The Impact of Artificial Intelligence on the Sustainability of Regional Ecosystems: Current Challenges and Future Prospects

Abstract

The integration of artificial intelligence (AI) technologies is reshaping diverse domains of human activity, including natural resource management, urban and rural planning, agri-food systems, industry, energy, education, and healthcare. However, the impact of AI on the sustainability of local ecosystems remains insufficiently systematized. This highlights the need for a comprehensive review that considers spatial, sectoral, and socio-economic characteristics of regions, as well as interdisciplinary approaches to sustainable development. This study presents a scoping review of 198 peer-reviewed publications published between 2010 and March 2025, focusing on applied cases of AI deployment in local contexts. Special attention is given to the role of AI in monitoring water, forest, and agricultural ecosystems, facilitating the digital transformation of businesses and territories, assessing ecosystem services, managing energy systems, and supporting educational and social sustainability. The review includes case studies from Africa, Asia, Europe, and Latin America, covering a wide range of technologies—from machine learning and digital twins to IoT and large language models. Findings indicate that AI holds significant potential for enhancing the efficiency and adaptability of local systems. Nevertheless, its implementation is accompanied by notable risks, including socio-economic disparities, technological inequality, and institutional limitations. The review concludes by outlining research priorities for the sustainable integration of AI into local ecosystems, emphasizing the importance of cross-sectoral collaboration and scientific support for regional digital transformations.

1. Introduction

In an era of rapid technological advancement, artificial intelligence has become an integral component of digital transformation across both global and local systems. Its implementation influences a wide array of domains, including natural resource management, agriculture and forestry, industrial processes, energy infrastructure, as well as educational and social services. Notably, the local scale—encompassing cities, rural areas, agroecosystems, and small- and medium-sized enterprises—serves as a critical arena for the practical application of sustainable development principles. These applications require careful consideration of regional specificity, socio-economic context, and ecological sensitivity.

In this review, regional ecosystems are defined as spatially bounded socio-ecological systems that integrate natural resources, infrastructure, and human communities under shared governance or land-use regimes. Within these systems, AI interventions refer to the deployment of machine learning models, digital twins, IoT-based monitoring, and related data-driven technologies designed to sense, predict, or manage environmental and socio-economic dynamics. The impact on sustainability is assessed in terms of how such interventions influence ecological resilience, resource efficiency, and social equity across the environmental, economic, and social pillars of sustainable development.

These technologies encompass a wide range of applications: machine learning algorithms are employed for classification, prediction, and anomaly detection across agricultural and environmental datasets; digital twins provide real-time virtual representations of physical assets such as irrigation systems or urban infrastructure; IoT sensors enable continuous environmental monitoring and data collection, while remote sensing technologies support large-scale land-cover analysis and resource tracking. This technological landscape forms the basis for AI interventions across multiple ecosystem types and sustainability goals.

Recent studies confirm the high applicability of AI in digital soil mapping [1], urban planning [2], agriculture [3], industrial process optimization [4], and energy systems [5]. Particular attention has been paid to the role of AI in mitigating climate risks [6], supporting the digital transformation of food systems [7,8], and contributing to ecological meteorology and ecosystem services monitoring [9,10]. There is also a growing interest in the integration of AI for carbon flow assessment, environmental risk prediction, and the adaptation of agroecosystems to climate change [11,12,13].

However, despite the rapid expansion of research on artificial intelligence, there remains a lack of studies that systematically integrate cross-sectoral experiences of AI application in local ecosystems while accounting for territorial characteristics and the interdisciplinary context of sustainability. This gap hinders a comprehensive understanding of how digital transformation driven by AI affects sustainability at regional and local levels, where technological implementation often encounters social, institutional, and environmental barriers. The absence of a holistic perspective limits the development of strategies for equitable digitalization and adaptive governance in the face of escalating climate, economic, and demographic challenges.

This article seeks to bridge this gap by providing a scoping review of 198 peer-reviewed sources published between 2010 and March 2025. The aim is to systematize practical AI-based solutions implemented across various types of local ecosystems, including aquatic, forest, agricultural, industrial, urban, and social systems.

Particular attention is devoted to the role of AI in enhancing the resilience of agricultural regions, where digital technologies not only improve production efficiency, but also influence broader socio-environmental dynamics. Innovations such as precision farming, crop yield forecasting, and water resource optimization have demonstrated the potential to reduce environmental pressure while contributing to food security [14]. Yet, the long-term ramifications of technological transformation for agri-industrial territories remain largely underexplored.

This review aims to synthesize and systematize the existing body of scholarly literature to illustrate how AI and machine learning (ML) methods contribute to the sustainable development of regional ecosystems. It also seeks to identify key thematic directions, integration barriers, and success factors associated with AI deployment in agricultural regions.

The analysis is intended to uncover research gaps and provide recommendations for the responsible implementation of AI in regional ecosystems, with a view to ensuring a fair and sustainable technological transition.

The specific objectives of this review are fourfold. First, it aims to identify the main research areas addressing the impact of AI on the sustainability of regional ecosystems. Second, it seeks to analyze the current state of academic knowledge concerning the role of AI in agri-industrial regions and to reveal existing knowledge gaps. Third, the review focuses on identifying potential opportunities and challenges associated with AI integration into regional processes, including its effects on environmental, economic, and social stability. Finally, it proposes directions for future research aimed at developing strategies for the responsible and context-sensitive application of AI in regional ecosystems.

This analysis contributes to a deeper scientific understanding of the role of artificial intelligence in supporting sustainable development. It also provides a foundation for formulating recommendations on the responsible integration of digital technologies into regional strategies, particularly in the context of agricultural ecosystems that require integrated resource management and mitigation of socio-economic inequality.

This review covers a broad spectrum of technologies—including machine learning algorithms, remote sensing, digital twins, and the IoT—and is structured according to the functional domains of AI implementation. It reflects both ecological and socio-economic implications of AI adoption. By integrating interdisciplinary approaches, the article offers a comprehensive perspective on the role of AI in transforming local ecosystems and lays a scientific basis for further research and practical applications in the field of regional sustainability.

Unlike most existing studies, this review pays particular attention to the regional and sectoral specifics of AI implementation and identifies key social, economic, and technological barriers to its adoption. The findings are intended to expand knowledge for both academic researchers and practitioners working in the field of sustainable development, especially within a regional context.

This article is organized into four main sections. The first outlines the relevance, objectives, and research questions. The second presents the methodology for selecting and analyzing peer-reviewed publications, ensuring a structured and replicable approach. The third section explores key research domains, including the impact of AI on water, forest, agricultural, industrial, urban, and social ecosystems. The final section summarizes the main findings and proposes directions for future research focused on achieving balanced AI integration and minimizing the asymmetry of its impacts.

2. Materials and Methods

This scoping review adopts a structured and exploratory approach to map existing research on the contribution of artificial intelligence and machine learning to the sustainable development of regional ecosystems. The goal is to uncover prevailing trends and critical challenges associated with AI integration at the local level, with a particular focus on its environmental, economic, and social implications.

The review is guided by the PRISMA framework, which ensures methodological transparency and reproducibility. Relevant peer-reviewed sources were identified through a stepwise search strategy in the Scopus database, refined iteratively to ensure thematic relevance and comprehensive coverage.

Initially, the following query string was constructed to identify relevant studies: TITLE-ABS-KEY ((“Regional Ecosystem”) AND (“Artificial Intelligence” OR “Machine learning”)) AND PUBYEAR > 2009 AND PUBYEAR < 2026. This query covered the period from 2010 to 2025 and targeted both foundational and recent research. However, as of 18 March 2025, it returned only 37 publications, which did not meet the requirements for a scoping review.

In the second iteration, the query was expanded by including additional keywords to broaden the scope: TITLE-ABS-KEY ((“Regional Ecosystem” OR “Local Ecosystem”) AND (“Artificial Intelligence” OR “Machine learning”)) AND PUBYEAR > 2009 AND PUBYEAR < 2026. This modification increased the number of retrieved records to 70, which still fell short of the necessary threshold for a comprehensive review.

During the third phase, the search terms “Regional Ecosystem” and “Local Ecosystem” were split to improve coverage, resulting in the following string: TITLE-ABS-KEY (“Ecosystem” AND (“Regional” OR “Local”) AND (“Artificial Intelligence” OR “Machine learning”)) AND PUBYEAR > 2009 AND PUBYEAR < 2026. While this search yielded 1728 documents, the results proved too broad, capturing a significant number of irrelevant publications and weakening the thematic focus.

To enhance the thematic focus and methodological clarity, an additional filter was applied to restrict the results to studies explicitly addressing sustainability. The final search string was: TITLE-ABS-KEY (“Ecosystem” AND (“Regional” OR “Local”) AND (“Artificial Intelligence” OR “Machine learning”) AND (“Sustainability” OR “Sustainable Development”)) AND PUBYEAR > 2009 AND PUBYEAR < 2026.

As a result of the refined search strategy, a total of 198 publications were identified, providing a balanced dataset in terms of both relevance and comprehensiveness. The selected keywords were carefully tailored to align closely with the aims and research questions of this study. Known constraints of the search strategy (database, language, and disciplinary coverage) were identified a priori and are discussed in detail in Section Discussions. As of 18 March 2025, a temporal analysis of publication trends revealed that 73% of the articles were published between 2022 and 2025, indicating a sharp increase in academic interest in this topic over recent years (Figure 1).

The number of publications has grown from 16 in 2021 to 57 in 2024, demonstrating exponential growth in the field.

An analysis of the geographical distribution of the reviewed publications revealed that the highest number of studies originated from China (n = 78), significantly outpacing all other countries. The United States ranked second with 27 publications, followed by India with 22.

A considerably smaller number of studies was recorded in Germany, the United Kingdom, Italy, South Africa, and Spain, where publication counts ranged between 8 and 16. Other notable contributors included Australia, France, Portugal, and Singapore, each represented by six publications (Figure 2).

The disciplinary classification of the reviewed literature (Figure 3) indicates that the largest proportion of studies falls under the field of Environmental Science (23%). A substantial number of publications also belong to the categories of Computer Science (13%) and Earth and Planetary Sciences (10%). Meanwhile, studies focused on the application of artificial intelligence within Agricultural and Biological Sciences account for 9% (n = 37).

In the next stage, a manual screening and content analysis of all 198 publications was conducted. To ensure comprehensive coverage and a structured approach, the methodology of a scoping review was applied in accordance with established guidelines [15]. This ensured transparency and academic rigor throughout the selection process.

Eligibility criteria: Records were eligible if they (i) reported an empirical, modeling-based, or conceptual application of AI/ML to a regional or local ecosystem; (ii) addressed at least one pillar of sustainability (environmental, social, economic); (iii) were peer-reviewed journal articles published in English between 2010 and 2025. Conference papers, non-English publications, and grey literature were excluded.

Following the exclusion of duplicate, irrelevant, and methodologically weak studies, a final set of 155 peer-reviewed articles was selected for thematic analysis. These publications formed the analytical foundation for the findings presented in the “Results” section. A PRISMA flow diagram was constructed to illustrate the document selection process, ensuring transparency and methodological rigor (Figure 4). The diagram outlines the number of records identified through database searches, screened for relevance, and included in the final synthesis.

The adopted approach enhances the reproducibility and reliability of the results while ensuring their relevance in both international and regional contexts. All data were derived from reputable open-access sources, thereby ensuring the objectivity of the analysis and the validity of the conclusions.

