Spatio-temporal modelling of wind speed using machine learning with a custom Weibull deviance loss for XGBoost

Abstract With the rapid development of surveying and mapping technology, high-precision and high-temporal-resolution Earth rotation parameters play a crucial role in fields such as satellite navigation and deep space exploration. Currently, ultra-rapid data exhibits lower accuracy, while final solution data suffers from low temporal resolution. Interpolation methods can estimate the unknown values at unmeasured points based on known discrete data, thereby generating continuous and higher-resolution datasets. The Inverse Distance Weighted (IDW) method leverages a distance-based weighting mechanism to fully utilize the spatial distribution characteristics of the data, assigning greater weights to nearby points and producing smooth and highly accurate interpolation results, making it well-suited to the continuity features of polar motion data. In this study, the IDW method is applied to interpolate the IGS final solution polar motion data to obtain high-temporal-resolution interpolated components. The reliability of the high-resolution data is validated using Fast Fourier Transform and Dynamic Time Warping algorithms. Furthermore, forecasting is conducted by combining least squares extrapolation and a sliding autoregressive model, with the mean absolute error serving as the evaluation metric. The results indicate that the interpolated component data can improve the accuracy of medium-and long-term polar motion forecasts. In the PMX direction, the Z1 interpolated component over 10 years achieves improvement rates of 6.0% for short-term (30 days) and 5.5% for medium-term (180 days) forecasts. In the PMY direction, the Z1 interpolated component over 8 years results in MAE improvement rates of 5.5%, 4.8%, and 3.2% for forecast periods of 30, 60, and 90 days, respectively.

Recent advances in statistical and machine learning (ML) methods have improved the prediction of soil attributes at fine spatial scales, yet the comparative performance and reliability of these techniques remain unclear. This study compared Ordinary Kriging (OK), Inverse Distance Weighting (IDW), and ML algorithms in predicting and spatializing soil attributes, while also evaluating prediction uncertainty and computational processing time. Conducted in Minas Gerais State (Brazil), the analysis used Euclidean distance based predictors derived from X-Y coordinates and regular grids with 5, 7, and 10 divisions. Soil attribute maps (CEC, phosphorus, sand, and clay) were generated using OK, IDW, Random Forest (RF), Cubist, Support Vector Machine (SVM), and Earth. Model performance was assessed using R2, RMSE, MAE, and the coefficient of variation. IDW and OK showed the lowest predictive accuracy (R2 = 0.52–0.58), whereas ML methods, especially RF and SVM achieved superior performance (R2 = 0.62–0.70). Among ML algorithms, Earth performed worst, while RF produced the highest accuracy for all attributes except sand, for which SVM performed best. Processing time was shortest for IDW, followed by OK; among ML models, Earth was fastest, followed by RF, SVM, and Cubist. Larger regular grids improved ML prediction and spatialization but increased computational cost. ML methods thus outperform traditional geostatistical interpolators, benefiting from the use of numerous covariates and flexible algorithmic structures, although requiring greater computational time. These findings demonstrate the robustness and practical potential of ML approaches for soil attribute mapping.

Abstract. Gold mining plays a central role in Ghana’s national economy. However, conventional exploration approaches, particularly within artisanal and small‑scale mining sectors, often rely on trial‑and‑error methods that lead to extensive environmental degradation. The absence of systematic geological exploration before mining has contributed to deforestation, soil contamination, and landscape disturbance in many gold-bearing regions. This study introduces an eco-friendly and cost‑effective remote sensing-based approach, referred to as Green Gold Exploration, for identifying potential gold-rich zones before field excavation. Using Sentinel-2 surface reflectance imagery processed on the Google Earth Engine (GEE) platform, iron oxide and clay mineral spectral indices were derived to detect hydrothermal alteration features commonly associated with gold mineralization. The Wasa Amenfi District in the Western Region of Ghana, a historically active gold‑producing area, was selected as the study area. Field validation was conducted using approximately 2,000 soil samples collected at 40 cm depth within a 1 km2 sampling grid and analyzed using Inductively Coupled Plasma Mass Spectrometry (ICP-MS) to determine gold concentrations in the soil. Spatial interpolation of laboratory results was performed using Inverse Distance Weighting (IDW). Results demonstrate spatial correspondence between high index values and elevated gold concentrations, with most confirmed gold occurrences located within 1 km of identified alteration zones. The findings confirm the potential of satellite-based spectral analysis as a sustainable pre-exploration tool capable of reducing environmental impacts, lowering exploration costs, and supporting informed decision-making before mining activities.

