Search Results (613)

Sustainable land-use regulation in historical and cultural cities requires balancing heritage conservation, development demand, cropland retention, and urban–rural spatial restructuring. However, the stage-wise reorganization of urban–rural construction land under these coupled pressures remains insufficiently understood. Taking Jingzhou District, China, as a case study, this study uses land-use data from 2000, 2005, 2010, 2015, and 2020 and integrates stage-wise random-forest analysis, consistency-based interaction-network mining, and multi-scenario simulation within the intPLUS framework. Population, GDP, and areal-water distance layers were matched to the corresponding stage-terminal snapshots where applicable, whereas 2020 POI data were used as contemporary spatial-context proxies. From 2000 to 2020, urban industrial land (UIL) expanded from 16.63 to 46.42 km², increasing by approximately 179.1%, whereas rural settlements (RS) increased more moderately from 56.59 to 60.27 km², increasing by approximately 6.5%. The stage-wise RF and interaction-network results show that UIL and RS followed different spatial association structures, with stronger UIL self-reinforcement and stronger RS self-continuity in the later stage. Historical validation showed overall accuracy values of approximately 91% and Kappa values around 0.80, but FoM values remained relatively low, ranging from 0.098 to 0.176. Class-specific mapping accuracy was higher for RS (81.90–82.37%) than for UIL (55.20–66.93%), indicating a weaker performance in locating UIL change. Therefore, the 2035 simulations should be interpreted as parameter-conditioned regulatory comparisons rather than deterministic pixel-level forecasts. The scenario results indicate that the conservation-oriented limited growth was associated with the restricted UIL expansion and better cropland retention under the prescribed demand and constraint settings, while the RS reduction occurred only under explicit village-consolidation and construction-land quota reallocation assumptions. By distinguishing UIL and RS, this study provides differentiated regulation-oriented evidence for sustainable land-use governance in historical and cultural cities. Full article

This paper presents a multibeam hybrid beamforming (MHBF) architecture for microwave power transfer (MPT), enabling wireless power delivery to multiple receivers with a reduced number of RF chains. The proposed architecture decouples beam control into the horizontal and vertical dimensions, where horizontal multibeams are generated in the baseband through digital precoding, while the vertical beam direction is controlled by a Butler-matrix-based analog beamformer. In particular, multibeam transmission is achieved using multi-tone signals with distinct phase weights assigned to each tone, enabling beams to be steered toward different directions, while the Butler-matrix-based analog beamformer provides vertical beam-steering capability. Compared with fully digital beamforming (DBF), MHBF enables simultaneous multibeam formation in the horizontal domain with fewer RF chains, thereby reducing hardware overhead and system complexity. To validate the proposed architecture, a 5.8 GHz prototype was designed and fabricated. The experimental results demonstrate three-beam and four-beam operation under a transmit power of 30.57 dBm, while the average received RF power in the single-beam case was 12.11 dBm at a distance of 1 m. In the three-beam and four-beam cases, average received RF power levels of 7.3 dBm and 6.1 dBm per beam were achieved, respectively. RF-to-DC conversion measurements under 430 $\Omega$ and 680 $\Omega$ load conditions further showed average PCE values of up to 38.77% and 35.05% for the three-beam and four-beam cases, respectively. These results confirm the feasibility of simultaneous multibeam wireless power delivery and its potential as an effective solution for multi-receiver operation with reduced RF-chain requirements. Full article

Accurate path-loss models are essential for planning reliable wireless networks in underground mines, yet existing characterization studies rely on specialized channel sounders and vector network analyzers costing tens of thousands of dollars, placing them beyond the reach of most mine operators. This paper demonstrates that LoRa transceivers costing approximately US $15 per node can serve as a self-contained path-loss measurement instrument, logging the received signal strength indicator (RSSI) and signal-to-noise ratio (SNR) directly to a CSV file over a standard USB serial connection. A measurement campaign conducted at the Missouri S&T Experimental Mine on 31 March 2026 collected 4801 packets across four distinct underground canonical primitives: straight tunnel, T-junction, vertical shaft, and post-bend NLoS gallery at distances of 5 to 60 m using Waveshare Pico-LoRa-SX1262 boards operating at 915 MHz. The results reveal a pronounced two-zone propagation structure, including a line-of-sight (LoS) zone with a negative path-loss exponent of −0.34, confirming tunnel waveguide gain up to 25 m, followed by a steep NLoS zone with an exponent of 13.0 after a 24.0 dB bend diffraction loss. Environment-specific measurements quantify a 5.5 dB junction excess loss and a 29.5 dB shaft excess loss relative to a straight-tunnel reference. Spreading factor sensitivity tests across SF7, SF9, and SF12 confirm that RSSI measurements are consistent to within 2 dB across all SFs, validating the measurement methodology. The resulting four-zone path-loss model provides mine network planners with parameters sufficient for LoRa link budget design and relay node placement without any specialized RF instrumentation. Full article

