Abstract:Seamless hourly surface fine particulate matters (PM2.5) datasets are urgently needed for environmental and epidemiologic studies. Most existing models rely on satellite AOD with spatial gaps and the time-geolocation information not directly physically related to ground-level PM2.5 as input, leading to several problems, e.g., for daytime only, spatial discontinuity. By exploring spatiotemporal connection, here we propose a novel framework for improving reanalysis AOD with satellite AOD and retrieving seamless PM2.5 concentration. With the wavelet decomposition on the meteorological / AOD time-series and the machine learning algorithm 'Cat -Boost', the grid cells are well connected based on the meteorological/AOD similarity (quasi-Lagrangian description), rather than the spatial distance (Eulerian description). Our new AOD dataset (compared with AERONET AOD, R2 = 0.60, RMSE = 0.19) combines the advantages of reanalysis AOD (all-day hourly; seamless) and satellite AOD (high accuracy and spatial resolution). The new PM2.5 estimation strategy performed excellent in cross-validation (spatial CV R2/RMSE = 0.84/14.57 mu g/m3) and mapping (no spatial discontinuity). With the help of new AOD dataset, the PM2.5 estimation exhibits significant improvement (R2: increased from 0.60 to 0.65, RMSE: reduced from 21.52 to 20.10 mu g/m3) on retrieving PM2.5 at ground monitoring site sparse area (the spatial distance to the closest site is over 20 km). Such high temporal-resolved PM2.5 prediction improved daily esti-mates of PM2.5 as it leveraged the additional hourly information to well capture the transport pattern, with the increased R2 (from 0.72 to 0.77) and decreased RMSE (from 15.54 to 14.30 mu g/m3) at ground monitoring site sparse area. The annual averaged PM2.5 estimated by hourly model also exhibit noticeable differences (-6%) from that estimated with daily model, implying the latter might suffer overestimation (-3%) in population -weighted PM2.5 exposure across China. In addition, our new PM2.5 dataset shows the Chinese average PM2.5 is highest around 24:00, lowest around 17:00.Synopsis: By performing the wavelet decomposition on meteorological factors/AOD (independent variables), the hourly seamless PM2.5 concentration across China is well estimated with improved spatiotemporal connection.

Hourly PM2.5 concentration forecasting based on mode decomposition-recombination technique and ensemble learning approach in severe haze episodes of China

A hybrid Daily PM2.5 concentration prediction model based on secondary decomposition algorithm, mode recombination technique and deep learning

A novel hourly PM2.5 concentration prediction model based on feature selection, training set screening, and mode decomposition-reorganization

DESA: a novel hybrid decomposing-ensemble and spatiotemporal attention model for PM2.5 forecasting

Research and Application of the Mode Decomposition-Recombination Technique Based on Sample-Fuzzy Entropy and K-Means for Air Pollution Forecasting

Application of Decomposition-Ensemble Learning Paradigm with Phase Space Reconstruction for Day-Ahead PM2.5 Concentration Forecasting.

PM2.5 Prediction Based on the CEEMDAN Algorithm and a Machine Learning Hybrid Model

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

A Novel Short-Term PM2.5 Forecasting Approach Using Secondary Decomposition and a Hybrid Deep Learning Model

Three-hourly PM 2.5 and O 3 concentrations prediction based on time series decomposition and LSTM model with attention mechanism

Forecasting PM2.5 Using Hybrid Graph Convolution-Based Model Considering Dynamic Wind-Field to Offer the Benefit of Spatial Interpretability.

An enhanced interval PM2.5 concentration forecasting model based on BEMD and MLPI with influencing factors

Retrieving Hourly Seamless PM2.5 Concentration Across China with Physically Informed Spatiotemporal Connection

Prediction of PM2.5 Concentration Based on Deep Learning, Multi-Objective Optimization, and Ensemble Forecast

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

A hybrid model for enhanced forecasting of PM2.5 spatiotemporal concentrations with high resolution and accuracy

A Novel Hybrid Framework for Hourly PM2.5 Concentration Forecasting Using CEEMDAN and Deep Temporal Convolutional Neural Network

Hourly PM2.5 Concentration Prediction Based on Empirical Mode Decomposition and Geographically Weighted Neural Network

A Hybrid Forecasting Approach to Air Quality Time Series Based on Endpoint Condition and Combined Forecasting Model

A PM2.5 concentration prediction model based on multi-task deep learning for intensive air quality monitoring stations

Integration of complete ensemble empirical mode decomposition with deep long short-term memory model for particulate matter concentration prediction