3. Results

The analysis of 155 peer-reviewed publications, covering the period from 2010 to March 2025, enabled the identification of nine key research areas in which artificial intelligence technologies are being applied to enhance the sustainability of local ecosystems. Below is a brief overview of these thematic areas and their core focus points.

The first area centers on environmental monitoring and the mapping of land use, vegetation, and soil using AI and remote sensing data. These tools support high-resolution assessments of degradation processes and inform decision making in sustainable land management. Special emphasis is placed on digital soil mapping and modeling the spatiotemporal dynamics of ecosystems under the influence of urbanization and climate change.

The second area focuses on the application of AI for sustainable agriculture and the management of agroecosystems. Core themes include yield prediction and land suitability assessments, monitoring of soil and vegetation health, and optimization of water and energy use.

The third thematic area focuses on the application of AI for monitoring, forecasting, and the sustainable management of aquatic and coastal ecosystems. Key topics include the analysis of water quality and hydrological processes using machine learning, mapping of degradation and pollution risks, and forecasting water balances under climate change scenarios. AI tools improve the accuracy of environmental assessments, support ecosystem restoration efforts, and enable the development of local strategies in response to increasing anthropogenic pressures.

The fourth area pertains to the use of AI in the management and conservation of forest ecosystems. It emphasizes forest fire prediction and spatiotemporal modeling, biomass and carbon stock assessment for climate strategies, 3D modeling of forest structures, habitat classification for biodiversity monitoring, and tracking of forest degradation and regeneration.

The fifth area addresses carbon monitoring, vulnerability assessment, and ecosystem change forecasting using AI. Core themes include the integration of satellite data and AI to estimate greenhouse gas emissions and carbon balances, the prediction of climate risks and droughts, and the modeling of ecosystem dynamics under various environmental stressors.

The sixth area is concerned with the evaluation and management of ecosystem services through AI applications. Key aspects involve the use of machine learning and remote sensing for spatiotemporal modeling and land zoning, interpretable AI models to quantify ecosystem service losses, and the incorporation of AI into strategies for sustainable tourism, landscape restoration, and natural capital valuation.

The seventh thematic area explores the application of AI in regional contexts—specifically in business, industry, and agriculture. Key focal points include the transformation of small- and medium-sized enterprises (SMEs) and agricultural systems through AI, digital twins, and cloud platforms; the use of AI in precision agriculture, food system planning, and the assessment of the environmental impacts of agricultural practices; the integration of AI into industrial processes and sustainable manufacturing; and the deployment of AI for managing urban ecosystems, infrastructure, and waste.

The eighth area is dedicated to the application of AI in transforming local energy systems. It highlights the use of IoT and AI to improve the energy efficiency of buildings and urban infrastructure; the optimization of distributed and renewable energy management; the development of predictive models for energy consumption and generation; and the use of AI to assess climate impacts and support the digitalization of the energy sector at the regional level.

The ninth area concerns the use of AI to support local sustainable development in the domains of education, healthcare, and social inclusion. Central themes include the application of AI in environmental and climate education, personalized learning, and the engagement of vulnerable groups; the use of AI in digital healthcare, including disease monitoring, big data analysis, and telemedicine; and the integration of AI into social and scientific ecosystems to enhance community inclusivity, awareness, and resilience.

Taken together, these thematic areas illustrate that AI has become a key enabler not only for improving efficiency but also for enhancing the adaptability and sustainability of local ecosystems.

Table 1. Thematic analysis by years.

	Summary	≤2020	2021	2022	2023	2024	2025
AI for Environmental Monitoring and Mapping of Land Use, Vegetation, and Soils	15	2	0	5	1	6	1
AI for Sustainable Agriculture and Agroecosystem Management	13	1	3	0	4	1	4
AI for Monitoring, Forecasting, and Management of Aquatic Ecosystems	34	5	3	9	8	8	1
Artificial Intelligence in Forest Ecosystem Conservation and Management	15	2	1	2	3	5	2
Carbon Monitoring, Vulnerability Assessment, and Ecosystem Change Forecasting Using AI	14	2	1	2	2	5	2
AI in the Assessment and Management of Ecosystem Services	26	4	3	2	5	8	4
AI in Regional Contexts: Business, Industry, and Agriculture	34	4	7	4	8	11	0
AI in the Transformation of Local Energy Systems	10	3	0	1	1	5	0
AI in Supporting Local Sustainable Development: Education, Health, and Social Inclusion	13	2	1	1	3	2	4
Total		25	19	26	35	51	18

Source: compiled by authors.

presents the distribution of sources by key research themes and publication years.

This study showed that some topics were of higher interest than others. In Figure 5, we can see the distribution of topics by the number of mentions in sources.

The temporal dynamics of interest across thematic areas exhibit a heterogeneous pattern. For most topics, scholarly attention peaked in 2024, with the notable exception of “AI for Sustainable Agriculture and Agroecosystem Management” and “AI in Supporting Local Sustainable Development: Education, Health, and Social Inclusion”, both of which reached their highest publication levels in 2025—despite only covering the first three months of the year.

The topic “AI for Monitoring, Forecasting, and Management of Aquatic Ecosystems” experienced its peak in 2022 and has maintained a consistently high level of interest in the years since. Meanwhile, the “AI in Regional Contexts: Business, Industry, and Agriculture” category demonstrates the most dynamic growth trajectory, with a continuous upward trend culminating in 2024 (Figure 6).

Below, a detailed analysis is provided, presenting the distribution of sources across thematic areas.

3.1. AI for Environmental Monitoring and Mapping of Land Use, Vegetation, and Soils

Regional coverage: This sub-section includes 15 studies that apply AI to ecological monitoring and land-use/soil mapping. The evidence is overwhelmingly concentrated in Asia (n = 12)—with case studies drawn from China (e.g., the Chongming Dongtan wetlands, Yellow River Delta, Inner-Mongolia grasslands, and Poyang Lake Basin) and South/South-East Asia (e.g., rural Bangladesh). Two studies are conducted in Africa (n = 2), and two others adopt a global perspective (n = 2).

The advancement of artificial intelligence and remote sensing technologies has opened new frontiers in environmental monitoring and spatial analysis of terrestrial ecosystem dynamics. Of particular relevance are studies focused on land use, vegetation health, and soil conditions—critical components of sustainable natural resource management in local ecosystems.

The use of AI for assessing the condition and transformation of wetlands and vegetative communities has proven highly effective in analyzing the impacts of urbanization and climate change. For example, the K-ELM model applied to Landsat data from 1986 to 2013 enabled the analysis of spatiotemporal changes in the Chongming Dongtan wetlands, capturing the effects of urban expansion and generating forecasts of ecosystem shifts [16].

In another study, the spatiotemporal dynamics of groundwater-dependent ecosystems (GDE) were assessed using Landsat 8 imagery and the Random Forest algorithm [10]. This approach contributes to more accurate predictions of the implications of climate and land-use change for biodiversity.

The integration of satellite data, remote sensing techniques, and cloud-based platforms—such as Google Earth Engine—with AI algorithms significantly enhances the mapping of terrestrial groundwater-dependent vegetation (TGDV) [17]. These technologies enable the monitoring of both seasonal and long-term ecosystem changes, while improving the understanding of climatic and anthropogenic impacts.

Machine learning methods, including Extreme Gradient Boosting, have been applied to time series satellite data in the Yellow River Delta to precisely track the degradation and restoration of wetland ecosystems, supporting the development of evidence-based ecological management strategies [18]. The accuracy of spatiotemporal analysis of wetland plant communities has been further improved through the application of Random Forest and SHAP algorithms, facilitating sustainable management of ecosystem functions [19].

Additionally, Random Forest classifiers applied to Sentinel-1 and Sentinel-2 satellite data have been used to distinguish between rice paddies and wetland areas in the Poyang Lake Basin [20]. This approach allows for a more accurate interpretation of ecosystem phenological signals and the development of targeted land-use recommendations for sustainable resource management.

AI is also employed to assess pastureland changes in Inner Mongolia, analyzing their responses to climatic variability and anthropogenic pressures [21]. This is critical for developing adaptive strategies for the sustainable use of land in areas vulnerable to degradation.

A study on the spatiotemporal dynamics of oases in China over the period 1987–2017 leveraged AI tools to identify patterns of expansion and degradation. The research also assessed their impact on land use, including the increase in cropland and grazing areas. These insights are applicable for optimizing agricultural policy and mitigating resource depletion in arid regions [22]. Similarly, another study evaluated the impact of a large-scale infrastructure project (the Padma Bridge) on land-use change and urbanization in the rural ecosystems of Bangladesh [23]. The analysis used Random Forest and artificial neural networks combined with cellular automata to model satellite-derived land-use transitions up to 2033.

Land degradation mapping using machine learning algorithms and Landsat 8 satellite imagery enabled the classification of areas according to their degradation risk levels. The resulting degradation likelihood distribution (DLD) maps serve as valuable decision-support tools for regional natural resource management policies [24].

AI is widely used in digital soil mapping, significantly improving the accuracy of assessing physical and chemical properties. In Iran, the analysis of soil-forming factor similarity using Gower’s index helped to fill gaps in existing soil maps and enhanced resource management efficiency [1]. In India, algorithms such as Random Forest, support vector machines, decision trees, and logistic regression were employed to develop high-precision models of soil properties, contributing to more sustainable agricultural management practices [25].

Spatial mapping of available nitrogen and phosphorus in the alpine meadow soils of the Qingzang Plateau, using a Random Forest model, identified ecological drivers influencing nutrient dynamics [26]. This information is critical for pasture management and maintaining ecosystem productivity. Ensemble learning techniques have also been applied to estimate soil organic carbon (SOC) at multiple depths within the Heihe River Basin [27]. These models incorporate climate and landscape variables affecting SOC levels and support the development of agroecological strategies tailored to local environmental conditions.

In addition, recent studies incorporate hyperspectral remote sensing, proximal soil sensors, and AI-based methods—such as Random Forest, SVM, and neural networks—to account for spatial uncertainty in digital soil mapping [28]. This is particularly important for assessing the impacts of climate change and soil degradation, as well as for developing sustainable environmental protection strategies.

The body of research shows that integrating AI with remote sensing data enables the development of new, effective tools for monitoring and mapping key ecosystem components—such as wetlands, vegetation, soils, and land use. These technologies not only allow for detailed spatial-temporal assessments but also help identify causal relationships between anthropogenic pressures, climatic variables, and ecosystem degradation.

AI applications support the development of more accurate forecasts, inform science-based decision making in natural resource management, and contribute to strengthening the resilience of local ecosystems. These approaches are especially valuable for early detection of degradation processes, optimizing agricultural policy, protecting biodiversity, and enhancing climate adaptation strategies at the regional scale.

In total, according to the regional classification, the following references pertain to Asia [1,16,18,19,20,21,22,23,25,26,27] and Africa [10,17].

3.2. AI for Sustainable Agriculture and Agroecosystem Management

Regional coverage. This sub-section includes 13 studies. The majority of cases pertain to Asia (n = 6), with additional case studies identified in Europe (n = 2; e.g., Western Greece and Luxembourg). Two studies adopt a global perspective (n = 2).

The integration of artificial intelligence into agriculture is unlocking new possibilities for the sustainable management of agroecosystems in the face of climate change and resource constraints. Machine learning (ML) and deep learning (DL) techniques are being increasingly employed to monitor vegetation and soil conditions, predict land suitability, optimize water and energy use, and minimize the environmental footprint of agricultural practices.

