Forecasting PM_2.5 induced male lung cancer morbidity in China using satellite retrieved PM_2.5 and spatial analysis

The present study predicts a spatial distribution of lung cancer morbidity in Chinese males due to exposure to PM_2.5 concentration from 2010 to 2015. A spatial autocorrelation method was used to evaluate the spatial relationship between the lung cancer morbidities from 2006 to 2009 and satellite-derived PM_2.5 atmospheric levels. A comprehensive grey correlation degree analysis was carried out to assess the simultaneous and lag associations between the lung cancer morbidity and PM_2.5 concentration. These relationships were subsequently applied to predict male lung cancer morbidity in a specific year. Annual mean PM_2.5 levels in this specific year and previous 8years were used as 9 independent variables to establish four statistical models. These models include ridge regression (RR), partial least squares regression (PLSR), support vector regression (SVR), and the combined forecasting model (CFM) to predict the male lung cancer morbidity in China from 2010 to 2015. The model error evaluations suggested that the partial least squares regression model performed the best in the male lung cancer morbidity forecast. We calculated the male lung cancer morbidity by the optimal method among the established statistical forecasting models at 1948 sites in China. The gridded morbidity distribution from 2010 to 2015 across the country was obtained by Kriging interpolation method. Results showed that the male lung cancer morbidity increased significantly from western to eastern China, except for the far north region. This spatial pattern is in line with the spatial distribution of PM_2.5 concentration, manifesting a significant relationship between PM_2.5 concentration level and lung cancer morbidity in Chinese males.