [CrossRef] [PubMed]
130. Park, J.; Baek, J.; Kim, J.; You, K.; Kim, K. Deep Learning-Based Algal Detection Model Development Considering Field
Application. Water 2022, 14, 1275. [CrossRef]
131. Yi, H.-S.; Park, S.; An, K.-G.; Kwak, K.-C. Algal Bloom Prediction Using Extreme Learning Machine Models at Artificial Weirs in
the Nakdong River, Korea. Int. J. Environ. Res. Public Health 2018, 15, 2078. [CrossRef] [PubMed]
132. Singh, Y.; Walingo, T. Smart Water Quality Monitoring with IoT Wireless Sensor Networks. Sensors 2024, 24, 2871. [CrossRef]
[PubMed]
Biology 2025, 14, 520 34 of 41

133. Rodríguez-López, L.; Usta, D.B.; Duran-Llacer, I.; Alvarez, L.B.; Yépez, S.; Bourrel, L.; Frappart, F.; Urrutia, R. Estimation of
Water Quality Parameters through a Combination of Deep Learning and Remote Sensing Techniques in a Lake in Southern Chile.
Remote Sens. 2023, 15, 4157. [CrossRef]
134. Bekmurzaeva, R.; Kalimullin, R.; Aguzarova, F. Application of Drones and Artificial Intelligence to Monitor and Protect Natural
Ecosystems. BIO Web Conf. 2024, 140, 01008. [CrossRef]
135. Xiao, Y.; Guo, Y.; Yin, G.; Zhang, X.; Shi, Y.; Hao, F.; Fu, Y. UAV Multispectral Image-Based Urban River Water Quality Monitoring
Using Stacked Ensemble Machine Learning Algorithms—A Case Study of the Zhanghe River, China. Remote Sens. 2022, 14, 3272.
[CrossRef]
136. El Din, E.S.; Zhang, Y. Improving the Accuracy of Extracting Surface Water Quality Levels (SWQLs) Using Remote Sensing and
Artificial Neural Network: A Case Study in the Saint John River, Canada. Int. Arch. Photogramm. Remote Sens. Spat. Inf. Sci. 2017,
42, 245–249. [CrossRef]
137. Sánchez-Azofeifa, A.; Rivard, B.; Wright, J.; Feng, J.-L.; Li, P.; Chong, M.M.; Bohlman, S.A. Estimation of the Distribution of
Tabebuia Guayacan (Bignoniaceae) Using High-Resolution Remote Sensing Imagery. Sensors 2011, 11, 3831–3851. [CrossRef]
138. Uddin, M.G.; Nash, S.; Rahman, A.; Dabrowski, T.; Olbert, A.I. Data-Driven Modelling for Assessing Trophic Status in Marine
Ecosystems Using Machine Learning Approaches. Environ. Res. 2024, 242, 117755. [CrossRef] [PubMed]
139. Wei, S.; Hogue, D.; Mondal, I.; Xu, T.; Boyer, T.H.; Hamilton, K.A.; Richard, R.; Hamilton, K.A. High Resolution Data Visualization
and Machine Learning Prediction of Free Chlorine Residual in a Green Building Water System. Water Res. X 2024, 24, 100244.
[CrossRef] [PubMed]
140. Redoloza, F.S.; Williamson, T.N.; Headman, A.O.; Allred, B.J. Machine-Learning Model to Delineate Sub-Surface Agricultural
Drainage from Satellite Imagery. J. Environ. Qual. 2023, 52, 907–921. [CrossRef]
141. Tripathi, R.N.; Ramachandran, A.; Agarwal, K.; Tripathi, V.; Badola, R.; Hussain, S.A. UAV and Deep Learning: Detection of
Selected Riparian Species along the Ganga River. ISPRS Int. Arch. Photogramm. Remote. Sens. Spat. Inf. Sci. 2024, 48, 637–642.
[CrossRef]
142. Wang, H.; Feng, C.; Li, X.; Yang, Y.; Zhang, Y.; Su, J.; Luo, D.; Wei, D.; He, Y. Plant Species Diversity Assessment in the Temperate
Grassland Region of China Using UAV Hyperspectral Remote Sensing. Diversity 2024, 16, 775. [CrossRef]
143. Moura, M.M.; de Oliveira, L.E.S.; Sanquetta, C.R.; Bastos, A.; Mohan, M.; Corte, A.P.D. Towards Amazon Forest Restoration:
Automatic Detection of Species from Uav Imagery. Remote Sens. 2021, 13, 2627. [CrossRef]
144. Ou, J.; Tian, Y.; Zhang, Q.; Xie, X.; Zhang, Y.; Tao, J.; Lin, J. Coupling UAV Hyperspectral and LiDAR Data for Mangrove
Classification Using XGBoost in China’s Pinglu Canal Estuary. Forests 2023, 14, 1838. [CrossRef]
145. Nasim, S.; Oussalah, M.; Klöve, B.; Haghighi, A.T. Vegetation Height Estimation Using Ubiquitous Foot-Based Wearable Platform.
Environ. Monit. Assess. 2020, 192, 774. [CrossRef]
146. Abad-Uribarren, A.; Prado, E.; Sierra, S.; Cobo, A.; Rodríguez-Basalo, A.; Gómez-Ballesteros, M.; Sánchez, F. Deep Learning-
Assisted High Resolution Mapping of Vulnerable Habitats within the Capbreton Canyon System, Bay of Biscay. Estuar. Coast.
Shelf Sci. 2022, 275, 107957. [CrossRef]
147. Detka, J.; Coyle, H.; Gomez, M.; Gilbert, G.S. A Drone-Powered Deep Learning Methodology for High Precision Remote Sensing
in California’s Coastal Shrubs. Drones 2023, 7, 421. [CrossRef]
148. Xie, Y.; Xiang, J.; Li, X.; Yang, C. An Intelligent Fishery Detection Method Based on Cross-Domain Image Feature Fusion. Fishes
2024, 9, 338. [CrossRef]
149. Aguilar-Amuchastegui, N.; Riveros, J.C.; Forrest, J.L. Identifying Areas of Deforestation Risk for REDD+ Using a Species Modeling
Tool. Carbon Balance Manag. 2014, 9, 10. [CrossRef] [PubMed]
150. Yanes Luis, S.; Gutiérrez-Reina, D.; Toral Marín, S. Censored Deep Reinforcement Patrolling with Information Criterion for
Monitoring Large Water Resources Using Autonomous Surface Vehicles. Appl. Soft Comput. 2023, 132, 109874. [CrossRef]
151. Mert, B.K.; Kasapoğulları, D. A Case Study of Using Artificial Neural Networks to Predict Heavy Metal Pollution in Lake Iznik.