AI technologies play a crucial role in the spatiotemporal analysis of vegetation health. For instance, Ref. [29] applied multiple ML models—including Random Forest, AdaBoost, Back Propagation Neural Network (BPNN), and Stacking—to estimate grass height in the Eurasian steppe using MODIS data from 2001 to 2021. By incorporating meteorological and topographic variables, the study provided insights into pasture ecosystem health, biomass estimation, and carbon storage, thereby contributing to a scientific foundation for sustainable steppe management.

Additionally, a cubist regression model was employed to investigate global trends in vegetation greening and browning in response to climate change. The study by [30] demonstrated how AI can accurately attribute the effects of temperature, precipitation, nitrogen deposition, and atmospheric CO₂ concentration on vegetation dynamics and forecast ecosystem productivity under various climate scenarios.

One of the key research areas involves predicting zones suitable for sustainable agricultural use. In [31], machine learning algorithms—including support vector machine (SVM), Mixture Discriminant Analysis (MDA), Random Forest (RF), and Classification and Regression Trees (CART)—were employed to identify potential areas for cultivating Moringa peregrina, a tree species critical for land restoration and food security. By integrating climatic, geological, and morphometric data, the study identified the primary growth determinants and produced a suitability map for the species.

AI algorithms are also widely used for spatial analyses of land suitability across different land-use categories. For example, the study by [32] optimized an Agricultural Land Suitability Analysis (ALSA) by incorporating climatic, soil, and hydrological parameters. The model significantly improved the precision of agricultural potential assessments. However, the authors emphasize the need to account for socio-economic variables to ensure a more comprehensive approach to sustainability.

Artificial intelligence is increasingly applied in the management of agricultural resources under conditions of climatic instability. The study by [33] provides an overview of regional CROP-AP models developed for forecasting yield, water demand, greenhouse gas emissions, and climate change impacts. It highlights that AI integration substantially enhances model accuracy and adaptability, and it underscores the importance of open-source codes and shared datasets as priorities for the scientific community.

To assess the throughput capacity of interconnected water–land–energy systems in Heilongjiang Province, China, the RBMO-RF algorithm was developed, combining Random Forest with the Red-Billed Blue Magpie Optimization method. The model [34] revealed spatiotemporal patterns and key drivers—such as irrigation efficiency and balanced resource allocation—offering scientific support for sustainable agricultural planning.

AI also enhances soil monitoring and management. In [35], machine learning was applied to predict soil quality based on microbiome data and physico-chemical properties, enabling early detection of degradation and more effective agronomic decision making. For the assessment of heavy metal contamination (Cd, Pb, Cr, Hg), the study by [36] employed ensemble models including Random Forest and Gradient Boosting Machine, reducing prediction errors by 58% and improving food safety outcomes.

The application of artificial intelligence and deep learning in agroforestry presents new opportunities for evaluating and improving land-use systems. In [37], a DL model based on U-Net architecture was developed for tree component classification in agroforestry landscapes. This approach allowed for precise analysis of land-use types and provided robust support for sustainable development initiatives.

The integration of artificial intelligence and the Internet of Things (IoT) enables real-time monitoring of soil properties and adaptive management of water and fertilizer use. In the study by [38], these technologies contributed to increased crop yields, reduced operational costs, and minimized environmental impacts.

Additionally, smart farms leveraging cloud platforms and AI facilitate the automation of agricultural operations. Study [39] introduces the concept of a “smart cloud brain”, which coordinates the actions of unmanned agricultural machinery, IoT devices, and precision farming technologies, ensuring sustainable production with minimal reliance on pesticides and fertilizers.

To evaluate the agroecological impacts of agricultural practices in Luxembourg, agent-based modeling (ABM) was combined with life cycle assessment (LCA). The study by [40] illustrates how AI can forecast the effects of management changes—such as reduced livestock density or a shift to local feed—on emissions and economic performance, thereby supporting the development of sustainable agricultural strategies.

Finally, the study by [41] focuses on forecasting water usage and optimizing the distribution of water and energy resources to enhance the resilience of agroecosystems and reduce their carbon footprint. The authors emphasize the need to integrate digital, robotic, and biological approaches in the future of agriculture.

The cumulative analysis of the reviewed studies demonstrates that AI has become an integral component of modern agriculture. Its implementation not only improves the precision of monitoring and forecasting but also transforms approaches to managing land, water, and soil resources. This enables adaptation to the impacts of climate change, enhances productivity, and supports food security while reducing anthropogenic pressure on the environment.

AI plays a pivotal role in strengthening the sustainability of agroecosystems at both regional and global scales—from steppe pastures and degraded lands to high-intensity agricultural areas. Ongoing advancements in AI algorithms, their localization to specific conditions, and integration with other digital technologies lay the foundation for transitioning toward more sustainable and efficient models of agricultural production in the future.

In total, according to the regional classification, the following references pertain to Asia [29,31,34,36,37,39] and Europe [38,40].

3.3. AI for Monitoring, Forecasting, and Management of Aquatic Ecosystems

Regional coverage: This sub-section includes 34 studies. The majority of cases pertain to Asia (n = 20; e.g., China, Japan, Indonesia, Taiwan, India, Bangladesh, Cambodia), with additional case studies identified in Africa (n = 2; Algeria and South Africa), Europe (n = 1), and North America (n = 4, e.g., USA). Eight studies adopt a global perspective (n = 8).

AI is gradually assuming a central role in environmental monitoring and the sustainable management of aquatic and coastal ecosystems. In the face of global climate change, urbanization, increasing freshwater demand, and water pollution, there is a growing need for precise, scalable, and adaptive tools that can provide timely and reliable data for informed decision making.

The study by [42] presents an integrated water resource management system that combines natural process modeling with AI tools and hierarchical optimization strategies. This approach accounts for the interactions between groundwater and agriculture in arid coastal regions. AI contributes to multi-criteria decision optimization, ensuring a sustainable balance among water quality, availability, and regional socio-economic conditions.

One of the most prominent areas for AI application remains coastal zone monitoring, where anthropogenic pressures such as aquaculture, urban expansion, recreational use, and pollution are particularly intense. Study [43] introduced an object-based model for identifying aquaculture facilities in Asia using SAR and optical satellite imagery. This model effectively tracked spatiotemporal changes and assessed the degradation levels of coastal ecosystems. In a related study, Ref. [44] used Landsat imagery and machine learning algorithms on Google Earth Engine to map aquaculture ponds and their impact on mangrove forests. The study identified symbiotic zones and priority areas for ecosystem restoration, with seasonal water level variations improving the accuracy of anthropogenic impact assessments.

Using machine learning techniques, study [45] analyzed the impact of climate change and land use on phytoplankton and cyanobacteria in lakes, drawing on data from 1971 to 2016. The findings reveal that climatic factors dominate in remote areas, whereas urbanization emerges as the key driver in urban lakes, underscoring the need for locally tailored management approaches.

Study [46] emphasizes the shift from empirical models to AI-driven approaches for processing remote sensing data, which enables the analysis of long-term and large-scale changes in lakes, including phytoplankton blooms and eutrophication. The advancement of interpretable machine learning (IML) and explainable artificial intelligence (XAI) offers a significant leap in freshwater ecosystem monitoring. These technologies enhance predictive modeling and facilitate the integration of heterogeneous data streams to support sustainable water management decisions.

The study by [47] focused on coastal green infrastructure in Puducherry, India, utilizing Sentinel-2 imagery and ALTM laser topography. A Random Forest model was used to assess the effectiveness of ecosystems in mitigating climate risks such as flooding, as well as their role in biodiversity conservation and carbon sequestration.

In coastal Louisiana, the study by [48] employed XGBoost and Random Forest models to simulate land loss and predict vulnerable areas, offering restoration scenarios for future planning. Studies [49,50] focused on the analysis of marine pollution. The former examined marine debris concentrations in Japan using satellite and drone imagery processed through machine learning algorithms, while the latter developed AI-based models for automatic recognition of electronic waste, enabling proactive removal of toxic and non-biodegradable pollutants. Together, these studies highlight the potential of AI in combating marine pollution.

The study by [51] addressed the impact of unregulated tourism on marine ecosystems by developing an AI-driven decision support system (DSS) aimed at minimizing environmental damage. The WILDetect platform [52] leverages aerial imagery and AI to monitor seabird populations, generating biodiversity data and enabling the detection of ecological trends for the formulation of effective marine conservation strategies.

The study by [53] applied neural networks, geographical detectors, spatial correlation analysis, and trend analysis to assess pollution in the Yangtze River Basin, focusing on NPPE (nitrogen and phosphorus) emissions and their dynamics across municipalities. Additionally, the study by [16] contributed to wetland monitoring by introducing AI-based methods for analyzing spatiotemporal changes and mapping aquatic ecosystem structures using remote sensing data.

Considerable attention is given to the assessment of water quality and the forecasting of its changes. The study by [54] tested five machine learning algorithms d tree, K-nearest neighbors, discriminant analysis, support vector machines, and extra trees) and found that SVM achieved the highest accuracy (95.4%) in predicting the Water Quality Index (WQI) in Algeria, thereby reducing the need for costly laboratory analyses. The study by [55] utilized support vector machines to estimate nutrient concentrations based on Landsat satellite data in Yueqing Bay (China), identifying both natural and anthropogenic influences.

The use of Random Forest models and Landsat data in the study by [10] enabled the identification of high-risk zones and the prediction of aquatic ecosystem degradation, particularly in areas surrounding reservoirs, rivers, and agricultural landscapes. The study by [56] developed an LSTM model to reconstruct missing nitrate concentration data in watercourses, a method particularly relevant for regions with infrequent monitoring. The study by [57] applied decision tree and KNN algorithms to model groundwater vulnerability to nitrate and salinity contamination in the Indo-Gangetic Plain, highlighting zones of high toxicity.

The study by [58] conducted a long-term analysis (1985–2019) of groundwater levels using satellite data and AI techniques, revealing a persistent decline in 44% of the examined regions, posing a threat to the hydrological integrity of local watersheds. In the study by [59], deep learning (VMD-iTransformer) was employed to accurately forecast groundwater levels in the arid Kubuki Desert region. The model effectively addressed the non-stationarity of time series data and maintained high predictive accuracy even under data-scarce conditions.

The study by [60] applied the XGBoost machine learning algorithm to improve the accuracy of groundwater quality (GWQ) assessment in a coastal region of Bangladesh. The XGBoost model achieved high sensitivity (R² = 0.97), significantly reducing uncertainty in forecasting the Water Quality Index.

In the study by [61], water turbidity in the Yangtze River Delta was analyzed over the period 1990–2020 using remote sensing techniques and big data analytics. Satellite imagery from Landsat was processed to assess the impact of urbanization, ecological engineering interventions, and the Three Gorges Dam. The study by [62] applied deep learning to correct anomalies in satellite time series data, enabling improved ecological assessment of the Miyun Reservoir in China.

Study [63] employed a multi-model AI approach to analyze the impacts of land-use change on hydrological processes, including the prediction of emissions and water consumption. In the study by [64], AI was used to identify the key natural and anthropogenic drivers influencing water conservation in the tropical forests of Hainan, China—factors such as precipitation, evaporation, and agricultural activity were highlighted. The study by [65] utilized NDVI data and machine learning methods to forecast the impacts of urbanization on water conservation zones in the Funiu Mountains, focusing on spatiotemporal changes in hydrological regulation and soil erosion.

Equally important is the application of AI in water balance forecasting and resource management under changing climate conditions. The study by [32] modeled water availability and irrigation demand, contributing to drought and water scarcity risk reduction. The study by [41] explored the use of digital twins, AI, and parallel intelligence for analyzing water use in arid ecosystems and improving the efficiency of resource distribution. The research emphasized the importance of international cooperation and virtual water trade as mechanisms to support sustainable water management strategies.

