Spatio-temporal dynamics and drivers of carbon storage in arid ecosystems: Integrated analysis using InVEST and PLUS models with machine learning.

Integrating horizontal-vertical heterogeneity and environmental factors to unravel heavy metal distributions and phytoaccumulation in a soil-maize system.

Women's micronutrient deficiency, also named hidden hunger, remains a serious public health concern of low- and middle-income countries. To reduce micronutrient deficiencies, Sustainable Development Goal was designed to provide universal access to essential health and nutrition services to women of reproductive age. Recent evidence indicates that multiple micronutrient deficiencies often co-occur. But little is known about their extent, geographic variation, and underlying determinants among non-pregnant women of reproductive age, a critical gap that limits targeted interventions. Therefore, mapping the most affected women is critical to tailor interventions targeting needy segments of the population. We analyzed individual record and coordinates data from the Ethiopian National Micronutrient Survey 2015 to examine the spatial distribution of multiple micronutrient deficiencies. Vitamin A, folate, vitamin B12, zinc, iron, and iodine were micronutrients considered in the analysis. Weighted sample of 1450 non-pregnant women aged 15-49 years was included. Global spatial autocorrelation and hotspot analysis were performed to display geographic equity in multiple micronutrient deficiencies. A multi-scale geographical weighted regression analysis was also performed to identify factors explaining geographic disparities in the distribution of multiple micronutrient deficiencies. Two-fifths (41.3%; 95% CI 37.9-44.8) of women were found to have multiple micronutrient deficiencies in Ethiopia. There were geographical variations in the distribution of multiple micronutrient deficiencies among women in Ethiopia (Global Moran's I value of 0.11 (p < 0.0087)). Higher proportions of multiple micronutrient deficiencies were observed in the Eastern parts of the country. Dire Dawa City Administration, Harari, and the Northern, Eastern, and Southwest parts of the Somali region had higher predicted multiple micronutrient deficiency levels. Hotspot areas with significantly higher rates of multiple micronutrient deficiencies were observed in southern Afar, Dire Dawa, Harari, Eastern Oromia, and Southern Somali regions. Multiple micronutrient deficiencies affect a substantial proportion of non-pregnant women of reproductive age in Ethiopia, highlighting a critical public health concern. The findings demonstrate significant regional variation and identify key predictors, including rural residence, breastfeeding status, and lack of multivitamin supplementation. These findings underscore public health efforts must integrate multiple micronutrient supplementation into maternal health services. In addition, designing targeted interventions that can reduce the high burden of hidden hunger and geographic inequity. Furthermore, addressing the micronutrient requirements of breastfeeding women and rural communities are critical step toward reducing multiple micronutrient deficiencies across Ethiopia.

As a critical ecological barrier in eastern China, the Huai River urban agglomeration (HRUA) necessitates research on the spatiotemporal dynamics, drivers, and future scenarios of Habitat Quality (HQ). This study established an integrated "Scenario Simulation-HQ Assessment-Dual perspective Factor Analysis" framework. We integrated Multiscale Geographically Weighted Regression (MGWR) with explainable machine learning to explore the influencing factors, as these two methods are highly complementary. Key findings indicate: (1) The area with median-or-higher HQ decreased by 3% (2000-2020). Multi-scenario simulations (MS) project the 2035 Urban Development Scenario (UDS) will cause the most severe HQ degradation. HQ's overall spatial pattern remained stable, with persistent, significant hot spot clustering and strong habitat resilience in the southwest. (2) Elevation, NDVI, slope, and population density were key factors influencing HQ spatial differentiation, showing significant effect heterogeneity. (3) Urban expansion and cropland loss were the primary land-use drivers of HQ degradation, jointly accounting for 85.7% of land-use change impact and exhibiting a significant synergistic effect. The research framework provides practical value for exploring the changes of HQ in the HRUA, exploring the influencing factors of HQ from multiple perspectives, and carrying out zoning planning.