Environ. Monit. Assess. 2024, 196, 586. [CrossRef]
152. Shao, W.; Liu, J.; Zhang, H.; Yan, D.; Li, W. Countermeasure Analysis on Promoting Drinking Water Safety in Shanshan County,
Xinjiang Autonomous Region, China. Water 2018, 10, 1022. [CrossRef]
153. Perivolioti, T.-M.; Tušer, M.; Frouzova, J.; Znachor, P.; Rychtecký, P.; Mouratidis, A.; Terzopoulos, D.; Bobori, D. Estimating
Environmental Preferences of Freshwater Pelagic Fish Using Hydroacoustics and Satellite Remote Sensing. Water 2019, 11, 2226.
[CrossRef]
154. Coffer, M.M.; Nezlin, N.P.; Bartlett, N.; Pasakarnis, T.; Lewis, T.N.; DiGiacomo, P.M. Satellite Imagery as a Management Tool for
Monitoring Water Clarity across Freshwater Ponds on Cape Cod, Massachusetts. J. Environ. Manag. 2024, 355, 120334. [CrossRef]
155. Liao, K.; Song, Y.; Nie, X.; Liu, L.; Qi, S. Suspended Sediment Concentrate Estimation From Landsat Imagery and Hydrological
Station in Poyang Lake Using Machine Learning. IEEE Access 2024, 12, 85411–85422. [CrossRef]
Biology 2025, 14, 520 35 of 41

156. Liang, C.-P.; Sun, C.-C.; Suk, H.; Wang, S.-W.; Chen, J.-S. A Machine Learning Approach for Spatial Mapping of the Health Risk
Associated with Arsenic-Contaminated Groundwater in Taiwan’s Lanyang Plain. Int. J. Environ. Res. Public Health 2021, 18, 11385.
[CrossRef]
157. Bourillon, B.; Feunteun, E.; Acou, A.; Trancart, T.; Teichert, N.; Belpaire, C.; Dufour, S.; Bustamante, P.; Aarestrup, K.;
Walker, A.; et al. Anthropogenic Contaminants Shape the Fitness of the Endangered European Eel: A Machine Learning Approach.
Fishes 2022, 7, 274. [CrossRef]
158. Arrighi, C.; Castelli, F. Prediction of Ecological Status of Surface Water Bodies with Supervised Machine Learning Classifiers. Sci.
Total Environ. 2023, 857, 159655. [CrossRef] [PubMed]
159. Nasiri Khiavi, A.; Tavoosi, M.; Khodamoradi, H.; Kuriqi, A. Integration of Watershed Eco-Physical Health through Algorithmic
Game Theory and Supervised Machine Learning. Groundw. Sustain. Dev. 2024, 26, 101216. [CrossRef]
160. Al-Najjar, H.; Ceribasi, G.; Ceyhunlu, A.I. Effect of Unconventional Water Resources Interventions on the Management of Gaza
Coastal Aquifer in Palestine. Water Supply 2021, 21, 4205–4218. [CrossRef]
161. Chen, K.; Wang, H.; Valverde-Pérez, B.; Zhai, S.; Vezzaro, L.; Wang, A. Optimal Control towards Sustainable Wastewater
Treatment Plants Based on Multi-Agent Reinforcement Learning. Chemosphere 2021, 279, 130498. [CrossRef]
162. Laneve, G.; Téllez, A.; Kallikkattil Kuruvila, A.; Bruno, M.; Messineo, V. Eutrophication and HAB Occurrence Control in Lakes of
Different Origins: A Multi-Source Remote Sensing Detection Strategy. Remote Sens. 2024, 16, 1792. [CrossRef]
163. Al-Sulttani, A.O.; Al-Mukhtar, M.; Roomi, A.B.; Farooque, A.A.; Khedher, K.M.; Yaseen, Z.M. Proposition of New Ensemble
Data-Intelligence Models for Surface Water Quality Prediction. IEEE Access 2021, 9, 108527–108541. [CrossRef]
164. Khan, Y.; See, C.S. Predicting and Analyzing Water Quality Using Machine Learning: A Comprehensive Model. In Proceedings of
the 2016 IEEE Long Island Systems, Applications and Technology Conference (LISAT), Farmingdale, NY, USA, 29 April 2016.
165. Liu, Y.; Tian, W.; Xie, J.; Huang, W.; Xin, K. LSTM-Based Model-Predictive Control with Rationality Verification for Bioreactors in
Wastewater Treatment. Water 2023, 15, 1779. [CrossRef]
166. Ahmed, M.F.; Lim, C.K.; Mokhtar, M.B.; Khirotdin, R.P.K. Predicting Arsenic (As) Exposure on Human Health for Better
Management of Drinking Water Sources. Int. J. Environ. Res. Public Health 2021, 18, 7997. [CrossRef] [PubMed]
167. Aslam, B.; Maqsoom, A.; Cheema, A.H.; Ullah, F.; Alharbi, A.; Imran, M. Water Quality Management Using Hybrid Machine
Learning and Data Mining Algorithms: An Indexing Approach. IEEE Access 2022, 10, 119692–119705. [CrossRef]
168. Landuyt, D.; Lemmens, P.; D’hondt, R.; Broekx, S.; Liekens, I.; De Bie, T.; Declerck, S.A.J.; De Meester, L.; Goethals, P.L.M.
An Ecosystem Service Approach to Support Integrated Pond Management: A Case Study Using Bayesian Belief Networks—
Highlighting Opportunities and Risks. J. Environ. Manag. 2014, 145, 79–87. [CrossRef]
169. Mpakairi, K.S.; Muthivhi, F.F.; Dondofema, F.; Munyai, L.F.; Dalu, T. Chlorophyll-a Unveiled: Unlocking Reservoir Insights
through Remote Sensing in a Subtropical Reservoir. Environ. Monit. Assess. 2024, 196, 401. [CrossRef]
170. Chang, N.-B.; Wei, X.; Mostafiz, C.; Yang, Y.J.; Weiss, J.; Belavel, M. Reconstruction of Sea-Land Interactions between Terrestrial
Vegetation Cover and Water Quality Constituents in the Mattapoisett Harbor Area during the 1991 Hurricane Bob Event. Int. J.
Appl. Earth Obs. Geoinf. 2019, 83, 101929. [CrossRef] [PubMed]
171. Yanes Luis, S.; Shutin, D.; Marchal Gómez, J.; Gutiérrez Reina, D.; Toral Marín, S. Deep Reinforcement Multiagent Learning
Framework for Information Gathering with Local Gaussian Processes for Water Monitoring. Adv. Intell. Syst. 2024, 6, 202300850.
