IGARSS 2019 - 2019 IEEE International Geoscience and Remote Sensing Symposium | 2019
Downscaling Ocean Surface Net Radiation at Global Scales with Random Forest
Abstract
The ocean surface net radiation (Rn) characterizing the ocean surface radiation budget, is a key variable of interest in ocean climate modeling and analysis, which determines the climate model of ocean surface with heat, freshwater and momentum flux. Most available ocean surface Rn data face various issues in accuracy, spatiotemporal patterns, and coarse spatial resolutions. J-OFURO3 (The third generation of Japanese Ocean Flux Datasets with Use of Remote Sensing Observations) released recently is considered to be one of the best datasets in ocean Rn, but its spatial resolution is also at 0.25°. In this study, we downscaled ocean surface Rn from J-OFURO3 using AVHRR TOA observations with Random Forest (RF) method. The accuracy of the downscaling daily surface ocean Rn is satisfactory with a root mean square error (RMSE) of 29Wm−2, which is better than other reanalysis data and comparable to that of CERES-SYN.