Son preference remains a key driver of gender inequality in India, yet most studies treat its determinants as uniform across space, obscuring critical subnational variation. Addressing this gap, this study investigates the geographic heterogeneity of son preference and examines how its predictors vary spatially. Data and methods Using district-level data on 102,045 ever married women aged 15-49 from the National Family Health Survey-5 (2019-2021), we applied a spatially explicit analytical framework, including choropleth mapping, Global Moran's I, hotspot analysis, Local Indicators of Spatial Association (LISA) cluster mapping, kriging interpolation, and Geographically Weighted Regression (GWR). The use of GWR was justified by significant spatial non-stationarity (Koenker BP = 32.78, p < 0.001) detected in OLS diagnostics. Son preference prevalence ranged 8.2%-42.3%. Global Moran's I (0.397, z = 57.86, p < 0.001) confirmed significant clustering. OLS identified five significant predictors: parity 2, no mass media exposure, household size 5-8, illiterate mothers, and younger maternal age, explaining 64% variance. GWR demonstrated superior fit (AICc: 4459.98 vs 4478.46; adjusted R2: 0.65 vs 0.64). Local R2 ranged 0.45-0.71, highest in northern/central districts. Women's illiteracy (β: 0.40-0.62), large household size (β: 0.35-0.83), and younger mothers showed strongest associations in northern/central India, but negligible effects in southern regions, confirming spatial heterogeneity. Son preference is a spatially embedded social process, shaped by localized patriarchy, economy, and institutions. Geographically targeted, gender-transformative policies such as conditional incentives for girls' schooling and localized media campaigns effectively address the structural devaluation of daughters.

ABSTRACT Although traditional villages and intangible cultural heritage (ICH) have been widely studied, integrated quantitative evidence of their spatial coordination and directional mismatch remains limited, especially in high-altitude transition regions. This study examines the spatial relationship between traditional villages and ICH in Qinghai Province, China, and proposes an integrated planning framework. Using coupling coordination analysis, a Spatial Mismatch Index (SMI), GeoDetector, and Multiscale Geographically Weighted Regression (MGWR), we assess distribution patterns and associated factors. Results show highly uneven coupling: strong alignment in agro-valley counties but a pronounced ICH-dominant mismatch in pastoral highlands. Accessibility and socio-economic structure are the main associated factors, with spatially varying effects; road-network conditions are strongly associated with the co-location of tangible and intangible heritage. Based on these diagnostics, we propose a rule-based zoning instrument with four policy toolkits: Synergistic Conservation, Ecological-Cultural Corridors, Mobile Heritage, and General Transition zones. The framework provides an operational pathway from settlement-centric protection toward connectivity-sensitive, adaptive landscape governance, and supports county-level zoning, project screening, and implementation prioritization.

This study assessed the spatial distribution of HIV test non-uptake among pregnant women who attended antenatal care (ANC) in sub-Saharan Africa. Cross-sectional study design. Sub-Saharan Africa (SSA) region. 24 SSA countries were included in this study. Demographic and Health Survey (DHS), 2016-2024. 82 397 women who were pregnant in the last 2 years preceding the survey. HIV test non-uptake, which is a legacy indicator of HIV test among pregnant women. The HIV test non-uptake among ANC attending pregnant women was 39.6% (95% CI 39.27% to 39.93%). The spatial autocorrelation test revealed that HIV testing non-uptake among pregnant women was clustered. The global Moran's I value was 0.48 with a p value <0.01. Hotspot areas were those with high rates of HIV testing non-uptake. These hotspot areas were located in most parts of Mali, Mauritania, Madagascar, Guinea, Senegal, central parts of Angola, eastern part of Senegal and northern parts of Ethiopia. The Multiscale Geographical Weighted Regression (MGWR) model explained 91% of the spatial variation of HIV test non-uptake among pregnant women, who were in the age group 15-19 years, had no formal education and no health insurance, and less than four antenatal care contacts were significantly associated with HIV test non-uptake. There was a significant geographical variation in HIV test non-uptake among pregnant women attending antenatal care (ANC) in sub-Saharan Africa. Prioritising hotspot areas with high rates of HIV test non-uptake for spatially targeted interventions is essential. Policymakers, health professionals, and other stakeholders should focus on improving women's formal education, expanding health insurance coverage, and increasing ANC contacts to ensure that each visit includes HIV screening. Moreover, special attention should be given to younger women to enhance HIV testing uptake among those attending ANC in sub-Saharan Africa.

Causal inference in geographical sciences faces the challenge of isolating treatment effects from high-dimensional observational data, complicated by spatial non-stationarity and persistent confounding. Double machine learning (DML) offers a powerful solution for high-dimensional debiasing through orthogonalization and cross-fitting, but traditional variants overlook spatial heterogeneity by treating space as a simple covariate. To address this, we introduce DML-Geo, an ensemble extension of DML for estimating spatially varying causal effects. Retaining the orthogonalization procedure of DML at its first stage, DML-Geo augments the second stage with three complementary estimators, namely a linear regression model for covariate-driven effects, a generalized additive model (GAM) for spatially smoothed additive effects, and geographically weighted regression (GWR) for localized patterns. Robustness is further enhanced by an adaptive weighting scheme based on inter-model correlations to aggregate outputs from these variants, complemented by a bootstrap procedure for significance testing. Extensive simulations confirm DML-Geo’s superior precision and stability relative to its component models and competing baselines. In real-world applications to housing prices and mental health outcomes, DML-Geo uncovers interpretable spatial causal effect patterns, offering place-specific insights to support policy decisions. DML-Geo provides a flexible toolkit for geospatial causal inference that does not require causal graphs or strong structural assumptions.

