Subregional Model Research Articles

Near-term fire weather forecasting in the Pacific Northwest using 500-hPa map types

Background Near-term forecasts of fire danger based on predicted surface weather and fuel dryness are widely used to support the decisions of wildfire managers. The incorporation of synoptic-scale upper-air patterns into predictive models may provide additional value in operational forecasting. Aims In this study, we assess the impact of synoptic-scale upper-air patterns on the occurrence of large wildfires and widespread fire outbreaks in the US Pacific Northwest. Additionally, we examine how discrete upper-air map types can augment subregional models of wildfire risk. Methods We assess the statistical relationship between synoptic map types, surface weather and wildfire occurrence. Additionally, we compare subregional fire danger models to identify the predictive value contributed by upper-air map types. Key results We find that these map types explain variation in wildfire occurrence not captured by fire danger indices based on surface weather alone, with specific map types associated with significantly higher expected daily ignition counts in half of the subregions. Conclusions We observe that incorporating upper-air map types enhances the explanatory power of subregional fire danger models. Implications Our approach provides value to operational wildfire management and provides a template for how these methods may be implemented in other regions.

Spatial distribution of near-fault landslides along Litang fault zones, eastern Tibetan Plateau

Active fault zones are areas impacted by intense tectonic activity, where geohazards such as landslides occur and their distribution characteristics vary across segments. Using the Litang fault zone in the eastern Tibetan Plateau as an example, this study cataloged landslides within a 10 km range on both sides of the fault zone. The Litang fault zone was divided into four sections based on landform characteristics and the spatial distribution of active faults. A comprehensive analysis was conducted on the relationship between landslide distribution and topography, lithology, and active faults in different sections. Additionally, the study employed a random forest model to compare landslide susceptibility between the whole-region and its sub-regions. The results show that lithology and topography significantly influence the distribution of landslide types. The Upper Triassic sandstone, Slate exhibit more debris flows in areas with slopes ranging from 5° to 35°, while slopes >35° in the Upper Triassic sandstone, Slate are prone to unstable slopes. Rock avalanche occur in Upper Triassic sandstone, Slate, and Quaternary sand and gravel. Active faults play a key role in control ling landslides distribution through distance, hanging wall, locked segment, and direction effects. The number of landslides is positively correlated with the activity of faults. Additionally, the dip angle and motion characteristics of faults have significant influence on the distribution of landslides. The results of landslide susceptibility evaluation show that sub-region models have higher accuracy than the whole-region model. This study provides a new method for exploring the spatial distribution of landslides and can provide scientific reference for reducing landslide disasters in active fault zones.

The impact of green finance on green economy development efficiency: based on panel data of 30 provinces in China

Within the framework of China’s socioeconomic transition from a phase of rapid growth to one of high-quality development, it becomes crucial to focus on advancing the green economy to sustain economic progress. Green finance plays a pivotal role in underpinning green industries and fostering the progression of a green economy. Under the auspices of green finance, social capital is increasingly directed towards industries that prioritize energy efficiency, low carbon emissions, and environmental friendliness, thereby spurring technological innovation and industrial metamorphosis in businesses. In this paper, data envelopment approach (DEA) is used to measure the green economic development efficiency of 30 Chinese provinces from 2008 to 2018 in general and by sub-region. The two-way fixed-effects model (Two-way FE) and difference-in-difference (DID) model are established to empirically analyze the effect of green finance on green economy development efficiency and the influence of green polices on these two, with sub-region models examining the heterogeneity of this impact in eastern, western and central regions. The findings suggest that green finance significantly enhances a positive influence on green economic development efficiency, albeit with regional variations. Furthermore, the implementation of green policies facilitates green finance and green economic development.