Dehydrated alcohol and lauromacrogol are the predominant sclerosing agents employed in cyst ablation. A clear determination of the therapeutic effects comparison between the two is still lacking. This study evaluates the therapeutic efficacy of dehydrated alcohol versus lauromacrogol sclerotherapy in the management of simple renal cysts (SRCs). Descriptive statistics for baseline characteristics and outcomes were gathered for each group, covering clinical data such as sex, age, location, position, baseline and post-sclerotherapy volumes, percentage, and level of disappearance, and adverse reaction (AR). The primary analysis used a propensity score model with covariates like age, cyst location, position, and baseline volume. We also utilized the relatively novel statistical methodologies of propensity score (PS)-based inverse probability weighting (IPW) and overlap weight (OW) adjustments to compare outcomes between the dehydrated alcohol and lauromacrogol groups. Patients were stratified according to baseline characteristics, and there was no significant difference between the two groups. After application of IPW and OW, all these baseline differences were well balanced between the two groups (i.e., standardized differences < 10%). Overall, there was no significant difference in cyst disappearance and level of disappearance in post-sclerotherapy follow-up between the two groups with IPW analysis (p = 0.61, p = 0.43). There was no significant difference in cyst disappearance post-sclerotherapy follow-up between the two groups with OW analysis (p = 0.57) and in the level of disappearance post-sclerotherapy between the two groups with OW analysis (p = 0.50). Our study found that dehydrated alcohol has a sclerosing effect on simple renal cysts similar to lauromacrogol, based on regression rates. Compared to open surgery, this method reduces tissue damage and speeds up recovery, providing surgeons with better decision-making information.

CBFI: A multi-layer gene network key node identification algorithm integrating structural-biological features and dynamic regulatory weights.

Proportional, fair scheduling in OFDM-based vehicle-to-everything (V2X) uplink causes the resource-block allocation of each vehicle to vary from slot to slot, yet conventional semantic encoders produce a fixed number of output tokens regardless of the instantaneous channel capacity. When the encoder output exceeds the slot budget, transmitted features are truncated and the resulting federated learning gradient is corrupted—a problem that affected 23% of training rounds for non-line-of-sight vehicles in our experiments. The difficulty is worsened by a spatial pattern common in urban deployments: vehicles at congested intersections suffer the poorest propagation conditions while carrying the training data most relevant to safety, and throughput-driven client selection excludes them in favor of vehicles with strong channels but uninformative scenes. We address both issues within a single framework for OFDM-based V2X federated learning. On the transmission side, a Sensing-Guided Adaptive Modulation (SGAM) module derives a per-slot token budget from the current resource-block allocation and selects tokens through differentiable Gumbel-TopK pruning with a hard capacity clip, so the transmitted token count stays within the slot budget. On the scheduling side, a Channel-Decoupled Federated Learning (CDFL) module partitions clients independently by channel quality and data complexity, selects diverse representatives per partition via facility location optimization, and corrects for partition-size imbalance through inverse propensity weighting during model aggregation. Experiments on NuScenes with 20 non-IID vehicular clients under realistic OFDM channel simulation demonstrate a Macro-F1 of 0.710 (+8.7 points over the Oort-adapted baseline), zero budget violations throughout training, and a 75% reduction in training variance; the worst-class F1 more than doubles relative to FedAvg.

