Air Quality Prediction Research Articles

An enhanced interval-valued PM2.5 concentration forecasting model with attention-based feature extraction and self-adaptive combination technology

Accurate predictions of air quality enable governments and relevant authorities to take promptly measures for protecting public health. With the increasing time-varying nature of air pollutants, predicting daily average concentrations alone is no longer sufficient for environmental management and risk warning. Hence, this paper proposes a multi-resolution interval-valued PM2.5 concentration combination prediction model, which based on interval decomposition and attention mechanism reconstruction. Firstly, the interval-valued time series (ITS) was decomposed and adaptively reconstructed using the binary empirical mode decomposition (BEMD) algorithm and attention-based reconstruction. Subsequently, multi-resolution linear projection layers were applied to extract temporal features from the time series. Finally, a hybrid prediction module was implemented that combines CNN and LSTM to predict each subsequence and integrate them to derive the final interval prediction values for PM2.5. In the proposed framework, the reconstruction technique effectively resolved the issue of inconsistent numbers of different feature decomposition subsequences, while the linear projection layer fully captured the multi-resolution characteristics of the time series. Empirical studies conducted in three districts of Beijing showed that, compared to state-of-the-art baseline models, the framework reduced the average values of five interval evaluation metrics by 11.2%, 17.4%, 11.7%, 10.5%, and 14.8%, respectively. This interval-valued prediction framework can effectively assist urban air quality management and warning.

Regional PM2.5 prediction with hybrid directed graph neural networks and Spatio-temporal fusion of meteorological factors

Traditional statistical prediction methods on PM2.5 often focus on a single temporal or spatial dimension, with limited consideration for regional transport interactions among adjacent cities. To address this limitation, we propose a hybrid directed graph neural network method based on deep learning, which utilizes domain features to quantify the influence of neighboring cities and construct a directed graph. The model comprises a historical feature extraction module and a future transmission prediction module, and each module integrates a Graph Neural Network (GNN) and a Long Short-Term Memory Network (LSTM) for spatiotemporal encoding. Compared to other neural network models, our model improves the prediction accuracy of PM2.5 concentration and demonstrates superior performance for 48-h prediction in the North China Plain. For 3- to 48-h prediction tasks, the proposed model achieves mean absolute error (MAE) at 7.64 − 14.04 μg/m3. In addition, by expanding the modeling scope from different directions and integrating domain information, the model significantly enhances its ability to predict PM2.5 trends, seasonal variations, and PM2.5 exceedances in heavily polluted urban areas. The proposed model represents a promising advancement in optimizing air quality forecasting and management.

Current Situation and Prospect of Geospatial AI in Air Pollution Prediction

Faced with increasingly serious environmental problems, scientists have conducted extensive research, among which the importance of air quality prediction is becoming increasingly prominent. This article briefly reviews the utilization of geographic artificial intelligence (AI) in air pollution. Firstly, this paper conducts a literature metrology analysis on the research of geographical AI used in air pollution. That is, 607 documents are retrieved from the Web of Science (WOS) using appropriate keywords, and literature metrology analysis is conducted using Citespace to summarize research hotspots and frontier countries in this field. Among them, China plays a constructive role in the fields of geographic AI and air quality research. The data characteristics of Earth science and the direction of AI utilization in the field of Earth science were proposed. It then quickly expanded to investigate and research air pollution. In addition, based on summarizing the current status of Artificial Neural Network (ANN), Recurrent Neural Network (RNN), and hybrid neural network models in predicting air quality (mainly PM2.5), this article also proposes areas for improvement. Finally, this article proposes prospects for future research in this field. This study aims to summarize the development trends and research hotspots of the utilization of geographic AI in the prediction of air quality, as well as prediction methods, to provide direction for future research.

Predicting air quality index using machine learning: a case study of the Himalayan city of Dehradun

Predicting air quality index using machine learning: a case study of the Himalayan city of Dehradun

Evaluation of machine learning and deep learning models for daily air quality index prediction in Delhi city, India.

