| Particulate matter 2.5 (aerodynamic diameter less than 2.5 μm) is the major pollutant of haze, numerous epidemiological studies have shown that PM2.5 may be associated with adverse health effects. With rapid development of social economy, air pollution has gradually became a severe environmental problem in China, the government and public has paid great attention to it. China has not established regulatory PM2.5 monitoring network until 2012. The lack of spatially and temporally continuous ground PM2.5 measurements makes it difficult to support the epidemiological and health effects study of PM2.5 in China.PM2.5 was associated with the morbidity and mortality from respiratory and cardiovascular diseases. The hospital admissions for respiratory diseases could reflect the acute effects of atmospheric pollutants on population health. However, the lack of appropriate statistical methods and difficulty in acquiring diagnosis data, limited the researches on association between PM2.5 and hospital admissions of respiratory disease.In our study, we built a high-accuracy PM2.5 land use regression model, combined with the ground level monitor data and the estimate PM2.5 data to explore the spatiotemporal PM2.5 concentrations in China. We used a time-series analysis to explore the relationship between PM2.5 and hospital admissions of respiratory disease in 2014-2015 in Yinzhou district, Ningbo city.Materials and methodTo improve the power in the model, we used the 2 million PM2.5 concentration data from 1385 Chinese ground PM2.5 monitors during 2014 to 2015 from Harvard GIS analysis center, we applied the Globe land cover dataset at 30 meter, which used remote sensing data and had overall accuracy of 83.51%, as the land use data. In addition, we used the OpenStreetMap, LandScan, ASTER GDEM, Landsat8 datasets to abstract the latest road data, population density, sea elevation, and NDVI data, respectively. We used a generalized additive model to build the GIS-derived PM2.5 land use model, and then we checked the stability and accuracy of the model by using handout cross validation, leave one out cross validation, and K fold cross validation. Kriging interpolation method was used to build a similar model to predict the PM2.5 concentrations, we compared the estimated accuracy in these two models. Finally, based on the estimated data from the model, we described the spatiotemporal distribution of PM2.5 in China.In this study, we also evaluated the relationship between the ground monitors’ PM2.5 concentration and hospital admissions for respiratory disease. We used the distributed lag-time models to explore the lag effects of all four air pollutants on hospital admissions for respiratory disease. Stratified analyses were conducted by age, gender and season. We also checked the threshold effect of PM2.5 using exposure-response analysis.ResultsThe model R-squared is 0.7 and the cross validation R-squared was 0.65 for the land use regression model. Our results showed that temperature, relative humidity, wind speed, population density, altitude, distance to nearest road had significant effect on PM2.5 concentrations. In our study, we used Kriging interpolation method to build Chinese PM2.5 concentration model, the handout cross validation R-squared was 0.48, K fold cross validation R-squared was 0.57, the stability and accuracy of the kriging model was lower than the land use model. The North China, middle and lower Yangtze River, Sichuan Basin, and the central Shaanxi Plain districts were the most PM2.5 polluted area, we found the PM2.5 concentration was the highest in winter, and the PM2.5 was significantly lower in 2015 than that in 2014.We found that the PM2.5 concentration distribution increased gradually from southeast to northwest, and we used time series analysis method to explore the relationship between PM2.5 and hospital admissions for respiratory disease. The lag period for PM2.5 on hospital admissions of respiratory disease was 5 days, the total hospital admissions for respiratory disease were significantly increased by 5.59% (95%CI 3.0-8.25%) per IQR increased in PM2.5, the estimated effect changed small after adjusted for other gaseous pollution. We also found the significant association for PM2.5-10 among children (EER:6.51,95%CI:2.31-10.88%) and the association for PM2.5 among adults (15-64 years) (EER:7.1,95%CI 3.27-11.08%). The acute upper respiratory tract cases elevated 11.46%(95%CI 8.16-14.86%) per IQR increase in PM2.5 concentration on commutative lag days of 0-6. In addition, we found that an IQR increase of PM2.5-10 was significantly associated with acute upper respiratory infection in warm seasons (EER:15.01,95%CI 8.17-22.27%) after adjustment for PM2.5. We did not observe any significant association between 3 days cumulative concentrations of PM2.5 and total hospital admissions for pneumonia. In stratified analyses, we found that after adjusted for PM2.5, an IQR increased in PM2.5-10 had higher estimated effects in children (EER:7.04,95%CI 0.95-13.5%). An IQR increase in 3 days cumulative concentration of PM2.5 was associated with the 3.27%(95%CI 0-6.66%) increase in asthma hospital admission, but after adjusting for the PM2.5-10 the estimated effect became no significant, however, we found that the PM2.5-10 has higher effects on asthma hospital admission in children (EER:7.34,95%CI 1.75-13.24%) and cold season (EER: 7.78,95%CI 2.55-13.29%)groups. We found that an IQR increase of PM2.5-10 was associated with 5.21%(95%CI 1.66-8.89%) increase of hospital admissions for COPD. In this study, we did not find any threshold for PM2.5 or PM2.5-10 on respiratory hospital admissions by exposure-response analysis. We found that there were different lag effects for different respiratory diseases.ConclusionOur study showed that PM2.5 was related to plat terrain, population density, and higher degree of social economy, the winter central heating, and certain meteorological factors. We also found the relationship between PM2.5 and hospital admissions for respiratory disease, and there were different lag periods for disparate respiratory diseases. Our results indicated that ground monitor data based on PM2.5 land use regression model provided a feasible way to fill the temporal and spatial gaps left by monitoring network in China. Besides, PM2.5 data estimated in this study can be served as exposure assessment in future environmental and epidemiological studies in China. |
PDF Full Text Request