Does digital transformation matter to improve sugarcane agroindustry? Techniques, challenges and future trends

Sugarcane is a strategic agricultural commodity that plays a pivotal role in the global economy, serving as the primary raw material for sugar production and a wide range of bio-based products, including renewable energy, bioethanol, and industrial materials (Salatein et al., 2024). Globally, production is concentrated in a few major economies such as Brazil, India, and China collectively and contributes more than 60% of total output underscoring the strategic importance and geographic specialization of this crop (Almeida et al., 2022). Despite its critical economic and industrial value, the sugarcane sector faces persistent structural and technological challenges. Many agroindustrial facilities remain dependent on outdated technologies and traditional production practices, resulting in low levels of mechanization, high production costs, and inefficiencies across the value chain (Molin et al., 2024).

These challenges are compounded by additional pressures such as the reduction of arable land, limited research and innovation capacity, and the increasing global demand for sugar and its derivatives (Toharisman and Triantarti, 2016). Consequently, productivity growth has stagnated, posing a significant threat to the industry's long-term competitiveness and sustainability. Addressing these constraints requires a strategic transformation driven by the adoption of digital technologies, data-driven decision-making, and automation. Such advancements have the potential to optimize cultivation, processing, and distribution processes thereby enhancing efficiency, resilience, and value creation across the sugarcane agroindustrial system.

Emerging technologies such as machine learning, internet of things (IoT), and digital twins show strong potential in enhancing sugarcane agroindustry performance, with precision agriculture in Russia increasing productivity by 32% (Beksultanova et al., 2021). In another study, Alemán-Montes et al. (2023) leveraged remote sensing techniques to model crop yield forecasts within sugarcane production systems. Moreover, recent applications of digital twin in the agri-food sector have focused on the development of personalized nutrition strategies (Gkouskou et al., 2020).

This study integrates bibliometric analysis and systematic literature review to examine digital technology adoption in the sugarcane agroindustry, addressing a major gap in current research. Previous studies remain fragmented, focusing mainly on other commodities such as rice and cocoa (Apicella, 2025; Yusuf et al., 2024) or offering broad analyses of the agri-food sector (Romanello and Veglio, 2022), with limited insight into sugarcane systems. Unlike earlier works, this study provides a comprehensive, theory-driven synthesis of digital transformation across the entire sugarcane value chain. By combining bibliometric mapping and systematic content analysis within the TCM–ADO framework, it captures conceptual, methodological, and adoption perspectives to identify technological trends, implementation challenges, and emerging opportunities for sustainable digital transformation, while guiding future research and stakeholder strategies. Accordingly, the study aims to inform stakeholders and addresses the following research questions:

This study adopts an integrated methodological design combining three complementary approaches: bibliometric analysis to quantitatively map research patterns and intellectual structures across the literature, systematic review to provide interpretive depth through rigorous content analysis of eligible studies, and the Theory–Context–Method (TCM) and Antecedents–Decisions–Outcomes (ADO) integrated framework to facilitate comprehensive conceptual synthesis and theoretical generalization. Bibliometric analysis provides structural breadth by identifying dominant themes, geographic distributions, temporal trends, and keyword co-occurrence networks. The systematic review follows the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) 2020 protocol (Page et al., 2021) to ensure transparency and replicability, while the five-step procedure of Wolfswinkel et al. (2013) structures data selection, coding, and synthesis. Building upon these two analytical layers, the integrated TCM–ADO framework is applied to structure a comprehensive conceptual synthesis. Originally proposed as separate frameworks, TCM by Paul et al. (2017) and ADO by Paul and Benito (2018), and subsequently combined by Lim et al. (2021) and Jaziri et al. (2025). The TCM component organizes the analytical dimensions along theoretical foundations (T), contextual settings (C), and methodological approaches (M), while the ADO component maps the drivers of adoption (Antecedents), implementation choices (Decisions), and resulting impacts (Outcomes). This triangulated and iterative approach ensures coherence across methodological layers, enhancing both the empirical rigor and theoretical robustness of the study (Jaziri et al., 2025).

The literature search was conducted from 1 January 2020 through 1 March 2026 to capture the most recent acceleration of digitalization in agriculture and agroindustrial systems. The final database search was executed on 25 March 2026, following the recommended practice of conducting searches no more than three months prior to submission (Rethlefsen et al., 2021). This period reflects the post-COVID-19 surge in automation, IoT, and AI adoption, coinciding with global initiatives such as the UN Food Systems Summit (United Nations, 2021) and the FAO Digital Agriculture Transformation Framework (FAO, 2022). Together, these policy-driven movements signify a pivotal phase in embedding digital innovation within agroindustrial practices, ensuring that this review synthesizes the most current and policy-relevant evidence.

Data were collected from Scopus and Web of Science (WoS), which together represent the two largest multidisciplinary abstract and citation databases for peer-reviewed scientific literature (Mongeon and Paul-Hus, 2016; Pranckutė, 2021). Their combined coverage encompasses over 90% of journals relevant to agricultural engineering, food technology, and agroindustrial systems, making them the most appropriate sources for a bibliometric-systematic review in this domain (Donthu et al., 2021). Scopus and WoS were specifically preferred over alternative databases for the following reasons: (1) PubMed was excluded as its indexing is primarily oriented toward biomedical and clinical sciences, with limited coverage of agroindustrial engineering and supply chain management literature (Falagas et al., 2008), (2) Google Scholar was excluded due to its lack of advanced Boolean search operators, inconsistent metadata quality, and absence of structured export functionality required for reproducible bibliometric analysis (Haddaway et al., 2015), and (3) the use of two complementary databases mitigates single-source indexing bias, as Scopus provides broader geographic coverage while WoS offers stronger impact-factor filtering (Martín-Martín et al., 2021). This dual-database approach is consistent with established protocols in agri-food systematic reviews (Apicella, 2025; Molin et al., 2024) and bibliometric studies in management and engineering (Donthu et al., 2021).