The air quality index (AQI), based on criteria for air contaminants, is defined to provide a shared vision of air quality. As air pollution continues to rise in global cities due to urbanization and climate change, air pollution monitoring and forecasting models for effective air quality monitoring that gather and forecast information about air pollution concentration are essential in every city. Air quality predictions have evolved to be more helpful for management. Recently, better performance and ability have developed due to the involvement of machine learning (ML) and artificial intelligence (AI) in forecasting air quality in urban cities in India. This paper focuses on air pollution as a significant ecological problem that directly impacts human health and the distribution of an environmental system in urban areas. Hence, we have developed advanced models for daily AQI forecasting to understand the air effluence level in the upcoming days. In this research, six data-driven models have been developed and implemented for daily AQI forecasting in the study area; it is crucial for understanding the future air pollution levels to plan and control air pollution in the entire city. The developed model is applied to air quality datasets. A comparison of the performance of ML models tested here indicates that the XGBoost algorithm achieves the highest coefficient of determination (R2) and root-mean-square deviation (RMSE) value of 0.99 and lower values value of 4.65 than other models in the testing phase. The results of the artificial neural network (ANN) algorithm are slightly lower than the extreme gradient boosting (XGBoost model); the ANN model results are as R2, mean squared error (MSE), and RMSE values of 0.99, 13.99, and 198.88, respectively. All the models were subjected to a ten-fold cross-validation model. However, the RF cross-validation model outperforms other models; the RF model result shows the R2, RMSE, and MSE values of 0.99, 3.64, and 4.12, respectively. This study also employed two interpretable models, namely feature importance analysis and Shapley additive explanation (SHAP), to evaluate both the global and local methods in a manner that is independent of specific ML models. The feature importance shows that particle matter (PM) 2.5, PM10, carbon monoxide (CO), and nitrogen oxides (NOx) were the most influential variables. The results determined that such novel DL and ML models may improve the accuracy of AQI forecasts and understanding of air pollution, particularly in metropolitan cities.

Artificial intelligence in environmental monitoring: in-depth analysis

This study provides a comprehensive bibliometric and in-depth analysis of artificial intelligence (AI) and machine learning (ML) applications in environmental monitoring, based on 4762 publications from 1991 to 2024. The research highlights a notable increase in publications and citations since 2010, with China, the United States, and India emerging as leading contributors. Key areas of research include air and water quality monitoring, climate change modeling, biodiversity assessment, and disaster management. The integration of AI with emerging technologies, such as the Internet of Things (IoT) and remote sensing, has significantly expanded real-time environmental monitoring capabilities and data-driven decision-making. In-depth analysis reveals advancements in AI/ML methodologies, including novel algorithms for soil mapping, land-cover classification, flood susceptibility modeling, and remote sensing image analysis. Notable applications include enhanced air quality predictions, water quality assessments, climate impact forecasting, and automated wildlife monitoring using AI-driven image recognition. Challenges such as the “black-box” nature of AI models, the need for high-quality data in resource-constrained regions, and the complexity of real-time disaster management are also addressed. The study highlights ongoing efforts to develop explainable AI (XAI) models, which aim to improve model transparency and trust in critical environmental applications. Future research directions emphasize improving data quality and availability, fostering interdisciplinary collaborations across environmental and computer sciences, and addressing ethical considerations in AI-driven environmental management. These findings underscore the transformative potential of AI and ML technologies for sustainable environmental management, offering valuable insights for researchers and policymakers in addressing global environmental challenges.

Reduced aerosol transport from South Asia to the Tibetan Plateau following the January 2021 sudden stratospheric warming event

Aerosols from South Asia directly enhance glacier melt over the Tibetan Plateau. While the transboundary transport of aerosols from South Asia towards the Tibetan Plateau has been extensively investigated from a tropospheric perspective, less focus has been given to the stratospheric dimension. Here we examined the impact of the sudden stratospheric warming in January 2021 on aerosol transport from South Asia towards the Tibetan Plateau via numerical simulation. The results revealed that the aerosol transport from South Asia to the Tibetan Plateau reduced by 30%–40% following the January 2021 sudden stratospheric warming event. The eastward-propagating wave train stimulated by the stratospheric sudden warming induced an anticyclonic anomaly from the Persian Gulf to northern China, south of which easterly anomalies hindered the aerosol transport from South Asia to the Plateau. This study provides valuable insights for predicting air quality over the Tibetan Plateau.

