Downscaling of the European Space Agency's CCI Soil Moisture Product Based on Artificial Neural Network
Listed in
This article is not in any list yet, why not save it to one of your lists.Abstract
The European Space Agency's CCI soil moisture (SM) product spans from 1978 to now with 0.25° scale. Downscaling of CCI SM can estimate high resolution data, but it is easily affected by the scale invariance assumption. The applicability of this assumption requires further exploration at global scale. The artificial neural network (ANN) method is used to downscale daily CCI SM in 2020 from 0.25° to 0.05° under scale invariance assumption in the study. It shows that the downscaled SM (DSM) provides more abundant detailed spatial information and decreases the data gaps by 20% compared with CCI SM. The evaluations against in-situ data demonstrate that the temporal accuracy of DSM is not inferior to CCI SM with global average accuracy of r = 0.580, rmse = 0.091 m 3 /m 3 , bias=-0.039 m 3 /m 3 and ubrmse = 0.057 m 3 /m 3 . Moreover, the 100 downscaling fitting formulas with different accuracies are constructed by ANN and then the downscaling performances between them are analyzed. It suggests that there is a very good positive linear relationship between accuracy of downscaling model and accuracy of DSM verifying the applicability of scale invariance assumption. Therefore, the study will play an important role in promoting the application and research of CCI SM.