Liver flukes (Opisthorchis viverrine, OV) infections in water sources continue to persist in Sakon Nakhon Province, which is linked to the Mekong River. The agency's traditional infection data comprises the locations of infected water sources. However, this data is insufficient for developing a predictive model for infections within the sub-basin. When analyzed alongside independent variables, represented as identical points, it lacks the necessary information to generate a trend line that produces a reliable coefficient. This study implemented a spatial model that integrates a geographic-weighted regression (GWR) framework with appropriate weighting as a prototype. This approach improves the selection of independent variables by shifting from a point-based methodology to a weighted hexagonal grid. A set of eight independent variables land use, soil drainage, road network, water sources, streamlines, surface temperature, NDMI (Normalized Difference Moisture Index), and NDVI (Normalized Difference Vegetation Index) was initially weighted. This study developed three linear models based on the Geographically Weighted Regression (GWR) model. It demonstrates the advantages of utilizing a hexagonal grid instead of a point grid. The three alternative models were tested with various independent variables and employed a factor-by-factor averaging approach, which necessitates the hexagonal grid size as a counterweight to ensure fairness across the entire grid, rather than relying solely on point data. A mathematical model was developed to calculate the average of each factor in order to achieve equality across a hexagonal grid area. Subsequently, the correlation was tested, and the alternative models were grouped. The resulting dendrogram includes three models. The results of the GWR comparison test were derived from both infected and hexagonal water source data. Models constructed from hexagonal grids consistently outperformed all alternative models, with R2 values improving to 58.7%, 41.1%, and 53.2% for Model-1, Model-2, and Model-3, respectively. The RMSE also showed significant improvement, decreasing to 77.1%, 60.2%, and 67.2%. Additionally, the model's accuracy was evaluated using AUC values of 0.725, 0.652, and 0.707, indicating that the developed model can effectively predict water source infections. Model-1 emerged as the most representative across all tests, incorporating soil drainage factors and road proximity as key influences on water source infection. Finally, the results are presented as infection prediction maps for each grid, highlighting areas of both overestimation and underestimation. The most accurate prediction model identified that over 95% of grids had a high degree of accuracy. This study is anticipated to be applicable to infections caused by other water-mediated parasites.

Study on spatiotemporal evolution and regional control of financial risks in the Yangtze river delta based on SDM-GWR

This study investigates whether long-term changes in neighborhood racial and ethnic diversity are associated with adult diabetes prevalence across the contiguous United States. We calculated diversity indexes for US census tracts (n = 72,033) using decennial census data from 1990 to 2020. Five diversity trajectory clusters were identified using K-means clustering. Diabetes prevalence in 2019 was obtained from the CDC's PLACES dataset. Linear mixed models (LMMs) assessed global associations between diversity trajectories and diabetes prevalence, adjusting for age, sex, poverty, marital status, and public insurance. Geographically weighted regression (GWR) examined local variations in these associations. Compared to the reference group (low and stable diversity), tracts with increasing or high diversity had significantly lower diabetes prevalence. In fully adjusted LMMs, diversity trajectory clusters characterized by large increases or high diversity by 2020 showed the largest negative associations with diabetes prevalence (e.g., cluster 4: 0.71, 95% CI: 0.76 to -0.65). GWR revealed spatial heterogeneity in these relationships: tracts with significant negative associations were most concentrated in the South, but also appeared in urban areas of the Midwest and Northeast. Neighborhoods with increasing racial and ethnic diversity over time were associated with lower diabetes prevalence, independent of key socioeconomic factors. These findings suggest that historical trajectories of integration may shape present-day diabetes prevalence at the neighborhood level and highlight the value of incorporating demographic change into spatially aware public health strategies.

