<p>Near surface ozone pollution has become one of the biggest challenges in China’s air quality management. In order to better study the spatiotemporal heterogeneity of near surface ozone pollution, this paper takes the East China region as an example, uses a spatiotemporal weighted variable coefficient regression model, and uses ozone measurement data from the summer of 2021 to investigate the spatiotemporal evolution and heterogeneity of ozone concentration. In this work, we propose an innovative method to achieve better regression fitting results. We utilize the excellent properties of wavelet functions to create different wavelet kernels and develop wavelet kernel estimation packages for use. Due to the diversity of wavelet kernels, the most suitable wavelet kernel can be iteratively selected to achieve better model fitting based on the unique characteristics of different spatiotemporal data. We found that based on local weighting of spatial and temporal dimensions, this method can handle complex multidimensional spatiotemporal domains. As shown in simulation studies, the goodness of fit and mean square error of wavelet kernel regression for simulated data may be superior to the regression results of Gaussian kernel and double square kernel in certain situations. We further emphasized the potential of environmental problem methods by applying wavelet kernel estimation of spatiotemporal data to the analysis of ozone concentration, and found that longitude, latitude, and time can affect the relationship between ozone concentration, air quality, and nitrogen dioxide.</p>