Tooth extraction and maximum anchorage assessment are key decision points in orthodontic treatment planning, yet existing machine learning models for orthodontic decision support often lack transparency, limiting their clinical interpretability and trustworthiness. In this study, we developed and externally validated an explainable orthodontic treatment decision-support model that integrates expert-derived Analytic Hierarchy Process (AHP) weighting with machine learning. A diagnostic indicator framework comprising 18 orthodontic variables was established through a literature review, clinical data analysis, and two rounds of expert surveys. A retrospective cohort of 485 patients receiving fixed-appliance orthodontic treatment was used for model development and internal validation. AHP-derived composite scores were incorporated into the machine learning models for two prediction tasks, namely tooth extraction and maximum anchorage requirement, and an expert-informed fuzzy-rule score was calculated from pretreatment indicators for the maximum anchorage task to capture clinically interpretable anchorage tendencies. Model performance was evaluated using ROC-AUC, F1 score, precision, recall, PR-AUC, calibration analysis, and decision curve analysis, while SHAP was applied to interpret feature contributions. The AHP-RF extraction model and AHP-enhanced LR maximum anchorage model achieved the highest AUCs among the compared models (0.864 and 0.822, respectively), although paired DeLong tests showed no significant differences from the closest competing models. SHAP analysis identified lower lip-to-E-line distance, U1-NA, and the AHP composite score as important predictors, indicating consistency between model outputs and clinical reasoning. In the external validation cohort, the extraction model correctly classified 57 of 74 cases, and the maximum anchorage model correctly classified 24 of 29 cases, supporting the preliminary transportability of the proposed framework. These results suggest that integrating AHP-derived expert knowledge with machine learning provides an explainable and clinically interpretable decision-support model for orthodontic treatment planning, with potential value in improving standardized, evidence-informed, and patient-specific orthodontic decision-making. Full article

High-precision geomagnetic reference maps are essential for reliable geomagnetic field modeling and accurate geomagnetic matching navigation, especially in regions with sparse observations and complex magnetic anomaly variations. However, conventional map construction methods often exhibit limited precision and robustness, particularly when geomagnetic observations are sparse or spatial variations are complex. To address these challenges, this study proposes an improved artificial bee colony-optimized random forest model (IABC-RF) for reconstructing geomagnetic reference maps using magnetic anomaly data. The proposed method integrates an enhanced artificial bee colony strategy to optimize the hyperparameters of the random forest model, improving its predictive accuracy and stability in nonlinear geomagnetic environments. The experiments conducted on geomagnetic anomaly data from the South China Sea region, specifically between 5–25′ N and 100–120′ E, derived from the World Digital Magnetic Anomaly Map, show that the IABC-RF method outperforms traditional approaches. The IABC-RF method achieves the lowest root mean square error (RMSE) of 1.46 nT and the smallest standard deviation of 1.58 nT, while also maintaining a competitive computational time of 3.4 s. In comparison, Kriging interpolation produces an RMSE of 2.47 nT, inverse distance weighting (IDW) results in an RMSE of 14.45 nT, and improved Shepard interpolation gives an RMSE of 11.68 nT. The IABC-RF method excels at preserving global geomagnetic trends and accurately recovering localized anomaly details, offering enhanced robustness to outliers. Further evaluation of the IABC-RF method under noisy conditions (5% and 10% noise) revealed that although all methods experienced a decrease in performance due to the added noise, the IABC-RF method continued to show superior robustness. These findings demonstrate that the IABC-RF method provides a highly effective and reliable solution for constructing high-precision geomagnetic reference maps, with strong performance even in noisy environments. The method is particularly valuable for improving geomagnetic matching navigation in complex operational settings. Full article