The search employed Boolean logic using the following structured string applied to Title, Abstract, and Keywords fields: TITLE-ABS-KEY ((“Sugarcane” OR “Sugar Cane”) AND (“Industry” OR “Agroindustry” OR “Agro-industry” OR “Value Chain” OR “Supply Chain”) AND (“Machine Learning” OR “Deep Learning” OR “Artificial Intelligence” OR “Internet of Things” OR “IoT” OR “Digital Twin” OR “Digital Transformation” OR “Algorithm” OR “Remote Sensing” OR “Precision Agriculture” OR “Automation” OR “Data-Driven”)).The expanded search string was developed through an iterative scoping process that tested multiple keyword combinations against a set of known relevant papers to maximize both recall and precision. The final string yielded 515 publications, capturing additional relevant studies on AI-driven agri-systems and IoT-based monitoring.

Duplicate records were eliminated using the bibliometrix package (v4.1.4) (Aria and Cuccurullo, 2017) in RStudio (v2023.12.1), which provides a comprehensive suite of tools for bibliometric analysis including data import, deduplication, co-occurrence network generation, and thematic mapping. Following deduplication, a unified dataset of 461 unique records was constructed by merging exports from both databases in CSV format. The bibliometric analysis of these 461 studies captured the overall research landscape, including geographical distribution, temporal trends, thematic evolution, and keyword co-occurrence networks, providing an evidence-based overview of digital technology applications in the sugarcane agroindustry. Keyword co-occurrence network visualization was performed using VOSviewer software (v1.6.20) (van Eck and Waltman, 2010), enabling identification of dominant research clusters and inter-topic linkages across the literature. Thematic mapping and cluster analysis were subsequently performed using the Bibliometrix graphical interface (Biblioshiny), with the Leiden Eigenvalue algorithm applied for thematic quadrant classification to distinguish motor, niche, emerging, and basic themes within the intellectual structure of the field (Öztürk et al., 2024).

Building on the bibliometric mapping described above, the systematic review phase further refined the dataset through the four-stage PRISMA screening protocol (Page et al., 2021) and the five-step framework of Wolfswinkel et al. (2013), as introduced in the methodological design above. In the first step, the research scope and questions were established around digital transformation in the sugarcane agroindustry, encompassing upstream cultivation, processing, and downstream supply chain operations, with four research questions formulated to guide subsequent screening and synthesis decisions. In the second step, the search string described above was applied across Scopus and Web of Science, yielding 436 and 79 records respectively (n = 515 total). Following deduplication as described above, 54 duplicate records were removed, producing a unified dataset of 461 unique records that formed the basis of the bibliometric mapping analysis.

In the third step, records were screened by title and abstract against predefined inclusion and exclusion criteria, retaining only empirical, English-language studies focusing on digital technologies in the sugarcane agroindustry and available in full open access, while excluding review articles and conceptually vague or non-empirical works, at which point 192 records were excluded, leaving 269 records for full-text assessment. In the fourth step, full-text screening was applied to the remaining 269 records, where 159 records were excluded due to inaccessibility (closed access), non-English language, or lack of direct relevance to digital transformation in the sugarcane agroindustry, and a further 16 records were excluded for falling outside the defined study period of 1 January 2020 to 25 March 2026, resulting in 94 articles eligible for qualitative synthesis. In the fifth and final step, the 94 eligible studies underwent qualitative synthesis using a structured coding protocol encompassing bibliographic information, research objectives, applied technologies, implementation levels, methods, and key results. Data extraction was performed independently by two coders to minimize selection bias, with inter-rater reliability assessed using Cohen's Kappa coefficient (κ = 0.84), indicating almost perfect agreement (McHugh, 2012), and all disagreements resolved through discussion and consensus. Quantitative and qualitative syntheses were conducted using VOSviewer and the bibliometrix package to producing comprehensive thematic insights across the theoretical, contextual, methodological, and implementation dimensions of agroindustrial digitalization.

Building upon these results, the integrated framework combines bibliometric mapping of 461 studies for structural breadth, systematic review of 94 eligible studies for interpretive depth, and the TCM–ADO framework for conceptual synthesis. Together, these approaches establish a coherent, multi-layered methodology where bibliometric analysis identifies macro-level research patterns and intellectual clusters, the systematic review deepens thematic interpretation through rigorous content analysis, and the TCM–ADO organizes theoretical, contextual, and methodological dimensions to guide future research trajectories. This integrative design enhances the study's robustness, transparency, and interpretability by bridging quantitative evidence, qualitative insights, and conceptual theorization, addressing a methodological gap identified in prior agroindustrial digital transformation studies that have relied predominantly on single-method approaches (Apicella, 2025; Romanello and Veglio, 2022). The PRISMA flow diagram illustrating this methodological workflow is presented in Figure 1, and the full list of included articles along with their summary characteristics is provided in Supplementary Material (Table S1).

The bibliometric analysis of 461 unique records refined to 94 eligible studies through systematic PRISMA screening, reveals a dynamic and expanding research landscape in digital transformation within the sugarcane agroindustry. As summarized in Table 1, the dataset spans the active publication period of 2020–2026, contributed by 417 authors across 77 journal sources, accumulating 696 total citations with an average of 7.404 citations per document. The annual growth rate of 24.57% confirms a rapid and sustained acceleration in scholarly attention toward this domain, reflecting the field's strong momentum particularly in the post-COVID-19 era of agroindustrial digitalization. The international co-authorship rate of 28.72% and co-authors per document of 4.44 indicate that research in this domain is predominantly conducted through collaborative and cross-border partnerships, with zero single-authored publications among the 94 included studies. The breadth of intellectual engagement is further evidenced by 373 author keywords and 7,370 references across the corpus, with a document average age of 3.06 years confirming the recency and currency of the included literature. These bibliometric characteristics collectively establish the intellectual context and research momentum within which the thematic and technological findings presented in the following sub-sections are situated. The consistent upward trajectory in publication output further underscores the growing academic and industrial attention toward digitalization as a key enabler of efficiency, sustainability, and competitiveness in the sugarcane sector. To further elucidate the intellectual structure of this field, a keyword co-occurrence analysis using VOSviewer was employed to map relationships among dominant terms in the literature (Figure 2). Following Zhou et al. (2022), such visualization facilitates the identification of prevailing research themes and emerging knowledge domains, as presented in the following paragraphs.