This study presents a spatially explicit, seasonally resolved analysis of the intra-urban thermal heterogeneity in Chennai, a rapidly urbanizing tropical megacity along India's southeast coast. Leveraging mobile environmental surveys across 81 georeferenced sites spanning six land-use zones, data on temperature, humidity, PM2.5, PM10, CO₂, and formaldehyde were collected during nighttime in summer and winter seasons. Thermal comfort was assessed using the thermal humidity index (THI), while spatial variability was visualized using GIS-based heat maps and inverse distance weighting (IDW) interpolation. Results revealed a pronounced summer intra-urban thermal contrast, with air temperatures in urban cores exceeding 32.5°C compared to 31°C or lower in vegetated suburban zones. In winter, central hotspots remained elevated at ~ 28.9°C relative to peripheral regions (~ 25-26°C). PM2.5 concentrations were significantly higher in summer (p = 0.00082), reflecting enhanced photochemical activity and dust resuspension under drier conditions. CO₂ showed a moderate positive correlation with temperature (R2 = 0.096, p = 0.0052), suggesting a potential climate-pollution feedback linked to anthropogenic heat emissions and increased energy demand. Analysis of thermal comfort revealed that 63% of surveyed sites were in the "torrid" discomfort category during summer, while the remaining 37% were "very hot." Even in winter, 98% of sites were classified as "hot," indicating persistent nocturnal thermal stress across the city. PCA indicated that temperature and pollution gradients jointly shaped the spatial clustering of intra-urban thermal hotspots, particularly in industrial and commercial zones. The study emphasizes the compounded impact of heat and pollution in shaping Chennai's urban microclimates and highlights the need for climate-sensitive planning, urban greening, and adaptive infrastructure for tropical coastal Indian cities.

Rainfall regionalization plays an essential role in identifying homogeneous rainfall patterns and supporting hydrological and climate analyses. In Thailand, the regionalization adopted by the Thai Meteorological Department (TMD) is primarily based on monsoon wind systems and broad geographic boundaries, which may not adequately represent sub-regional variability in rainfall behavior. This study proposes a data-driven framework to identify homogeneous rainfall regions using monthly rainfall observations from 67 stations across Thailand over the period 1983-2018. Two representations are examined: a standardized representation and a principal component analysis (PCA)-based standardized representation. K-means clustering is applied to both representations, and the resulting clusters are then evaluated using L-moment homogeneity testing, principal component visualization, and statistical and spatial validity indices. The PCA-standardized dataset produces clusters with improved separation and stronger homogeneity relative to the standardized dataset, and the resulting rainfall regions are further interpreted in terms of their physical rainfall characteristics. The practical relevance of the identified regions is further demonstrated through leave-one-out cross-validation comparing inverse distance weighting (IDW), K-nearest-neighbor IDW, and cluster-based IDW approaches using the proposed regions, existing data-driven regions, and TMD regions. The proposed cluster-based IDW approach achieves interpolation error reductions of approximately 9.18-11.55% compared with conventional IDW- and TMD-based alternatives, while providing performance comparable to that obtained using other recent data-driven regional classifications.

Double potential gradient analysis for critical isthmus detection in scar-related atrial tachycardia.