Small, dispersed, and vegetated creeping landslides in hilly areas of eastern China hinder traditional remote sensing and detection. To address this, this study takes Yixian County (Anhui Province) as a representative area. Based on high-resolution GF-2 satellite images, three deep learning models embedded with the Squeeze-and-Excitation (SE) attention mechanism (ResNet18-SE, VGG13-SE, UNet-SE) were developed and compared with the original UNet model. Combined with field investigation, landslide mapping and accuracy assessment were conducted to evaluate the feature extraction capabilities of four models. The results indicate that the UNet-SE model achieved optimal performance (Precision: 0.911, Recall: 0.685, F1-score: 0.782, Kappa: 0.730, IoU: 0.643). Its F1-score exceeds ResNet18-SE, VGG13-SE, and the original UNet by 8%, 3%, and 5%, respectively, proving superior regional adaptability and generalization performance. Additional verification on creeping landslides in Kecun Town (Yixian County) and post-earthquake landslides in Lushan County (Sichuan Province) further confirms the reliability of the UNet-SE model. Furthermore, Frequency Ratio (FR), Random Forest (RF), and SHapley Additive exPlanations (SHAP) were adopted to reveal the correlation between landslide occurrence and seven geological-environmental factors. Landslides are most susceptible to develop at elevations of 400–500 m, NDVI values of 0.4–0.5, slopes below 10°, east and northeast aspects, 300–500 m away from rivers, 500–1000 m away from faults, and areas covered by soft sedimentary lithology. Distance from faults, distance from rivers, and elevation are identified as the three favorable conditional factors. In conclusion, the proposed landslide detection framework can provide reliable spatial data and technical references for regional geological hazard prevention, ecological conservation and sustainable development in hilly areas. Full article

Background/Objectives: To evaluate the associations and concurrent validity between baseline functional and morphofunctional assessments in patients with cardiovascular disease participating in a Phase II cardiac rehabilitation program, as a basis for informing individualized exercise prescription. Methods: We conducted an observational retrospective cross-sectional study of patients enrolled in a Phase II outpatient cardiac rehabilitation program (January 2021–December 2023, Málaga). Functional assessments included handgrip strength (HGS), isometric biceps and quadriceps dynamometry, and direct assessment of 20-repetition maximum (20RM) through dynamic resistance exercises using external loads (defined as the maximum load allowing approximately 20 repetitions to near muscular fatigue). Aerobic capacity was evaluated using the 6-min walk test (6 MWT) and a modified Bruce exercise stress test with estimated METs. Morphofunctional assessment included vector bioimpedance analysis (phase angle [PhA], fat-free mass [FFM], body cell mass [BCM]) and rectus femoris ultrasound (cross-sectional area [RF-CSA] and contracted diameter [RF-CON]). Correlation and linear regression analyses were performed. Results: The sample included 223 participants (78.0% male; age 57.7 ± 8.6 years). HGSmax correlated strongly with 20RM biceps (r = 0.89) and moderately with quadriceps (r = 0.72). 6 MWT distance and speed correlated with ergometry-derived METs (r = 0.38–0.40; p < 0.001), whereas Borg ratings correlated inversely with METs and exercise time (r = −0.32 to −0.34; p < 0.001). PhA, BCM, FFM, and rectus femoris ultrasound measures correlated with both strength and aerobic outcomes (ρ ≈ 0.33–0.50; all p < 0.001). In regression analyses, HGSmax was the main predictor of 20RM biceps (R² = 0.792) and showed moderate predictive capacity for quadriceps performance (R² = 0.521). The MET model demonstrated limited explanatory capacity (R² = 0.288). Conclusions: The integration of simple, accessible, and reproducible tools such as HGS and the 6 MWT with morphofunctional parameters may provide a pragmatic approach to support individualized exercise prescription in cardiac rehabilitation. While stronger associations were observed for upper-limb resistance performance, the predictive capacity for lower-limb strength and aerobic exercise intensity was more moderate and should be interpreted cautiously. These findings support the potential clinical utility of combining functional and morphofunctional assessments in routine cardiac rehabilitation practice. Full article