The keyword co-occurrence analysis reveals that the most frequent terms in the literature are “sugar cane” (occurrences = 48, total link strength = 293, representing the merged forms of “sugar cane” and “sugarcane”), “machine learning” (occurrences = 37, total link strength = 251, representing the merged forms of “machine learning” and “machine-learning”), “deep learning” (occurrences = 15, total link strength = 115), “crops” (occurrences = 14, total link strength = 119), and “remote sensing” (occurrences = 17, total link strength = 93), collectively reflecting the dominant technological and sectoral focus of recent investigations. Supporting technologies such as “convolutional neural networks” (occurrences = 8, total link strength = 68), “precision agriculture” (occurrences = 9, total link strength = 50), “unmanned aerial vehicles” (occurrences = 6, total link strength = 59), and “computer vision” (occurrences = 6, total link strength = 33) further underscore the field's strong orientation toward AI-driven image analysis and remote monitoring applications. This pattern is consistent with findings in adjacent agri-food bibliometric studies, where machine learning and remote sensing technologies consistently emerge as dominant intellectual clusters (Romanello and Veglio, 2022; Yusuf et al., 2024), reflecting the field's broader trajectory toward precision and data-driven agriculture (Bacco et al., 2019).

Building on this keyword structure, the co-occurrence network further reveals three distinct thematic clusters. The red cluster emphasizes algorithmic refinement for precise disease detection; the green cluster centers on intelligent decision-making across production and distribution processes; and the blue cluster focuses on remote sensing for yield forecasting and land monitoring, as illustrated in Figure 2. When examined through the TCM–ADO framework, these clusters predominantly occupy the technological and decision-making dimensions, revealing an analytical bias that warrants greater integration of contextual, organizational, and institutional perspectives to achieve a more holistic and sustainable digital transformation pathway in the sugarcane sector.

Addressing research question Q1, this section presents a structured overview of digital transformation research in the sugarcane agroindustry across three analytical dimensions: the distribution of research focus along the value chain from upstream cultivation to downstream operations, the profile of dominant technological approaches applied across the literature, and the geographic concentration of research activities among major sugarcane-producing nations.

Findings from the systematic literature review indicate a markedly uneven distribution of research focus across the sugarcane value chain. Of the 94 eligible studies, 69.47% are concentrated on upstream activities encompassing cultivation, field management, soil monitoring, and crop health assessment, 29.47% address midstream processing operations including milling, extraction, and manufacturing quality control, while only 1.05% address downstream post-harvest operations including logistics, distribution, and supply chain management. This three-level distribution reveals that digital transformation research in the sugarcane agroindustry remains overwhelmingly field-centric, with midstream processing receiving moderate research attention and downstream operations remaining virtually unexplored as a domain of digital innovation.

The upstream concentration is not incidental but reflects the strategic and economic primacy of the cultivation stage. From an industrial standpoint, inefficiencies or crop losses at the upstream level propagate irreversibly throughout the entire value chain, as a degraded harvest cannot be compensated at the milling or distribution stage. Sugarcane cultivation is highly susceptible to abiotic and biotic stresses including water scarcity, climatic variability, and pest infestations, all of which directly affect biomass accumulation and sucrose content (Mehdi et al., 2024). These vulnerabilities create strong institutional incentives for upstream digitalization, as precision technologies such as remote sensing, soil sensors, and UAVs enable real-time monitoring, adaptive management, and risk mitigation at the point of highest value creation (Molin et al., 2024). From the farmers' perspective, upstream digitalization delivers measurable returns including higher crop yields, reduced input costs, and a comparatively rapid return on investment relative to downstream infrastructure projects. Precision agriculture in major producing nations has demonstrated productivity gains of up to 32%, further reinforcing the economic case for field-level digital investment (Beksultanova et al., 2021).

The 29.47% of studies addressing midstream processing operations represent a domain of critical industrial significance, encompassing key manufacturing stages including milling, extraction, clarification, evaporation, crystallization, and centrifugation, all of which demand high precision, real-time monitoring, and adaptive process control to maintain product quality and operational stability. Measurable progress has been demonstrated in this domain, as Kusoncum et al. (2022) optimized machine allocation in milling through metaheuristic algorithms, while Meng et al. (2021) developed predictive monitoring systems for juice color and purity using deep KELM and multi-layer ELM models, collectively confirming that AI-driven process intelligence can substantially enhance manufacturing consistency and resource efficiency across sugarcane industrial operations.

In stark contrast, downstream operations represent the most critically underserved level of the sugarcane digital transformation landscape, with only a single study identified in the systematic review, accounting for 1.05% of the total corpus. This study, examining an adaptive fuzzy multi-criteria model for sustainability assessment of the sugarcane agroindustry supply chain, demonstrates that computational decision-support frameworks can be effectively applied to evaluate and optimize supply chain sustainability across multiple performance dimensions (Yani et al., 2022). The near-complete absence of downstream digitalization research is particularly striking given the strategic importance of logistics, procurement, and distribution efficiency in determining the final economic value and market competitiveness of sugarcane products. This finding represents one of the most significant structural gaps identified in the current literature, as end-to-end value chain integration cannot be achieved without commensurate research investment across all three operational levels.

