Abstract:High-resolution column-averaged dry air mole fraction of CO 2 (XCO 2 ) data is crucial for understanding the spatiotemporal patterns of XCO 2 and for mitigating carbon emissions. Due to the limited scanning range of sensors and strict inversion conditions, satellite-retrieved XCO 2 data are often significantly incomplete. Machine learning models are widely used to fill these gaps in satellite XCO 2 data. However, the limitations of individual machine learning models and the complexity of the spatial distribution of XCO2 mean that the accuracy of XCO 2 predictions still needs improvement. In this study, a new spatiotemporal stacked ensemble learning model (STEL) was developed by combining random forest (RF), extremely randomized trees (ERT), extreme gradient boosting (XGBoost), optical gradient boosting (LightGBM), and categorical boosting (CatBoost) using the stacking ensemble learning methodology. Considering the spatiotemporal heterogeneity of XCO 2 , a novel spatiotemporal weighting feature was constructed as part of the model's input parameters. Finally, the XCO 2 observed by Orbiting Carbon Observatory 2 (OCO-2) was reconstructed using STEL, and a monthly mean XCO 2 dataset covering China from 2015 to 2020 was generated at a spatial resolution of 0.1°. The results show that STEL exhibits superior performance and generalization capabilities compared to individual machine-learning models. R 2 RMSE and MAPE were 0.9624, 1.0023 ppm, and 0.1583 % on the test set, and 0.8970, 1.4213 ppm, and 0.2475 % for R 2 , RMSE, and MAPE in ground validation, respectively. In 10-fold cross-validation, STEL's RMSE was reduced by 9.52 % compared to the best-performing single model (RF). The spatiotemporal trend of CO 2 in China from 2015 to 2020 was analyzed using STEL XCO 2 data. The results indicate that this dataset accurately reflects the spatiotemporal heterogeneity of XCO 2 distribution at a fine scale. Overall, XCO 2 exhibited a spatiotemporal pattern of "high in the east and low in the west" and "high in spring and low in summer." Except in summer, high XCO2 values were mainly distributed in the North China Plain. XCO 2 trends and hotspots showed considerable spatial variation. The Pearl River Delta and Yangtze River Delta urban agglomerations have the fastest XCO 2 growth rates, and the distribution of XCO 2 hotspots is consistent with the distribution of population and economic centers. In the sparsely populated northwest of China, XCO 2 is growing rapidly due to increased thermal power generation and coal mining. XCO 2 hotspots in Northwest China are mainly located in Xinjiang, Ningxia, and Inner Mongolia. The methodology and data presented are useful for further research on carbon emissions, carbon sinks, and climate change.

Using Multisource Data and Time Series Features to Construct a Global Terrestrial CO2 Coverage by Deep Learning

Generating Daily High-Resolution Regional XCO2 by Deep Neural Network and Multi-Source Data

Spatio-temporal modeling of satellite-observed CO 2 columns in China using deep learning

Multi-sensor integrated mapping of global XCO 2 from 2015 to 2021 with a local random forest model

A Full-Coverage Satellite-Based Global Atmospheric CO2 Dataset at 0.05° Resolution from 2015 to 2021 for Exploring Global Carbon Dynamics

Reconstructing global daily XCO2 at 1° × 1° spatial resolution from 2016 to 2019 with multisource satellite observation data

High Spatial-Resolution Classification of Urban Surfaces Using a Deep Learning Method

Mapping Contiguous XCO2 by Machine Learning and Analyzing the Spatio-Temporal Variation in China from 2003 to 2019

Full-coverage estimation of CO 2 concentrations in China via multisource satellite data and Deep Forest model

XCO2 Super-Resolution Reconstruction Based on Spatial Extreme Random Trees

Deriving full‐coverage and fine‐scale XCO2 across China based on OCO‐2 satellite retrievals and CarbonTracker output

Estimation of daily XCO2 at 1 km resolution in China using a spatiotemporal ResNet model

A Remote Sensing-Based Method for Generating a Global Continuous Carbon Dioxide Concentration Dataset

Near Real-time CO$_2$ Emissions Based on Carbon Satellite and Artificial Intelligence

Global spatiotemporal estimation of daily high-resolution surface carbon monoxide concentrations using Deep Forest

Global Land Mapping of Satellite-Observed CO2 Total Columns Using Spatio-Temporal Geostatistics

Fast retrieval of XCO2 over east Asia based on Orbiting Carbon Observatory-2 (OCO-2) spectral measurements

Improved Consistency of Satellite XCO2 Retrievals Based on Machine Learning

Mapping high-resolution XCO 2 concentrations in China from 2015 to 2020 based on spatiotemporal ensemble learning model

Reconstructing Global Daily CO2 Emissions via Machine Learning

Enhancing Carbon Cycle Understanding through Deep Learning: Development and Validation of the Global Carbon Fluxes Dataset (GCFD)