Abstract
Models have historically been the source of global soil moisture (SM) analyses and estimates of land–atmosphere coupling, even though they are usually calibrated and validated only locally. Satellite-based analyses have grown in fidelity and duration, offering an independent observationally based alternative. However, satellite-retrieved SM time series include random and periodic errors that degrade estimates of land–atmosphere coupling, including correlations with other variables. This study proposes a mathematical approach to adjust daily time series of the European Space Agency (ESA) Climate Change Initiative (CCI) satellite SM product using information from physically based land surface model (LSM) datasets using a Fourier transform time-filtering method to match the temporal power spectra locally to the LSMs, which tend to agree well with in situ observations. When the original and timely adjusted SM products are evaluated against ground-based SM measurements over the conterminous United States, Europe, and Australia, results show the adjusted SM has significantly improved subseasonal variability. The skill of the adjusted SM is increased in temporal correlation by ∼0.05 over all analysis domains without introducing spurious regional patterns, affirming the stochastic nature of noise in satellite estimates, and skill improvement is found for nearly all land cover classes, especially savannas and grassland. Autocorrelation-based soil moisture memory (SMM) and the derived random component of soil moisture error (SME) are used to investigate the improvement of SM features. The time filtering reduces the random noise from the satellite-based SM product that is not explainable by physically based SM dynamics; SME is usually diminished and the increased SMM is generally statistically significant.
© 2022 American Meteorological Society. For information regarding reuse of this content and general copyright information, consult the AMS Copyright Policy (www.ametsoc.org/PUBSReuseLicenses).