The overall three-level distribution, when examined through the TCM–ADO framework, reveals a structurally fragmented digitalization landscape. At the antecedent level, operational necessity and demonstrable ROI at the upstream stage drive disproportionate research investment, while institutional barriers, high infrastructure costs, and weaker financial incentives constrain midstream and downstream adoption. At the decision level, most digital interventions remain reactive and domain-specific rather than adaptive and integrated across the value chain, with the singular downstream study signaling an emergent but nascent recognition of supply chain-level digitalization as a research priority. At the outcome level, empirical evidence of end-to-end efficiency gains, sustainability improvements, and supply chain resilience remains virtually absent, underscoring the urgent need for future research to prioritize holistic, cross-stage digitalization frameworks that bridge upstream precision agriculture with midstream process intelligence and downstream supply chain optimization.

This study identifies seven dominant technological categories underpinning digital transformation research in the sugarcane agroindustry, namely machine learning, deep learning, spatial analysis, metaheuristics, Internet of Things, reinforcement learning, and transfer learning. The distribution across 94 eligible studies reveals a clear hierarchical pattern: machine learning accounts for the largest share (33.33%), followed by deep learning (21.14%), spatial analysis (21.14%), metaheuristics (14.63%), IoT applications (8.13%), and reinforcement learning and transfer learning each representing an emerging presence at 0.81% respectively. It is important to note that the total number of technological instances across the corpus (n = 123) exceeds the number of eligible studies (n = 94), indicating that a substantial proportion of included studies employ multiple technological approaches in combination rather than relying on a single method. This pattern of methodological integration reflects the growing recognition among researchers that the operational complexity of sugarcane agroindustrial systems necessitates hybrid computational architectures that combine the complementary strengths of different technological paradigms. This profile indicates that the field is predominantly characterized by algorithmic and predictive orientations, with data-driven modeling constituting the primary mode of technological inquiry across the sugarcane value chain, while the emergence of reinforcement learning and transfer learning signals an incipient diversification toward more adaptive and knowledge-transferable computational paradigms.

Machine learning emerges as the most extensively adopted approach, owing to its proven capacity to capture nonlinear, high-dimensional relationships inherent in agroindustrial datasets encompassing soil properties, climatic variables, crop physiology, and process parameters (Khatoon et al., 2024). Its methodological versatility across regression, classification, and anomaly detection tasks renders it applicable at multiple stages of the value chain, from upstream cultivation monitoring to midstream quality assurance. Deep learning and spatial analysis, each representing 21.14% of studies, constitute the second tier of technological adoption. Deep learning extends machine learning capability through hierarchical feature extraction from raw image and sensor inputs via multilayer neural architectures, enabling autonomous pattern recognition at a level of granularity that conventional approaches cannot achieve (Thite et al., 2024), while spatial analysis underpins field-scale environmental monitoring through the integration of satellite imagery, geographic information systems, and remote biophysical modeling (Hochmair et al., 2023). The equal representation of these two categories reflects a complementary dynamic in the literature, where deep learning addresses within-field diagnostic precision and spatial analysis addresses landscape-scale monitoring and yield estimation.

Metaheuristic optimization, representing 14.63% of studies, addresses the combinatorial complexity of scheduling, resource allocation, and multimodal logistics problems that are analytically intractable through exact methods. Algorithms such as genetic algorithms, particle swarm optimization, and non-dominated sorting genetic algorithms navigate large, multi-objective solution spaces to derive near-optimal operational decisions within practical computational constraints (Filip et al., 2020). IoT applications, representing 8.13% of studies, fulfill a strategically critical foundational role as the real-time data acquisition infrastructure upon which higher-order analytical technologies depend, enabling continuous field monitoring, process telemetry, and sensor-driven decision support at the operational level.

Two emerging technological categories warrant particular scholarly attention despite their current marginal representation. Reinforcement learning, identified in a single study, was applied to optimize sugarcane bale logistics operations through dynamic multi-fleet management and multi-period scheduling under machine breakdown constraints, demonstrating the technology's distinctive capacity to learn adaptive decision policies through environmental interaction rather than static training datasets. This application represents a paradigmatic departure from conventional supervised learning approaches, as reinforcement learning agents continuously refine operational strategies in response to real-time system feedback, making them particularly well-suited to the dynamic, uncertainty-laden scheduling environments characteristic of sugarcane harvest and transportation logistics (Pitakaso et al., 2025). Transfer learning, also identified in a single study, was applied to the detection of sugarcane leaf diseases through a customized weighted ensemble of modified pre-trained deep learning models, leveraging knowledge representations acquired from large-scale image datasets to overcome the labeled data scarcity that constrains conventional deep learning model training in agricultural contexts (Hu et al., 2024). The emergence of both categories signals an important directional shift in the field toward more data-efficient, adaptive, and generalizable computational approaches that address fundamental limitations of the dominant supervised learning paradigm.

Collectively, the technological profile of the included studies reveals three structural characteristics of the current research landscape. First, there is a pronounced concentration in upstream, AI-driven applications relative to downstream process and logistics domains, consistent with the value chain distribution identified. Second, the IoT adoption rate, while improved at 8.13%, continues to signal a persistent infrastructural deficit that constrains the real-world deployment of AI systems at scale, as effective machine learning and deep learning implementations require continuous, high-quality data streams that pervasive IoT networks are designed to provide. Third, the emergence of reinforcement learning and transfer learning, while currently marginal in volume, represents a qualitatively significant frontier that addresses the adaptability and data efficiency limitations of the dominant technological paradigm. The convergence of these seven technological categories within increasingly hybrid, multi-technology architectures reflects the field's trajectory toward integrated digital intelligence, where sensing, learning, optimization, and adaptive decision-making capabilities are orchestrated in concert to address the multidimensional complexity of sugarcane agroindustrial systems. Examined through the TCM–ADO framework (Jaziri et al., 2025), this technological distribution reflects a predominance of methodological exploration over theoretical and contextual development, with digital tools oriented primarily toward operational decision-making rather than systemic outcome-driven transformation, highlighting a critical gap between technological capability and institutional readiness that future research must urgently address.