Microwave wireless power transfer (MWPT) is a promising technology for powering dedicated industrial Internet of Things (IoT) devices, enabling battery-free operation. However, in realistic MWPT deployments, the received RF signals fluctuate drastically due to varying transmission distances and multipath fading. Additionally, the equivalent impedance of sensor nodes varies significantly during duty cycles, shifting between a low-resistance active state and a high-resistance sleep state. Consequently, maintaining high rectification efficiency under these dynamic conditions remains a critical challenge. This paper proposes a high-efficiency rectifier with a wide input power and load range based on the suppression of second and third harmonics. The rectifier adopts a dual-diode parallel configuration. By leveraging the impedance compensation characteristics of two short-circuited stubs with distinct electrical lengths, it simultaneously achieves fundamental-frequency impedance matching and harmonic suppression without the need for an additional matching network. Validated through theoretical derivation, simulation analysis, and physical prototype testing, the proposed 2.45 GHz rectifier realizes high-efficiency rectification over a wide dynamic range. Experimental results demonstrate that the power dynamic range reaches 10 dB when the rectification efficiency exceeds 70%, and extends to 17 dB when the efficiency is above 60%. Furthermore, the rectification efficiency is insensitive to load variations (100–1200 Ω), making it highly suitable for powering wireless sensor nodes with varying operating modes in complex electromagnetic environments. Full article

Accurate assessment of soil moisture is essential for enhancing irrigation efficiency and promoting sustainable agriculture. This study was conducted at Punjab Agricultural University, Ludhiana (PAU), to investigate the spatial and depth-wise variability of soil moisture across 30 field sites by using field measurements, geospatial-based (inverse distance weighting: IDW) interpolation, and machine learning techniques. Soil moisture was recorded at four depth intervals, including 0–15 cm, 15–30 cm, 30–45 cm, and 45–60 cm. The surface layer (0–15 cm) exhibited the highest variability due to evaporation and irrigation timing, with values ranging from 4.5% to 16.0%. Deeper layers showed more stable moisture retention, particularly at sites with intensive irrigation and crop cover, such as L11 (wheat), L22 (Gobhi Sarson), and L25 (wheat), where the moisture levels exceeded 14% at 45–60 cm depth, supporting suitability for deep-rooted crops. Supervised machine learning models, namely decision tree (DT), random forest (RF), and logistic regression (LR), were employed to classify soil moisture into low, medium, and high categories. The highest classification accuracy (88.9%) was achieved by the decision tree at 30–45 cm and logistic regression at 15–30 cm. Shallow layers exhibited frequent misclassification between medium and high classes, indicating surface-induced variability. Unsupervised clustering using K-means (k = 4) and hierarchical methods effectively delineated distinct soil moisture zones aligned with land use, irrigation history, and crop cover. The combination of geospatial analysis, depth-specific field data, and machine learning models provides an integrated framework for precision soil moisture assessment. This approach supports site-specific irrigation scheduling and water resource optimization, which are particularly critical for groundwater-stressed regions like Punjab. The novelty of this study lies in integrating depth-specific field-based soil moisture observations with geospatial interpolation and machine learning-based classification and clustering approaches to improve subsurface moisture characterization for precision irrigation management. Full article

Accurate and timely monitoring of soil salinity is of great significance for the ecological restoration of saline-alkali land and precision agricultural management. In this study, a typical coastal saline-alkali farmland located in Huanghua City, Hebei Province, China, in the Bohai coastal region, was selected as the study area. High-resolution images were acquired using an unmanned aerial vehicle (UAV) equipped with a multispectral sensor, and ground soil salinity samples were collected synchronously. Based on the construction of a feature library comprising spectral reflectance, vegetation indices, and salinity indices, three algorithms, PSO-SFLA, MultiSURF, and VIP, were employed for feature selection. Subsequently, an ensemble model was established, utilizing Ridge Regression (Ridge), Random Forest (RF), and Extra Trees (ET) as primary base learners, and Extreme Gradient Boosting (XGBoost) as the secondary meta-learner. This ensemble model was applied for soil salinity inversion. Furthermore, the coefficient of determination (R²), standardized root mean square error (S_RMSE), and the ratio of performance to interquartile distance (RPIQ) were introduced to comprehensively evaluate the accuracy of the models. Finally, the intrinsic physical responses of the features were explored through SHAP. The results showed that the optimization by the PSO-SFLA effectively reduced the impact of spectral multicollinearity, and 11 core features highly sensitive to salinity were selected from a vast number of indices. The ensemble model showed better predictive performance on the independent test set, achieving an R² of 0.758, an S_RMSE of 0.285, and an RPIQ of 3.382, outperforming the single Ridge, RF, and ET models under the current experimental conditions. Based on this model, the spatial distribution map of soil salinity in the experimental area was generated. The integrated and interpretable workflow proposed in this study, combining UAV multispectral imagery, PSO-SFLA-based feature selection, ensemble learning, and SHAP interpretation, provides a practical approach for accurate soil salinity inversion and dynamic agricultural monitoring in coastal saline-alkali lands. Full article

