| Carbon dioxide(CO2)is the most important greenhouse gas in anthropogenic emissions,and its rapid growth has led to global climate change.China contributes large carbon emissions,but also incorporates"carbon peaking" and "carbon neutrality" into the national strategy,and"carbon reduction and pollution reduction",has been continuously coordinated and promoted.The study of China’s CO2 spatio-temporal variation and its influencing factors is a basic issue in the "dual carbon" target.Among the methods for CO2 quantification,satellite observation has the advantages of long term and large spatial coverage.However,due to factors including low swath and inversion,the data is very sparse.Spatial interpolation ignores CO2s influencing factors and data characteristics.It is difficult to know atmospheric CO2 spatiotemporal distribution because of low resolution and even unreliable analysis conclusions.Understanding how factors such as emissions,vegetation,and climate affect China’s atmosphere CO2 is the basic scientific basis for studying carbon cycle.Machine learning method has the advantages of being convenient and fast for multi-source data fusion,and has been widely used in atmospheric environment research.However,there are still few studies on the use of machine learning for CO2 multi-source data spatio-temporal fusion modeling.This paper takes China’s atmosphere CO2 as the research object(the concentration measure is the column-averaged of carbon dioxide in dry air,ie XCO2).This paper uses machine learning methods to fuse multiple satellite data,multiple factor data,and multiple types of data,establishes a multi-source data fusion model based on machine learning,and analyzes the spatio-temporal distribution of CO2 in China from 2003 to 2021.The main research results and work of this paper are as follows:(1)Focusing on the problem of sparse carbon monitoring satellite data and difficulty in analyzing the temporal and spatial distribution of atmospheric CO2 a machine learning model based on convolutional neural network was designed,which integrated the satellite XCO2 data of SCIAMACHY,GOSAT and OCO-2,combined with various influences factor data.The model generates a spatiotemporal continuous XCO2 dataset of 0.25° on a monthly scale from 2003 to 2019(long-term monthly scale model).In the process of studying the above problems,the importance of the proxy was found,and carbon monoxide(CO)was used as the proxy.In this study,a random forest model was used to fuse ground and satellite observations to generate a 0.10° dataset on a daily scale from 2019 to 2021(short-term daily scale model).These results have been systematically validated and evaluated with satellites,atmospheric models,and the ground observation.Overall,the model results performed well.(2)In terms of the temporal and spatial distribution of China’s atmosphere CO2 it reflects the spatial characteristics dominated by anthropogenic emissions,the seasonal variation characteristics dominated by vegetation activities,and the long-term growth fluctuation characteristics dominated by climate.Spatially,the areas with dense human activities such as the North China region and the Yangtze River Delta are relatively high,especially the enhancement of urban areas,and the areas such as the Northeast and the Qinghai-Tibet Plateau are relatively low;In the time dimension,for seasonal variation,CO2 is highest in April and lowest in September,forming a seasonal cycle.The seasonal cycle amplitude(SCA)shows the obvious characteristics of climate division as the boundary line.The higher the latitude,the greater the SCA,especially with the Qinling-Huaihe line as the boundary;In the long-term trend,from 2003 to 2021,China’s atmospheric CO2 increased by an average of 2.26 ppmv per year,of which the fastest growth occurred in 2016,which is consistent with the global level.The spatial distribution of the growth rate shows slight differences,which is larger in the area of anthropogenic emissions.(3)This paper mainly analyzes the impact of three factors:emissions,vegetation and climate.Emissions and XCO2 anomalies are positively correlated,both spatially and temporally.In the decoupling relationship between carbon emissions and economic development,most provincial-level administrative regions are still in a weak decoupling relationship,and after 2012 it has become a stronger decoupling relationship.Vegetation activity is negatively correlated with CO2 and there is a certain time lag effect,and the seasonal fluctuation range is closely related to the net ecosystem productivity.The climate mainly affects the CO2 interannual variability of the atmosphere on a global scale.The CO2 growth rate is higher during several El Ni(?)o periods,which may be due to the weak carbon sink effect at this time.And there is a coupling relationship between CO2 growth and climate,the doubling of CO2 concentration will lead to the warming of the troposphere and the cooling of the stratosphere,which will bring about 3.4 W/m2 of the tropopause of radiative forcing. |
PDF Full Text Request