Statistical and machine learning approaches for estimating pollution of fine particulate matter (PM2.5) in Vietnam

This study aims to predict fine particulate matter (PM2.5) pollution in Ho Chi Minh City, Vietnam, using autoregressive integrated moving average (ARIMA), linear regression (LR), random forest (RF), long short-term memory (LSTM), bidirectional LSTM (Bi-LSTM), and convolutional neural network (CNN) combining Bi-LSTM (CNN+Bi-LSTM). Two experiments were set up: the first one used data from 2018–2020 and 2021 as training and test data, respectively. Data from 2018–2021 and 2022 were used as training and test data for the second experiment, respectively. Consequently, ARIMA showed the worst performance, while CNN+Bi-LSTM achieved the best accuracy, with an R² of 0.70 and MAE, MSE, RMSE, and MAPE of 5.37, 65.4, 8.08 µg/m³, and 29%, respectively. Additionally, predicted air quality indexes (AQIs) of PM2.5 were matched the observed ones up to 96%, reflecting the application of predicted concentrations for AQI computation. Our study highlights the effectiveness of machine learning model in monitoring of air pollution.

Systematic Review of Machine Learning and Deep Learning Techniques for Spatiotemporal Air Quality Prediction

Background: Although computational models are advancing air quality prediction, achieving the desired performance or accuracy of prediction remains a gap, which impacts the implementation of machine learning (ML) air quality prediction models. Several models have been employed and some hybridized to enhance air quality and air quality index predictions. The objective of this paper is to systematically review machine and deep learning techniques for spatiotemporal air prediction challenges. Methods: In this review, a methodological framework based on PRISMA flow was utilized in which the initial search terms were defined to guide the literature search strategy in online data sources (Scopus and Google Scholar). The inclusion criteria are articles published in the English language, document type (articles and conference papers), and source type (journal and conference proceedings). The exclusion criteria are book series and books. The authors’ search strategy was complemented with ChatGPT-generated keywords to reduce the risk of bias. Report synthesis was achieved by keyword grouping using Microsoft Excel, leading to keyword sorting in ascending order for easy identification of similar and dissimilar keywords. Three independent researchers were used in this research to avoid bias in data collection and synthesis. Articles were retrieved on 27 July 2024. Results: Out of 374 articles, 80 were selected as they were in line with the scope of the study. The review identified the combination of a machine learning technique and deep learning techniques for data limitations and processing of the nonlinear characteristics of air pollutants. ML models, such as random forest, and decision tree classifier were among the commonly used models for air quality index and air quality predictions, with promising performance results. Deep learning models are promising due to the hyper-parameter components, which consist of activation functions suitable for nonlinear spatiotemporal data. The emergence of low-cost devices for data limitations is highlighted, in addition to the use of transfer learning and federated learning models. Again, it is highlighted that military activities and fires impact the O3 concentration, and the best-performing models highlighted in this review could be helpful in developing predictive models for air quality prediction in areas with heavy military activities. Limitation: This review acknowledges methodological challenges in terms of data collection sources, as there are equally relevant materials on other online data sources. Again, the choice and use of keywords for the initial search and the creation of subsequent filter keywords limit the collection of other relevant research articles.

Urban form and seasonal PM2.5 dynamics: Enhancing air quality prediction using interpretable machine learning and IoT sensor data

This study investigates the critical issue of how urban form characteristics influence PM2.5 concentrations, a key concern for public health in densely populated cities. Traditional monitoring methods have faced data gaps and methodological limitations. To address this, we employed interpretable machine learning (ML) models with data from 1,069 Internet-of-Things (IoT) sensors across Seoul, South Korea (September 2020–August 2023). Over 80 urban form variables—including density, transportation, road design, building morphology, and land use—were analyzed using Recursive Feature Elimination to identify key factors affecting PM2.5 concentrations within three buffer zones (300-m, 500-m, 1-km). The random forest model demonstrated the highest accuracy, with an R² of 95 % for autumn and 96 % for spring. Our findings show higher PM2.5 levels in colder months, driven by road width and building density in autumn and traffic and industrial activity in winter. In summer, green spaces and meteorological conditions were primary factors, while spring air quality was notably impacted by localized traffic emissions around highways and bus stops. This study offers robust predictions and actionable insights for urban planning and air quality management. Future research could integrate additional environmental variables and expand sensor coverage to further refine predictive models.