The geographic distribution of digital transformation research in the sugarcane agroindustry exhibits a pronounced concentration among the world's leading producing nations. Analysis of the 94 eligible studies identifies 22 contributing nations, with India contributing the largest share of publications (n = 26, 27.66%), followed by China (n = 16, 17.02%), Brazil (n = 13, 13.83%), Thailand (n = 7, 7.45%), and Australia (n = 6, 6.38%), collectively accounting for 72.34% of all included studies. Beyond this dominant quintet, a secondary tier of contributing nations has emerged, including Iran (n = 5, 5.32%), Indonesia (n = 3, 3.19%), Sri Lanka (n = 2, 2.13%), Colombia (n = 2, 2.13%), Pakistan (n = 2, 2.13%), and the United States (n = 2, 2.13%), alongside single-study contributions from Cuba, South Africa, Costa Rica, Malaysia, Philippines, Ukraine, United Kingdom, Ethiopia, Egypt, and Uganda. This expanded geographic footprint represents a meaningful diversification relative to prior bibliometric analyses of the field, suggesting that awareness of digitalization imperatives in the sugarcane agroindustry is gradually diffusing beyond the traditional research centers of the Asia-Pacific and Latin American production regions.

The concentration among the five leading nations is not coincidental but reflects their dominant position in global sugarcane production. According to FAO (2021), Brazil leads global output at 715.6 million tons, followed by India (405.4 million), China (107.3 million), Thailand (66.2 million), and Australia (31.1 million), together representing over 80% of total world production. The positive correlation between production volume and research output observed across these nations suggests that institutional research investment in agroindustrial digitalization is driven by economic necessity, as sugarcane contributes 2–3% of national GDP in each of these economies (Solomon, 2016), creating strong incentives for productivity-enhancing technological adoption to sustain competitive advantage in global commodity markets.

A more granular examination of country-specific research contexts reveals, however, that digitalization imperatives, while universally present, are shaped by distinct agronomic, economic, and infrastructural conditions that differ substantially across geographies. India, despite leading in publication volume, confronts a complex constellation of upstream challenges including chronic water scarcity, government price controls, market volatility, and stagnant yield growth, which collectively constrain the agroindustry's capacity to finance and sustain large-scale digital transformation initiatives. Brazil, the world's largest producer, faces structurally similar upstream vulnerabilities compounded by land access inequities and the dual pressure of serving both domestic sugar markets and global bioethanol demand, necessitating precision agriculture solutions that optimize resource use under variable climatic and market conditions (Almeida et al., 2022). Thailand's primary digitalization imperative is concentrated at the supply chain level, where harvest scheduling inefficiencies between field operations and mill processing create significant productivity losses, a challenge that has been systematically addressed through metaheuristic optimization approaches generating near-optimal harvesting schedules under multi-objective constraints (Kusoncum et al., 2021). China and Australia present a distinct digital transformation profile driven primarily by disease management imperatives, as smut disease infestations have caused documented yield losses of 14.89% in China and between 26 and 62% in Australia (Jiangjiang et al., 2020; Magarey et al., 2010), creating urgent demand for AI-driven diagnostic systems capable of early detection at field scale, with both nations pioneering digital innovations in processing domains including neural network-based yield prediction (Han et al., 2022) and machine learning-based clarification process optimization (Meng et al., 2021).

The emergence of Iran (n = 5, 5.32%) as a notable contributor warrants specific scholarly attention, as it represents the most significant non-traditional research presence in the dataset. Iranian contributions are concentrated in precision irrigation and evapotranspiration modeling, reflecting the country's acute water resource constraints and the strategic imperative to optimize agricultural water use efficiency under arid and semi-arid climatic conditions. The presence of Indonesia (n = 3, 3.19%) is equally noteworthy given its status as a major sugarcane producer facing severe digitalization challenges related to fragmented smallholder land tenure, limited rural connectivity, and constrained institutional capacity for technology adoption (Toharisman and Triantarti, 2016). Single-study contributions from geographically diverse nations including Uganda, Ethiopia, Egypt, Costa Rica, and Ukraine indicate that awareness of digitalization imperatives in sugarcane systems is beginning to permeate regions where sugarcane production is economically significant but research infrastructure remains nascent.

Despite this expanded geographic footprint, a critical systemic gap persists in the current literature. Major producing nations including several African and Central American economies remain severely underrepresented, and the collective weight of the five leading nations (72.34%) continues to reflect a research landscape disproportionately shaped by high-resource, large-scale production contexts. This geographic bias constitutes a significant limitation of the existing evidence base, as the technological solutions developed and validated in these contexts may not be directly transferable to smallholder-dominated systems characterized by fragmented land tenure, limited digital infrastructure, and constrained access to capital and technical expertise. Maintaining production sustainability across all major producing nations is increasingly urgent given mounting threats from climate variability, pest dynamics, and supply chain disruption (Waykar and Yambal, 2025). Examined through the TCM–ADO framework, the geographic distribution of research reveals a contextual dimension characterized by persistent scale heterogeneity and institutional diversity that current digitalization frameworks have yet to adequately address, underscoring the need for future research to engage comparatively with underrepresented producing nations and smallholder agricultural systems to develop contextually appropriate, inclusive, and scalable digital transformation pathways.

Building on the technological profile identified, this section systematically examines how digital and computational tools are deployed across specific operational domains in the sugarcane agroindustry, addressing research question Q2. The systematic literature review indicates that digital transformation spans six operational domains, with crop cultivation and field optimization emerging as the most intensively explored area (31.91%), followed by process optimization (27.66%), disease detection (15.96%), harvesting (13.83%), supply chain management (7.45%), and other emerging applications (3.19%). This distribution reflects a predominant research emphasis on upstream field-level precision management, with cultivation and process optimization collectively accounting for 59.57% of all identified applications, confirming that agronomic performance at the field and manufacturing levels constitutes the primary focus of digital innovation in the sugarcane agroindustry. This pattern is consistent with the value chain distribution, where upstream activities represented 69.47% of included studies, underscoring the structural alignment between research focus and operational domain emphasis across the corpus.