The study by [66] demonstrated how the integration of AI, IoT, and satellite data improves the accuracy of hydrological forecasts and water cycle management by reducing uncertainty in precipitation and runoff modeling. In the study by [67], the Dikrong River system was assessed using ADCP and machine learning techniques, revealing the impact of anthropogenic pressures on flow parameters, current velocity, and forest cover change.

The study by [68] applied GRACE satellite data and machine learning methods to examine how agricultural greening and increasing NDVI contribute to the depletion of terrestrial water storage in Northwest China. Mahdian [69] analyzed the dynamics of the Caspian Sea, linking wetland loss to precipitation levels, sea level changes, and runoff patterns.

In the study by [70], a comprehensive investigation of the Mekong region in Cambodia combined remote sensing, geographic information systems (GIS), and AI to detect trends in declining water resources. The analysis considered both climatic and anthropogenic drivers, including irrigation practices and land-use changes.

The study by [71] proposed a cloud-based platform that integrates AI and remote sensing technologies, designed to support environmental monitoring in rural and vulnerable regions. This initiative ensures cost efficiency, scalability, and the active involvement of local communities in the protection and management of water resources, including in villages and municipalities of developing countries.

Thus, the analysis of an extensive body of contemporary research highlights the increasing integration of AI into tasks related to the monitoring, assessment, and management of aquatic and coastal ecosystems. AI not only reduces operational costs and enhances precision but also enables predictive modeling, early risk detection, and the formulation of locally adapted strategies.

The contribution of AI to the resilience of local water ecosystems is particularly significant. Its application enhances the effective protection of biodiversity, supports the maintenance of water balances, and facilitates adaptation to both climatic and anthropogenic pressures. By combining algorithmic precision, real-time data processing, and the richness of remote sensing information, AI is transforming water management approaches—from reactive to predictive, from fragmented to systemic—thereby reinforcing the resilience of natural environments and social infrastructure.

In total, according to the regional classification, the following references pertain to Asia [16,43,44,47,49,51,53,55,57,59,60,61,62,63,64,65,67,68,69,70]; Africa [10,54]; Europe [45] and North America [45,48,56,58].

3.4. AI in Forest Ecosystem Conservation and Management

Regional coverage: This sub-section includes 15 studies. The majority of cases pertain to Asia (n = 9; e.g., India, China, Iran, Indonesia), with additional case studies identified in Africa (n = 1; Morocco), Europe (n = 3; e.g., Romania, Italy), and Latin America (n = 1, Colombia). One study adopts a global perspective (n = 1).

Forest ecosystems are fundamental to global biospheres’ equilibrium, contributing to carbon sequestration, biodiversity conservation, and the protection of soils and water resources. However, escalating anthropogenic pressures and climate change are increasing the risks of forest degradation, thereby necessitating the development of more precise, scalable, and adaptive management tools. In this context, AI—including both ML and DL techniques—has emerged as a key enabler in advancing forest monitoring, forecasting, and sustainable land-use planning.

A critical application of AI lies in wildfire prediction and risk mitigation. For instance, the study by [72] employed the MaxEnt model, leveraging climatic and biophysical variables to assess fire susceptibility in Indian forest landscapes. This model supports early warning systems by identifying high-risk zones with notable spatial accuracy. Similarly, Ref. [73] implemented ensemble models such as XGBoost, GBM, Random Forest, and deep neural networks (DNN) to improve wildfire probability prediction. The integration of explainable AI techniques, such as SHAP and LIME, enabled the identification of key contributing factors, including precipitation, evapotranspiration, and proximity to infrastructure.

Further advancements are illustrated in the work of [74], which applied reinforcement learning in combination with Markov decision processes to simulate spatial-temporal scenarios for forest management. This included timber harvesting, fire prevention, and pest control. Such models allow for the dynamic evaluation of alternative strategies and facilitate adaptive decision making, aligning forest governance with evolving environmental and socio-economic conditions.

Another critical research direction involves the assessment of forest biomass, carbon stocks, and deadwood dynamics. For instance, Ref. [75] combined remote sensing data with forest inventory records to estimate biomass trends across European forests. This approach enables the identification of spatial and temporal patterns and supports the formulation of evidence-based sustainable forest management policies. In a related study, Ref. [76] employed AI to model deadwood distribution under future climate scenarios, predicting potential losses of up to 13% by the mid-21st century. The models incorporate a range of biophysical and socio-economic variables, such as logging intensity and adaptive forest governance strategies.

The quantification of blue carbon reserves in mangrove forests has also benefited from AI applications. In [77], Random Forest and XGBoost models were used alongside satellite imagery to estimate carbon storage, improving the accuracy of carbon sequestration assessments and supporting the integration of ecosystem services into carbon markets and emission reduction strategies.

AI technologies further facilitate the development of three-dimensional ecological models. A notable contribution is the CPH-Fmnet model introduced in [78], which integrates Vision Transformer architecture and attention mechanisms to enable 3D reconstruction of forest environments using imagery from standard handheld devices such as smartphones. This high-precision model is well suited for forest cover assessment, monitoring, and inventory tasks. Additionally, Ref. [79] utilized machine learning algorithms and Sentinel-2 data to classify 24 forest habitat types with an accuracy of 87%, thereby enhancing biodiversity monitoring and informing national conservation policies.

Artificial intelligence is increasingly employed to detect long-term changes in forest cover and assess ecosystem resilience. For example, Ref. [80] conducted a spatiotemporal analysis of tropical forest decline in Jinghong, China, revealing a 25% decrease in forest cover between 1989 and 2018. Similarly, Ref. [81] applied support vector machine (SVM) algorithms and Landsat satellite imagery to assess the reduction in argan forests in Morocco from 1985 to 2017. This study provided a quantitative evaluation of ecosystem degradation, including losses in biodiversity and cultural heritage. In Iraq, Ref. [82] used an ensemble of deep learning models—comprising convolutional and fully connected neural networks—to forecast changes in forest cover, achieving the highest accuracy through integration of anthropogenic, climatic, soil, and topographical variables.

Mangrove forests, as particularly vulnerable ecosystems, have also become a focal point for AI-driven analysis. In mountainous regions of China, Ref. [83] employed Sentinel-1 and Sentinel-2 data combined with ensemble learning techniques to map forest cover and assess the influence of climatic and topographic factors. On Tuanaku Island in Indonesia, Ref. [84] utilized spectral indices and geospatial data to track mangrove dynamics over the 2010–2020 period, identifying phases of both expansion and degradation.

In Colombia, Ref. [85] leveraged AI and satellite imagery to estimate above-ground biomass of mangroves along the Pacific coast, thereby improving predictions of climate change impacts. Furthermore, Ref. [86] applied XGBoost and multiple satellite datasets (Landsat 8, Sentinel-2, WorldView-2, and Zhuhai-1) to calculate the leaf area index (LAI), enabling effective monitoring of mangrove ecosystem recovery and resilience under changing climatic conditions.

Thus, the integration of artificial intelligence into forest management encompasses a wide range of tasks, from fire prevention and degradation monitoring to biomass estimation and carbon stock assessment. AI enables a shift from fragmented interventions to comprehensive ecosystem management by incorporating ecological, climatic, and socio-economic parameters.

The use of explainable AI techniques and advanced deep learning architectures enhances the transparency and adaptability of forest governance. Importantly, AI contributes to the resilience of both terrestrial and coastal forest ecosystems by supporting the development of strategies that align biodiversity conservation, reduction in anthropogenic pressures, and the fulfillment of climate commitments.

In total, according to the regional classification, the following references pertain to Asia [72,73,77,78,80,82,83,84,86]; Africa [81]; Europe [75,76,79] and Latin America [85].

3.5. Carbon Monitoring, Vulnerability Assessment, and Ecosystem Change Forecasting Using AI

Regional coverage: This sub-section includes 14 studies. The majority of cases pertain to Asia (n = 9; e.g., China and Iran), with additional case studies identified in Europe (n = 1; Italy). Four studies adopt a global perspective (n = 4).

Artificial intelligence technologies and satellite-based remote sensing have become critical tools in ensuring the accuracy and timeliness of environmental monitoring, particularly for assessing climate-related risks, carbon balance, and ecosystem resilience. The integration of AI into climate research enables the development of more reliable forecasts, a detailed analysis of spatiotemporal dynamics, and the adaptation of conservation strategies to rapidly changing environmental conditions.

One of the most promising directions involves the monitoring of greenhouse gases and carbon fluxes within ecosystems. For example, Ref. [11] highlights the role of the Chinese satellite TanSat-2 in providing high-precision estimates of atmospheric concentrations of CO₂, CH₄, and other pollutants. AI is employed to integrate “bottom-up” (ground-based) and “top-down” (satellite-derived) data, enhancing the accuracy of emissions inventories in line with the Paris Agreement. In another study, Ref. [87] applies a Random Forest model to assess carbon stocks in the Yangtze River Basin from 2001 to 2021. Through machine learning methods, the researchers identified carbon distribution patterns, distinguished between climatic and anthropogenic influences, and evaluated the role of forest ecosystems in carbon sequestration.

In [88], the integration of system dynamics models, patch-generating land-use simulation (PLUS), and Random Forest algorithms was employed to assess carbon sequestration by vegetation in Hubei Province, China. Climate scenario modeling under the SSP-RCP framework (CMIP6) revealed that the SSP1-1.9 scenario offers the highest potential for carbon sequestration through 2060.

A significant contribution to the understanding of carbon cycle instability is presented in [89], where machine learning algorithms were applied via the Google Earth Engine platform to evaluate landslide susceptibility in Italy. The findings indicated that large-scale landslides can shift ecosystems from functioning as carbon sinks to becoming carbon sources, thereby undermining the stability of biogeochemical cycles.

In [90], deep learning methods were used to enhance the estimation accuracy of gross primary productivity (GPP) by integrating outputs from CMIP6 climate models with satellite-derived data. This approach improved the precision of carbon balance projections and enabled more reliable assessment of climate change impacts on ecosystem productivity in China through the end of the 21st century.

Additionally, Ref. [91] introduced a CatBoost-based model incorporating topographic variables to improve GPP simulations in Fujian Province, China. The proposed model reduced prediction error by 16% compared to conventional approaches, contributing to the development of more robust and location-sensitive carbon monitoring systems.

Artificial intelligence is increasingly employed to assess climate-related risks and ecosystem vulnerability. In [6], a Random Forest model was used to identify the drivers of drought by integrating satellite observations, climate model outputs, and reanalysis data. This approach significantly improved the accuracy of Strategic Environmental Assessment (SEA), making it more data-driven and quantitative, and thereby strengthening its role in climate adaptation planning.

In Golestan Province, Iran, a study [92] utilized a maximum entropy (MaxEnt) machine learning model to assess both current and projected drought susceptibility for the period 2030–2050. The analysis incorporated CMIP6 climate projections and socio-economic pathways (SSP1-2.6, SSP2-4.5, SSP5-8.5), using 14 explanatory variables, including precipitation, temperature, soil moisture, and NDVI-based vegetation indices. The model demonstrated high predictive accuracy and identified key high-risk areas, highlighting the compound influence of land-use change, urbanization, and climate variability on regional hydrological vulnerability.

In a related study, Ref. [69] employed long short-term memory (LSTM) neural networks to forecast the decline of wetlands in Iran under the influence of climatic and anthropogenic stressors. The results indicated a high likelihood of seasonal desiccation by 2058, emphasizing the urgency of integrated climate adaptation measures for water-sensitive ecosystems.