A Numerical and Experimental Study to Compare Different IAQ-Based Smart Ventilation Techniques

Maintaining indoor environmental quality in residential buildings is essential for occupants’ comfort, productivity, and health, with effective mechanical ventilation playing a key role in removing or diluting indoor pollutants. A two-week experimental campaign was conducted in an apartment in Lyon, France, known for its poor urban air quality, assessing temperature, relative humidity, CO2, and PM2.5 concentrations. A model verification study was performed to compare experimental measurements against numerical modeling in the living room and bedroom, leading to errors in the accuracy of the sensors. In addition, this study also investigates the impact of different ventilation strategies on indoor air quality. This research evaluates a baseline mechanical exhaust-only ventilation approach with constant air volume against two innovative smart ventilation approaches: mechanical exhaust-only ventilation with humidity control and mechanical exhaust-only ventilation with room-level CO2 and humidity control. A key contribution of this research is the novel coupling of multizone simulation models (DOMUS and CONTAM) with a CFD tool to refine pressure coefficients on the building façade, which enhances the accuracy of indoor air quality predictions. The smart ventilation strategies showed improvements, including a 20% reduction in CO2 concentration and a 5% reduction in the third-quartile PM2.5 concentration, highlighting their effectiveness in enhancing ventilation and pollutant dilution. This research provides valuable insights into advanced ventilation strategies and modeling techniques in urban environments.

Multi-classification prediction of PM2.5 concentration based on improved adaptive boosting rotation forest

Developing efficient, stable, and user-friendly methods and technologies for predicting air quality has contributed to environmental research and management. Most traditional machine learning (ML) models often struggle to efficiently process extensive air quality data and grapple with the challenge of imbalanced data distributions. To this end, we introduced a novel multi-strategy collaborative approach that incorporates weighted feature selection, an adaptive enhanced rotation forest algorithm, and Bayesian Optimization for parameter tuning. Moreover, to improve the transparency in black box ML models, the novel Shapley Additive Explanations (SHAP) method was applied to interpret the outputs and analyze the importance of individual variables. Through experiments on real air quality datasets, we have verified the accuracy of our proposed method in classifying air quality levels. Notably, our model demonstrated an average test set accuracy improvement of approximately 10 % in cities like Beijing, Tianjin, Shijiazhuang, and Baoding after hyperparameter optimization using Bayesian methods. Furthermore, compared to alternative algorithms, our model showed an improvement of 1–2 % in evaluation metrics. By introducing novel methodologies and tools, our research contributes significantly to the advancement of air quality classification technology and holds profound implications for informing environmental policy decisions.

Air Quality Prediction Using Machine Learning Models: A Predictive Study in the Himalayan City of Rishikesh

Air Quality Prediction Using Machine Learning Models: A Predictive Study in the Himalayan City of Rishikesh

The Global Forest Fire Emissions Prediction System version 1.0

Abstract. The Global Forest Fire Emissions Prediction System (GFFEPS) is a model that estimates biomass burning in near-real time for global air quality forecasting. The model uses a bottom-up approach, based on remotely sensed hotspot locations, and global databases linking burned area per hotspot to ecosystem-type classification at a 1 km resolution. Unlike other global fire emissions models, GFFEPS provides dynamic estimates of fuel consumption, fire behaviour and fire growth based on the Canadian Forest Fire Danger Rating System, plant phenology as calculated from daily global weather and burned-area estimates using near-real-time Visible Infrared Imaging Radiometer Suite (VIIRS) satellite-detected hotspots and historical burned-area statistics. Combining forecasts of daily fire weather and hourly meteorological conditions with a global land classification, GFFEPS produces fuel consumption and emission predictions in 3 h time steps (in contrast to non-dynamic models that use fixed consumption rates and require a collection of burned area to make post-burn estimates of emissions). GFFEPS has been designed for use in operational forecasting applications as well as historical simulations for which data are available. A study was conducted showing GFFEPS predictions through a 6-year period (2015–2020). Regional annual total smoke emissions, burned area and total fuel consumption per unit area as predicted by GFFEPS were generated to assess model performance over multiple years and regions. The model's fuel consumption per unit area results clearly distinguished regions dominated by grassland (Africa) from those dominated by forests (boreal regions) and showed high variability in regions affected by El Niño and deforestation. GFFEPS carbon emissions and burned area were then compared to other global wildfire emissions models, including the Global Fire Assimilation System (GFAS), the Global Fire Emissions Database (GFED4.1s) and the Fire INventory from NCAR (FINN 1.5 and 2.5). GFFEPS estimated values lower than GFAS and GFED (80 % and 74 %) and had values similar to FINN 1.5 (97 %). This was largely due to the impact of fuel moisture on consumption rates as captured by the dynamic weather modelling. Model evaluation efforts to date are described – an ongoing effort is underway to further validate the model, with further developments and improvements expected in the future.