The large-scale nonferrous metal smelting activities involved in industrialization processes have raised serious environmental concerns, particularly heavy metals (HMs) pollution in soil, which poses significant threats to ecosystems and human health. This study comprehensively examined the horizontal and vertical spatial distribution patterns of HMs in soils at the 0-100 cm depth in the historical slag storage area of Laibin Chemical Plant, Guangxi Province. The site is representative of HM pollution resulting from long-term inadequately managed smelting activities in slag accumulation areas in southern China. The inverse distance weighting (IDW) interpolation, geostatistical techniques (Voronoi method, Moran's I), and Pearson correlation analysis were integrated to quantify the horizontal and vertical spatial distribution, variability, and migration characteristics of the target HMs. The spatial analysis of HMs (and metalloids) shows that soil quality is severely threatened by Cd, Cr, Pb, Zn, As, and Ni. Cd, Pb, Zn, As are the main pollutants. All HMs (and metalloids) exhibit marked spatial heterogeneity and a "patchy aggregation" pattern due to intense human and industrial activities. The vertical movement of HMs suggests pollutants are primarily concentrated in the 0-10 cm soil layer. Different contaminants had various migration depths, with migration mobility ranking as follows: As > Zn > Cd > Pb > Cr > Ni. Spatial variability analysis reveals that Zn, Pb, and As have substantial local spatial variation. Their average vertical variation coefficients are 0.55, 0.884, and 0.607, respectively. This study introduces a novel method that combines horizontal-vertical spatial heterogeneity, variability, and migration depth quantification, differing from conventional approaches that typically focus on single-dimensional distribution-providing a comprehensive analysis of HM (and metalloid) pollution in nonferrous metal smelting sites and supporting targeted layered remediation and precise hotspot pollution control.

To investigate the characteristics and health risks of heavy metal (HM) contamination in soils of typical industrial sites during urban renewal, this study selected Pudong New District, Shanghai, as a case. Seven HMs (Cd, Pb, Cu, Zn, Ni, Hg, and As) were analyzed for their concentrations, ecological risks, spatial patterns, and potential sources. Inverse Distance Weighted (IDW) interpolation was used to assess spatial distribution, Random Forest (RF) regression to predict HM concentrations, and a two-dimensional Monte Carlo simulation to evaluate human health risks. The results showed that all HMs except As exceeded Shanghai background values in surface soils, with varying levels observed in subsoil and saturated layers. The Index of Geoaccumulation (Igeo) and Risk Index (RI) indicated low contamination and moderate ecological risk. Pearson correlation combined with Positive Matrix Factorization (PMF) identified four major sources: traffic emissions dominated by Cd and Zn, combustion-related sources dominated by Pb and Hg, industry-related inputs dominated by Cu and Ni, and a natural source dominated by As. The RF model demonstrated strong predictive accuracy for Cd, Pb, Hg, and As (R2 = 0.80-0.94), and predicted values were consistent with observations. Monte Carlo results showed that non-carcinogenic risks for children and adults were within acceptable limits, while carcinogenic risks reached "notable" levels with probabilities of 62.06%, 55.65%, and 22.49% for children, adult females, and adult males, respectively. Cd and As were identified as key contributors. This work provides scientific support for soil pollution prevention and remediation during urban renewal.

ABSTRACT As important carriers of heritage activation, traditional villages contain cultural genes of the human-nature relationship and integrate tangible and intangible cultural heritage, whose systematic heritage corridor construction is significant for rural revitalization. Taking 792 traditional villages in Guangxi as the research object, this study uses the optimal parameter geographical detector to identify the impact of geological disasters and other factors on village spatial distribution, supplemented by kernel density estimation and Inverse Distance Weighting (IDW) interpolation to analyze provincial cultural concentration. Circuit Theory is introduced to construct a traditional village heritage corridor network, with analyses of accessibility and cultural cohesion. The results show that land cover is the dominant factor affecting village distribution, with earthquakes and other geological disasters having significant complex impacts under multi-factor interaction. The corridor network presents a “one axis, two rings, multiple points” spatial pattern; corridors under constraints constructed based on Circuit Theory have strong spatial accessibility in high cultural concentration areas. The identified theme-based tour routes connect networks with different degrees of cultural cohesion, and constructing health preservation, experience and study tours can promote coordinated traditional village protection, tourism development and cultural communication, providing spatial path support for the rural revitalization strategy.

Employing K-means clustering to reconstruct missing water surface elevations in LiDAR digital elevation models for hydrodynamic simulation.