Crop cultivation and precision field management constitutes the most extensively digitalized domain, reflecting the upstream sector's critical and foundational influence on overall value chain performance. Zhao et al. (2023) predicted soil organic carbon using a hybrid multiple linear regression and digital soil mapping approach, reducing assessment costs by 36% compared to conventional methods, thereby enabling cost-effective precision fertilization at field scale. Veysi et al. (2023) enhanced irrigation decision-making by estimating evapotranspiration through Landsat 8-derived land surface temperature using a single-channel algorithm, providing actionable water management guidance without requiring expensive ground instrumentation. Tanut et al. (2021) employed an ensemble of machine learning models including k-nearest neighbors, random forest, decision trees, and multilayer perceptron to classify unfit sugarcane prior to harvest, with the random forest algorithm achieving 98.69% accuracy. Gopikrishnan et al. (2022) integrated IoT sensors with machine learning to monitor soil salinity and optimize irrigation scheduling in real time, demonstrating that sensor-driven precision management can simultaneously improve production efficiency and ensure sustainable water use under saline soil conditions. Som-ard et al. (2024) further demonstrated the effectiveness of combining UAV and satellite imagery with machine learning for multi-scale yield estimation, confirming the scalability of spatial analysis approaches across heterogeneous field environments. Collectively, these studies establish that digital precision agriculture delivers measurable upstream performance gains while progressively reducing dependence on resource-intensive conventional management practices.

Process optimization represents the second most intensively digitalized domain, driven by the precision demands of key manufacturing stages including milling, extraction, clarification, evaporation, crystallization, and centrifugation, all of which require real-time monitoring and adaptive control to maintain product quality and operational stability. Kusoncum et al. (2022) optimized machine allocation in the milling process through metaheuristic algorithms, achieving consistent throughput quality across variable input conditions. Meng et al. (2021) developed predictive monitoring systems for juice color and purity using deep kernel extreme learning machine and multi-layer ELM models, enabling non-invasive quality assessment at high processing speeds. Yang et al. (2023) further advanced process intelligence by developing a hybrid convolutional neural network and KELM model for juice quality prediction, achieving near-perfect accuracy under unsupervised training conditions, while Boskabadi et al. (2024) demonstrated the applicability of machine learning-based soft sensors in replacing costly laboratory measurements with continuous in-line crystallization predictions. Collectively, these studies demonstrate that AI-driven process optimization can substantially reduce operational variability, minimize energy consumption, and enhance product consistency across the sugarcane manufacturing chain, confirming the viability of data-driven process intelligence as a scalable alternative to conventional laboratory-based quality control.

Disease detection constitutes a domain of critical strategic importance given that infections caused by bacterial or viral pathogens can reduce productivity by up to 40% and lower juice quality during milling by approximately 30% (Kong et al., 2025). To mitigate these losses, recent studies have employed advanced deep learning architectures for early and accurate diagnosis at field scale. Vivekreddy et al. (2024) achieved state-of-the-art accuracy in detecting red rot and red rust using DenseNet201 (97.94%) and ResNet50 (99.7%), while Kumar et al. (2023) developed a CNN-based detection model using VGG16 and VGG19 architectures, with VGG19 yielding the best accuracy of 92%. Mishika et al. (2024) combined IoT-based imaging with CNN deep learning to detect smut disease on sugarcane leaves with 95% accuracy, demonstrating the operational viability of integrated sensing and inference pipelines for field-deployable diagnostics. Collectively, these findings demonstrate that AI-driven image recognition has achieved sufficient technical maturity to function as a viable component of mainstream agronomic decision support, with deep learning models consistently exceeding 90% classification accuracy across diverse disease categories, representing a transformative departure from labor-intensive, subjective visual inspection methods (Dolatabadian et al., 2025; Kamilaris and Prenafeta-Boldú, 2018). The effectiveness of AI diagnostics extends beyond classification accuracy along three reinforcing dimensions: interpretability, enhanced through explainability mechanisms such as gradient-weighted class activation mapping that build stakeholder trust in model outputs (Biswas et al., 2024); data efficiency, improved through transfer learning approaches that overcome labeled dataset scarcity in agricultural contexts (Daphal and Koli, 2023); and deployment scalability, advanced through federated learning architectures that enable collaborative model training without centralized data sharing, addressing connectivity constraints in rural producing regions (Mamba Kabala et al., 2023). Notwithstanding these advances, generalizability to heterogeneous real-world field conditions and the development of edge-deployable architectures suited to smallholder environments remain priorities for future research.

Yield prediction and harvest management has emerged as an equally dynamic domain, with digital transformation enabling a fundamental shift from reactive, experience-based scheduling toward proactive, data-driven operational planning. Galphade et al. (2022) developed a machine learning model integrating climatic, soil, and vegetation index data including the normalized difference vegetation index, achieving 91.5% accuracy with a decision tree algorithm for pre-harvest yield forecasting. Krishnan et al. (2024) advanced this capability by integrating IoT technologies with deep learning, where a temporal-based Swin Transformer Network optimized via the Nutcracker Optimizer Algorithm achieved 94.4% accuracy, demonstrating the potential of hybrid intelligent systems to capture temporal dynamics in sugarcane growth patterns for enhanced predictive precision. Jarumaneeroj et al. (2021) extended the planning horizon beyond individual field estimation by designing a multi-objective supply chain system that simultaneously maximized output, improved schedule equity among growers, and enhanced supply chain efficiency through a metaheuristic approach using non-dominated sorting genetic algorithm III, establishing a systems-level framework for integrated harvest and logistics planning. Collectively, these studies demonstrate that digital yield prediction and harvest planning systems can substantially improve operational efficiency, grower equity, and supply chain coordination, establishing data-driven scheduling as a critical enabler of agroindustrial competitiveness in the sugarcane sector.