PR-FCNN: a data-driven hybrid approach for predicting PM2.5 concentration

The atmosphere’s fine articulate Matter (PM2.5) poses various health-related risks. Even though multiple efforts have been made to lower the emissions of these substances, the mortality rate is continuously increasing, requiring immediate inclination of the scientific community towards the design and development of advanced predictive models. Conventional statistical approaches have become dormant due to their limitations in capturing the innate relationships between the pollutants, particularly for predicting PM2.5 concentrations. In contrast, machine and deep learning techniques have shown great potential for forecasting air quality, providing more accuracy than its predecessor techniques. The present study investigates the utilization of hybrid approaches by integrating machine learning models with deep learning models to improve the prediction capabilities of PM2.5 concentration. It uses datasets from the World Air Quality Index (WAQI) and the State of Global Air (SOGA) to analyze the performance of the models on both the daily and annual data, respectively. This ensures the model’s effectiveness on a diversified dataset. The present study implements Random Forest (RF), Polynomial Regression (PR), XGBoost, and Extra Tree Regressor (ETR) coupled with Fully Connected Neural Network (FCNN), Long Short-Term Memory (LSTM), and Bi-directional LSTM (Bi-LSTM) for obtaining optimized results. Finally, after a thorough investigation, the hybrid PR model coupled with FCNN (PR-FCNN) is found to be the best model with improved R-squared (R2) values, portraying its potential for predicting PM2.5 concentration accurately. Based on the experimentation, the preset study recommends implementing hybrid approaches, offering better predictive accuracy in forecasting air pollutants, especially PM2.5.

The role of vertical grid resolution and turbulent diffusion uncertainty on chemical transport modeling

Chemical transport models (CTM) tend to perform poorly in simulating pollution processes under weak turbulent diffusion conditions. In this study, we address this issue from the perspectives of vertical grid resolution and vertical mixing schemes. Three vertical grid resolution configurations (L4, L12, L40) with the CHIMERE model are evaluated during a winter episode, which includes a heavy pollution episode (PE) in the Paris region. The results emphasize the significance of vertical grid resolution, particularly noticeable during nighttime, and consequently impacts CHIMERE simulations under nocturnal stable conditions. Consistent improvement in CTM modeling is observed with refined vertical resolutions and the first layer height based on a simple linear vertical diffusion scheme defined as the initial Kz diffusion (IKD) scheme. Compared to the other two configurations, the finest configuration (referred to as L4-IKD) demonstrates an average improvement in root mean square error of 23.26 % and 25.09 % on regular days (RD) and 62 % and 129 % during PE, respectively. However, simulations using the 1.5-order turbulence kinetic energy (TKE) based eddy diffusivity closure scheme, named the new eddy diffusion (NED), are more sensitive to the first layer height setup. Excessively fine first-layer heights can lead to inaccurate TKE calculations. Generally, models with low vertical grid resolution can reasonably predict air quality on RD or during light pollution events but struggle with heavy PEs. One straightforward enhancement strategy involves adding an extra fine first layer height in CTM simulations, resulting in an average 50.10 % improvement from L4-IKD to L12-IKD during PE. Another strategy is enhancing the model's vertical diffusion scheme, which improved the CTM modeling by 26.67 % compared with IKD during PE under identical vertical grid resolution.

Prediction of Jakarta's Air Quality Using a Stacking Framework of CLSTM, CatBoost, SVR, and XGBoost

Air quality prediction, particularly in estimating PM10 particle concentration, is a significant challenge in major cities like Jakarta, which experience high levels of air pollution. This study aims to develop an air quality prediction model using an innovative stacking framework that combines several machine learning algorithms, namely ConvLSTM, CatBoost, SVR, and XGBoost. The methodology employed in this research is an experimental approach, where each model is trained and tested individually before being integrated into the stacking framework. The dataset used was sourced from the Kaggle platform, containing historical air quality data in Jakarta. Performance evaluation was conducted by measuring the Root Mean Squared Error (RMSE) for each model. The results of the study showed that the ConvLSTM model produced an RMSE of 13.5168, CatBoost with an RMSE of 13.4113, and SVR with an RMSE of 14.2725. To improve prediction accuracy, the researchers employed a stacking approach of the four models (ConvLSTM, CatBoost, SVR, and XGBoost), which yielded a much lower RMSE of 0.8093. Thus, this stacking framework has proven to significantly enhance air quality prediction performance, particularly in predicting PM10 concentrations in Jakarta.