Artificial intelligence is also playing an increasingly important role in environmental meteorology, particularly in analyzing the complex interrelationships between meteorological conditions and ecosystem health, as well as in forecasting natural disasters. As highlighted by [9], these capabilities are essential for supporting ecosystem resilience in a changing climate. In a recent study, Ref. [93] implemented a deep neural network to model ecological deficit in Henan Province, China, using indicators such as net primary productivity (NPP), aerosol optical depth (AOD), and population density. The deep learning approach enabled accurate forecasting of ecological dynamics, providing a robust scientific foundation for sustainable regional development and monitoring anthropogenic pressures on the environment.

To assess climate resilience in the Yangtze River Delta urban agglomeration, an integrated DPSIR (Driving forces–Pressures–State–Impact–Responses) framework was developed, augmented with machine learning techniques and spatial autocorrelation analysis [94]. The study employed geographically and temporally weighted regression and a super-efficiency slacks-based measure (super-SBM) model to evaluate spatial heterogeneity and improvement potential in climate resilience, offering valuable insights for evidence-based regional governance.

Additionally, Ref. [95] demonstrated how AI integration into life cycle assessment (LCA) frameworks can be used to evaluate the climate impact of buildings, including CO₂ emissions and natural resource consumption. Such approaches are critical for advancing the development of climate-neutral architecture and environmentally sustainable construction practices.

Special attention has been given to the application of AI in integrating localized data into climate monitoring systems. For instance, Ref. [96] introduced a spatial analysis approach based on citizen complaints, processed using natural language processing (NLP). By segmenting the text and applying geospatial tagging, the study identified areas of air pollution that closely aligned with data from official monitoring stations in Shandong Province, China. This method demonstrates the potential of AI to expand the scope of environmental surveillance and promote citizen engagement in environmental quality control.

In summary, AI has proven highly effective in advancing climate monitoring and forecasting—from global carbon inventories to vulnerability assessments of local ecosystems. Its ability to process vast and multidimensional datasets enhances the precision of climate modeling and the prediction of environmental risks.

Moreover, AI facilitates the integration of local and global scales of analysis, thereby supporting the development of adaptive strategies, sustainable resource management, and the fulfillment of international climate commitments.

In total, according to the regional classification, the following references pertain to Asia [69,87,88,90,91,92,93,94,96] and Europe [89].

3.6. AI in the Assessment and Management of Ecosystem Services

Regional coverage: This sub-section includes 26 studies. The majority of cases pertain to Asia (n = 14; e.g., India, China, Taiwan), with additional case studies identified in Africa (n = 3; South Africa), Europe (n = 1; Italy), and Latin America (n = 2; e.g., Colombia). Six studies adopt a global perspective (n = 6).

The application of artificial intelligence methods in the assessment, monitoring, and forecasting of ecosystem services (ES) presents new opportunities for sustainable management of natural areas and the maintenance of ecosystem functional stability. Recent research in this domain demonstrates the potential of AI to support automatic land zoning, spatiotemporal modeling, analysis of anthropogenic pressures, and the integration of ES into conservation planning and economic development strategies.

Several studies emphasize the role of AI in spatial modeling and assessment of ES. For instance, Ref. [97] employed Self-Organizing Feature Maps (SOFM) and support vector machines (SVM) to analyze the spatial heterogeneity of ecological functions, enabling automatic delineation of zones based on ecosystem characteristics. In a related study, Ref. [10] applied AI for spatially explicit monitoring of groundwater-dependent ecosystem (GDE) services, including vegetation variation indices, which facilitated the evaluation of biodiversity potential and livelihood support capacity.

Additionally, Ref. [98] used Rotation Forest (RF) models to reclassify satellite imagery and assess the impact of land-use changes on the economic value of ecosystem services. This approach enables the analysis of landscape configuration, the identification of trends in forest and pasture expansion, and the support of balanced natural resource planning at the regional level.

Interpretable machine learning models combined with spatiotemporal analysis were employed in [99] to quantitatively assess ES losses—measured using the GEP indicator—caused by coal mining activities. This approach enabled the identification of industrial activity thresholds compatible with sustainable development. In a case study of Shuozhou city (China), Ref. [100] demonstrated the application of a geographically weighted artificial neural network (GWANN) to evaluate the contribution of anthropogenic and climatic factors to ecosystem health, with a particular focus on the impacts of coal extraction. The use of AI facilitated a detailed analysis of spatiotemporal dependencies and helped identify zones most negatively affected by mining operations. Similarly, Ref. [101] applied the XGBoost–SHAP framework to model the spatiotemporal evolution of ecosystem services in a karst region of China, revealing the key determinants of degradation and potential mitigation strategies.

According to [102], artificial intelligence and remote sensing technologies provide effective tools for monitoring spatial changes in ecosystem services driven by both climate change and human activity, thus supporting the development of conservation strategies. Human-induced changes in the value of ecosystem services can also be tracked through the analysis of socioeconomic influences. For instance, Ref. [65] highlights that demographic and economic drivers are becoming dominant factors in shaping ES dynamics, necessitating adaptive management responses to ensure their preservation.

In industrial and mining regions, artificial intelligence facilitates the assessment of synergies and trade-offs between environmental and economic priorities. The study by [103] explores the coordination between ecosystem services and regional development in the Qinling Mountains (China), accounting for the impacts of urbanization, tourism, and industrialization. This approach enables the identification of zones with weakened linkages, highlights regional specificities, and supports the design of context-sensitive sustainable development strategies.

Special attention has been given to cultural and recreational ecosystem services. In [104], the XGBoost, Random Forest, and Naïve Bayes models were applied to analyze land degradation in the Cradle of Humankind and its implications for tourism attractiveness. The study revealed that the decline in vegetation cover, habitat destruction, and the loss of aesthetic appeal pose significant threats to the sustainability of ecotourism, thereby necessitating more precise forecasting and adaptive conservation strategies.

Additionally, Ref. [105] employed social media data to automate the assessment of cultural ecosystem services in national parks. AI enabled the automation of data collection and analysis, enhancing the understanding of visitor preferences while incorporating intangible values such as aesthetic appeal, recreational significance, and educational potential.

User-generated content (UGC) analysis has been proposed in [106] as a method for assessing the value of nature in tourism and informing spatial planning. In [107], the InVEST model was used to evaluate the tourism attractiveness of the Basilicata–Apulia region. A related approach is presented in [2], where automated analysis of urban greenery was applied to preserve ecosystem services such as microclimate regulation and air quality improvement.

The role of AI and IoT in promoting sustainable ecotourism and forecasting environmental risks is examined in [12,108]. The former highlights the benefits of AI for generating personalized tourist routes, managing visitor flows, and minimizing environmental impacts on natural ecosystems. The latter emphasizes the potential of AI to enhance monitoring accuracy, develop early warning systems, and establish sustainable models for managing ecological tourism.

System dynamics (SD) modeling combined with the STELLA decision support system (DSS) platform was utilized in [51] to conduct scenario-based assessments of tourism’s impact on coastal ecosystems, aiming to balance economic development with environmental risk mitigation. Furthermore, Ref. [109] demonstrates how a hybrid intelligent (HI) system, integrating citizen science and machine learning, can be employed to monitor marine biodiversity in tourist areas and strengthen the protection of marine ecosystems.

AI tools are also employed to assess natural capital and support the development of sustainable economic strategies. In [110], an artificial neural network (ANN) was applied to estimate carbon sequestration in Indian wetlands, identifying key climatic determinants such as temperature and dissolved oxygen levels. The findings contribute to the optimization of ecosystem service management, particularly by reducing methane emissions and enhancing ecological restoration efforts.

In a study of ecological restoration on the Yunnan–Guizhou Plateau during the period 2000–2019, Ref. [111] utilized remote sensing techniques and the Random Forest algorithm to identify factors influencing ecosystem recovery. Machine learning enabled a quantitative assessment of restoration across three dimensions—macrostructure, quality, and ecosystem services—while revealing the primary natural and anthropogenic drivers.

Bayesian learning was applied in [112] to analyze the dependence of vulnerable social groups on natural resources, providing a more effective basis for ecosystem service management strategies under conditions of limited data availability. Using AI, Ref. [13] uncovered complex interrelations among land-use change, food production, emissions, and agricultural sustainability.

An AI-based system integrating deep learning and blockchain was developed in [113] to support ecosystem management in the Páramo region of the Andes. This system enables the monitoring of threats such as deforestation, mining, and agricultural pressure, facilitates data analysis, and enhances decision-making processes. In [114], a hybrid model combining AdaBoost, XGBoost, and CatBoost was proposed to estimate the gross ecosystem product (GEP) using environmental data, achieving high interpretability and robust regional-level forecasting.

The study in [47] employed the Google Earth Engine cloud platform and machine learning algorithms to map coastal green infrastructure in Puducherry, India. By combining Sentinel-2 satellite imagery and ALTM laser topography, a Random Forest model was developed to effectively monitor ecosystems that provide essential services, including flood protection, carbon sequestration, and biodiversity conservation.

In [115], AI was used to analyze soil resource changes and assess the carbon storage potential of ecosystem services in an Indian watershed using the InVEST model. Land-use mapping and carbon stock estimation were based on remote sensing data and ground control points processed through machine learning algorithms. The model enabled a quantitative assessment of current ecosystem service conditions and simulated two land-use change scenarios: forest expansion and agricultural land expansion. The results highlight the potential of machine learning and spatial modeling for evaluating ecosystem functions.

Overall, the analysis of these studies demonstrates the growing significance of AI in the quantitative assessment and spatiotemporal modeling of ecosystem services across diverse socio-ecological contexts. AI tools enable the integration of remote sensing data, field observations, and spatial and socio-economic information, facilitating a comprehensive approach to evaluating natural capital. These tools contribute to the development of adaptive management strategies that consider local ecosystem characteristics, anthropogenic pressures, climate risks, and social vulnerability. Thus, the reviewed research underscores AI’s emerging role as a key enabler of local ecosystem resilience and the continued provision of essential ecosystem services under global environmental change.

In total, according to the regional classification, the following references pertain to Asia [47,51,65,97,98,99,100,101,102,103,109,110,111,115]; Africa [10,104,112]; Europe [107] and Latin America [106,113].

3.7. AI in Regional Contexts: Business, Industry, and Agriculture

Regional coverage: This sub-section includes 33 studies. The majority of cases pertain to Asia (n = 6; e.g., China, India), with additional case studies identified in Africa (n = 5; e.g., Tunisia, Morocco, South Africa), Europe (n = 4; e.g., Italia), North America (n = 1, USA), and Latin America (n = 1). Sixteen studies adopt a global perspective (n = 16).

The application of artificial intelligence in regional contexts is becoming a critical factor for sustainable development across various sectors—from business and agri-food production to urban and industrial planning. AI contributes to the transformation of traditional economic models by enhancing adaptability, decision-making precision, and resource-use efficiency.

Recent studies highlight how digital technologies—including machine learning, IoT, digital twins, and cloud platforms—are being integrated into business operations, agriculture, and infrastructure management at both local and regional levels. The use of AI enables the consideration of spatial and socio-economic disparities, the identification of hidden dependencies, and the formulation of comprehensive solutions aimed at increasing the resilience and competitiveness of local ecosystems.

The adoption of AI in small- and medium-sized enterprises (SMEs) and industrial systems plays a vital role in enhancing sustainability, environmental performance, and digital maturity. As noted in [116], AI and cloud computing offer significant potential for SMEs, particularly in supply chain management, improving business model transparency, and optimizing resource use. The study emphasizes the importance of technological transformation in SMEs as a pathway to sustainable development and balanced regional growth.

