Spatial downscaling of surface ozone concentration calculation from remotely sensed data based on mutual information

Accurate near surface ozone concentration calculation with high spatial resolution data is very important to solve the problem of serious pollution and health impact assessment. However, existing remotely sensed products cannot meet requirements monitoring. In this study, O 3 (at 30 km resolution) was extracted from daily TROPOMI profile products. Meanwhile, study improved downscaling algorithm based on mutual information applied it mapping in China. Combined (with 5 obtained by using Light Gradient Boosting Machine (LightGBM) AOD 1 MODIS, ground has been achieved study. The downscaled were subsequently validated an independent dataset. main conclusion that entropy between bottom layer resolution), LightGBM MCD19A2 can accurately reduce layer. procedure not only resulted increase over whole area but also significant improvements precision coefficient determination ( R 2 ) increased 0.733 0.823, mean biased error decreased 7.905 μg/m 3.887 , root-mean-square 14.395 8.920 for concentration.