Public transportation and environmental sustainability nodes in TOD-based Mahakam River city planning

1 Introduction

Urban development in Indonesia faces major challenges in achieving an efficient, equitable, and sustainable transportation system. Uncontrolled urbanization has led to spatial inequality, congestion, and environmental degradation (Avashia et al., 2021), prompting the adoption of Transit Oriented Development (TOD) as a strategy for inclusive, low-emission cities. TOD integrates transport, land use, and socio-economic functions to reduce private vehicle dependence (Zhao et al., 2021), with success hinging on the balance of social, economic, and ecological dimensions (Alexandrescu et al., 2021) and integration with urban ecological policies (Liu et al., 2020). Beyond physical integration, TOD also functions as a multidimensional urban system in which transportation accessibility, ecological sustainability, socio-economic activity, and governance legitimacy continuously interact and influence urban development outcomes. In Samarinda’s Mahakam River area, rapid urbanization has strained infrastructure and ecosystems, necessitating TOD as a tool for adaptive, ecological, and participatory revitalization aligned with SDGs 11 and 13 (Hu et al., 2023). Weak intermodal integration and policy coordination persist (Anik et al., 2020), deepening socio-spatial inequality (Owusu et al., 2021; Yang et al., 2020) and ecological decline (Ghadami et al., 2021; Zhao et al., 2022). Strengthening community participation and collaborative governance (Enoch and Potter, 2023; Guterbock, 2021) supported by empirical approaches like (Biau and Scornet, 2016; Breiman, 2001), which is designed to model complex non-linear relationships through ensemble learning and bootstrap aggregation, is vital for sustainable Mahakam River revitalization. This study is based on residents’ perceptions collected through a structured survey, and therefore the findings reflect perceived social, economic, environmental, and policy conditions rather than objective macroeconomic or financial indicators. The analysis does not measure direct economic performance variables such as income growth, investment value, or fiscal statistics, but instead captures community-level assessments of TOD-related impacts and governance dynamics.

Most TOD research in Indonesia and other developing countries emphasizes physical and spatial aspects like land use and accessibility, while social, economic, and environmental factors are treated separately and linearly (Magalhães and Rivera-Gonzalez, 2021). Yet, TOD operates as a non-linear system (Bautista-Puig et al., 2022), especially in complex urban river contexts such as Mahakam, with distinct ecological and sociocultural dynamics (GUO et al., 2023; Ma et al., 2021). RFR offers methodological advantages by modeling non-linear relationships and measuring variable importance (Setthasuravich and Kato, 2020; Yang et al., 2023). This study bridges the gap by using RFR to analyze dynamic interactions among social, economic, environmental, and policy dimensions in TOD-based river revitalization. Theoretically, it expands TOD research beyond physical aspects (Habibi et al., 2022) and contributes to capturing non-linear interdependencies (Jin et al., 2021), while practically, it offers context-specific empirical evidence derived from community survey data that can inform discussions on sustainable and low-carbon urban mobility planning, (Chen et al., 2022), aligned with SDGs 11 and 13 (Bittencourt and Giannotti, 2021; Liu et al., 2020) through AI-driven sustainable urban planning (San-Akca et al., 2020).

2 Methods

This study employs an explanatory quantitative approach with a non-experimental survey design to examine predictive and reciprocal relationships among four key TOD dimensions: transportation accessibility, socio-economic impact, environmental sustainability, and policy satisfaction (Stadtländer, 2009). The Mahakam River area in Samarinda, East Kalimantan, serves as a representative case for TOD-based urban river revitalization. The conceptual framework emphasizes that transportation systems, socio-economic conditions, environmental sustainability, and public policy support operate as mutually reinforcing components within an integrated urban system. Rather than assuming a linear causal structure, the framework proposes reciprocal and adaptive interactions among the four dimensions. For instance, improved environmental sustainability may strengthen public satisfaction and policy legitimacy, while stronger policy support may reinforce transportation integration and ecological management. Similarly, improved transportation accessibility may stimulate socio-economic activity, which in turn can influence environmental conditions and governance responses. This reciprocal structure reflects the non-linear dynamics of TOD implementation in river-based urban revitalization contexts, as illustrated in Figure 1.