The built environment is a critical component in shaping urban areas and their resilience. While existing studies frequently discuss how different built environment elements can enhance a city's capacity to respond quickly and effectively to external shocks, limited research has validated these effects through actual disaster event outcomes. This study aims to empirically investigate the relationship between the built environment and urban resilience, focusing on the context of the COVID-19 pandemic in Singapore. Using urban vitality, approximated by public transit passenger data, as an indicator of the resilience process, we quantified urban resilience through three metrics: robustness, recovery degree, and total performance loss, derived from temporal changes in urban vitality. Multiple linear regression (MLR), spatial lag model (SLM), and geographically weighted regression (GWR) were applied to examine how built environment factors relate to these metrics. The results across models indicate that higher residential density, more diverse land use, and greater distance to CBD are positively associated with resilience, whereas greater transit service is associated with lower resilience. Moreover, GWR explains the highest variations in all resilience metrics compared to MLR and SLM. By mapping spatially varying associations, the findings offer insights to support localized and data-driven planning for building resilient urban environments capable of withstanding and adapting to future shocks. • A new approach assesses built environment's impact on urban resilience. • Urban vitality from human mobility effectively captures resilience processes. • Proposed resilience metrics reveal intra-city disparities in resilience. • Higher density and land use mix are linked to better resilience outcomes. • Proximity to CBD and transit service relate to lower resilience during COVID-19.

Improving predictive accuracy for soil cadmium distribution in highly heterogeneous region based on a synergistic dual-stage feature selection strategy.

Using spatially explicit machine learning to enhance assessment of the Global Gravity-based Groundwater Product for groundwater storage change in Germany

A construction of statistical inferences in geographically weighted univariate log-gamma regression\u2606\u2606\u2606

Prostate cancer incidence varies markedly across the United States (U.S.), yet the broad contextual predictors associated with this variation and the spatial scales at which these associations operate remain insufficiently characterized within a unified spatial framework. This study aimed to examine annual state-level associations between prostate cancer incidence and selected contextual predictors across the U.S. from 2018 to 2022 using Multiscale Geographically Weighted Regression (MGWR). Five annual state-level datasets covering the 50 U.S. states were assembled for 2018-2022. After exploratory spatial analysis, correlation screening, collinearity diagnostics, theory-informed re-evaluation, and representative-year sensitivity analyses, six predictors were retained for the final annual models: elevation, rainfall, atmospheric pressure, obesity prevalence, the percentage of males aged 65 years and older among the male population, and the percentage of the Black or African American alone population. Ordinary Least Squares (OLS), Geographically Weighted Regression (GWR), and MGWR were fitted separately for each annual dataset. Model performance was compared using R², adjusted R², residual sum of squares, corrected Akaike information criterion (AICc), bandwidth diagnostics, and residual spatial autocorrelation. Across all five annual analyses, MGWR showed the most favorable overall performance, with adjusted R² values ranging from 0.555 to 0.618 and AICc values ranging from 124.934 to 132.779. MGWR also identified variable-specific adaptive bandwidths ranging from 24 to 40 nearest neighbors, indicating spatial scale heterogeneity among predictors. Local coefficient surfaces showed clear spatial non-stationarity in both coefficient magnitude and, for some predictors, coefficient direction. Residual Moran's I tests were non-significant for all five annual models, suggesting that substantial residual spatial autocorrelation had been reduced after model fitting. Multivariate clustering of standardized local coefficients further revealed recurring regional groupings in coefficient-pattern similarity. At the state level in the U.S., prostate cancer incidence was associated with geographically differentiated contextual patterns rather than a single spatially uniform relationship structure. In this setting, the main value of MGWR lay in its ability to characterize multi-scale and regionally varying association patterns across repeated annual spatial analyses. These findings may support geographically differentiated surveillance and future hypothesis generation while underscoring the importance of considering spatial heterogeneity in ecological studies of prostate cancer incidence.

ABSTRACT Synergistic governance of air pollution control (AP) and carbon mitigation (CM) is a pivotal strategy for China. This study evaluates provincial-level AP-CM synergistic governance performance from 2013 to 2020, using an indicator evaluation system and a coupling coordination degree model. To understand the spatially varying relationships underlying these dynamics, it compares three models’ performance, including OLS (Ordinary Least Squares), GWR (Geographically Weighted Regression), and MGWR (Multi-scale Geographically Weighted Regression). The results indicate that China’s synergistic governance performance improved significantly over the study period, primarily driven by stringent environmental policies and enhanced coordination mechanisms. Spatially, a distinct gradient was observed, where the eastern, central, and southern regions demonstrated higher synergy levels, while western regions exhibited lower levels. This evolution was shaped by the varying performances of the individual AP and CM subsystems across provinces and time periods. Moreover, driving factors such as economic development, energy utilization, the industrial production structure, and public travel modes significantly influenced the synergistic governance performance. The MGWR analysis revealed substantial spatial heterogeneity in these effects across China. The findings of this paper underscore the importance of developing locally tailored strategies that address specific contexts and spatially varying driver impacts, rather than applying a “one-size-fits-all” strategy.