A Multi-Scenario analysis method for collapse risks due to initial local failure of building structures

Traditional progressive failure analysis is usually computationally expensive and can only be used to analyze a limited number of cases. The resulting mechanical data on structural components are not suitable for displaying collapses or providing an intuitive reference for emergency management. In addition, numerical models do not incorporate uncertain factors in failure development, which makes it difficult to deduce multiple scenarios of future. This paper proposes a multi-scenario progressive failure analysis method based on a sub-region model for emergency management. Firstly, the material model is augmented with a possibility factor to evaluate the failure of components based on internal forces, and thereby deduce different progressive collapse scenarios. Secondly, the building structure is divided into several sub-regions based on the damage correlation of components, so that the degree of regional collapse can be demonstrated at a more macroscopic scale, which is more suitable for emergency management. By analyzing typical failure scenarios, the collapse evolution rules at the sub-region level can be summarized, which can be used for deducing the results of multiple collapse scenarios rapidly. The proposed method can identify the weak positions of buildings, deduce highly probable progressive damage scenarios, and exhibit macro collapse processes. The multi-scenario analysis results can serve as a reference for collapse safety management in risk pre-assessment and help to reconstruct the collapse process in post-disaster analysis.

Subregion Based Prediction of Residual States in Friction Stir Welding of Dissimilar Metals

Mechanical property changes in friction stir welding can directly affect the rebalance of the stress field in friction stir welding. This means that it reveals a high relevance with the residual states of friction stir welding. Here, we propose a subregion model in which the mechanical property changes are considered to predict the residual states in friction stir welding of dissimilar metals. Results indicate that the accuracy of the predicted distortion can be greatly increased when the different mechanical properties are considered in friction stir welding of 2024-T3 and 6061-T6. The final mechanical property is determined by the mixture of the materials at retreating and advancing sides. The final mechanical property in the stirring zone can be increased to 171 MPa for yield strength and 194 MPa for tensile strength when the strength of the advancing side material is higher. The shrinkage of material in the stirring zone during the cooling stage is the key reason for the formation of the tensile residual stress and the V-shape distortion on the cross-section in the as-weld state.

Integrated modeling of microstructure-property-distortion relationship in friction stir additive manufacturing

A Monte Carlo model is established to simulate grain morphologies and combined with a precipitate evolution model to predict the mechanical properties of different zones in different layers of friction stir additive manufacturing (FSAM). Subsequently, a subregion model is established by considering the different mechanical properties at different locations to predict the residual states in FSAM. Results indicate that equiaxed grains are formed in the stirring zone and that the average grain size can be decreased by increasing the build height. The yield strengths in the stirring and heat-affected zones increase along the build direction and decrease with increasing temperature. The error between the predicted and experimental yield strengths is 3.3%. By considering changes in the mechanical properties in the stirring and heat-affected zones, the distortion error between the experimental test and numerical model can be reduced by 25% compared with that between the experimental test and the conventional model. The predicted residual stresses can be reduced by considering the strength reduction caused by the friction stirring effect. As the number of built layers increases, the maximum distortion in FSAM increases.

ALASKA NATIVE SUCCESSFUL AGING IN NORTHWEST ALASKA: AGING IN A GOOD WAY STARTS WITH YOUR FAMILY

Abstract Alaska Natives (AN) view aging from a holistic perspective. One of the challenges of researching with cultural groups is the lack of data, or research, on culture and aging. This research explored successful aging from an AN perspective. A community-based participatory research (CBPR) model was used to engage participants at every stage of the research process. Semi-structured interviews were conducted with 16 AN men and 25 women. Kleinman’s Explanatory Model of Illness was adapted to gain a sense of the beliefs about aging and to guide the data analysis to establish an AN understanding of successful aging or attaining “Eldership” in Northwest Alaska. The foundation of the Norton Sound southern sub-region Model of Successful Aging is family, which contributed to Elders’ feelings of emotional well-being, ability to engage in their Native Way of life, maintain their physical health, and continue spiritual practices.