Harmful algal blooms (HABs) are increasing in frequency in the Chesapeake Bay, posing risks to marine ecosystems, water quality, and public health. Chlorophyll-a (Chl-a) is a widely used indicator of algal biomass, and satellite observations such as Sentinel-3 Ocean and Land Color Instrument (OLCI) enable large-scale monitoring of bloom dynamics. However, cloud cover and atmospheric interference frequently introduce missing pixels in daily satellite products, reducing temporal continuity and limiting monitoring reliability. Satellite-derived chlorophyll-a (Chl-a) data exhibit substantial missingness, with daily pixel gaps ranging from approximately 52.30% to 100% (mean ≈ 88.95%). This study evaluates spatial interpolation, EOF-based, supervised machine-learning, deep-learning, and convolutional autoencoder approaches for reconstructing missing Chl-a values. Sentinel-3 OLCI Chl-a data from 2023–2024 were used for model training, while data from 2025 served as a temporally independent test set to avoid spatiotemporal leakage. To simulate cloud-induced data gaps, artificial missingness scenarios ranging from 50% to 90% were applied for the Inverse Distance Weighting (IDW) and Data Interpolating Empirical Orthogonal Functions (DINEOF) baseline approaches, while machine-learning, deep-learning, and convolutional autoencoder models were evaluated using real satellite-derived missing observations. The evaluated models include IDW, DINEOF, K-Nearest Neighbors (KNN), Random Forest (RF), Extra Trees (ET), XGBoost, a Long Short-Term Memory (LSTM) network, and a Temporal Data Interpolating Convolutional Autoencoder (Temporal DINCAE). Model performance was assessed using Root Mean Square Error (RMSE), Mean Absolute Error (MAE), prediction bias, and the coefficient of determination (R²). Results indicate that tree-based ensemble models outperform spatial interpolation and EOF-based methods, with XGBoost achieving the best overall performance (R² ≈ 0.86; RMSE ≈ 9.61 mg m⁻³). The LSTM model achieved lower prediction errors (RMSE ≈ 5.87 mg m⁻³; MAE ≈ 2.16 mg m⁻³), highlighting the benefit of incorporating temporal dependencies, although with slightly reduced variance capture. The convolutional autoencoder-based Temporal DINCAE model achieved strong reconstruction performance (R² ≈ 0.84; RMSE ≈ 11.15 mg m⁻³). Uncertainty quantification shows that Extra Trees tends to underestimate uncertainty with narrower prediction intervals, whereas XGBoost provides better-calibrated but wider intervals. Full article

The reinforcement and retrofit of damaged steel buildings has emerged as a primary focus in civil engineering. It should be noted that completing the reasonable strengthening design for avoiding the sudden collapse of a structure in extreme engineering conditions was an urgent task, while the existing method required a long time which significantly influenced the reinforcing practice. In the present study, an improved explainable machine learning (ML) framework was developed for the rapid assessment of the bending property of repaired laminated channel beams. Firstly, a comprehensive database of 192 samples combining experimental and finite element data was established. The Mahalanobis distance analysis and Pearson correlation analysis were sequentially performed to evaluate the singularity of the samples and the dependencies between the variables. Secondly, the adversarial tests were conducted on the randomly selected 10 pairs of training and testing sets to determine the database with the best distribution consistency. Then, three machine-learning models of artificial neural networks (ANN), random forest (RF), and extreme gradient boosting tree (XGBoost) were respectively trained and validated. Finally, the explainability analysis of the XGBoost model was carried out in the global and local perspectives based on the SHAP method. The prediction accuracy (R²) of all ML models exceeded 90%, demonstrating good accuracy and providing a useful reference within the current database for the reinforcement design of damaged steel beams in emergency situations. In addition, the XGBoost model achieved superior prediction accuracy (R² = 97.98%) and stability (CoV = 0.82%) compared to ANN and RF. The explainability analysis revealed that boundary conditions and load type had the most significant influence on bending capacity. The proposed ML approach enabled efficient and reliable bending capacity estimation, supporting rapid decision-making in emergency reinforcement scenarios for damaged steel structures. Full article