The RFR models were implemented using the following hyperparameters: number of trees (n_estimators = 500), maximum tree depth (max_depth = None), minimum samples per split (min_samples_split = 2), and minimum samples per leaf (min_samples_leaf = 1). The number of trees was set sufficiently high to stabilize error convergence and variable importance rankings. Hyperparameters were selected based on iterative testing and performance comparison using out-of-bag (OOB) error estimation and cross-validation. Model configurations were evaluated by comparing MSE and R ² stability across runs to ensure robustness. To mitigate overfitting beyond OOB estimation, the ensemble structure of RFR (bootstrap aggregation) was combined with performance monitoring across repeated runs and comparison of training and OOB errors. The relatively shallow effective tree structures and aggregation across 500 trees reduced model variance and enhanced generalization.

Figure 1 describes the mutual interactions among transportation accessibility, socio-economic impact, environmental sustainability, and policy satisfaction, highlighting the non-linear linkages between variables that shape TOD outcomes. Figure 1 illustrates the reciprocal conceptual relationships among transportation accessibility, socio-economic impact, environmental sustainability, and policy satisfaction within the TOD revitalization framework. The figure also presents the analytical workflow of Random Forest Regression, where each dimension is alternately modeled as a dependent variable to capture non-linear interactions among TOD dimensions. The analysis employs the RFR method (Gosiewska and Biecek, 2020), chosen for its ability to capture complex, non-linear interactions between social, economic, and ecological variables (Table 1). Although the variables were measured using Likert-scale perceptions, Random Forest Regression remains appropriate because it does not require strict distributional assumptions and can effectively handle ordinal variables transformed into numerical representations.

TABLE 1. Summary of variable operational definitions.

Source: Adapted and synthesized from TOD, urban governance, and sustainable urban revitalization literature (2025).

Although Extreme Gradient Boosting (XGBoost) was considered as an alternative ensemble learning method, Random Forest Regression was selected due to its greater robustness and stability when applied to relatively small-to-moderate sample sizes. RFR relies on bootstrap aggregation (bagging), which effectively reduces variance and minimizes overfitting without requiring extensive hyperparameter tuning. In contrast, boosting-based methods such as XGBoost are more sensitive to hyperparameter configuration and may be prone to overfitting in limited datasets if not carefully optimized. Therefore, RFR was deemed more appropriate for the exploratory and predictive objectives of this study, particularly in identifying stable variable importance across TOD dimensions. Primary data were collected via a structured questionnaire comprising 20 Likert-scale statements (1–5) from 105 respondents living around the Mahakam River. The questionnaire items were originally developed to capture perception-based manifestations of each TOD dimension and remained unchanged throughout the study, ensuring consistency between data collection, construct measurement, and subsequent analysis. The 20 questionnaire items were proportionally distributed across the four TOD dimensions based on conceptual relevance and operational indicators derived from previous validated studies.

Each TOD dimension was operationalized using multiple observable and measurable indicators adapted from previous TOD, urban sustainability, and public policy studies to ensure construct validity and contextual relevance. Transportation accessibility was measured through indicators related to multimodal connectivity, travel efficiency, infrastructure adequacy, and ease of access. Socio-economic impact included perceived employment opportunities, local business activity, household welfare improvement, and quality of life enhancement. Environmental sustainability was measured through pollution reduction, ecological preservation, green public space availability, and community participation in environmental management. Policy satisfaction included perceived policy effectiveness, transparency, public trust, and support for TOD continuity. Composite scores for each dimension were calculated by averaging all validated item responses following reliability and construct validity assessment.

The survey was conducted in accordance with institutional research guidelines. Participation was voluntary, no personal identifiers were collected, and informed consent was obtained from all respondents prior to participation. This sample size meets recommendations for multivariate analysis (Jussi Klemelä, 2018) and Cochran’s sample adequacy criteria (±9% sampling error). Although the sample size of 105 respondents may be considered modest for machine-learning applications, RFR is relatively robust for small-to-moderate datasets due to its bootstrap aggregation (bagging) mechanism and ensemble structure, which reduce overfitting and variance. Unlike parametric regression models that rely heavily on distributional assumptions, RFR is non-parametric and can maintain stable predictive performance even with limited observations, particularly when the number of predictors is moderate. However, when compared to more conservative population estimation formulas such as Slovin’s or Cochran’s formula with lower margins of error (e.g., 5%), the sample size may not fully represent the broader population of the Mahakam River area. Therefore, the findings should be interpreted as indicative of community-level perceptions rather than statistically definitive population estimates. Although the sample size was relatively modest, Random Forest Regression remains appropriate due to its ensemble-based robustness, bootstrap aggregation mechanism, and internal validation through out-of-bag estimation. Nevertheless, the findings should be interpreted as exploratory and context-specific within the Mahakam River TOD area.