[CrossRef]
172. Brehob, M.M.; Pennino, M.J.; Handler, A.M.; Compton, J.E.; Lee, S.S.; Sabo, R.D. Estimates of Lake Nitrogen, Phosphorus,
and Chlorophyll-a Concentrations to Characterize Harmful Algal Bloom Risk Across the United States. Earth’s Future 2024, 12,
e2024EF004493. [CrossRef]
173. Kimothi, S.; Thapliyal, A.; Singh, R.; Rashid, M.; Gehlot, A.; Akram, S.V.; Javed, A.R. Comprehensive Database Creation for
Potential Fish Zones Using IoT and ML with Assimilation of Geospatial Techniques. Sustainability 2023, 15, 1062. [CrossRef]
174. El Morabet, R.; Barhazi, L.; Bouhafa, S.; Dahim, M.A.; Khan, R.A.; Khan, N.A. Geospatial Distribution and Machine Learning
Algorithms for Assessing Water Quality in Surface Water Bodies of Morocco. Sci. Rep. 2023, 13, 20599. [CrossRef]
175. Booth, N.L.; Everman, E.J.; Kuo, I.-L.; Sprague, L.; Murphy, L. A Web-Based Decision Support System for Assessing Regional
Water-Quality Conditions and Management Actions. J. Am. Water Resour. Assoc. 2011, 47, 1136–1150. [CrossRef]
176. Tal-maon, M.; Ostfeld, A. Developing Water Quality Formulations for a Semi-Distributed Rainfall–Runoff Model. Water 2024,
16, 2072. [CrossRef]
177. Falconí-López, A.; Mitesser, O.; Martin Schaefer, H.; Blüthgen, N.; Busse, A.; Feldhaar, H.; Freile, J.; Gelis, R.; Grella, N.;
Heibl, C.; et al. Habitat Niches of Bird Species along a Recovery Gradient in the Chocó Tropical Forest. Ecol. Indic. 2024,
166, 112260. [CrossRef]
178. Liu, H.; Jin, Y.; Roche, L.M.; O’Geen, A.T.; Dahlgren, R.A. Understanding Spatial Variability of Forage Production in California
Grasslands: Delineating Climate, Topography and Soil Controls. Environ. Res. Lett. 2021, 16, 014043. [CrossRef]
179. Dinar, A.; Quinn, N.W.T. Developing a Decision Support System for Regional Agricultural Nonpoint Salinity Pollution Manage-
ment: Application to the San Joaquin River, California. Water 2022, 14, 2384. [CrossRef]
Biology 2025, 14, 520 36 of 41

180. Cojbasic, S.; Dmitrasinovic, S.; Kostic, M.; Sekulic, M.T.; Radonic, J.; Dodig, A.; Stojkovic, M. Application of Machine Learning in
River Water Quality Management: A Review. Water Sci. Technol. 2023, 88, 2297–2308. [CrossRef]
181. Derdour, A.; Jodar-Abellan, A.; Pardo, M.Á.; Ghoneim, S.S.M.; Hussein, E.E. Designing Efficient and Sustainable Predictions of
Water Quality Indexes at the Regional Scale Using Machine Learning Algorithms. Water 2022, 14, 2801. [CrossRef]
182. Mallya, G.; Hantush, M.M.; Govindaraju, R.S. A Machine Learning Approach to Predict Watershed Health Indices for Sediments
and Nutrients at Ungauged Basins. Water 2023, 15, 586. [CrossRef]
183. Di, Z.; Chang, M.; Guo, P. Water Quality Evaluation of the Yangtze River in China Using Machine Learning Techniques and Data
Monitoring on Different Time Scales. Water 2019, 11, 339. [CrossRef]
184. Lundström, J.; Öhman, K.; Laudon, H. Comparing Buffer Zone Alternatives in Forest Planning Using a Decision Support System.
Scand. J. For. Res. 2018, 33, 493–501. [CrossRef]
185. Zia, H.; Harris, N.R.; Merrett, G.V.; Rivers, M. A Low-Complexity Machine Learning Nitrate Loss Predictive Model-Towards
Proactive Farm Management in a Networked Catchment. IEEE Access 2019, 7, 26707–26720. [CrossRef]
186. Li, R.; Yan, T.; Zhao, Y.; Zhang, J.; Zhang, L. A Reinforcement Learning-Based Approach for Active Optimization of High
Proportional Distributed Power Supply Access to Rural Distribution Networks. J. Phys. Conf. Ser. 2023, 2465, 012033. [CrossRef]
187. Lima Silva, S.L.D.; Leite, M.S.; Fernandes, T.C.R.L.; Silva, S.K.D.; Araújo, A.C.B.D. Optimal operation of domestic and industrial
sewage treatment plants using machine learning methods. Rev. Gest. Soc. Ambient. 2023, 17, e04124. [CrossRef]
188. Doherty, K.E.; Evans, J.S.; Coates, P.S.; Juliusson, L.M.; Fedy, B.C. Importance of Regional Variation in Conservation Planning: A
Rangewide Example of the Greater Sage-Grouse. Ecosphere 2016, 7, 1462. [CrossRef]
189. Er-Rousse, O.; Qafas, A. Artificial Intelligence for the Optimization of Marine Aquaculture. E3S Web Conf. 2024, 477, 00102.
[CrossRef]
190. Fischer, A.; ter Laak, T.; Bronders, J.; Desmet, N.; Christoffels, E.; van Wezel, A.; van der Hoek, J.P. Decision Support for Water
Quality Management of Contaminants of Emerging Concern. J. Environ. Manag. 2017, 193, 360–372. [CrossRef] [PubMed]
191. Wei, H.; Huang, Y.; Santiago, P.J.; Labachyan, K.E.; Ronaghi, S.; Magana, M.P.B.; Huang, Y.-H.; Jiang, S.C.; Hochbaum, A.I.;
Ragan, R. Decoding the Metabolic Response of Escherichia Coli for Sensing Trace Heavy Metals in Water. Proc. Natl. Acad. Sci.
USA 2023, 120, 2210061120. [CrossRef] [PubMed]
192. Zhang, Q.; Bostic, J.T.; Sabo, R.D. Regional Patterns and Drivers of Total Nitrogen Trends in the Chesapeake Bay Watershed:
Insights from Machine Learning Approaches and Management Implications. Water Res. 2022, 218, 118443. [CrossRef]
193. Duan, L.; Yang, L.; Guo, Y. SIAlex: Species Identification and Monitoring Based on Bird Sound Features. Ecol. Inform. 2024,
81, 102637. [CrossRef]
194. Mäder, P.; Boho, D.; Rzanny, M.; Seeland, M.; Wittich, H.C.; Deggelmann, A.; Wäldchen, J. The Flora Incognita App–Interactive
Plant Species Identification. Methods Ecol. Evol. 2021, 12, 1335–1342. [CrossRef]
195. Nokelainen, O.; Schiestl-Aalto, P.; Lauha, P.; Somervuo, P.; Andrejeff, S.; Sundell, J.; Hänninen, J.; Talaskivi, J.; Inkinen, J.;
Vallinmäki, M.; et al. A Mobile Application–Based Citizen Science Product to Compile Bird Observations. Citiz. Sci. Theory Pract.