This review presents a comprehensive and thorough evaluation of recent developments in physical vapour deposition (PVD) radiofrequency (RF)-sputtered β-Ga₂O₃ thin-film-based solar-blind ultraviolet (UV) photodetectors (SB-UVPDs), emphasizing their potential for next-generation optoelectronic applications. The review highlights different photodetector architectures, the performance characteristics of SB-UVPDs, and an overview of the attributes of β-Ga₂O₃ that make it a promising wide-bandgap semiconductor for next-generation devices. Additionally, the working principle of the PVD RF magnetron sputtering technique is discussed briefly, with a particular focus on the influence of deposition parameters, including sputtering power, gas pressure, deposition time, target-to-substrate distance, and substrate temperature, on the resulting film’s crystallinity and morphology and the optical quality of SB-UVPDs. Moreover, the impact of post-deposition treatments, such as post-annealing and elemental doping, is also discussed here for SB-UVPDs. And finally, the electrical performance characteristics of SB-UVPDs are discussed categorically based on deposition parameters. Overall, this review establishes that PVD RF magnetron sputtering is a highly versatile and controllable technique for fabricating high-quality β-Ga₂O₃ thin film-based SB-UVPDs. By carefully optimizing deposition and post-processing parameters, the optoelectronic performance of β-Ga₂O₃-based SB-UVPDs can be effectively tuned, enabling their integration into next-generation high-performance optoelectronic and photonic systems. Full article

The brain and testes are connected via the hypothalamic–pituitary–gonadal (HPG) axis. Both are vulnerable to radiofrequency electromagnetic radiation (RF-EMR). However, no comprehensive study had evaluated the effects of RF-EMR on key hormones along this axis. Hereby, this study evaluated the effect of RF-EMR on the hormonal changes along the axis, including the neuropeptide kisspeptin. A total of 18 (N = 18) adult Sprague–Dawley rats were divided into three groups (n = 6): Control, 4 h, and 24 h. The Control group was sham-exposed to an inactive router. The exposed groups were subjected to 2.45 GHz RF-EMR for 4 and 24 h daily, for 60 days at a 20 cm distance. The power density was 0.141 W/m² with a whole-body specific absorption rate (SAR) of 0.41 W/kg. No significant changes were observed in hypothalamic Kiss1 gene expression or serum kisspeptin levels. GnRH levels increased significantly in both exposed groups, while FSH and LH remained unchanged. Testicular testosterone was significantly reduced in the 24 h group, while serum testosterone was elevated in the 24 h group compared to the 4 h group. In conclusion, prolonged 2.45 GHz RF-EMR exposure caused selective changes in components of the HPG axis, particularly involving GnRH and testosterone, suggesting potential endocrine effects on male reproductive regulation. Full article

Urban park use is a key indicator of sustainable urban development, reflecting the accessibility and social value of urban green infrastructure. However, existing studies often struggle to distinguish stable spatial differences from short-term temporal dynamics. Using monthly data for 125 urban parks in Las Vegas from 2022 to 2024, this study examines how park visitation is shaped by spatial, temporal, and contextual factors. It addresses three objectives: identifying cross-park determinants of visitation, examining within-park monthly dynamics, and assessing spatial variation in key relationships. Park visitation is measured using observed visit counts, with dwell time and travel distance used as alternative behavioral outcomes for robustness tests. To address these research questions, this study asks: (1) what structural and contextual factors explain cross-park differences in park visitation; (2) how park visitation responds to changing contextual conditions within parks over time at the monthly scale; and (3) whether the relationships between park visitation and its key determinants vary across space. To answer these questions, the analysis combines annual cross-sectional ordinary least squares (OLS) regression, monthly panel models, Random Forest analysis, robustness tests, and geographically weighted regression. This study employs a triangulated analytical framework combining cross-sectional ordinary least squares (OLS) regression monthly fixed-effects (FE) panel models, and Random Forest (RF) analysis. These factors function as stable support for sustainable park use. Crime exposure shows no stable global linear effect, but its association with visitation appears conditional on temporal and spatial context. Overall, the findings suggest that park visitation is shaped by the interaction of physical design, safety conditions, and urban context. By explicitly separating cross-sectional spatial and economic inequalities from within-park temporal dynamics, this study offers policy-relevant evidence for urban planners and park managers seeking to promote more inclusive, efficient, and sustainable urban park systems through integrated design, economic activation, and safety-oriented interventions. Full article