Each dimension was alternately modeled as a dependent variable to examine reciprocal effects within the RFR framework. It is important to clarify that each Random Forest model was estimated independently, and no simultaneous or structural equation system was specified. The reciprocal testing of dimensions was intended to explore predictive interrelationships rather than to establish formal bidirectional causality. As a non-parametric ensemble learning method, RFR focuses on prediction accuracy and variable importance rather than structural coefficient estimation. Nevertheless, because the dimensions are conceptually interrelated, the possibility of latent endogeneity or omitted variable bias cannot be entirely excluded. Therefore, the findings should be interpreted as indicative of associative and predictive patterns rather than definitive causal pathways. Future research employing structural equation modeling or longitudinal data may provide stronger causal identification. Prior to model estimation, all variables were standardized, and model performance was evaluated using MSE, RMSE, MAPE, and R ². Correlation analysis among features was also conducted to provide an initial understanding of inter-variable relationships before non-linear modeling. To enhance the transparency of model evaluation, graphical visualization of MAPE values across models has also been included, allowing clearer comparison of predictive accuracy among the four TOD dimensions. Nevertheless, the limited sample size may restrict the generalizability of the findings beyond the Mahakam River context and may affect the stability of variable importance rankings. Therefore, the results should be interpreted as context-specific and exploratory, and future studies with larger samples are recommended to validate the robustness of the identified interdimensional relationships.

Table 1 presents the operationalization of each TOD dimension using measurable and observable indicators adapted from previous validated studies. Each construct was measured using multiple Likert-scale items to strengthen analytical consistency between theoretical constructs and empirical modeling. The operational indicators were adapted from established TOD, urban governance, and sustainability literature to strengthen the analytical linkage between theoretical constructs and empirical measurement. Accessibility indicators were adapted from multimodal transportation integration studies, while socio-economic and environmental indicators were derived from sustainable urban revitalization literature. Policy satisfaction indicators were adapted from public governance and policy acceptance studies.

The RFR model identifies the relative importance of each dimension using feature importance measures, revealing the non-linear interactions that underpin sustainable TOD implementation in the Mahakam River revitalization context. It is important to clarify that the conceptual refinement of TOD dimensions introduced in this revised manuscript did not involve any modification to the original survey instrument, questionnaire items, measurement scales, or empirical indicators. The four dimensions—Transportation Accessibility and Integration, Socio-Economic Impact, Environmental and Ecological Sustainability, and Satisfaction and Support for Policy—remain identical to those measured during the original data collection process. The revisions were intended to strengthen conceptual precision and theoretical grounding by providing clearer operational definitions and stronger justification for the selected indicators, rather than altering the empirical operationalization itself. Therefore, no changes were made to construct meaning, coding procedures, or analytical variables used in the Random Forest Regression models.

Prior to Random Forest modeling (Table 2), statistical validation procedures were conducted to ensure reliability and construct validity of the survey instrument (n = 104 valid responses). Internal consistency was confirmed using Cronbach’s Alpha and McDonald’s Omega (α = 0.831; ω = 0.831), both exceeding the recommended 0.70 threshold with stable confidence intervals.

TABLE 2. Integrated reliability and construct validity assessment.

Source: Survey data analysis (2025).

Construct validity was assessed through Exploratory Factor Analysis (EFA) with promax rotation, yielding acceptable sampling adequacy (KMO = 0.782) and a significant Bartlett’s Test of Sphericity (χ² = 169.387, df = 6, p < 0.001). Construct validity was assessed through Exploratory Factor Analysis (EFA) with promax rotation, yielding acceptable sampling adequacy (KMO = 0.782) and significant Bartlett’s Test results (χ² = 169.387, df = 6, p < 0.001). One dominant latent factor emerged (eigenvalue = 2.312), explaining 57.8% of total variance, with all factor loadings exceeding 0.60. These results indicate acceptable construct validity and internal coherence among TOD dimensions for predictive modeling purposes.