The study in [117] examines digital transformation across Moroccan industries, identifying a lack of managerial knowledge and the need to develop flexible innovations, ethical AI frameworks, and workforce training programs. These findings are supported by earlier research in [118], which highlights key barriers to AI adoption in Africa, including data scarcity, limited funding, and infrastructural fragmentation. AI integration is viewed as a driver of sustainable economic growth for local enterprises and ecosystems. Strategic approaches to AI adoption in developing economies are further explored in [119], where the authors propose a conceptual framework aimed at reducing social inequality, fostering inclusivity, and promoting sustainable growth through business model adaptation and job creation.

In [120], AI is employed to analyze the activities of over 1200 companies in the Piedmont region of Italy, using text-mining algorithms on LinkedIn profiles to map the diffusion of AI and digital technologies. This methodology enables the assessment of AI’s potential as a tool for monitoring regional innovation ecosystems and supporting decision making for sustainable technological development. The study in [121] presents an innovative combination of website text analysis from metallurgical companies with remote sensing data to detect discrepancies between stated environmental initiatives and actual pollution levels. This approach offers a valuable instrument for addressing greenwashing and advancing standards for sustainable reporting.

Artificial intelligence is also actively utilized in applied business processes. In [122], examples of logistics optimization through AI are explored, with a focus on adaptability to market conditions within specific local contexts. The study in [123] applies clustering algorithms (K-means) for customer segmentation and inventory optimization in the retail sector, demonstrating AI’s potential to increase profitability and reduce costs by accounting for regional consumer preferences.

The development of artificial intelligence has a profound impact on local agricultural and agri-food systems, promoting their sustainability, digital transformation, and climate resilience. AI is applied across all stages of agricultural production—from soil analysis and water resource management to logistics, ecosystem monitoring, and crop yield forecasting.

One of the key areas of application is digital soil mapping and condition monitoring. Studies [25,35] show how Random Forest models, support vector machines, and decision trees are used to enhance soil resource management, thereby improving the efficiency of agricultural enterprises. AI enables farmers to plan fertilizer and water use with greater precision and facilitates the integration of these processes into digital precision farming platforms—an especially important factor for enhancing the resilience of local agroecosystems to climatic and economic challenges.

AI and automation in eco-friendly unmanned farms, as shown in [39], support the transition from traditional agricultural labor to intelligent management systems. These technologies combine ecological practices with digital solutions, enabling the implementation of smart agriculture under conditions of limited resources and increased climate instability.

AI-driven solutions for vertical farming also make a significant contribution to the resilience of food systems, as emphasized in [8]. The application of AI in resource management—including water, light, and nutrients—and predictive analytics helps to optimize production processes, reduce losses, and enhance local food security.

The study in [3] demonstrates the integration of AI, sensors, drones, and IoT in smart agriculture. These technologies facilitate rapid adaptation to environmental changes, improving crop yields while reducing ecological impacts. In [31], AI models such as Random Forest (RF), support vector machines (SVM), Multiple Discriminant Analysis (MDA), and Classification and Regression Trees (CART) are applied to identify optimal cultivation zones for plants like Moringa peregrina. This contributes to the development of sustainable income sources for local enterprises, reduces risks for farmers, strengthens local value chains, and supports efforts to prevent land degradation.

The sustainable food planning model presented in [7] employs AI to analyze urbanization, resource demand, and availability within agri-food subsystems. This approach supports the reduction in food waste, improves land-use efficiency, balances production and consumption, and enhances the resilience of food ecosystems.

AI is also applied in studies evaluating the impact of technological innovation and land-use management on crop yields in BRICS countries, as shown in [13]. AI helps identify correlations between productivity, land use, and emissions, which is essential for adapting agriculture to climate challenges and strengthening its sustainability.

In [124], AI is proposed for monitoring livestock density and assessing the environmental impacts of cattle farming. Using Random Forest algorithms and Sentinel-2 satellite data, the study provides the first regional-level assessment of cattle stocking intensity—an essential input for sustainable pasture management.

A systems-based approach to agroecosystems is demonstrated in [40], where agent-based modeling and life cycle assessment are used to forecast the environmental and economic outcomes of management decisions, based on a case study of agriculture in Luxembourg. These digital models offer valuable tools for designing sustainable agribusiness strategies that account for subsidies, market dynamics, and environmental constraints.

Moreover, AI and machine learning are being integrated into digital platforms such as DigiGram, which is designed to support the digital transformation of rural ecosystems in India [125]. This platform facilitates the holistic development of rural regions—spanning agriculture, trade, education, and healthcare—and underscores the importance of digital solutions in advancing smart villages and building resilient and inclusive local ecosystems.

Thus, the application of AI in agriculture goes beyond technological modernization; it establishes a new paradigm of sustainability at the local level. Through AI integration, local agroecosystems become more adaptive, environmentally sustainable, and economically efficient, supporting their development amid global change.

AI is also emerging as a key driver in the transformation of industrial and business ecosystems, particularly in the context of the Fourth Industrial Revolution. Its adoption enables small, medium, and large enterprises to enhance resilience, improve environmental performance, and strengthen competitiveness, while adapting to the digital environment and meeting the demands of sustainable development.

The integration of AI into industry enhances automation and robotics, fostering hyper connectivity in production processes and improving operational efficiency. As emphasized in [126], the interaction between AI and digital twins enables precise modeling and optimization, supports the development of distributed infrastructure, promotes servitization, and facilitates the transition to a symbiotic economy. A similar approach is demonstrated in [4], where the integration of AI with virtual and augmented reality strengthens experiential learning, particularly in industry, urban planning, and healthcare, creating potential for cross-sectoral synergies.

AI is also increasingly used as a tool for assessing and analyzing business sustainability. In [127], cluster analysis and principal component methods were applied to evaluate the resilience of business ecosystems in border regions, identifying key internal and external determinants of local sustainability. The study in [128] explores the main factors contributing to business resilience in a turbulent world, while [129] presents a clustering analysis of agri-enterprises.

AI applications extend to specific sectors as well. For instance, Ref. [130] examines its use in the sustainable production of cultivated meat (CM), where Industry 4.0 technologies help overcome infrastructural barriers and enhance the potential of start-ups.

Artificial intelligence is also playing a transformative role in the development of urban ecosystems by reshaping the management of natural, infrastructural, and social resources in response to ongoing urbanization. The application of AI and machine learning to assess and enhance the environmental sustainability of urban environments has been demonstrated in a number of studies. For example, Ref. [131] illustrates how the AHP, GCA, and BPNN models can support the development of sustainable urban infrastructure within the framework of sponge city concepts. These models contribute to the prediction of hydrological risks and facilitate the adaptation of urban development to extreme weather conditions.

The study in [132] investigates the impact of urban green infrastructure (UGI) on urban temperature dynamics, employing remote sensing and digital modeling methods to quantitatively assess the cooling effect across different spatial scales, from plot level to citywide. This supports the optimization of UGI planning strategies, improves urban ecosystems, and fosters sustainable development in metropolitan areas under climate change conditions. Similarly, AI is applied to analyze data on trees located on private land, assisting municipalities in developing more effective urban greening strategies and sustainable management of green spaces, as shown in [2]. Machine learning enables the automated collection and processing of data on tree size and location, reducing costs and increasing the accuracy of urban vegetation monitoring.

In the context of waste and pollution management, AI models are employed to assess risks associated with landfill leachate and pollutant gas emissions. The study in [133] demonstrates the potential of AI in forecasting the spread of contaminants, optimizing waste recycling processes, and designing sustainable strategies for urban environmental safety.

At the intersection of natural resource management and disaster response, Ref. [134] demonstrates the use of AI and machine learning to analyze the impacts of natural disasters—such as typhoon—on forest ecosystems near urban areas. This enables municipal and governmental agencies to plan the restoration of affected territories, reduce the risk of infrastructure damage, and improve preparedness for future extreme events. In [135], an environmental monitoring methodology is introduced, based on the digital information-analytical system “Ecological Barometer”, which is designed to assess anthropogenic pressure and prevent environmental threats. The system supports environmental quality management and sustainable territorial development by providing strategic decision-making data for public authorities.

In the context of monitoring progress toward Sustainable Development Goals 13 (climate action) and 15 (life on land) in Tunisia, the study proposes integrating Earth observation (EO) data with deep learning (DL) architectures [136]. A DL-based model is used to quantify progress in SDG implementation, demonstrating AI’s potential in analyzing large-scale environmental datasets and informing decision making at the level of local ecosystems.

Particular attention is given to the application of AI in the context of smart cities in developing countries. As noted in [137], AI, ML, IoT, and blockchain play a crucial role in addressing the challenges faced by African cities, including inadequate infrastructure, unstable access to resources, and the need for adaptive solutions. The deployment of AI under such conditions enables the development of innovative ecosystems that are resilient to both economic and environmental challenges. At the same time, Ref. [118] highlights institutional and human capital barriers to AI adoption in Africa, stressing the importance of national digitalization strategies and capacity-building initiatives to support sustainable urban growth.

An example of digital twin implementation for urban infrastructure management is presented in [138], which introduces a digital village framework that leverages machine learning for environmental monitoring and enhanced emergency preparedness. These solutions strengthen the resilience of small settlements and have the potential to be scaled for urban applications.

The integration of machine learning with GIS methods for synchronizing biophysical and socio-economic data provides deeper insights into the causal relationships affecting ecosystems [139]. The application of such approaches to the study of urban system resilience is particularly important, as it promotes more comprehensive and sustainable urban planning.

In total, according to the regional classification, the following references pertain to Asia [25,31,39,125,131,134]; Africa [117,118,119,136,137]; Europe [8,40,120,127]; North America [121] and Latin America [124].

3.8. AI in the Transformation of Local Energy Systems

Regional coverage: This sub-section includes 10 studies. The majority of cases pertain to Europe (n = 2; Belgium and Italy), with additional case studies identified in North America (n = 1, USA). Seven studies adopt a global perspective (n = 7).

The application of artificial intelligence in the energy sector is driving the transformation of local energy systems by enabling resource consumption optimization, emission reduction, and enhanced resilience [140]. Research highlights a wide range of solutions in which AI contributes to improving energy efficiency—at the level of individual buildings, as well as across cities, regions, and the energy sector as a whole.

In [95], the use of AI and digital construction resources—such as connected devices, semantic models, and IoT—is examined for optimizing the life cycle assessment (LCA) of buildings. AI-based analysis of energy consumption, emissions, and environmental impacts enables the development of strategies for reducing energy demand and integrating renewable energy sources, thereby supporting sustainable construction and urban planning.

The study in [141] focuses on enhancing energy efficiency in urban infrastructure through the use of IoT device data. It proposes a method for appliance identification by converting one-dimensional signals into two-dimensional images, achieving an accuracy of 99.68%. This approach enables a more detailed understanding of consumption behavior and the creation of a taxonomy of data fusion strategies, contributing to more sustainable energy consumption monitoring and control.

The analysis of household electricity consumption data in Los Angeles [142] enables the identification of microclimatic zones that reflect the region’s topography and climate. The integration of AI with socio-economic data facilitates accurate forecasting and management of energy resources, supporting sustainable urban energy policies.

To ensure sustainable data collection and processing in distributed systems, the LSCEA-AIoT framework [143] has been developed, incorporating AI and machine learning methods, including Monte Carlo tree search. This architecture reduces energy consumption, extends network lifespan, and outperforms existing solutions by optimizing processes within AIoT ecosystems.