2024, 9, 710. [CrossRef]
196. Mahecha, M.D.; Rzanny, M.; Kraemer, G.; Mäder, P.; Seeland, M.; Wäldchen, J. Crowd-Sourced Plant Occurrence Data Provide a
Reliable Description of Macroecological Gradients. Ecography 2021, 44, 1131–1142. [CrossRef]
197. Thornhill, I.; Batty, L.; Death, R.G.; Friberg, N.R.; Ledger, M.E. Local and Landscape Scale Determinants of Macroinvertebrate
Assemblages and Their Conservation Value in Ponds across an Urban Land-Use Gradient. Biodivers. Conserv. 2017, 26, 1065–1086.
[CrossRef]
198. Zhang, Y.; Qin, J.; Li, M.; Han, Q.; Gruen, A.; Li, D.; Zhong, J. Advancing Coral Structural Connectivity Analysis through Deep
Learning and Remote Sensing: A Case Study of South Pacific Tetiaroa Island. ISPRS Int. Arch. Photogramm. Remote. Sens. Spat. Inf.
Sci. 2024, XLVIII-2-2, 471–476. [CrossRef]
199. Uddin, M.G.; Nash, S.; Rahman, A.; Olbert, A.I. A Sophisticated Model for Rating Water Quality. Sci. Total Environ. 2023,
868, 161614. [CrossRef]
200. Jakovljevic, G.; Álvarez-Taboada, F.; Govedarica, M. Long-Term Monitoring of Inland Water Quality Parameters Using Landsat
Time-Series and Back-Propagated ANN: Assessment and Usability in a Real-Case Scenario. Remote Sens. 2024, 16, 68. [CrossRef]
201. Thorburn, P.J.; Biggs, J.S.; McCosker, K.; Northey, A. Assessing Water Quality for Cropping Management Practices: A New
Approach for Dissolved Inorganic Nitrogen Discharged to the Great Barrier Reef. J. Environ. Manag. 2022, 321, 115932. [CrossRef]
202. Yao, S.; Zhang, Y.; Wang, P.; Xu, Z.; Wang, Y.; Zhang, Y. Long-Term Water Quality Prediction Using Integrated Water Quality
Indices and Advanced Deep Learning Models: A Case Study of Chaohu Lake, China, 2019–2022. Appl. Sci. 2022, 12, 11329.
[CrossRef]
203. Duttagupta, S.; Bhanja, S.N.; Dutta, A.; Sarkar, S.; Chakraborty, M.; Ghosh, A.; Mondal, D.; Mukherjee, A. Impact of COVID-19
Lockdown on Availability of Drinking Water in the Arsenic-Affected Ganges River Basin. Int. J. Environ. Res. Public Health 2021,
18, 2832. [CrossRef] [PubMed]
Biology 2025, 14, 520 37 of 41

204. Wei, Z.; Wu, N.; Zou, Q.; Zou, H.; Zhu, L.; Wei, J.; Huang, H. Data Modeling of Sewage Treatment Plant Based on Long Short-Term
Memory with Multilayer Perceptron Network. Water 2023, 15, 1472. [CrossRef]
205. Sáinz-Pardo Díaz, J.; Castrillo, M.; López García, Á. Deep Learning Based Soft-Sensor for Continuous Chlorophyll Estimation on
Decentralized Data. Water Res. 2023, 246, 120726. [CrossRef]
206. Utku, A.; Ayaz, Z.; Ciftci, D.; Ali Akcayol, M. Deep Learning Based Classification for Hoverflies (Diptera: Syrphidae). J. Entomol.
Res. Soc. 2023, 25, 529–544. [CrossRef]
207. Pérez-Girón, J.C.; Díaz-Varela, E.R.; Álvarez-Álvarez, P.; Hernández Palacios, O.; Ballesteros, F.; López-Bao, J.V. Linking Landscape
Structure and Vegetation Productivity with Nut Consumption by the Cantabrian Brown Bear during Hyperphagia. Sci. Total
Environ. 2022, 813, 152610. [CrossRef]
208. Pukhrambam, B.; Sahayadhas, A. Advanced Medicinal Plant Classification and Bioactivity Identification Based on Dense Net
Architecture. Int. J. Adv. Comput. Sci. Appl. 2022, 13, 104–109. [CrossRef]
209. Rahman, D.A.; Sitorus, A.B.Y.; Condro, A.A. From Coastal to Montane Forest Ecosystems, Using Drones for Multi-Species
Research in the Tropics. Drones 2022, 6, 6. [CrossRef]
210. Shao, X.; Chen, H.; Magson, K.; Wang, J.; Song, J.; Chen, J.; Sasaki, J. Deep Learning for Multilabel Classification of Coral Reef
Conditions in the Indo-Pacific Using Underwater Photo Transect Method. Aquat. Conserv. Mar. Freshw. Ecosyst. 2024, 34, 4241.
[CrossRef]
211. Li, S.; Xu, S.; Song, K.; Kutser, T.; Wen, Z.; Liu, G.; Shang, Y.; Lyu, L.; Tao, H.; Wang, X.; et al. Remote Quantification of the Trophic
Status of Chinese Lakes. Hydrol. Earth Syst. Sci. 2023, 27, 3581–3599. [CrossRef]
212. Saha, G.K.; Rahmani, F.; Shen, C.; Li, L.; Cibin, R. A Deep Learning-Based Novel Approach to Generate Continuous Daily Stream
Nitrate Concentration for Nitrate Data-Sparse Watersheds. Sci. Total Environ. 2023, 878, 162930. [CrossRef]
213. Xu, X.; Yang, C.-C.; Xiao, Y.; Kong, J.-L. A Fine-Grained Recognition Neural Network with High-Order Feature Maps via
Graph-Based Embedding for Natural Bird Diversity Conservation. Int. J. Environ. Res. Public Health 2023, 20, 4924. [CrossRef]
[PubMed]
214. Das, N.; Padhy, N.; Dey, N.; Bhattacharya, S.; Tavares, J.M.R.S. Deep Transfer Learning-Based Automated Identification of Bird
Song. Int. J. Interact. Multimed. Artif. Intell. 2023, 8, 33–45. [CrossRef]
215. Kaufman, D.E.; Shenk, G.W.; Bhatt, G.; Asplen, K.W.; Devereux, O.H.; Rigelman, J.R.; Ellis, J.H.; Hobbs, B.F.; Bosch, D.J.; Van
Houtven, G.L.; et al. Supporting Cost-Effective Watershed Management Strategies for Chesapeake Bay Using a Modeling and
Optimization Framework. Environ. Model. Softw. 2021, 144, 105141. [CrossRef]
216. Rasheed Abdul Haq, K.P.; Harigovindan, V.P. Water Quality Prediction for Smart Aquaculture Using Hybrid Deep Learning
Models. IEEE Access 2022, 10, 60078–60098. [CrossRef]
217. Zhang, Z.; Hörmann, G.; Huang, J.; Fohrer, N. A Grid-Based Interpretable Machine Learning Method to Understand the Spatial
Relationships between Watershed Properties and Water Quality. Ecol. Indic. 2023, 154, 110627. [CrossRef]
218. Krivek, G.; Gillert, A.; Harder, M.; Fritze, M.; Frankowski, K.; Timm, L.; Meyer-Olbersleben, L.; von Lukas, U.F.; Kerth, G.; van
Schaik, J. BatNet: A Deep Learning-Based Tool for Automated Bat Species Identification from Camera Trap Images. Remote Sens.
