Spatial And Temporal Response Of Lung Cancer Incidence To Typical Air Pollutants In China

Lung cancer is a serious health threatening disease,often ranking first among all cancers in terms of incidence and mortality.Despite the development of medical technology,the survival rate of lung cancer patients has not improved significantly,currently,lung cancer accounts for nearly twenty percent of deaths due to cancer worldwide.The causative factors of lung cancer have been discussed extensively in traditional medical research,and as the research progresses,many scholars have found that long-term exposure to air pollution and other harsh environments is an important factor affecting the development of lung cancer,and there is a significant association between it and lung cancer disease and death.On the basis of satellite remote sensing technology,which can obtain large-scale and long-term pollutant atmospheric column concentration data,many studies have attempted to obtain air pollutant ground-level concentrations in different regions based on satellite data inversion.In the current background of rapid development of big data technology,it has become possible to obtain the lung cancer incidence rate and air pollutant concentration and study the response relationship between them.In this paper,the characteristics of spatial and temporal variation in lung cancer incidence in China were firstly analyzed with mainland China as the study area.Then,using meteorological elements,air pollutant atmospheric column concentrations,vegetation indices,and population as input elements and air pollutant ground monitoring concentrations as output,a back propagation（BP）neural network and a random forest model were built to invert the NO₂ and SO₂ ground-level concentrations over a long time series in China,and to study the spatial and temporal response of lung cancer incidence to cumulative concentrations of various air pollutants.The spatial and temporal responses of lung cancer incidence to the accumulated concentrations of various air pollutants were investigated.The main studies are as follows.（1）Lung cancer incidence rates at the prefecture-level city scale in China were calculated using lung cancer and population data from the 2015-2019 China Tumor Registry Annual Report,and thematic mapping of lung cancer incidence rates in China was completed.Based on the global Moran index and cold hotspot analysis,it was found that there was a significant positive spatial correlation of lung cancer incidence rates in Chinese prefecture-level cities,and the local spatial aggregation of lung cancer incidence rates in China increased significantly after 2015,manifested by a significant increase in the number of cold hotspot cities.In terms of temporal changes,the overall lung cancer incidence rate in China remained on an increasing trend from 2012 to 2016,among which,East China had the most significant increase and only North China had a decreasing trend,while the overall lung cancer incidence rate in the remaining regions remained stable,with the highest incidence rate in Northeast China.（2）The multi-source data sets were pre-processed to complete the fusion of meteorological parameters,atmospheric column concentrations of air pollutants,vegetation indices,and population data sets.BP neural network,random forest,support vector machine and other inversion models are built,and the inversion models are trained using the fused datasets,and the Coefficient of Determination（R²）,Root Mean Square Error（RMSE）and Mean Absolute Error（MAE）were used to assess the inversion accuracy of all models.The optimal inversion models were generated through iterative training and hyperparameter tuning.Finally,the BP neural network model was chosen to invert the ground-level concentration of NO₂ with the test sets of RMSE,MAE and R² of5.79ug/m³,4.16ug/m³ and 0.79,respectively,and the random forest model was used to invert the near-ground concentration of SO₂ with 8.97ug/m³,4.9ug/m³ and 0.75,respectively.（3）The spatial and temporal distribution characteristics and changing trends of different pollutant concentrations are studied,and it is finally found that both NO₂ and SO₂ show obvious spatial aggregation and seasonal characteristics,and the pollution levels of NO₂ and SO₂ are the highest in central and eastern China.The seasonal concentrations of both pollutants are highest in winter and lowest in summer,and the temporal change curve of NO₂ concentration shows an increase followed by a decrease,while SO2 shows a fluctuating increase.From 2005 to 2012,NO₂ concentrations in most regions of China were on a significant or very significant upward trend,while SO₂concentrations did not change significantly.2013-2016 NO₂ and SO₂ concentrations basically showed strong and continuous changes,with NO₂ decreasing continuously in North China and East China,while other regions remained stable,and SO₂ remained stable overall,indicating that air pollution in key regions of China has been controlled.air pollution has been managed.（4）The spatial and temporal response of lung cancer incidence to air pollutants is studied based on the inverse pollutant concentrations.By replacing different cumulative concentrations of pollutants as the core explanatory variables of the model,we found that there was a lag in the effect of all three pollutants on lung cancer incidence rate according to the change of the spatio-temporal geographically weighted model fit.Under the single-pollutant condition,NO₂ 4-year cumulative concentration,SO₂ 4-year cumulative concentration and PM_2.5 5-year cumulative concentration had the greatest effect on lung cancer incidence.Under the condition of two pollutants,the interaction of PM_2.5 and SO₂had the strongest effect on lung cancer incidence.the two-two interaction of NO₂ and other pollutants was weaker and the model fit was less enhanced compared with that of single pollutant,and the interaction of three pollutants had the strongest effect on lung cancer incidence.