Supply chain management and transportation optimization, while collectively representing 7.45% of identified applications, constitute areas of high strategic leverage given their potential to unlock systemic efficiency gains across the entire value chain. Kurade et al. (2024) employed a fuzzy non-dominated sorting genetic algorithm for multi-objective route optimization, evaluating multiple transportation scenarios involving carts, trucks, and tractors to identify cost-efficient, operationally feasible delivery configurations under real-world constraints. Parkhan et al. (2023) demonstrated that applying a vehicle routing problem approach to logistics operations can reduce distribution costs by up to 45.02%, underscoring the substantial economic returns achievable through algorithmic supply chain optimization even at modest implementation scales. The remaining 3.19% of applications classified under other emerging domains reflects early-stage exploratory research that does not yet constitute a statistically dominant category but signals the gradual broadening of digital transformation inquiry beyond the established operational core. The comparatively limited research attention directed toward supply chain and downstream domains, relative to their economic significance, represents a structural gap in the current literature that future investigations should urgently prioritize, particularly given the growing imperative for end-to-end digital integration across the sugarcane value chain.

Digital transformation in the sugarcane agroindustry presents substantial benefits and opportunities across operational, economic, and sustainability dimensions. However, its implementation also faces significant challenges related to technological readiness, infrastructure, and stakeholder adaptation. Addressing research question Q3, this section examines both the enabling and constraining factors that influence the success of digital transformation, as well as the tangible benefits realized when these initiatives are effectively implemented, offering a balanced understanding of opportunities and obstacles within the sector. From the perspective of the TCM–ADO framework, these issues represent critical antecedents that shape digital adoption decisions, while the resulting efficiency, traceability, and competitiveness gains correspond to outcome dimensions that have yet to be systematically assessed.

Operational and technical challenges in sugarcane digital transformation stem from readiness levels and the technological framework employed. A key issue is the lack of system integration across stakeholders from small farmers to processing units which creates fragmented workflows and data silos (Hildbrand and Bodhanya, 2014). According to Gajdić et al. (2022) and Assis et al. (2023) collaboration and trust have a significant influence on the food and agricultural commodity supply chain, especially sugarcane. End-to-end integration demands a robust infrastructure and coordinates system governance. Data interoperability is another barrier, as variability in format, quality, and volume from diverse sources sensors, satellites, and manual records can slow progress (Molin et al., 2024). Connectivity limitations in remote plantations further hinder real-time communication, data exchange, and monitoring (Bolfe et al., 2020). Overcoming these obstacles requires integrated strategies that combine infrastructure investment, standardization, and improved digital literacy to ensure effective, scalable, and sustainable technology adoption.

Implementing digital transformation in the sugarcane agroindustry is inherently complex, requiring careful consideration of technology procurement, adaptability, and stakeholder readiness. High initial capital expenditure remains a major barrier (Carrer et al., 2022). Adoption rates vary by scale: large-scale landowners, supported by higher education and technical assistance, tend to adopt digital technologies more readily due to clear economic benefits, whereas smallholders despite comparable educational levels often lack technical support and hesitate to invest because of cost, uncertainty, and limited access to credit. The complexity of digital tools presents another challenge, compounded by a persistent digital skills gap among rural workforces that constrains effective implementation and long-term maintenance (Bacco et al., 2019). Furthermore, systemic pressures from large-scale retailers, who prioritize cost efficiency over transparency and traceability, often discourage upstream actors from investing in traceable digital systems (Romeo et al., 2025). Digital transformation thus requires not only technological investment but also a cultural and institutional shift from historical, intuition-driven decision-making toward data and AI-based approaches (Biswas et al., 2024).

Digital transformation allows the sugarcane industry to operate more agilely in dynamic global markets. Technologies such as IoT, machine learning, and integrated supply chain platforms enhance quality control, reduce costs, and meet traceability requirements for export markets. These capabilities support compliance with standards such as ISO and HACCP, expanding market access and building consumer trust. Enhanced transparency also improved the sector's position in high-value markets. Ciasullo et al. (2025), noted that digital innovation boosts competitiveness through faster, data-driven decisions and attracts investment via greater efficiency. Additionally, Verhoef et al. (2021) add that effective digital transformation modernizes traditional agroindustries, fostering resilience, sustainability, and long-term value creation by reshaping business models, workflows, and stakeholder engagement across the sugarcane value chain.

Efficiency is central to creating digital value in the sugarcane agroindustry, where resource optimization is critical. Integrating IoT sensors, precision irrigation using real-time evapotranspiration data, automated milling controls, and AI-based decision support systems shifts operations from reactive to predictive, reducing uncertainty and minimizing losses. Parkhan et al. (2023) reported that applying the VRP approach to logistics can cut distribution costs by up to 45.02%. Achieving such efficiency requires not only technology but also organizational transformation aligning digital workflows with human skills, institutional readiness, and adaptive governance. As Rasputina (2022) noted, efficiency arises from integrated system architectures that enable real-time monitoring, transparency, and agile response, positioning digital transformation as a holistic reconfiguration of production and management models.

This subsection addresses research question Q4, examining the evolution of themes and conceptual structures guiding future research on digital transformation in the sugarcane agroindustry. Beyond keyword frequency, the analysis explores thematic maturity and centrality using the Leiden Eigenvalue algorithm (Öztürk et al., 2024). The resulting diagram categorizes the intellectual landscape into four quadrants (motor, niche, emerging/declining, and basic themes) that illustrating varying levels of conceptual development and integration within the broader research ecosystem of digital transformation. Figure 3 presents the thematic map generated from this analysis, providing a visual depiction of the positioning and interrelations among these thematic clusters.