Abstract Digital Elevation Models (DEMs) are fundamental to engineering projects, influencing the accuracy of hydrologic modeling, earthwork calculations, and infrastructure design. The resolution and quality of a DEM are primarily determined by the density of survey points and the interpolation algorithm used. This study presents a comparative evaluation of four common interpolation techniques—Natural Neighbor (NN), Kriging, Inverse Distance Weighting (IDW), and Spline—to generate a high-accuracy local DEM for a bare land area, typical of civil engineering project sites, on the Najran University campus, Saudi Arabia. A total of 7,026 high-precision GPS points were collected and divided into training (80%) and validation (20%) datasets. The vertical accuracy was assessed using Root Mean Square Error (RMSE) and the coefficient of determination (R 2 ). The results demonstrated that the Natural Neighbor interpolation method achieved superior performance with the lowest RMSE of 0.124 m and the highest R 2 of 0.969. Critically, the study evaluated the impact of data density by thinning the training dataset by 0% to 75%. It was found that a 75% reduction in data points—which equates to a significant saving in surveying time and cost—increased the RMSE by only ~ 2 cm when using the NN algorithm. This finding indicates that the Natural Neighbor method is not only the most accurate but also the most robust and cost-effective solution for generating reliable DEMs. The outcomes of this research provide a practical framework for engineers to optimize surveying efforts and produce high-fidelity terrain models essential for precise earthwork volume calculation, drainage design, and flood risk assessment in local-scale projects.

Oligohydramnios is a clinically relevant but understudied pregnancy complication. This study evaluated the association between maternal exposure to mixed ambient air pollutants and the risk of oligohydramnios, focusing on identifying critical exposure windows and pollutant-specific concentration thresholds. We conducted a retrospective cohort study of 7,608 singleton live births from a tertiary hospital in northwestern China (2015-2019). Individual-level air pollution exposure was estimated by inverse distance weighting. Weighted quantile sum (WQS) and lagged WQS (lWQS) models were used to assess mixture effects and time-specific susceptibility. Restricted cubic spline models were applied to estimate concentration-response relationships and preventive thresholds of representative weeks and corresponding key pollutants. The WQS index showed a significant joint effect for daily average exposure during whole pregnancy (odds ratio = 1.204, 95% confidene interval 1.049, 1.285), mainly driven by NO2 and O3. The lWQS model identified the early and late pregnancy as critical exposure windows. As representative time points for early, mid, and late pregnancy, estimated O3 thresholds were 49.28 μg/m3 (week 4), 36.28 μg/m3 (week 16), and 37.40 μg/m3 (week 32); the NO2 threshold at week 32 was 37.41 μg/m3. Maternal exposure to mixed air pollutants, particularly O3 and NO2, increases the risk of oligohydramnios. Findings highlight gestational timing and pollutant-specific targets for prenatal environmental protection.

Multi-source precipitation fusion for hydrological models: Correction and metrics importance analysis

This study has explored the effectiveness of machine learning methods for high-precision spatial interpolation of meteorological variables, aiming to provide accurate atmospheric delay corrections for high-precision edge and corner nets observation in complex-terrain environments such as the Xiluodu Hydropower Station, thereby enhancing the accuracy of deformation monitoring. Considering the significant limitations of traditional interpolation methods such as Inverse Distance Weighting (IDW) and Ordinary Kriging (OK) in capturing spatial variability under complex topographic conditions, we systematically introduced machine learning algorithms including Random Forest (RF)and eXtreme Gradient Boosting (XGBoost, XGB) to compare their performance with traditional methods for high-density interpolation of sparsely distributed temperature, relative humidity, and surface pressure, respectively. Concurrently, we proposed an enhanced XGB model incorporating center-point features (XGB-C) which frames spatial interpolation as a supervised learning problem that learns physical mapping from synoptic backgrounds to local microclimates instead of relying on geometric distances alone. The interpolation performance indices (RMSE, MAE, and R2) were evaluated with daily meteorological observations from 47 stations (38 for training, 9 for testing) during 2023-2024. Results demonstrate that machine learning methods significantly outperform traditional approaches, with XGB-C achieving the highest accuracy (R2 ≈ 1.00 for pressure, 0.97 for humidity, 0.83 for temperature). Moreover, the interpolation performance also exhibits a dependence on seasons and the station location. Greater challenges are shown in the summer season and in the "Urban and Built-Up" and "Croplands" areas. These findings highlight the substantial advantages of machine learning, particularly the proposed XGB-C, for meteorological interpolation in mountainous hydropower station environments where accurate atmospheric correction is crucial for deformation monitoring. This also lays a solid foundation for developing operational ML-based interpolation models trained with high-quality labels derived from unmanned aerial vehicle (UAV) remote sensing data.