A New Mathematical Model in Three-dimensional Variational Data Assimilation for Air Quality Prediction

Abstract The three-dimensional variational (3DVar) data assimilation is a critical component of air quality prediction, and the forecast state in air quality prediction is inferred to be approximately low rank and heavily laden with low-frequency noises through the application of tensor decomposition. To reduce the effects of these uncertainties and achieve a more precise analysis state, we establish a new objective function that consists of the weighted sum of three terms: the first term is the norm difference between the gradients of the forecast and analysis states, the second term is the norm difference between the observation and the analysis state in observational space, and the third term is the rank of the analysis state. Numerical results show that our treatment is superior to other methods, such as the classic 3DVar method, the En3DVar method, and the method of 3DVar with forecast gradient.

A Comprehensive Analysis of Santiago's Air Quality Forecasts Using SARIMA Model

Abstract. This study analyses Santiagos air quality by forecasting key pollutant concentrations- PM2.5, PM10, NO, and O-over June 2024 to May 2025 using the Seasonal Autoregressive Integrated Moving Average (SARIMA) model. Historical data from 2016 to 2024 shows seasonal peaks, particularly in winter, driven by residential heating and atmospheric conditions. A notable decrease in annual peak NO level, from 71 g/m in 2019 to 23 g/m in 2020, likely due to the COVID-19 pandemic and air quality initiatives. The SARIMA forecasts indicate that PM2.5 levels could reach up to 80 g/m, NO around 28 g/m, O around 47ppb and PM10 around 45g/m during peak periods, highlighting ongoing seasonal pollution challenges. To address these, policymakers should enforce stricter emissions regulations, promote cleaner heating, expand public transportation, and improve air quality monitoring. Additional strategies like traffic management, urban greening, and preparedness for high-pollution days are also recommended. This study demonstrates the SARIMA models effectiveness in forecasting air quality trends, offering valuable insights for policy development to protect public health in Santiago.

Long-term urban air quality prediction with hierarchical attention loop network

Air pollution poses severe threats to human health, socioeconomic development, and natural environment, making it one of the most serious environmental issues. Accurate long-term regional air quality prediction plays a significant role in mitigating the severity of pollution events and effectively suppressing the intensification of air pollution, benefiting both the environment and societyAccurate regional air quality prediction contributes to the governance of atmospheric pollution, benefiting both the environment and society In this paper, we propose a novel Hierarchical Attention Loop Network (HALN) comprising four key components for long-term prediction of PM2.5 concentrations across regional multiple target stations, achieving effective prediction horizons of up to 120 h. Instead of using all monitoring stations within a region, the MIC-H selector selects station data exhibiting strong spatiotemporal correlations at each hierarchical level. HALN then injects additional spatial information through annular grid positional encoding. The hierarchical attention feature match layer refines the model’s focus on data with a more substantial predictive impact. As the core integrative component of HALN, attention loop block reinforces the influence of historical outputs, enhancing the model’s ability to capture long-term dependencies.Instead of using all monitoring stations within a cluster, the hierarchical MIC selector utilizes the maximal information coefficient (MIC) to select station data exhibiting strong spatiotemporal correlations at each hierarchical level. HALN then details the relative distances and azimuths between stations through annular grid positional encoding to inject additional spatial information into the selected input data. Subsequently, the hierarchical attention feature match layer employs multi-head attention to independently adapt the input weights at each level, refining the model’s focus on station data with a more substantial predictive impact. As the core integrative component of HALN, attention loop block employs gated recurrent units (GRUs) and a feedback attention mechanism to reinforce the influence of historical outputs, thereby enhancing the model’s ability to capture long-term dependencies. Extensive real-world experimental results demonstrate that the proposed model exhibits exceptional robustness in regional predictions and achieves high accuracy in long-term forecasting. The coefficient of determination (R2) reaches 0.793 and 0.636 for 24-hour and 120-hour predictions, respectively.