In [144], machine learning and natural language processing (NLP) techniques are applied to calculate China’s Energy Digitalization Index. By integrating spatial econometric models and investment analysis in the digital industry, the study assesses the impact of digital technologies on carbon productivity, identifies key drivers of low-carbon transformation, and formulates recommendations relevant for China’s local ecosystems as well developing countries.

The study in [145] presents a regional support system for photovoltaic energy distribution in Emilia-Romagna. The integration of big data, climate models, and AI algorithms enables accurate forecasting of energy production and storage. Such approaches enhance energy resilience, support climate policy goals, and promote the expansion of renewable energy sources.

The development of local energy markets (LEM) using AI is described in [146]. Multilayer neural networks and Gradient Boosting models are employed to forecast energy consumption and optimize the use of battery energy storage systems (BESS), thereby strengthening the resilience of local energy ecosystems and reducing risks for market participants.

In [147], examples are provided of micro grids managed by AI technologies to achieve Sustainable Development Goal 7 (SDG7). AI technologies improve the efficiency of managing distributed generation systems across various regions, including developing areas. This not only increases the share of renewable energy but also strengthens social capital, expands access to energy, and fosters trust in emerging technological solutions.

In [148], GIS and AI are used to assess the potential of non-wood biomass in Flanders and to optimize biogas supply chains. The OPTIMASS model demonstrated a positive energy balance for anaerobic digestion. However, its economic viability remains limited without government support. This highlights the need for a comprehensive approach when transitioning to bioenergy systems.

The application of the ACO-TCN-Attention model in [149] enables the assessment of the environmental and socio-economic impacts of marine renewable energy projects. The use of AI enhances the accuracy of evaluating effects on marine ecosystems and coastal communities, supporting evidence-based management and minimizing potential damage.

Thus, artificial intelligence is transforming local and global energy systems by promoting sustainable resource management, reducing emissions, and ensuring energy security. From the level of individual buildings and households to regional and marine energy networks, AI enables accurate forecasting, efficient energy distribution, and the integration of renewable energy sources.

In total, according to the regional classification, the following references pertain to Europe [145,149] and North America [142].

3.9. AI in Supporting Local Sustainable Development: Education, Health, and Social Inclusion

Regional coverage: This sub-section includes 13 studies. The majority of cases pertain to Asia (n = 4; e.g., Brunei, Philippines, Singapore), with additional case studies identified in Africa (n = 3) and Europe (n = 2; e.g., UK). Four studies adopt a global perspective (n = 4).

Artificial intelligence is increasingly recognized as a powerful enabler of sustainable development, contributing across environmental, economic, and social dimensions—such as education, healthcare, and inclusion. Recent research illustrates how AI can support the analysis of public attitudes, enhance engagement among diverse demographic groups, inform educational strategies, monitor biodiversity, and help build resilient healthcare systems, particularly within localized settings.

For instance, Ref. [150] applied machine learning to analyze over 25,000 newspaper articles focused on human–nature relationships, demonstrating how AI can be leveraged to explore public discourse around ecosystems and sustainability. This type of analysis helps uncover dominant narratives and informs the development of educational and outreach campaigns that align with public perceptions of ecological threats.

In a related study, Ref. [151] combined machine learning and social network analysis to assess the inclusivity of the Global Landscapes Forum (GLF), tracking the participation of youth, women, and African leaders in international land-use debates. The findings highlight AI’s capacity to support the design of inclusive educational platforms and decision-making processes that empower underrepresented groups in the governance of natural resources.

The study by [152] highlights the use of artificial intelligence in combination with passive acoustic monitoring (PAM) within biosphere reserves, contributing both to scientific biodiversity monitoring and the development of educational initiatives. AI-powered tools enable the automated processing of acoustic data, offering students and researchers practical access to nature-based learning and analytical methods.

In [153], the application of large language models (LLMs) and AI-driven chatbots is explored in the context of high school environmental education. Drawing on the value-sensitive design (VSD) framework, the study demonstrates how students can build causal understanding of the links between climate change and marine ecosystem health—fostering not only critical thinking but also ethical awareness in technologically mediated learning environments.

The contribution of AI to the advancement of environmental meteorology is presented in [9], which also emphasizes its role in training professionals in sustainability and climate strategies. Machine learning and big data analytics support the modeling of climate scenarios, facilitate the study of natural disaster impacts, and underpin adaptive planning based on scientifically grounded models.

The study by [154] focuses on the role of AI in transforming higher education in Asian countries, with a specific case from Brunei. AI helps bridge digital divides, promotes personalized learning, and enhances scientific collaboration. However, it also underscores the need to improve access to global research platforms and better integrate local institutions into the international academic ecosystem.

The study by [155] describes how China is leveraging artificial intelligence for science diplomacy and the promotion of sustainable climate solutions in Africa. The GeoGPT tool, developed under the Deep-time Digital Earth initiative, supports African researchers in natural resource management and climate adaptation, contributing to the creation of resilient scientific ecosystems.

In [156], the impact of digital healthcare ecosystems—comprising AI, wearable technologies, and telemedicine—on public health outcomes across 54 African countries is examined. Machine learning is used to analyze investment priorities and optimize resource allocation, thereby informing decision-making processes within local healthcare ecosystems.

The study by [157] applies interpretable machine learning (iML) to examine dengue fever incidence in the Philippines. Using NDVI and precipitation data, the study identifies key drivers of disease spread, including seasonality and urbanization, demonstrating AI’s potential in unified digital health systems that integrate environmental and medical data.

The role of AI in supporting molecular tumor boards (MTBs) is discussed in [158], where large volumes of clinical and genetic data are combined to enable personalized oncology treatment. These systems enhance the scalability of precision medicine approaches across local and national healthcare infrastructures, improving both the resilience and accessibility of medical services.

The study by [159] explores AI-based technological solutions for elderly care, focusing on the concept of “aging in place”. Through continuous health monitoring and adaptive home environments, these systems support the autonomy of older adults while reducing the burden on healthcare infrastructure.

In [160], a framework integrating the Internet of Medical Things (IoMT) with federated learning (FL) is proposed to ensure data privacy during edge-device processing. This approach contributes to the ethical and sustainable development of digital healthcare ecosystems within smart cities.

The importance of AI and network science in examining the link between marine environmental conditions and human health is emphasized in [161]. The study advocates for the integration of digital tools for collecting and analyzing data on coastal ecosystems, enhancing monitoring accuracy and enabling public health interventions aligned with the One Health approach.

Taken together, these findings underscore the versatility of AI in advancing sustainable development across education, healthcare, and social domains—all closely interconnected with the preservation and resilience of local ecosystems. By facilitating a deeper understanding of human–environment interactions, AI helps improve inclusivity and the effectiveness of educational and governance strategies, while also supporting the growth of digital resilience infrastructure at both local and global levels.

In total, according to the regional classification, the following references pertain to Asia [150,154,157,159]; Africa [151,155,156] and Europe [153,158].

3.10. Conceptual Synthesis: How AI Interventions Reinforce the Three Pillars of Regional Sustainability

This scoping review reveals a coherent pattern: regardless of the domain, AI operates as a data-driven feedback loop that (i) senses the state of the system, (ii) learns causal relations, and (iii) delivers prescriptions that translate into environmental, economic, and social co-benefits.

Table 2. Qualitative mapping of AI interventions in regional ecosystems to the three pillars of sustainability.

AI Domain/Section	Representative AI Techniques and Data	Environmental Impact	Economic Impact	Social Impact
Ecological monitoring and land use (3.1)	Remote sensing (Landsat, Sentinel, GEE); K-ELM, Random Forest, XGBoost, SHAP	Early detection of wetland and vegetation degradation; soil-nutrient and land-cover maps for conservation	Optimized zoning and infrastructure siting; lower restoration costs	Lower flood/erosion risk for communities; protection of local ecosystem services
Sustainable agriculture and agro-ecosystems (3.2)	ML Stacking, Cubist, IoT sensors, U-Net DL, agent-based crop models	Reduced fertilizer, water and pesticide loads; healthier soils	Higher and more stable yields; input–cost savings; precision-agriculture markets	Enhanced food security; stronger rural livelihoods; less agro-chemical exposure
Water and coastal ecosystems (3.3)	LSTM, object-based ML on SAR + optical, XAI, Random Forest	Improved water quality; mangrove and wetland protection; eutrophication control	Efficient irrigation and aquaculture siting; lower treatment costs; sustained fisheries	Flood-risk reduction; safer drinking water; coastal livelihood security
Forest ecosystems (3.4)	Fire-risk ensembles (XGBoost, GBM), Vision-Transformer 3D reconstructions, MaxEnt, SHAP	Fire prevention; carbon-sink conservation; deforestation monitoring	Verified carbon credits; sustainable timber harvest; eco-tourism planning	Health protection from smoke; recreation and cultural services for local communities
Carbon monitoring and climate resilience (3.5)	TanSat-2 GHG data, Random Forest, CatBoost GPP, CMIP6 + MaxEnt	High-resolution emission inventories; climate-risk hotspot mapping	Lower compliance/offset costs; climate-smart investment guidance	Reduced losses from extremes; policy transparency; cleaner air
Ecosystem services and natural capital (3.6)	XGBoost-SHAP, GWANN, InVEST, Rotation Forest	Preservation of regulating services (water, climate, habitat)	Valuation for PES schemes; eco-tourism revenue; balanced land-use trade-offs	Safeguarded livelihoods; cultural-heritage and recreation benefits; public preference integration
Regional business and industry (3.7)	NLP/text-mining of disclosures, digital twins, LinkedIn analytics, clustering	Detection of greenwashing; industrial-process optimization → lower emissions	Higher SME productivity and innovation; supply-chain transparency	New digital jobs; skills development; reduced regional inequality
Smart energy and infrastructure (3.8)	NILM classifiers, PV-yield ML, micro-grid and BESS optimization, OPTIMASS biomass logistics, LSCEA-AIoT	CO2 reduction; greater resource-use efficiency; higher RES penetration	Lower energy costs; reliable local grids; emergent local-energy markets	Broader energy access; community resilience; health gains from cleaner energy
Education, health and inclusion (3.9)	LLM-based curricula, interpretable ML disease maps, IoMT + federated learning, GeoGPT	Heightened public environmental awareness and stewardship	Green-skills human–capital formation; preventive-care cost savings	Inclusive education; personalized healthcare; civic engagement and social equity

Source: compiled by authors.

maps every empirical study from Section 3.1 to Section 3.9 to the sustainability pillars.

Cross-cutting enablers: Across all clusters, success hinges on open data infrastructures, interpretable AI, and local capacity building. These factors articulate why AI, when responsibly governed, can act as a unifying lever for the three sustainability pillars in regional ecosystems.

4. Discussions

The present scoping review underscores the accelerating importance of artificial intelligence in advancing the sustainability of regional and local ecosystems. Drawing on the analysis of 155 peer-reviewed publications, this study identifies nine functional domains in which artificial intelligence and machine learning methods are being deployed: environmental monitoring, climate-impact assessment, natural-resource optimization, smart agriculture, sustainable industry and manufacturing, urban infrastructure, logistics and mobility, health and education services, and broader socio-economic governance. Collectively, these domains reveal not only the breadth of applications but also the depth to which AI is penetrating decision-making processes that were once constrained by fragmented data streams and reactive management practices.

To interpret these findings within a unifying conceptual lens, we adopt a socio-technical-systems perspective [162] in which AI serves as an intermediary layer that links biophysical feedbacks to institutional response capacity; this framing helps explain why identical algorithms may yield divergent outcomes under different regulatory, cultural or infrastructural conditions.