Alaska Native successful ageing in Northwest Alaska: how family impacts how one ages in a good way

ABSTRACT Since time immemorial Alaska Natives (AN) lived and aged well, yet today they experience high rates of illness and lower access to care because of colonisation. Aand this research explores successful ageing from an AN perspective or what it means to achieve “Eldership” in the rural Northwest Alaska. A community-based participatory research approach was used to engage participants at every stage of the research process. Semi-structured interviews were conducted with 16 AN men and 25 women and the interviews were professionally transcribed. Kleinman’s explanatory model served as the foundation of the questionnaire to gain a sense of the beliefs about ageing and guide the thematic analysis to establish an AN understanding of successful ageing. The foundation of the Norton Sound southern sub-region Model of Successful Ageing is the reciprocal relationship between Elders and family which enables Elders to access meaningful activities, including Native ways of life, physical health, spirituality, and emotional well-being. Community-based interventions should foster opportunities for Elders to share their Native way of life alongside family and community members, which will enable them to remain physically active, maintain healthy emotional well-being, continue engaging in spiritual practices, and contribute to the health and well-being of families.

Economic Contributions of Forestry Service Providers in Mississippi, USA

We used three comprehensive datasets to assess economic activities at as granular a level as possible for Support Activities for Forestry industry in Mississippi in 2019. The labor market company Emsi® provided the most current employment estimates by 6-digit North American Industrial Classification System (NAICS) codes along with a geographic listing of businesses by NAICS codes that contained employment, earnings, and sales data. Esri® further categorized the 6-digit codes into 8-digits, thus providing a more detailed view of this sub-sector’s businesses. However, since business listings do not comprehensively tabulate all businesses of each classification within a region, we developed a weighting method to estimate output, jobs, and earnings for the following industries using Emsi® data- Foresters Consulting, Government – Forestry Services, and Forest Restoration. A statewide input–output model, along with sub-regional models, were estimated using the IMPLAN® software to identify economic contributions to the state and regional economies. Total annual sales across Mississippi were $84.5 million; sales were greatest in the Central subregion, with the South, North close behind. Consulting Foresters was the largest industry by sales, jobs, and earnings. Total economic contributions were 1,140 jobs and $59.79 million in value added on total sales of $121.99 million. South Mississippi received the greatest regional contributions from Foresters Consulting, while Central Mississippi received the greatest benefits from Government – Forestry Services and Forest Restoration.

Collaborative care: Primary health workforce and service delivery in Western New South Wales-A case study.

ObjectiveTo explore how four small towns in rural New South Wales known as the 4Ts are addressing challenges accessing quality care and sustainable health services through a collaborative approach to workforce planning using the collaborative care framework.DesignDescriptive case study approach.SettingThe collaborative care project was developed as a result of ongoing partnerships between 2 rural Local Health Districts, 2 Primary Health Networks and a non‐governmental health workforce organisation. The collaboration works with 5 subregions each comprising 2 or more rural communities. This paper focuses on the 4Ts subregion.ParticipantsStakeholders of the collaborative design including organisations and the community.InterventionA place‐based approach to co‐designing health services with community in one sub‐region of Western New South Wales.Main outcome measuresA synthesis of field observations and experiences of community and jurisdictional partners in implementation of the 4Ts subregional model. Mapping of implementation processes against the collaborative care framework.ResultsThe collaborative care framework is a useful planning and community engagement tool to build health workforce literacy and to impact on system change at the local level. We identify key elements of effectiveness in establishing the 4Ts model, including the need for coordinated health system planning, better integrating existing resources to deliver services, community engagement, building health workforce literacy and town‐based planning.ConclusionThis study adds to the body of knowledge about how to successfully develop a collaborative primary health care workforce model in practice. The findings demonstrate that the implementation of a collaborative primary health care workforce model using the collaborative care framework can improve service access and quality, which in turn might facilitate workforce sustainability.

