首页|Surface UV-assisted retrieval of spatially continuous surface ozone with high spatial transferability
Surface UV-assisted retrieval of spatially continuous surface ozone with high spatial transferability
扫码查看
点击上方二维码区域，可以放大扫码查看
原文链接
NSTL
Elsevier
With the rapid development of artificial intelligence, machine-learning has been used broadly in surface ozone retrievals. However, the accuracies of current machine-learning-based models tend to rely largely on the spatial and temporal patterns of the site observations of surface ozone instead of the internal physical and chemical mechanisms of ozone, which also limit the transferability of the retrieval models. In this study, the surface ultraviolet (UV) at 380 nm, which is an important component of the rate-determining step of ozone photochemical production, is involved as an indicator in the newly developed surface ozone retrieval algorithm. And the imputation of missing satellite observations of ozone nitrogen dioxide (NO2) column and surface UV is conducted to obtain spatially continuous surface ozone. Different validation schemes and case studies are used to comprehensively evaluate the new algorithm. With the involvement of the surface UV, the new retrieval algorithm shows the high accuracy (R2 = 0.853 and RMSE = 17.09 mu g/m3) in spatially continuous surface ozone estimation. More significantly, the new algorithm shows outstanding spatial transferability, which has been a critical challenge for machine-learning models on surface ozone estimation.
NSTL
Elsevier