• Quadrant 1 encompasses research domains with both high centrality and density, signifying conceptually mature and influential areas that drive the field forward. Central themes such as deep learning, remote sensing, and estimation models dominate this quadrant, representing the technological nucleus of the digital transformation discourse. These topics underpin predictive systems for crop yield estimation, disease detection, and stress monitoring (Das et al., 2023). The integration of remote monitoring, fuzzy inference, and machine learning has fostered non-invasive, cost-effective, and data-driven management frameworks. Collectively, these developments embody a convergence between artificial intelligence and precision agriculture, establishing a robust foundation for sustainable intensification and smart decision-making in sugarcane production systems.

• Quadrant 2 comprises specialized but relatively isolated areas such as supply chain management, decision-making systems, and genetic algorithms. Although theoretically grounded, these themes exhibit lower connectivity to mainstream digital transformation research. Strengthening their linkage with the dominant motor themes could accelerate the development of adaptive and real-time decision-support mechanisms for optimizing harvesting, processing, and logistics operations. For instance, the integration of optimization algorithms and data analytics could enhance operational efficiency, reduce energy consumption, and promote sustainability within sugar mills (Boskabadi et al., 2024).

• Quadrant 3 identifies research areas with low density yet considerable potential for renewal. Topics such as harvesting, decision support, and satellite imagery demonstrate an ongoing shift toward predictive and context-aware agricultural systems. The increasing adoption of satellite and UAV-based remote sensing enables the timely monitoring of crop health, evapotranspiration, and environmental stress (Daphal and Koli, 2023). These emerging applications point to future research trajectories focusing on climate-smart decision-support systems and resilience-oriented digital ecosystems.

• Quadrant 4 represents foundational yet less conceptually developed areas, including the sugar industry, sugarcane, and machine learning. Although these serve as entry points to the broader digitalization discourse, they often remain descriptive and fragmented. Integrating them with higher-order digital frameworks such as AI-driven process control, IoT-enabled traceability, and blockchain-based transparency could significantly enhance their theoretical and practical relevance to adaptive industrial transformation.

To bridge the thematic findings with broader conceptual and methodological perspectives, the results were synthesized through the TCM–ADO. This integration unifies quantitative insights from the bibliometric analysis and qualitative depth from the systematic review into a coherent analytical structure, illustrating how theoretical foundations, contextual factors, methodological approaches, and implementation outcomes interact in the digital transformation of the sugarcane agroindustry. Table 2 presents this integrative mapping across the TCM–ADO dimensions, linking empirical evidence to theoretical, contextual, and methodological implications. Building on this synthesis, the TCM–ADO is further employed to outline future research trajectories that extend beyond current technological and analytical boundaries. By operationalizing each domain, the framework supports the formulation of a structured research agenda. Table 3 summarizes these proposed directions, highlighting strategic opportunities for advancing scholarly inquiry and industrial innovation in the digital transformation of the sugarcane sector.

This review acknowledges three primary limitations that should be considered when interpreting its findings. First, reliance on Scopus and Web of Science, while ensuring comprehensive coverage of high-impact peer-reviewed literature, may exclude relevant gray literature, regionally indexed journals, and non-English publications, potentially introducing publication bias toward well-resourced and institutionally supported research contexts. Second, the absence of systematic cross-country comparative analysis limits the contextual generalizability of findings, particularly with respect to smallholder-dominated producing nations including Indonesia, Pakistan, Colombia, and several African economies that remain underrepresented in the current corpus, and whose digitalization challenges may differ substantially from those documented in high-resource production contexts. Third, the study period of January 2020 to March 2026, while capturing the most recent acceleration of agroindustrial digitalization, may not fully reflect longer-term adoption trajectories or pre-COVID-19 baseline conditions that provide important comparative context for understanding the pace and drivers of digital transformation in the sugarcane sector. Future studies should incorporate additional databases, employ systematic cross-country comparative designs, and extend temporal coverage to address these gaps and enhance the breadth and generalizability of evidence on which digitalization strategies and policies are formulated.

This study integrates bibliometric mapping, systematic review, and the TCM–ADO framework to examine digital transformation in the sugarcane agroindustry, providing empirical and conceptual insights across technological, contextual, and methodological dimensions. Findings reveal that machine learning, deep learning, and remote sensing dominate upstream applications supporting cultivation optimization, yield estimation, and disease detection, while processing and downstream adoption remains constrained by outdated infrastructure, weak interoperability, high investment costs, and low digital literacy. These barriers reflect antecedent–decision–outcome gaps identified in the TCM–ADO mapping, emphasizing that digital transformation must be approached as a socio-technical process requiring inclusive and adaptive strategies across stakeholders.

The practical implications extend across multiple time horizons. In the short term (1–3 years), decision-makers should prioritize interoperable IoT infrastructure, standardized data protocols, and digital literacy programs targeting smallholder farmers and field operators. In the medium term (3–7 years), agroindustries should invest in AI-powered adaptive decision-support systems, digital twin technologies, and blockchain-enabled supply chain visibility, supported by targeted government incentive schemes that promote inclusive adoption across large estates and smallholder systems alike. Beyond sugarcane, the frameworks and lessons identified here offer transferable insights for other labor-intensive tropical agroindustries including palm oil, cocoa, and rubber.

Failure to accelerate digital transformation carries compounding risks. Continued reliance on legacy systems will exacerbate productivity stagnation, increase vulnerability to climate-induced yield losses, and widen the competitiveness gap with digitally advanced commodity sectors. Given sugarcane's contribution of 2–3% to national GDP in Brazil, India, and Thailand, inaction carries significant macroeconomic consequences including reduced export competitiveness and declining farmer welfare. Future research should advance integrated decision-support systems, extend AI and remote sensing models to supply chain domains, and validate digitalization frameworks across underrepresented producing nations to ensure equitable and scalable transformation pathways for the global sugarcane agroindustry.

The supplementary material for this article can be found online.