A Transfer Learning Method for Meteorological Visibility Estimation Based on Feature Fusion Method

Meteorological visibility is an important meteorological observation indicator to measure the weather transparency which is important for the transport safety. It is a challenging problem to estimate the visibilities accurately from the image characteristics. This paper proposes a transfer learning method for the meteorological visibility estimation based on image feature fusion. Different from the existing methods, the proposed method estimates the visibility based on the data processing and features’ extraction in the selected subregions of the whole image and therefore it had less computation load and higher efficiency. All the database images were gray-averaged firstly for the selection of effective subregions and features extraction. Effective subregions are extracted for static landmark objects which can provide useful information for visibility estimation. Four different feature extraction methods (Densest, ResNet50, Vgg16, and Vgg19) were used for the feature extraction of the subregions. The features extracted by the neural network were then imported into the proposed support vector regression (SVR) regression model, which derives the estimated visibilities of the subregions. Finally, based on the weight fusion of the visibility estimates from the subregion models, an overall comprehensive visibility was estimated for the whole image. Experimental results show that the visibility estimation accuracy is more than 90%. This method can estimate the visibility of the image, with high robustness and effectiveness.

Partition satellite derived bathymetry for coral reefs based on spatial residual information

ABSTRACT Satellite imagery with high spatial and temporal resolutions provides strong support for remote sensing bathymetry. Heterogeneous underwater environments such as complex bottom types represent the main factor affecting accuracy of satellite derived bathymetry. Segmenting imagery can enhance geospatial adaptability and improve the accuracy. Conventionally, satellite imagery can be subdivided based on bottom types confirmed by a priori knowledge gained from in situ observations (e.g. video and photos); however, this process poses challenges concerning some politically sensitive or unreachable areas. Besides, water depth in coral reef areas is generally shallow (depth range: 0 to 5 m) and complex, and it has proved difficult to improve bathymetric mapping accuracy with passive multispectral imagery in such areas using traditional models. With the spatial distribution of the residuals of discrete points to realize regional segmentation, this study developed an effective partition model for satellite derived bathymetry without the need for a priori knowledge of bottom types. WorldView-2 high resolution (2 m) imagery of Dongdao Island (China) was used in the study. The results of a practical application of the model revealed the following. (1) The developed partition bathymetry model reflected underwater topographical changes and features more clearly than a comparative global model, especially in certain areas of complex terrain. (2) The accuracy of the residual partition model (Mean Absolute Error (MAE): 0.96 m, Mean Relative Error (MRE): 11.04%, and Root Mean Square Error (RMSE): 1.36 m) is greater than the global model and the conventional partition model. (3) In different water depth intervals, the residual partition model has always the best accuracy. Subregion models corresponding to the proposed models are more stable and more accurate than those to the conventional partition model. In areas of shallow water (depth: 0 to 5 m), the MRE of a subregion-A model using the residual partition method is 14.09%, i.e. better than the global model (19.45%). The findings demonstrate that the developed model can largely overcome the problem of poor precision in shallow water associated with conventional models.

Subregional statistical shape modelling identifies lesser trochanter size as a possible risk factor for radiographic hip osteoarthritis, a cross-sectional analysis from the Osteoporotic Fractures in Men Study