Ecol. Conserv. 2023, 9, 759–774. [CrossRef]
219. Mahbub, T.; Bhagwagar, A.; Chand, P.; Zualkernan, I.; Judas, J.; Dghaym, D. Bat2Web: A Framework for Real-Time Classification
of Bat Species Echolocation Signals Using Audio Sensor Data. Sensors 2024, 24, 2899. [CrossRef]
220. Liu, M.; Min, L.; Shen, Y.; Wu, L. Evaluating the Impact of Alternative Cropping Systems on Groundwater Consumption and
Nitrate Leaching in the Piedmont Area of the North China Plain. Agronomy 2020, 10, 1635. [CrossRef]
221. Niyongabo, A.; Zhang, D.; Guan, Y.; Wang, Z.; Imran, M.; Nicayenzi, B.; Guyasa, A.K.; Hatungimana, P. Predicting Urban
Water Consumption and Health Using Artificial Intelligence Techniques in Tanganyika Lake, East Africa. Water 2024, 16, 1793.
[CrossRef]
222. Guo, Y.; Gao, X.; Jiang, B. An Empirical Study on JIT Defect Prediction Based on BERT-Style Model. arXiv 2024, arXiv:2403.11158.
[CrossRef]
223. Yang, Q.; Liu, Y.; Chen, T.; Tong, Y. Federated Machine Learning: Concept and Applications. ACM Trans. Intell. Syst. Technol. 2019,
10, 12. [CrossRef]
224. Saha, S.; Ahmad, T. Federated Transfer Learning: Concept and Applications. Intell. Artif. 2021, 15, 35–44. [CrossRef]
225. Masood, A.; Niazkar, M.; Zakwan, M.; Piraei, R. A Machine Learning-Based Framework for Water Quality Index Estimation in
the Southern Bug River. Water 2023, 15, 3543. [CrossRef]
226. Soriano, M.A.; Deziel, N.C.; Saiers, J.E. Regional Scale Assessment of Shallow Groundwater Vulnerability to Contamination from
Unconventional Hydrocarbon Extraction. Environ. Sci. Technol. 2022, 56, 12126–12136. [CrossRef]
227. Sun, Y.-P.; Jiang, Y.; Wang, Z.; Zhang, Y.; Zhang, L.-L. Wild Bird Species Identification Based on a Lightweight Model with
Frequency Dynamic Convolution. IEEE Access 2023, 11, 54352–54362. [CrossRef]
228. Chen, Q.; Fu, S. Quantitative Analysis and Management of Sustainable Development of Ecological Water Resources and Digital
Financial System Based on an Intelligent Algorithm. Water Supply 2023, 23, 2881–2898. [CrossRef]
Biology 2025, 14, 520 38 of 41

229. Mounce, S.R.; Ellis, K.; Edwards, J.M.; Speight, V.L.; Jakomis, N.; Boxall, J.B. Ensemble Decision Tree Models Using RUSBoost for
Estimating Risk of Iron Failure in Drinking Water Distribution Systems. Water Resour. Manag. 2017, 31, 1575–1589. [CrossRef]
230. Khatun, R.; Talukdar, S.; Pal, S.; Saha, T.K.; Mahato, S.; Debanshi, S.; Mandal, I. Integrating Remote Sensing with Swarm
Intelligence and Artificial Intelligence for Modelling Wetland Habitat Vulnerability in Pursuance of Damming. Ecol. Inform. 2021,
64, 101349. [CrossRef]
231. Wege, M.; Salas, L.; LaRue, M. Citizen Science and Habitat Modelling Facilitates Conservation Planning for Crabeater Seals in the
Weddell Sea. Divers. Distrib. 2020, 26, 1291–1304. [CrossRef]
232. Tripathi, R.N.; Ramachandran, A.; Tripathi, V.; Badola, R.; Hussain, S.A. Optimizing Riparian Habitat Conservation: A Spatial
Approach Using Aerial and Space Technologies. IEEE J. Sel. Top. Appl. Earth Obs. Remote Sens. 2024, 17, 16932–16943. [CrossRef]
233. Hou, Z.; Ren, H.; Murray, C.J.; Song, X.; Fang, Y.; Arntzen, E.V.; Chen, X.; Stegen, J.C.; Huang, M.; Gomez-Velez, J.D.; et al. A
Novel Construct for Scaling Groundwater-River Interactions Based on Machine-Guided Hydromorphic Classification. Environ.
Res. Lett. 2021, 16, 104016. [CrossRef]
234. Xu, Y.; Zhang, D.; Lin, J.; Peng, Q.; Lei, X.; Jin, T.; Wang, J.; Yuan, R. Prediction of Phytoplankton Biomass and Identification of
Key Influencing Factors Using Interpretable Machine Learning Models. Ecol. Indic. 2024, 158, 111320. [CrossRef]
235. Kouadri, S.; Elbeltagi, A.; Islam, A.R.M.T.; Kateb, S. Performance of Machine Learning Methods in Predicting Water Quality Index
Based on Irregular Data Set: Application on Illizi Region (Algerian Southeast). Appl. Water Sci. 2021, 11, 190. [CrossRef]
236. Ahkola, H.; Kotamäki, N.; Siivola, E.; Tiira, J.; Imoscopi, S.; Riva, M.; Tezel, U.; Juntunen, J. Uncertainty in Environmental
Micropollutant Modeling. Environ. Manag. 2024, 74, 380–398. [CrossRef] [PubMed]
237. Achee, N.L. The Remote Emerging Disease Intelligence—NETwork. Front. Microbiol. 2022, 13, 961065. [CrossRef]
238. Erichsen, A.C.; Middelboe, A.L. Introduction to the Special Series, “The Future of Marine Environmental Monitoring and
Assessment”. Integr. Environ. Assess. Manag. 2022, 18, 888–891. [CrossRef]
239. Sakti, A.D.; Adillah, K.P.; Santoso, C.; Faruqi, I.A.; Hendrawan, V.S.A.; Sofan, P.; Fauzi, A.I.; Setiawan, Y.; Utami, I.;
Zain, A.F.M.; et al. Modeling Proboscis Monkey Conservation Sites on Borneo Using Ensemble Machine Learning. Glob. Ecol.