3 Results

Prior to conducting the Random Forest Regression analysis, descriptive statistics were computed to provide an overview of the distributional characteristics of the survey data across the four TOD dimensions. These statistics establish the empirical baseline of respondents’ perceptions before data standardization and algorithmic modeling. The descriptive results also provide an initial indication that respondents generally perceived the TOD revitalization process positively across transportation, socio-economic, environmental, and governance dimensions.

Table 3 indicates that all four dimensions exhibit mean values above the midpoint of the Likert scale, suggesting generally positive perceptions toward TOD implementation in the Mahakam River area. The relatively higher mean score for the Ecological Environment dimension suggests that respondents perceived environmental quality improvement, river ecosystem preservation, and green public space availability as relatively more satisfactory compared to transportation accessibility. The Ecological Environment dimension records the highest mean (3.801), followed by Socio-Economic Impact (3.719), Satisfaction & Policy Support (3.625), and Integrated Access Transport (3.231). Standard deviation values ranging between 0.494 and 0.652 indicate moderate variability in responses, reflecting perceptual diversity among residents. Notably, the minimum value of 1.000 in the Integrated Access Transport dimension suggests the presence of relatively critical evaluations, whereas the Satisfaction & Policy Support dimension shows a minimum of 3.000, indicating comparatively more consistent positive assessments. Overall, these descriptive statistics provide a transparent empirical foundation prior to the application of the RFR model.

TABLE 3. Descriptive statistics of TOD survey dimensions (n = 104).

The analysis employed four RFR models to evaluate interrelations among Integrated Access to Transport, Socio-Economic Impact, Ecological Environment, and Satisfaction and Policy Support (Table 4). To examine reciprocal predictive relationships among TOD dimensions, each variable was alternately modeled as a dependent variable within the RFR framework.

TABLE 4. Summary of RFR model performance and variable importance.

Source: analysis results, 2025.

The consistent prominence of the Ecological Environment variable across multiple models indicates that perceived environmental quality, ecological preservation, and sustainability-related conditions play an integrative role in shaping transportation effectiveness, socio-economic perceptions, and policy legitimacy within the TOD system. The RFR models demonstrated strong predictive performance: Integrated Access Transport (R ² = 0.868; MAPE = 7.44%), Socio-Economic Impact (R ² = 0.744; MAPE = 5.88%), Ecological Environment (R ² = 0.707; MAPE = 7.52%), and Satisfaction & Policy Support (R ² = 0.651; MAPE = 8.95%), reflecting consistent non-linear relationships across social, economic, environmental, and policy dimensions in the Mahakam River area. Therefore, the results do not explicitly identify the geographic origin of river transport users or quantify service coverage in terms of population served. Instead, the findings reflect perceived accessibility and utilization of the Mahakam River transport system within the surrounding urban area. Similarly, variables such as transport carrying capacity, service load, and ecological carrying capacity were not directly measured using physical or spatial indicators. In this study, ecological aspects are represented through perceived environmental quality, river condition, and sustainability-related indicators captured in the survey instrument. The strong influence of the Ecological Environment variable across multiple models therefore indicates that community perceptions of environmental conditions play a critical role in shaping transport effectiveness and policy acceptance. The reciprocal modeling structure demonstrates that no single TOD dimension operates independently; instead, transportation, socio-economic, environmental, and policy dimensions function as interconnected predictors within the urban revitalization system. The linkage to ecological considerations should thus be interpreted within a socio-ecological systems perspective, where environmental sustainability is understood not only in terms of biophysical capacity but also through public perception, participation, and policy support. Nevertheless, future research is needed to integrate spatial land-use data, passenger flow analysis, and quantitative measures of carrying capacity to provide a more comprehensive assessment of TOD implementation along the Mahakam River.

Model visualization (Figure 2) confirmed robust predictive accuracy and stability through Predictive Performance and OOB MSE plots, indicating minimal overfitting. The dominant role of the Ecological Environment in three of the four models reinforces its integrative position within the TOD system, linking ecology, mobility, and governance (Hongjun et al., 2023; Sun and Ertz, 2022). Conversely, the ecological model’s dependence on policy satisfaction (Felix et al., 2022; Ma et al., 2023) indicates a strong empirical linkage between policy perception and ecological sustainability within the TOD revitalization framework.