Rice cultivation in the Sultanpur district of Uttar Pradesh is increasingly constrained by soil salinization and nutrient depletion, particularly in alkali soils. This study aimed to assess the nutrient status and physicochemical properties of rice-growing alkali soils in the region. A total of 56 geo-referenced soil samples were collected and analyzed for pH, electrical conductivity (EC), organic carbon, available nitrogen (N), phosphorus (P), potassium (K), and sulphur (S). Spatial distribution patterns were evaluated using Inverse Distable Weighting (IWD). Results revealed that all soils were non-saline (EC &lt; 4 dS m-1), highly alkaline, with 42.85% of samples classified as very strongly alkaline (pH &gt; 9.0). Organic carbon levels were low to moderate (0.1-1.4%; mean 0.549%), while available nitrogen ranged from 12.54 to 301.06 kg ha-1 (mean 85.33 kg ha-1), indicating widespread N deficiency. In contrast, phosphorus and potassium were generally in the medium range, and sulphur was found to be medium to high. The low organic matter and nitrogen levels are likely attributed to intensive cropping, imbalanced fertilization, and high temperatures accelerating nutrient loss. These findings underscore the urgent need for site-specific nutrient management strategies to restore soil fertility and sustain rice productivity in the salt-affected landscapes of eastern Uttar Pradesh.

Soil salinity and nutrient availability are major constraints affecting crop productivity, soil quality, and agroecosystem sustainability, particularly in coastal regions vulnerable to seawater intrusion. This study provides a comprehensive spatial and temporal assessment of soil properties and quality dynamics on Hainan Island by integrating field observations and multi-temporal remote sensing (RS) datasets. In 2024, a total of 152 sampling sites were surveyed, with three topsoil soil samples collected at each location. Multi-year RS data (2024–2021), including soil salinity reflectance indices (SRSI and SI), the Normalized Difference Vegetation Index (NDVI), and land use and land cover (LULC), were analyzed to evaluate temporal and spatial variability. The soil fertility index was calculated using alkali-hydrolyzed nitrogen (AN), available phosphorus (AP), available potassium (AK), soil pH, and soil organic matter (SOM). The soil quality index was calculated using the same parameters with the addition of chromium (Cr) to account for potential heavy metal contamination. Furthermore, in this study the Inverse Distance Weighting (IDW) method was used for spatial distribution maps of soil properties and other indices. The results indicated that soils were predominantly acidic (pH &lt; 6.0) with generally low electrical conductivity (0.01–0.53 mS cm−1) across inland areas, whereas higher salinity levels (2.28–5.31 mS cm−1) were observed in southern and eastern coastal zones, suggesting potential seawater intrusion. Nutrient concentrations ranged from 60.1 to 150 mg kg−1 (AN), 4 to 332 mg kg−1 (AP), and 50.1 to 100 mg kg−1 (AK). NDVI values (0.70–0.94) indicated high vegetation density over most agricultural landscapes. Significant positive correlations were observed between soil EC and the SRSI (r = 0.781) and SI (r = 0.663; p &lt; 0.01), demonstrating the reliability of RS-derived indices for salinity assessment. The integrated indicator-based framework developed in this study provides a scientific basis for precision agriculture, soil health monitoring, and sustainable land management in coastal agroecosystems.