Conserv. 2024, 54, e03101. [CrossRef]
240. Veiga, R.J.M.; Rodrigues, J.M.F. Fine-Grained Fish Classification From Small to Large Datasets With Vision Transformers. IEEE
Access 2024, 12, 113642–113660. [CrossRef]
241. Szomolányi, O.; Clement, A. Use of Random Forest for Assessing the Effect of Water Quality Parameters on the Biological Status
of Surface Waters. GEM Int. J. Geomath. 2023, 14, 20. [CrossRef]
242. Derville, S.; Torres, L.G.; Iovan, C.; Garrigue, C. Finding the Right Fit: Comparative Cetacean Distribution Models Using Multiple
Data Sources and Statistical Approaches. Divers. Distrib. 2018, 24, 1657–1673. [CrossRef]
243. Chambault, P.; Fossette, S.; Heide-Jørgensen, M.P.; Jouannet, D.; Vély, M. Predicting Seasonal Movements and Distribution of the
Sperm Whale Using Machine Learning Algorithms. Ecol. Evol. 2021, 11, 1432–1445. [CrossRef]
244. Zheng, C.; Liu, S.; Wang, J.; Lu, Y.; Ma, L.; Jiao, L.; Guo, J.; Yin, Y.; He, T. Opening the Black Box: Explainable Deep-Learning
Classification of Wood Microscopic Image of Endangered Tree Species. Plant Methods 2024, 20, 56. [CrossRef]
245. Caniani, D.; Esposito, G.; Gori, R.; Mannina, G. Towards a New Decision Support System for Design, Management and Operation
of Wastewater Treatment Plants for the Reduction of Greenhouse Gases Emission. Water 2015, 7, 5599–5616. [CrossRef]
246. Shekar, P.R.; Mathew, A.; Pandey, A.; Bhosale, A. A Comparison of the Performance of SWAT and Artificial Intelligence Models for
Monthly Rainfall–Runoff Analysis in the Peddavagu River Basin, India. Aqua Water Infrastruct. Ecosyst. Soc. 2023, 72, 1707–1730.
[CrossRef]
247. Pang, P.K.; Lim, K.H. Review on Automatic Plant Identification Using Computer Vision Approaches. IOP Conf. Ser. Mater. Sci.
Eng. 2019, 495, 012032. [CrossRef]
248. Vilas, C.; Antelo, L.T.; Martin-Rodriguez, F.; Morales, X.; Perez-Martin, R.I.; Alonso, A.A.; Valeiras, J.; Abad, E.; Quinzan, M.;
Barral-Martinez, M. Use of Computer Vision Onboard Fishing Vessels to Quantify Catches: The iObserver. Mar. Policy 2020,
116, 103714. [CrossRef]
249. Nemčić-Jurec, J.; Ruk, D.; Oreščanin, V.; Kovač, I.; Ujević Bošnjak, M.; Kinsela, A.S. Groundwater Contamination in Public Water
Supply Wells: Risk Assessment, Evaluation of Trends and Impact of Rainfall on Groundwater Quality. Appl. Water Sci. 2022,
12, 172. [CrossRef]
250. Ruan, J.; Cui, Y.; Meng, D.; Wang, J.; Song, Y.; Mao, Y. Integrated Prediction of Water Pollution and Risk Assessment of Water
System Connectivity Based on Dynamic Model Average and Model Selection Criteria. PLoS ONE 2023, 18, 0287209. [CrossRef]
[PubMed]
251. Lobo, G.P.; Laraway, J.; Gadgil, A.J. Identifying Schools at High-Risk for Elevated Lead in Drinking Water Using Only Publicly
Available Data. Sci. Total Environ. 2022, 803, 150046. [CrossRef] [PubMed]
252. Cao, Z.; Li, C.; Wang, K.; He, K.; Wang, X.; Yu, W. A Fast and Accurate Identification Model for Rhinolophus Bats Based on
Fine-Grained Information. Sci. Rep. 2023, 13, 16375. [CrossRef]
Biology 2025, 14, 520 39 of 41

253. Santana, F.S.; Costa, A.H.R.; Truzzi, F.S.; Silva, F.L.; Santos, S.L.; Francoy, T.M.; Saraiva, A.M. A Reference Process for Automating
Bee Species Identification Based on Wing Images and Digital Image Processing. Ecol. Inform. 2014, 24, 248–260. [CrossRef]
254. Cheng, S.-T.; Tsai, W.-P.; Yu, T.-C.; Herricks, E.E.; Chang, F.-J. Signals of Stream Fish Homogenization Revealed by AI-Based
Clusters. Sci. Rep. 2018, 8, 15960. [CrossRef]
255. Chen, J.; Zhao, L.; Wang, B.; He, X.; Duan, L.; Yu, G. Uncovering Global Risk to Human and Ecosystem Health from Pesticides in
Agricultural Surface Water Using a Machine Learning Approach. Environ. Int. 2024, 194, 109154. [CrossRef]
256. Wieland, R.; Kuhls, K.; Lentz, H.H.K.; Conraths, F.; Kampen, H.; Werner, D. Combined Climate and Regional Mosquito Habitat
Model Based on Machine Learning. Ecol. Model. 2021, 452, 109594. [CrossRef]
257. Zheng, Y.; Wang, J.; Ryzhkov, S.; Nechai, O.; Topalov, A.; Zivenko, O.; Babchuk, S.; Harasymiv, T. Computer system for forecasting
water quality parameters based on machine learning. Comptes Rendus L’Academie Bulg. Sci. 2024, 77, 1629–1638. [CrossRef]
258. Choden, Y.; Chokden, S.; Rabten, T.; Chhetri, N.; Aryan, K.R.; Al Abdouli, K.M. Performance Assessment of Data Driven Water
Models Using Water Quality Parameters of Wangchu River, Bhutan. SN Appl. Sci. 2022, 4, 290. [CrossRef]
259. Wang, Y.; Liu, L. Research on Sustainable Green Building Space Design Model Integrating IoT Technology. PLoS ONE 2024, 19,
0298982. [CrossRef] [PubMed]
260. Guilbault, E.; Renner, I.; Beh, E.J.; Mahony, M. A Practical Approach to Making Use of Uncertain Species Presence-Only Data in
Ecology: Reclassification, Regularization Methods and Observer Bias. Ecol. Inform. 2023, 77, 102155. [CrossRef]
261. Mz, Y.; Rahayu, S. Decision Support System for Urban River Water Quality as a Source of Clean Water Using the SMART Method.