Figure 2 visualizes model accuracy and error stability for (a) Integrated Access Transport, (b) Socio-Economic Impact, (c) Ecological Environment, and (d) Satisfaction & Policy Support. The relatively narrow dispersion between observed and predicted values across the four models indicates satisfactory model generalization and limited overfitting during prediction. Specifically, the figure presents the comparison between predicted and observed values, along with the distribution of residual errors for each model, allowing assessment of prediction consistency and dispersion patterns. The relatively close alignment between predicted and actual values across the four models indicates stable predictive performance, while the spread of residuals suggests the presence of non-linear relationships captured by the Random Forest algorithm. The variation in predictor importance across models suggests that TOD dimensions interact adaptively rather than through fixed linear relationships, reinforcing the suitability of ensemble-based machine learning approaches for TOD analysis. This section reports the empirical findings of the Random Forest Regression models without theoretical interpretation. Analytical interpretation and linkage with existing literature are presented in the subsequent Discussion section.

4 Discussion

Building upon the empirical results presented in the previous section, this Discussion interprets the findings within the framework of sustainable urban mobility planning and TOD theory. The analysis reveals that Environmental and Ecological Sustainability is the most dominant factor in three of the four RFR models. It is important to emphasize that the ecological dimension in this study reflects residents perceived environmental conditions rather than direct biophysical measurements such as emission levels, water quality indices, or ecological carrying capacity. Therefore, the identified dominance of environmental sustainability should be interpreted as the perceived centrality of ecological quality within the TOD experience of local communities. This aligns with (Lu et al., 2023) who emphasize that environmental variables bridge mobility and public acceptance of sustainable transport, and with (Carrión-Mero et al., 2022) who confirm that river ecosystem health underpins water-based transport systems. The Integrated Transport Accessibility model achieved the highest performance, indicating that environmental quality enhances predictability and socio-economic synergy (Akudugu and Ogwu, 2024; Zhang et al., 2021). The Socio-Economic Impact model reinforces the link between ecological improvement and welfare gains (Wang et al., 2022; Strauss et al., 2021). Operationally, this relationship was reflected through measurable perception indicators including perceived employment opportunities, local business activity, household welfare improvement, and quality-of-life enhancement. The consistency of these indicators within the RFR models suggests that community welfare perceptions are strongly associated with environmental quality and accessibility integration. The Ecological Sustainability model highlights policy support as key for maintaining environmental quality (Elahi et al., 2024; Zhang et al., 2021). In this study, policy support was operationalized through observable indicators such as perceived policy effectiveness, implementation transparency, public trust, and willingness to support long-term TOD programs. These findings indicate that governance legitimacy functions not merely as an administrative factor, but as a predictive component shaping ecological sustainability perceptions. The findings demonstrate a reciprocal and systemic relationship among ecology, transport, and governance: improved environmental conditions are associated with more favorable socio-economic perceptions and transport efficiency, while effective mobility and participatory policy foster ecological protection, reflecting TOD’s integrative principle (Akudugu and Ogwu, 2024; Wang et al., 2022). Theoretically, this supports the argument that sustainable urban systems must balance mobility, ecology, and social wellbeing (Emberger and Pfaffenbichler, 2020; Langhans et al., 2022). The RFR approach contributes methodologically by uncovering non-linear interactions among social, economic, environmental, and policy dimensions, advancing the systems-thinking paradigm in adaptive urban governance (Kim, 2022; Waha et al., 2020). Unlike linear regression approaches that assume additive and proportional relationships, the RFR framework enabled the identification of recursive and non-linear interactions among transportation, ecological, socio-economic, and governance dimensions. This strengthens the analytical capability of the study in capturing the adaptive characteristics of river-based TOD systems.