The reviewed literature demonstrates, first and foremost, the growing sophistication with which AI is used to monitor and forecast environmental conditions. Algorithms ranging from deep convolutional networks to hybrid neuro-fuzzy systems are applied to spatial analyses of degradation processes across aquatic, forest, agricultural, and coastal ecosystems, enabling managers to pinpoint the primary drivers of change, model complex feedback loops, and design adaptive interventions that remain sensitive to regional ecological baselines [19,22,33]. In agricultural settings, for example, AI-enabled remote-sensing workflows now detect crop-stress signals weeks earlier than conventional field scouting, thereby reducing fertilizer and pesticide loads while safeguarding yield stability. Comparable advances are observed in coastal-zone management, where AI-integrated radar and satellite imagery help forecast shoreline retreat under compound pressures of sea-level rise and intensified storm surges.

These advances confirm rather than overturn existing knowledge about AI’s diagnostic power; their novelty lies in demonstrating how widely used techniques can be operationalized at sub-national scale and integrated into everyday resource-management routines.

Second, artificial intelligence is increasingly recognized as a platform technology whose impact transcends sectoral boundaries. Rural communities employ machine learning classifiers to optimize irrigation regimes at field scale, whereas megacities integrate multi-agent traffic models with real-time sensor networks to smooth congestion and cut particulate emissions [68,125,127,131,132]. By illuminating spatiotemporal dynamics that link biophysical processes with socio-economic behavior, AI tools help reveal causal relationships that had previously been masked by data heterogeneity and time-lag effects. The resulting situational awareness enhances the timeliness and robustness of policy choices, whether those choices concern adaptive zoning ordinances, drought contingency planning or the strategic placement of photovoltaic capacity within micro-grids.

Third, both agricultural producers and industrial operators exploit AI primarily to optimize natural and energy resource use, increase productivity, and shrink the environmental footprint of production processes [31]. In the manufacturing sector, digital twins of machinery and workflows feed data into reinforcement-learning agents that fine-tune process parameters on the fly, cutting material waste and energy intensity. Such AI-enabled optimization is becoming a pillar of corporate decarbonization strategies, particularly in the context of national commitments to climate neutrality and circular-economy targets [116,126]. By converting real-time operational data into prescriptive insights, artificial intelligence turns once-static life-cycle assessments into living management dashboards capable of guiding day-to-day resource allocation.

A fourth and equally significant finding is that AI can advance the social pillar of sustainability by expanding access to education, health care, and public services. Personalized learning algorithms adjust curricular pacing to individual student competencies, while diagnostic decision-support systems assist medical staff in underserved regions where specialist knowledge is scarce [153]. In several case studies from sub-Saharan Africa and South-East Asia, mobile health-monitoring platforms driven by AI analytics have improved early detection of vector-borne diseases, thereby strengthening local health-system resilience [156]. These examples illustrate how digital technologies, when appropriately adapted to local constraints, can bridge long-standing gaps in service delivery and enhance social inclusion.

Across all domains under review—agriculture, logistics, industry, urban infrastructure and natural-resource governance—AI adoption supports a paradigmatic shift from reactive to proactive management. Local context, however, remains a decisive variable. The adaptability of AI tools to idiosyncratic land-use patterns, climatic regimes, socio-economic structures and institutional capacities ultimately determines whether the promised efficiency gains translate into tangible ecosystem-service benefits. Where contextual alignment is strong, decision making becomes more transparent, environmental degradation risks are mitigated, and inclusive territorial development trajectories emerge. Conversely, misalignment often leads to stalled pilot projects, deskilling fears among workers, and the entrenchment of existing social inequities.

Cross-domain interactions reinforce these findings: AI-optimized irrigation schedules directly influence downstream water-quality models, while logistics algorithms that reroute freight traffic reshape urban air-pollution baselines. Such overlaps underscore the need for integrated governance frameworks capable of managing trade-offs between industrial efficiency, biodiversity conservation, and social equity.

Persistent trends detected in recent publications reinforce these observations. There is an unmistakable acceleration in the integration of AI into spatial modeling, risk-assessment methodologies, and early-warning systems, reflecting a broader scientific shift toward system-level analytics. Methodologically, the literature shows a move away from reliance on single machine learning algorithms toward more comprehensive, hybridized approaches that blend AI with digital-twin frameworks, IoT sensor webs, advanced geo-statistics, and multi-scale remote-sensing platforms. Such combinations not only improve predictive performance but also create operational feedback loops that can be actively steered toward sustainability outcomes.

The social dimension of AI applications likewise demonstrates positive momentum. Intelligent information systems facilitate citizen engagement in environmental governance by lowering the transaction costs of public participation. Open-access dashboards that visualize air-quality data or watershed health enable informed dialogue between civil-society groups and regulatory agencies, strengthening procedural justice in environmental decision making. Importantly, the literature shows that these technologies yield disproportionate benefits in regions that historically face deficits in institutional capacity, suggesting a democratizing potential when ethical safeguards and digital-literacy programmers accompany deployment.

Notwithstanding these encouraging developments, systemic challenges continue to constrain large-scale implementation. Digital inequality remains one of the most pressing barriers. Both cross-national and intra-national disparities in broadband access, high-performance computing resources, and data-science expertise hamper the diffusion of AI-based solutions, especially in rural and suburban settings [118]. Without concerted investment in digital infrastructure and human capital, AI may widen rather than narrow existing socio-economic divides.

Institutional and regulatory obstacles further complicate the landscape. The absence of coherent standards governing algorithmic transparency, data-protection protocols, and model interpretability is particularly problematic in safety-critical contexts such as health care, flood-risk management, and critical-infrastructure control [160]. Fragmented regulatory regimes not only heighten legal uncertainty for innovators but also erode public trust, which is essential for sustained technology uptake. Compounding this problem is the so-called black-box nature of many high-performance AI models; their internal logic is often opaque even to domain experts, limiting their suitability for public-sector decision making and undermining democratic accountability [46,73].

Recent practice points to several oversight mechanisms—independent algorithm audits, federated data spaces that preserve ownership rights, and public registers of high-impact models—that offer pragmatic pathways for aligning AI innovation with accountability and equity goals.

Limitations intrinsic to the present review must also be acknowledged when interpreting these findings. A substantial proportion of the analyzed literature focuses on case studies from Asia, whereas Africa and Europe remain comparatively under-studied; this regional concentration introduces a geographical bias that limits the generalizability of conclusions, particularly in view of the distinctive agro-ecological and infrastructural characteristics that prevail in North and Latin America. Moreover, the review did not identify eligible peer-reviewed studies from Australia or New Zealand, which may reflect either a true research gap or limitations inherent to the database selection and search strategy.

Furthermore, the corpus is heavily skewed toward technical research questions, while institutional, legal, and sociocultural dimensions of AI adoption are markedly underexplored. This imbalance restricts a holistic understanding of the enabling and constraining factors that determine success in specific locales. This review also relies almost entirely on peer-reviewed literature, thereby under-representing the practical insights embedded in grey literature such as policy briefs and implementation reports. Finally, the search strategy is confined to the Scopus database and to English-language publications; studies indexed in other repositories or written in other languages may therefore have been omitted, potentially introducing both thematic and disciplinary bias.

These limitations notwithstanding, the cumulative evidence paints a compelling picture of AI as a catalyst for transformative change in regional sustainability governance. By converting heterogeneous data streams into actionable intelligence, AI lessens the latency between environmental perturbation and managerial response, a capability that is indispensable in the face of accelerating climate risks and mounting resource pressures. Yet the trajectory from proof-of-concept to scaled deployment will depend on a balanced co-evolution of technological innovation, ethical regulation, and inclusive capacity building. Only under such conditions can artificial intelligence fulfil its promise as a cornerstone of adaptive, transparent, and socially equitable sustainability strategies.

5. Conclusions

This scoping review demonstrates that artificial intelligence (AI) is no longer an experimental add-on but an increasingly indispensable instrument for advancing sustainable development at regional and local scales. Across domains as diverse as environmental monitoring, climate-risk assessment, resource optimization, infrastructure management, and social service delivery, AI has begun to reshape decision making by elevating both the spatial precision and the temporal immediacy of available information. These technical gains translate into more adaptive and—in favorable institutional contexts—more inclusive development trajectories. Yet the benefits of AI are mediated by institutional readiness, regulatory clarity, and societal trust; technological sophistication alone is not sufficient.

Taken together, the studies reviewed here reveal a decisive shift from fragmented, post hoc interventions towards integrated, forward-looking governance. Well-designed AI applications widen the planning horizon, increase procedural transparency, and create novel channels for citizen participation in ecosystem stewardship. Their value is most pronounced where conventional governance tools are hampered by scarce data and limited administrative capacity. Nevertheless, the pathway from promising pilot projects to broad-based implementation remains uneven, and it is shaped as much by ethical and institutional considerations as by algorithmic performance. Within the socio-technical-systems frame adopted in the Discussion section, successful scaling will depend on the mutual alignment of technological design, regulatory oversight, and local knowledge systems.

Several gaps identified in the literature deserve systematic attention. First, empirical evidence remains geographically uneven. Regions such as Central and Eastern Europe, parts of North and South America, and many post-Soviet states are still under-represented, despite distinctive land-use regimes and policy environments that could critically affect AI performance. Expanding case-study coverage to these areas would test the transferability of current insights and refine our understanding of context sensitivity. Specific gaps include the absence of longitudinal data on the social impacts of AI-enabled precision agriculture in temperate regions and the lack of explainable-AI benchmarks for watershed management in humid-tropical catchments.

Second, the field would benefit from longitudinal, practice-oriented evaluations that scrutinize how AI systems perform under operational constraints. Issues of scalability, cost-effectiveness, risk resilience, and user acceptance are rarely addressed in depth, yet they determine whether experimental prototypes mature into reliable public or commercial services.

Third, promising opportunities lie in combining AI with complementary paradigms such as citizen science, blockchain-enabled transparency, nature-based solutions augmented by AI, and advanced Earth-system modeling. Such interdisciplinary approaches could yield governance frameworks that remain robust under deep uncertainty while enhancing accountability and public trust.

Equally important is a sustained focus on the social and ethical implications of large-scale AI deployment. Ensuring fairness, safeguarding personal data, and making algorithmic processes explainable are pre-conditions for durable societal endorsement. Research that engages vulnerable groups, measures public perceptions, and designs inclusive digital-literacy programmers will be essential for preventing the emergence of new forms of technological exclusion.

Finally, the development of universal yet locally adaptable metrics for tracing the environmental and social impacts of AI represents a critical frontier. A coherent indicator framework would supply policy makers with the feedback loops necessary to recalibrate sustainability strategies in real time and to monitor long-term progress.

Practical implications follow accordingly. Policy makers should mandate risk-proportionate algorithm audits and support open-data infrastructures that facilitate federated learning. Technology developers need to prioritize energy-efficient, transparently documented model architectures. Ecosystem managers are advised to embed AI deployment within capacity-building programmers and participatory design processes that surface local knowledge and diffuse ownership of outcomes.

In summary, artificial intelligence is poised to evolve from a sophisticated analytical tool into a structural pillar of sustainability governance. Realizing that potential will require a concerted effort to align technological innovation with ethical safeguards, institutional capacity-building, and context-aware policy design. Future research—interdisciplinary by necessity and oriented towards long-term impact—must therefore couple methodological rigor with a commitment to equity and transparency if AI is to fulfil its promise of fostering resilient and just socio-ecological systems.