Taiwan is one of the fastest-warming regions globally. As climate change intensifies, understanding how vulnerability influences health outcomes critical. This study aimed to identify regional vulnerability factors for temperature-related respiratory mortality and effective region-specific adaptation policies. A two-stage time-series study was conducted using daily respiratory mortality counts aggregated by county and day. This study employed a distributed lag non-linear model to estimate the temperature-attributable mortality burden from respiratory diseases across all counties and cities in Taiwan. A two-stage meta-analysis was conducted to estimate temperature-mortality associations and quantify cold- and heat-related mortality burdens by county. Meta-regression was used to identify regional vulnerability factors modifying temperature-related mortality risk, and geographically weighted regression (GWR) was applied to characterize the spatial heterogeneity of these effects across counties. Cold exposure was linked to a higher burden of respiratory disease mortality (attributable fraction [AF]: 2.03%, 95% CI: 1.10-2.95) than heat exposure (AF: 1.02%, 95% CI: 0.65-1.40). For cold-related AFs, higher proportions of Indigenous populations (3.27, 0.79-5.75), low-income populations (2.11, 0.67-3.55), greater population density (2.21, 0.46-3.96), and children (0.98, 0.35-1.61) were significantly associated with increased risk, suggesting vulnerability factors. GWR further showed that hospital bed availability had statistically significant protective effects against cold-related AF in 10 of 19 counties (β = - 5.24 to - 6.78), most pronounced in remote mountainous counties (Hualien, Taitung, Kaohsiung). Higher proportions of Indigenous populations, low-income population, and children amplify cold-related respiratory mortality. Hospital bed availability confers the strongest protection against cold-related mortality in remote, mountainous counties. Climate adaptation policies for cold-related respiratory health should therefore be tailored to local vulnerability profiles, prioritizing healthcare expansion in geographically remote counties rather than applying uniform investment across all regions.

ABSTRACT Changes in climatic characteristics increasingly intersect with socio-economic inequalities, producing spatially heterogeneous vulnerabilities that remain insufficiently explored at the subnational scale. This study examines the relationship between multiple deprivation and climatic and ecological variables – including land surface temperature (LST), precipitation, and vegetation cover (NDVI) – across 49 counties in northwestern Iran. Composite indices of deprivation were developed using exploratory factor analysis (EFA), while geographically weighted regression (GWR) was applied to capture spatial heterogeneity in deprivation–climate linkages. The results reveal strong spatial unevenness: education and housing deprivation emerged as the most climate-sensitive dimensions, whereas employment and accessibility displayed weak or non-significant associations. Critical hotspots such as Parsabad, Germi, Bileh Savar, and Ardabil were identified, where socio-economic disadvantages and climatic stressors converge most strongly. Building on earlier works that treated deprivation and climate independently, this research employs a spatial integration approach to better understand their intersection. Theoretically, this study reframes vulnerability to climatic conditions through a socio-spatial lens, emphasising the interplay between ecological stress and structural inequalities in semi-arid and mountainous contexts. Methodologically, it demonstrates the added value of spatially explicit models such as GWR over conventional global approaches, particularly for revealing localised patterns of vulnerability. Overall, the evidence shows that changes in climatic characteristics not only amplify existing socio-economic inequalities but also generate highly localised deprivation–climate interactions, demanding equitable and place-based policy responses.

ABSTRACT Cloud cover (CC) significantly impacts the effectiveness of optical remote sensing in Earth observation, particularly in tropical areas. However, pixel-level analyses of CC dynamics and their underlying mechanisms remain inadequate. Here, we examine pixel-based CC patterns in Mainland Southeast Asia (MSEA) based on 281,124 Sentinel-2 cloud scenes during 2017–2023. By integrating correlation analysis, Geographically Weighted Regression (GWR), and Geographical Detector (GD) methods, we then investigated the relationships between CC (246 billion pixels in total) and 20 environmental factors, including vegetation conditions (e.g. NDVI), underlying surfaces (e.g. land cover), and meteorological parameters (e.g. wind). Main conclusions include: (1) The average CC probability of Sentinel-2 pixels across MSEA reaches 0.54 (±0.19), exhibiting a decreasing gradient from northeast to southwest and from coastal zones to inland areas. (2) Probability of CC shows significant national and temporal variations across MSEA, with the lowest values observed in Myanmar of 0.48 (±0.08) and in February of 0.31 (±0.22). (3) Elevation and coastal proximity are the primary drivers of CC, outweighing the influence of land cover. We identified three spatial paradigms of CC variation: ocean-driven (coastal areas), topography-anchored (mountainous areas), and energy-convection (inland areas). Our findings provide practical guidance for optimizing satellite data acquisition and improving remote sensing applications in cloudy and rainy regions.