Operationally, the study provides an empirical basis for TOD-oriented planning that prioritizes ecological sustainability and welfare. Cross-sectoral policy integration is essential for river-based TOD (Alomari et al., 2025), and transportation planning should align with ecological carrying capacity (Kramer et al., 2021; Zhang et al., 2023). Strengthening public participation is equally critical, as satisfaction correlates strongly with environmental performance (Bloch and Martínez, 2020; Cao et al., 2023). The application of RFR supports evidence-based decision-making, improving policy accuracy and responsiveness (Benevenuto and Caulfield, 2022; Pais et al., 2021). The study proposes that Urban River Corridor Management integrates transport, economy, ecology, and disaster mitigation (Sha et al., 2020), positioning the Mahakam River as a potential national prototype for sustainable river cities. This reflects the principle of adaptive governance (Sheng et al., 2023), emphasizing integration of data, participation, and ecological sensitivity. Globally, the results support SDGs 11, 13, and 15, aligning with (Lavorel et al., 2020; Maihaiti et al., 2025; Yun et al., 2021), and reinforce the New Urban Agenda and the Paris Agreement through low-carbon, climate-resilient transport (Elbroch et al., 2023; Lam et al., 2022; Rendon et al., 2021; Villanueva-Merino et al., 2024). The findings position Indonesia within the global discourse on river-based sustainable urbanism, affirming the role of ecological integration in developing just, adaptive, and low-carbon cities (Tiller et al., 2022).

To extend the contribution beyond a localized case, the Urban River Corridor Management concept proposed in this study can be interpreted through established sustainable urban mobility indicator frameworks. Core indicators in sustainable mobility planning typically include accessibility integration, environmental impact reduction, socio-economic inclusivity, governance transparency, and system resilience. The dominance of the Environmental and Ecological Sustainability dimension across the RFR models aligns with environmental performance indicators that measure reduced ecological degradation and enhanced urban ecosystem services as prerequisites for sustainable transport systems. Similarly, the strong interaction between Accessibility, Socio-Economic Impact, and Policy Support corresponds with widely recognized mobility indicators related to equitable access, modal integration, and public legitimacy. By empirically demonstrating how ecological quality functions as a central node linking transport efficiency, welfare outcomes, and governance acceptance, the Mahakam River case provides a systems-based validation of integrated mobility frameworks in river-oriented urban contexts. Thus, Urban River Corridor Management is not positioned merely as a localized planning recommendation but as a transferable systemic model that operationalizes sustainable urban mobility principles within environmentally sensitive corridors. This framing situates the study within the broader scientific discourse on integrated, low-carbon, and adaptive mobility transitions.

In relation to conventional transport planning approaches, particularly the four-step model (trip generation, trip distribution, mode choice, and route assignment), this study offers a complementary perspective. Accordingly, the novelty of this study lies not only in the application of machine-learning techniques within TOD analysis, but also in the integration of socio-ecological governance variables into a reciprocal predictive framework. This approach expands conventional TOD studies that predominantly focus on spatial density, land-use configuration, and transport accessibility. Traditional TOD studies often rely on aggregate travel demand modeling to predict mobility patterns and infrastructure needs. While such approaches are effective in estimating traffic flows and network performance, they tend to treat environmental and governance dimensions as external or secondary factors. In contrast, the present study integrates environmental sustainability, socio-economic impact, and policy support as endogenous components within the analytical framework, allowing for a more holistic understanding of TOD dynamics in river-based systems. The findings demonstrate that, unlike the linear structure of the four-step model, TOD implementation in the Mahakam River context follows a non-linear and reciprocal pattern, where environmental quality not only results from transport policies but actively shapes accessibility, socio-economic outcomes, and public acceptance. This indicates that transport effectiveness in river-based TOD cannot be fully explained by movement-based models alone, but requires integration with socio-ecological system perspectives. Accordingly, the main novelty of this research lies in three aspects. First, it introduces a reciprocal multi-dimensional TOD framework in which transport, socio-economic, environmental, and governance variables are modeled as interdependent rather than hierarchical. Second, it applies Random Forest Regression to capture non-linear interactions and variable importance across TOD dimensions, extending beyond conventional parametric and linear TOD analyses. Third, it positions environmental sustainability as a central predictive driver rather than a peripheral outcome, providing empirical evidence that ecological conditions function as a key integrator of mobility performance and policy legitimacy in river-based urban systems. These contributions differentiate the study from existing TOD literature and provide a new analytical lens for understanding sustainable transport development in environmentally sensitive urban corridors.