SummaryObjectiveStatistical shape modelling (SSM) of hip dual-energy X-ray absorptiometry (DXA) scans has identified relationships between hip shape and radiographic hip OA (rHOA). We aimed to further elucidate shape characteristics related to rHOA by focusing on subregions identified from whole-hip shape models.MethodSSM was applied to hip DXAs obtained in the Osteoporotic Fractures in Men Study. Whole-hip shape modes (HSMs) associated with rHOA were combined to form a composite at-risk-shape. Subsequently, subregional HSMs (cam-type and lesser trochanter modes) were built, and associations with rHOA were examined by logistic regression. Subregional HSMs were further characterised, by examining associations with 3D-HSMs derived from concurrent hip CT scans.Results4,098 participants were identified with hip DXAs and radiographs. Composite shapes from whole-hip HSMs revealed that lesser trochanter size and cam-type femoral head are related to rHOA. From sub-regional models, lesser trochanter mode (LTM)1 [OR 0.74; 95%CI 0.63.0.87] and cam-type mode (CTM)3 [OR 1.27; 1.13.1.42] were associated with rHOA, associations being similar to those for whole hip HSMs. 515 MrOS participants had hip DXAs and 3D-HSMs derived from hip CT scans. LTM1 was associated with 3D-HSMs that also represented a larger lesser trochanter [3D-HSM7 (beta (β)-0.23;-0.33,-0.14) and 3D-HSM9 (β0.36; 0.27.0.45)], and CTM3 with 3D-HSMs describing cam morphology [3D-HSM3 (β-0.16;-0.25,-0.07) and 3D-HSM6 (β 0.19; 0.10.0.28)].ConclusionSubregional SSM of hip DXA scans suggested larger lesser trochanter and cam morphology underlie associations between overall hip shape and rHOA. 3D hip modelling suggests our subregional SSMs represent true anatomical variations in hip shape, warranting further investigation.

A Subregional Model of System Dynamics Research on Surface Water Resource Assessment for Paddy Rice Production under Climate Change in the Vietnamese Mekong Delta

Effective water management plays an important role in socioeconomic development in the Vietnamese Mekong Delta (VMD). The impacts of climate change and human activities (that is, domestic consumption and industrial and agricultural activities) vary in different subregions of the delta. In order to provide intersectoral data for determining the significantly impacted subregions of the VMD, the present study simulated interactions between local climatic patterns, human activities, and water resources using a system dynamics modeling (SDM) approach with each subregion as an agent of the developed model. The average rainfall and temperature of 121 subregions in the VMD were collected during 1982–2012, and the future changes of climate by provinces were based on the Representative Concentration Pathways (RCP) scenarios (RCP4.5 and RCP8.5) by the end of 21st century. The assessment was based on the levels of impact of various factors, including (1) water consumption, (2) differences between evapotranspiration and rainfall, and (3) spatial distribution of salinity intrusion over the delta scale. In the coastal areas, as well as the central and upstream areas, water resources were projected to be affected by environmental changes, whereas the former, characterized by the lack of surface freshwater, would be affected at a greater scale during the dry season. Besides, the sea level rise would lead to an increase in negative impacts in the eastern coastal areas, suggesting that water-saving techniques should be applied not only for agriculture, but also for industry and domestic water consumption during the dry season. In addition, the south subregions (that is, the western subregions of the Hau River except for An Giang) were likely to be flooded due to the simulated high rainfall and seasonal rises of sea level during the wet season. Therefore, the alternative forms of settlement and livelihood should be considered toward balance management with changing delta dynamics.

Construction of Path Selection Model for Economic Development of Coastal Rural Leisure Industry

Hou, P. and Wu, J., 2019. Construction of path selection model for economic development of coastal rural leisure industry. In: Guido-Aldana, P.A. and Mulahasan, S. (eds.), Advances in Water Resources and Exploration. Journal of Coastal Research, Special Issue No. 93, pp. 843–849. Coconut Creek (Florida), ISSN 0749-0208.Some achievements have been made in the economic development of coastal rural leisure industry in China, but there is still a certain gap between the overall level and the economic development level of international coastal rural leisure industry. This paper puts forward a path selection model of coastal rural leisure industry economic development based on grid sub-regional feature matching, adopts irregular triangulation network planning method to search the economic development path of coastal rural leisure industry oriented to Internet of things technology, and establishes a subregional feature matching model of coastal rural leisure industry economic development path. The adaptive grid block fusion method is used to plan and design the economic development path of coastal rural leisure industry, and the statistical characteristics of node positioning of coastal rural leisure industry economic development path are extracted. Combined with the relevant fuzzy constraint method, the economic development path selection of coastal rural leisure industry is realized, and the adaptive equilibrium ability of coastal rural leisure industry economic development path selection is improved. The simulation results show that this method has good adaptability and high equilibrium level in the economic development path selection of coastal rural leisure industry, and has good application value in the path selection and planning of coastal rural leisure industry economic development.

