Distribution Characteristics and Meteorological Prediction Model of Air Negative Oxygen Ions in Fujian

The concentration of negative oxygen ions in air is an important index to evaluate the freshness and cleanliness of air. In recent years, it has become one hot topic concerned by governments and the public. From 2018 to 2021, Fujian has set up a number of observation stations for negative oxygen ions and meteorological factors over the entire province including seashore, mountain, humanities landscape areas, with good representativeness, reliability and continuity. Using the local observations, the spatial and temporal variations of negative oxygen ions concentration in Fujian is analyzed, and the negative oxygen ions concentration and grade prediction models are established based on multiple linear regression method and LightGBM machine learning method. The results show that, negative oxygen ions in Fujian is very rich and is very good for human health. The annual average concentration is between 708-8315 cm^-3, which is highest in high altitude, next in low altitude, and the concentration in middle altitude is the smallest. Overall, the annual average concentration of negative oxygen ions of nearly 80% site is beyond the standard of fresh air defined by World Health Organization. The diurnal variation of the concentration of negative oxygen ions show the characteristics of a peak and a trough, with the peak value mainly occurring at 0400-0600 BT and the trough value at 1200-1300 BT. The seasonal variation of negative oxygen ions concentration is more complex. The seasonal variation in the middle altitude area is greater, the seasonal average concentration in descending order is spring, summer, winter and autumn, while the seasonal variation in the high and low altitude area is relatively small. The main meteorological factors affecting the concentration of negative oxygen ions are temperature, humidity, precipitation, wind speed, air pressure and visibility. The concentration of negative oxygen ions is significantly positively correlated with humidity, precipitation and visibility at different altitudes, while the concentration of negative oxygen ions is significantly correlated with air temperature, wind speed and air pressure, but the correlation is different at different altitudes. The comparisons indicate the effects of LightGBM machine learning model are better than those of the traditional multiple linear regression model at different altitudes. The overestimation of negative oxygen ions concentration prediction is significantly improved, and the prediction grade of negative oxygen ions concentration can be improved by up to 12%. The results of logistic regression show that the traditional logistic regression basically has no predictive ability for small samples, while the LightGBM method has good learning ability in the case of small samples or unbalanced samples.