While the findings contribute to broader sustainable mobility discourse, several methodological limitations should be acknowledged. First, the sample size (n = 105; valid n = 104) is modest, which may affect the stability of feature importance rankings in machine-learning modeling. Second, the study relies on self-reported perception data, which may introduce response bias and subjective interpretation of TOD impacts. Accordingly, the refined conceptual framing introduced in this study should be interpreted as theoretical clarification of measured constructs rather than substantive modification of the original operationalization. Moreover, several TOD dimensions were measured using aggregated perceptual indicators, which may not fully capture objective transport performance, ecological resilience, or spatial mobility dynamics. Consequently, the findings should be interpreted as perception-based systemic associations rather than direct measurements of physical urban performance. Third, the cross-sectional design limits the ability to infer temporal dynamics or causal relationships among dimensions. Fourth, the analysis does not incorporate spatial validation using GIS-based environmental or transport performance indicators, which could strengthen objective triangulation. Finally, the findings are context-specific to the Mahakam River corridor and may not be directly generalizable to other urban river systems without contextual adaptation. Future studies are also encouraged to integrate GIS-based transport accessibility metrics, environmental sensor data, passenger flow analysis, and spatial land-use indicators to strengthen triangulation between subjective perceptions and objective urban system performance.

5 Conclusion

This study demonstrates that the relationship among transportation accessibility, socio-economic impact, environmental sustainability, and policy support in the Mahakam River revitalization is multidimensional and non-linear. The RFR analysis shows that environmental and ecological sustainability is the most influential factor in three of the four models, underscoring its central role in determining transportation efficiency, social welfare, and policy legitimacy. The Transportation Accessibility and Integration model achieved the best performance (R ² = 0.868; MAPE = 7.44%), followed by Socio-Economic Impact (R ² = 0.744), Environmental Sustainability (R ² = 0.707), and Policy Support (R ² = 0.651). These findings directly address the research objective of examining reciprocal and non-linear relationships among key TOD dimensions in a river-based urban context. The results indicate that perceived ecological sustainability is consistently associated with transport accessibility, socio-economic evaluation, and policy legitimacy, suggesting that environmental perception functions as a central node within the integrated TOD system.These findings confirm that ecological sustainability acts as the key integrator of transportation, socio-economic, and governance dimensions within sustainable development planning, reinforcing the relevance of eco-integrated TOD in river-based urban areas such as the Mahakam River.

Rather than establishing causal relationships, the study identifies predictive associations derived from cross-sectional perception data using ensemble learning techniques. Methodologically, the application of Random Forest Regression contributes to TOD research by demonstrating how non-linear machine learning approaches can uncover complex interdependencies among environmental, social, economic, and governance dimensions in urban systems. The practical implications of this study lie in emphasizing the strategic role of ecological quality within river-oriented TOD planning. Integrated policy approaches that simultaneously address ecosystem management, accessibility enhancement, and participatory governance may strengthen public support and perceived system effectiveness. In this regard, Urban River Corridor Management can be interpreted as an operational framework that aligns transport planning with ecological resilience and governance legitimacy.

However, several limitations should be considered when interpreting these findings. The sample size is modest and based on self-reported perception data, which may introduce response bias. The cross-sectional design limits temporal inference, and the absence of spatial validation (e.g., GIS-based indicators) restricts objective triangulation. Furthermore, the findings are context-specific to the Mahakam River corridor and may not be directly generalizable without contextual adaptation.

Future research should incorporate spatial-temporal and geospatial analyses (e.g., GIS and remote sensing) to explore changes in land use, ecosystem health, and transport accessibility over time. Additionally, policy simulations and scenario modeling are recommended to evaluate the long-term impacts of low-emission transport, flood-resilient infrastructure, and green city initiatives. Further studies should also address behavioral and governance dimensions, including stakeholder collaboration and institutional capacity, and conduct cross-regional comparisons with other Southeast Asian river-based TOD projects.

By situating the Mahakam River case within broader sustainable urban mobility discourse, this study contributes a perception-based, systems-oriented perspective to environmentally sensitive TOD implementation. The findings underscore the importance of integrating ecological sustainability, accessibility, socio-economic inclusion, and governance responsiveness in advancing adaptive and climate-resilient urban mobility transitions.

These steps will support the development of a regional framework for sustainable, climate-resilient urban transportation, aligned with the SDGs, New Urban Agenda, and Paris Agreement.

Generative AI statement

The author(s) declared that generative AI was not used in the creation of this manuscript.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher’s note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Footnotes

Edited by: Diego Rossit, Universidad Nacional del Sur Departamento de Ingenieria, Argentina

Reviewed by: Randa Khalil, Future University in Egypt, Egypt

Endah Murtiana Sari, Universitas Sains Indonesia, Indonesia

Mudiyati Rahmatunnisa, Padjadjaran University, Indonesia