Probability of Streamflow Permanence Model (PROSPER): A spatially continuous model of annual streamflow permanence throughout the Pacific Northwest

The U.S. Geological Survey (USGS) has developed the PRObability of Streamflow PERmanence (PROSPER) model, a GIS raster-based empirical model that provides streamflow permanence probabilities (probabilistic predictions) of a stream channel having year-round flow for any unregulated and minimally-impaired stream channel in the Pacific Northwest region, U.S. The model provides annual predictions for 2004–2016 at a 30-m spatial resolution based on monthly or annually updated values of climatic conditions and static physiographic variables associated with the upstream basin. Predictions correspond to any pixel on the channel network consistent with the medium resolution National Hydrography Dataset channel network stream grid. Total annual precipitation and percent forest cover were consistently the most important predictor variables among global and most subregional models, which had error rates between 17 and 22%. Probabilities were converted to wet and dry streamflow permanence classes with an associated confidence. Wet and dry classifications were used to derive descriptors that characterize the statistical and spatial distribution of streamflow permanence in three focal basins. Predicted dry channel segments account for 52–92% of the stream network across the three focal basins; streamflow permanence decreased during climatically drier years. Predictions are publicly available through the USGS StreamStats platform. Results demonstrate the utility of the PROSPER model as a tool for identifying areas that may be resilient or sensitive to drought conditions, allowing for management efforts that target protecting critical reaches. Importantly, PROSPER’s successful predictive performance can be improved with new datasets of streamflow permanence underscoring the importance of field observations.

SUBREGIONAL MODELS OF THE CONFLICT PROCESS: RESEARCH METHODOLOGY

The paper presents the main methodological approaches to the study of sub-regional models of the conflict process in the South of Russia. The author substantiates and emphasizes the importance of scenario comparative method and social modeling, focuses on the use of interdisciplinary scientific tools, as well as the integrated concept of J. Galtung. The article discusses in detail the problems of studying and ensuring the security of the political and economic structure of the sub-region, pays attention to conceptual approaches to determining the degree of stability of its main functional components in crisis situations, as well as in the context of the escalation of geopolitical and socio-economic conflicts, the growth of migration flows and information warfare in the global space. In this study, the authors provide an argument for the causes of the emergence and escalation of the sub-regional conflict process on the basis of the analytical axiology of the multi-ethnic region. In the process of interpretation of the concept of “sub-regional model” the possibilities of scientific modeling and forecasting of scenarios of sub-region development on the basis of functional analysis of structural elements of this community are analyzed. The article also describes the main problems of implementation of the strategy of management of regional conflicts (including the role of regional ethnic elites in the formation of the conflict process), outlines the priorities and principles of this strategy, in particular the development of personnel policy and increasing the structural responsibility of the sub-regional authorities. The authors emphasize that on the basis of the most effective research practices and options for justifying the actualized evolutionary paradigms it is possible to implement the scientific and management model of correction and creation of a new, more equitable reality and to ensure the viability of the social system and favorable prospects for development

Improvements for the Eastern North Pacific ADCIRC Tidal Database (ENPAC15)

