Estimation of Near-surface NO2 Concentration in Guangdong Province Based on Catboost Model

Abstract

Nitrogen oxide （NOx） is an important air pollutant in the atmosphere, and nitrogen dioxide （NO2） is one of its main components. The monitoring and estimation of NO2 concentration is very important for environmental protection and public health. The near-real-time nitrogen dioxide concentration data （NRTI NO2）, ERA5 meteorological reanalysis data, and DEM data provided by Sentinel-5P atmospheric pollution monitoring satellite were used as estimation variables to estimate the near-surface NO2 concentration in Guangdong Province based on the Catboost model. The results showed that： ① The Catboost model estimated the near-surface NO2 concentration with the highest accuracy, with the coefficient of determination （R2）, root mean square error （RMSE）, and mean absolute error （MAE） of the model fit reaching 0.91, 4.89 μg·m-3, and 3.45 μg·m-3 and the cross-validated R2, RMSE, and MAE reaching 0.90, 4.91 μg·m-3, and 3.43 μg·m-3, with good stability on the monthly and quarterly scales. ② The monthly average NO2 concentration near the surface of Guangdong Province showed a U-shaped trend, with the highest value of 43.8 μg·m-3 in January and the lowest value of 14.37 μg·m-3 in July. The seasonal distribution of the near-surface NO2 concentration was characterized by "high during winter and low during summer and transitional during spring and autumn," and the NO2 concentration in each season was in the following order： winter （27.53 μg·m-3） > spring （20.77 μg·m-3） > autumn （18.77 μg·m-3） > summer （14.85 μg·m-3）. ③ From a spatial distribution perspective, areas with high near-surface NO2 values in Guangdong Province were mainly located in rapidly developing and densely populated areas, while areas with low values were mainly distributed in areas focusing on port economy, agriculture, and new energy sources.