IOP Conf. Ser. Earth Environ. Sci. 2024, 1372, 012038. [CrossRef]
262. Molina, J.-L.; Pulido-Velázquez, M.; Llopis-Albert, C.; Peña-Haro, S. Stochastic Hydro-Economic Model for Groundwater Quality
Management Using Bayesian Networks. Water Sci. Technol. 2013, 67, 579–586. [CrossRef]
263. Visser, H.; Evers, N.; Bontsema, A.; Rost, J.; de Niet, A.; Vethman, P.; Mylius, S.; van der Linden, A.; van den Roovaart, J.; van
Gaalen, F.; et al. What Drives the Ecological Quality of Surface Waters? A Review of 11 Predictive Modeling Tools. Water Res.
2022, 208, 117851. [CrossRef] [PubMed]
264. Li, Z.; Zhao, S.; Lu, Y.; Song, C.; Huang, R.; Yu, K. Deep Learning-Based Automatic Estimation of Live Coral Cover from
Underwater Video for Coral Reef Health Monitoring. J. Mar. Sci. Eng. 2024, 12, 1980. [CrossRef]
265. Jia, B.; Wang, C.; Qiu, E.; Qie, G. The Status and Trend on the Urban Tree Canopy Research. Shengtai Xuebao/Acta Ecol. Sin. 2013,
33, 23–32. [CrossRef]
266. Lee, J.M.; Ko, K.-S.; Yoo, K. A Machine Learning-Based Approach to Predict Groundwater Nitrate Susceptibility Using Field
Measurements and Hydrogeological Variables in the Nonsan Stream Watershed, South Korea. Appl. Water Sci. 2023, 13, 242.
[CrossRef]
267. Chen, K.; Shi, X.; Zhang, Z.; Chen, S.; Ma, J.; Zheng, T.; Alfonso, L. Using Unsupervised Learning to Classify Inlet Water for More
Stable Design of Water Reuse in Industrial Parks. Water Sci. Technol. 2024, 89, 1757–1770. [CrossRef]
268. Ewusi, A.; Ahenkorah, I.; Aikins, D. Modelling of Total Dissolved Solids in Water Supply Systems Using Regression and
Supervised Machine Learning Approaches. Appl. Water Sci. 2021, 11, 13. [CrossRef]
269. Ma, D.; Wei, J.; Zhu, L.; Zhao, F.; Wu, H.; Chen, X.; Li, Y.; Liu, M. Semi-Supervised Learning Advances Species Recognition for
Aquatic Biodiversity Monitoring. Front. Mar. Sci. 2024, 11, 1373755. [CrossRef]
270. Balajee, J.; Saleem Durai, M.A. Smart Survey on Recent Trends in Water Level, Drought and Water Quality Analysis System. IOP
Conf. Ser. Earth Environ. Sci. 2021, 1964, 042052. [CrossRef]
271. Sajib, A.M.; Diganta, M.T.M.; Moniruzzaman, M.; Rahman, A.; Dabrowski, T.; Uddin, M.G.; Olbert, A.I. Assessing Water Quality
of an Ecologically Critical Urban Canal Incorporating Machine Learning Approaches. Ecol. Inform. 2024, 80, 102514. [CrossRef]
272. Smedley, P.L.; Allen, G.; Baptie, B.J.; Fraser-Harris, A.P.; Ward, R.S.; Chambers, R.M.; Gilfillan, S.M.V.; Hall, J.A.; Hughes, A.G.;
Manning, D.A.C.; et al. Equipping for Risk: Lessons Learnt from the UK Shale-Gas Experience on Assessing Environmental Risks
for the Future Geoenergy Use of the Deep Subsurface. Sci. Total Environ. 2024, 921, 171036. [CrossRef] [PubMed]
273. Awwad, A.; Husseini, G.A.; Albasha, L. AI-Aided Robotic Wide-Range Water Quality Monitoring System. Sustainability 2024,
16, 9499. [CrossRef]
274. Panigrahi, S.; Maski, P.; Thondiyath, A. Real-Time Biodiversity Analysis Using Deep-Learning Algorithms on Mobile Robotic
Platforms. PeerJ Comput. Sci. 2023, 9, 1502. [CrossRef]
275. Reynaert, E.; Steiner, P.; Yu, Q.; D’Olif, L.; Joller, N.; Schneider, M.Y.; Morgenroth, E. Predicting Microbial Water Quality in On-Site
Water Reuse Systems with Online Sensors. Water Res. 2023, 240, 120075. [CrossRef]
276. Zia, H.; Harris, N.; Merrett, G. A Low Complexity Data Driven Model of Environmental Discharge Dynamics for Wireless Sensor
Network Applications. Procedia Eng. 2014, 87, 544–547. [CrossRef]
277. Chen, Z.; Xue, X.; Wu, H.; Gao, H.; Wang, G.; Ni, G.; Cao, T. Visible/near-Infrared Hyperspectral Imaging Combined with
Machine Learning for Identification of Ten Dalbergia Species. Front. Plant Sci. 2024, 15, 1413215. [CrossRef]
278. Wong, J.J.N.; Fadzly, N. Development of Species Recognition Models Using Google Teachable Machine on Shorebirds and
Waterbirds. J. Taibah Univ. Sci. 2022, 16, 1096–1111. [CrossRef]
Biology 2025, 14, 520 40 of 41

279. Marquez, L.; Fragkopoulou, E.; Cavanaugh, K.C.; Houskeeper, H.F.; Assis, J. Artificial Intelligence Convolutional Neural
Networks Map Giant Kelp Forests from Satellite Imagery. Sci. Rep. 2022, 12, 22196. [CrossRef] [PubMed]
280. Mazumder, A.; Engala, S.R.; Nallapuraju, A. Towards Sustainable Development: A Novel Integrated Machine Learning Model for
Holistic Environmental Health Monitoring. In Proceedings of the 2023 IEEE MIT Undergraduate Research Technology Conference
(URTC), Cambridge, MA, USA, 6–8 October 2023.