This research details the development and validation of the updated Eastern North Pacific (ENPAC) constituent tidal database, referred to as ENPAC15. The database was last updated in 2003 and was developed using the two-dimensional, depth integrated form of the ADvanced CIRCulation coastal hydrodynamic model, ADCIRC. Regional databases, such as ENPAC15, are capable of providing higher resolution near the coast, allowing users to more accurately define tidal forcing for smaller sub-regions. This study follows the same methodology as the EC2015 updates for the eastern coast of the United States and six main areas of improvement in the modeling configurations are examined: (1) placement of the open ocean boundary; (2) higher coastal resolution; (3) updated global bathymetry; (4) updated boundary forcing using two global tidal databases; (5) updated bottom friction formulations; and (6) improved model physics by incorporating the advective terms in ADCIRC. The skill of the improved database is compared to that of its predecessor and is calculated using harmonic data from three sources. Overall, the ENPAC15 database significantly (52% globally) reduces errors in the ENPAC03 database and improves the quality of tidal constituents available for sub-regional models in the ENPAC region.

Investigation of the Relationship Between Susceptibility Loci for Hip Osteoarthritis and Dual X-Ray Absorptiometry-Derived Hip Shape in a Population-Based Cohort of PerimenopausalWomen.

To examine relationships between known osteoarthritis (OA) susceptibility loci and hip shape in a population-based cohort of perimenopausal women in order to investigate whether hip shape contributes to OA development. Hip shape was measured, using statistical shape modeling, on dual x-ray absorptiometry scans of the hip from mothers in the Avon Longitudinal Study of Parents and Children (ALSPAC). The proximal femur and superior acetabulum were outlined, and independent hip shape modes were generated. In a subregional model, points were restricted to the acetabulum and superior femoral head. Associations between 11 OA-related single-nucleotide polymorphisms, identified by literature search, and shape modes were analyzed in a multivariate canonical correlation analysis. A total of 3,111 women (mean age 48 years) had genetic and hip shape data. The KLHDC5/PTHLH rs10492367 OA risk allele was associated with a wider upper femur in the whole shape model (P = 1 × 10-5 ). The DOT1L rs12982744 OA risk allele was associated with reduced superior joint space in the subregional shape model (P = 2 × 10-3 ). The COL11A1 rs4907986 OA risk allele was associated with lateral displacement of the femoral head relative to the acetabulum in the subregional shape model (P = 5 × 10-4 ). Regional association plots identified an additional COL11A1 locus in moderate linkage disequilibrium with rs4907986, which was more strongly associated with hip shape (rs10047217; P = 3 × 10-6 ). Colocalization analysis indicated sharing of genetic signals for hip shape and hip OA for the KLHDC5/PTHLH and COL11A1 loci. Hip OA susceptibility loci were associated with shape in this study, suggesting that these loci (and potentially yet-to-be-identified hip OA loci) could contribute to hip OA in later life via perturbing biologic pathways that mediate morphology development.

An Ensemble Machine-Learning Model To Predict Historical PM2.5 Concentrations in China from Satellite Data.

The long satellite aerosol data record enables assessments of historical PM2.5 level in regions where routine PM2.5 monitoring began only recently. However, most previous models reported decreased prediction accuracy when predicting PM2.5 levels outside the model-training period. In this study, we proposed an ensemble machine learning approach that provided reliable PM2.5 hindcast capabilities. The missing satellite data were first filled by multiple imputation. Then the modeling domain, China, was divided into seven regions using a spatial clustering method to control for unobserved spatial heterogeneity. A set of machine learning models including random forest, generalized additive model, and extreme gradient boosting were trained in each region separately. Finally, a generalized additive ensemble model was developed to combine predictions from different algorithms. The ensemble prediction characterized the spatiotemporal distribution of daily PM2.5 well with the cross-validation (CV) R2 (RMSE) of 0.79 (21 μg/m3). The cluster-based subregion models outperformed national models and improved the CV R2 by ∼0.05. Compared with previous studies, our model provided more accurate out-of-range predictions at the daily level ( R2 = 0.58, RMSE = 29 μg/m3) and monthly level ( R2 = 0.76, RMSE = 16 μg/m3). Our hindcast modeling system allows for the construction of unbiased historical PM2.5 levels.