281. Martínez, R.; Felis, I.; Navarro, M.; Sanz-González, J.C. Time Series Modelling and Predictive Analytics for Sustainable Environ-
mental Management—A Case Study in El Mar Menor (Spain). Eng. Proc. 2023, 58, 32. [CrossRef]
282. McLoughlin, C.A.; Riddell, E.S.; Petersen, R.M.; Venter, J. Adaptive and Transformative Learning in Environmental Water
Management: Implementing the Crocodile River’s Ecological Reserve in Kruger National Park, South Africa. Koedoe 2021,
63, 1663. [CrossRef]
283. Marre, G.; De Almeida Braga, C.; Ienco, D.; Luque, S.; Holon, F.; Deter, J. Deep Convolutional Neural Networks to Monitor
Coralligenous Reefs: Operationalizing Biodiversity and Ecological Assessment. Ecol. Inform. 2020, 59, 101110. [CrossRef]
284. McClanahan, T.R. Local Heterogeneity of Coral Reef Diversity and Environmental Stress Provides Opportunities for Small-Scale
Conservation. Divers. Distrib. 2023, 29, 1324–1340. [CrossRef]
285. Hee Shim, S.; Hyun Choi, J. Building an XGBoost Model Based on Landscape Metrics and Meteorological Data for Nonpoint
Source Pollution Management in the Nakdong River Watershed. Ecol. Indic. 2024, 165, 112156. [CrossRef]
286. McClanahan, T.R.; Sola, E. Comparing Modeled Predictions of Coral Reef Diversity along a Latitudinal Gradient in Mozambique.
Front. Ecol. Evol. 2024, 12, 1450383. [CrossRef]
287. Kadoya, S.-S.; Nishimura, O.; Kato, H.; Sano, D. Predictive Water Virology Using Regularized Regression Analyses for Projecting
Virus Inactivation Efficiency in Ozone Disinfection. Water Res. X 2021, 11, 100093. [CrossRef]
288. Yuan, J.; Wu, Z.; Li, S.; Kang, P.; Zhu, S. Multi-Feature-Based Identification of Subtropical Evergreen Tree Species Using Gaofen-2
Imagery and Algorithm Comparison. Forests 2023, 14, 292. [CrossRef]
289. Jongjaraunsuk, R.; Taparhudee, W.; Suwannasing, P. Comparison of Water Quality Prediction for Red Tilapia Aquaculture in an
Outdoor Recirculation System Using Deep Learning and a Hybrid Model. Water 2024, 16, 907. [CrossRef]
290. Carretta, J.V. A Machine-Learning Approach to Assign Species to “unidentified” Entangled Whales. Endanger. Species Res. 2018,
36, 89–98. [CrossRef]
291. Pittman, S.J.; Brown, K.A. Multi-Scale Approach for Predicting Fish Species Distributions across Coral Reef Seascapes. PLoS ONE
2011, 6, 0020583. [CrossRef]
292. Su, X.; Zhou, D.; Wang, H.; Xu, J. Research on the Scaling Mechanism and Countermeasures of Tight Sandstone Gas Reservoirs
Based on Machine Learning. Processes 2024, 12, 527. [CrossRef]
293. Dias, D.F.C.; Abily, M.; Ribeiro, J.M.; Jouhara, H.; Katsou, E. Screening Rainwater Harvesting Potentialities in the EU Industrial
Sector: A Framework for Site-Specific Assessment. Water 2024, 16, 1758. [CrossRef]
294. Eastwood, N.; Zhou, J.; Derelle, R.; Abdallah, M.A.-E.; Stubbings, W.A.; Jia, Y.; Crawford, S.E.; Davidson, T.A.; Colbourne, J.K.;
Creer, S.; et al. 100 Years of Anthropogenic Impact Causes Changes in Freshwater Functional Biodiversity. eLife 2023, 12, 86576.
[CrossRef] [PubMed]
295. Hu, C.; Cai, J.; Zeng, D.; Yan, X.; Gong, W.; Wang, L. Deep Reinforcement Learning Based Valve Scheduling for Pollution Isolation
in Water Distribution Network. Math. Biosci. Eng. 2020, 17, 105–121. [CrossRef]
296. Doña, C.; Chang, N.-B.; Caselles, V.; Sánchez, J.M.; Camacho, A.; Delegido, J.; Vannah, B.W. Integrated Satellite Data Fusion and
Mining for Monitoring Lake Water Quality Status of the Albufera de Valencia in Spain. J. Environ. Manag. 2015, 151, 416–426.
[CrossRef]
297. Dolph, C.L.; Cho, S.J.; Finlay, J.C.; Hansen, A.T.; Dalzell, B. Predicting High Resolution Total Phosphorus Concentrations for Soils
of the Upper Mississippi River Basin Using Machine Learning. Biogeochemistry 2023, 163, 289–310. [CrossRef]
298. Yan, Y.; Zhang, Y.; Yao, R.; Wei, C.; Luo, M.; Yang, C.; Chen, S.; Huang, X. Groundwater Suitability Assessment for Irrigation and
Drinking Purposes by Integrating Spatial Analysis, Machine Learning, Water Quality Index, and Health Risk Model. Environ. Sci.
Pollut. Res. 2024, 31, 39155–39176. [CrossRef]
299. Villa, F.; Bagstad, K.J.; Voigt, B.; Johnson, G.W.; Portela, R.; Honzák, M.; Batker, D. A Methodology for Adaptable and Robust
Ecosystem Services Assessment. PLoS ONE 2014, 9, 0091001. [CrossRef]
300. Egbemhenghe, A.U.; Ojeyemi, T.; Iwuozor, K.O.; Emenike, E.C.; Ogunsanya, T.I.; Anidiobi, S.U.; Adeniyi, A.G. Revolutionizing
Water Treatment, Conservation, and Management: Harnessing the Power of AI-Driven ChatGPT Solutions. Environ. Chall. 2023,
13, 100782. [CrossRef]
301. Alamanos, A. Simple Hydro-Economic Tools for Supporting Small Water Supply Agencies on Sustainable Irrigation Water
Management. Water Supply 2022, 22, 1810–1819. [CrossRef]
Biology 2025, 14, 520 41 of 41

302. Kadiyala, L.A.; Mermer, O.; Samuel, D.J.; Sermet, Y.; Demir, I. The Implementation of Multimodal Large Language Models for
Hydrological Applications: A Comparative Study of GPT-4 Vision, Gemini, LLaVa, and Multimodal-GPT. Hydrology 2024, 11, 148.
[CrossRef]
303. Barcala, V.; Rozemeijer, J.; Ouwerkerk, K.; Gerner, L.; Osté, L. Value and Limitations of Machine Learning in High-Frequency
Nutrient Data for Gap-Filling, Forecasting, and Transport Process Interpretation. Environ. Monit. Assess. 2023, 195, 892. [CrossRef]

Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual
author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to
people or property resulting from any ideas, methods, instructions or products referred to in the content.