Download the paper:

• Via DOI resolver
• From a password-protected webpage (send email to me at "rolf.reichle (at) nasa.gov")

Abstract:
Climate Data Records (CDR) that blend multiple satellite products are invaluable for climate studies, trend analysis and risk assessments. Knowledge of any inhomogeneities in the CDR is therefore critical for making correct inferences. This work proposes a methodology to identify the spatiotemporal extent of the inhomogeneities in a 36 year, global multisatellite soil moisture CDR as the result of changing observing systems. Inhomogeneities are detected at up to 24% of the tested pixels with spatial extent varying with satellite changeover times. Nevertheless, the contiguous periods without inhomogeneities at changeover times are generally longer than 10 years. Although the inhomogeneities have measurable impact on the derived trends, these trends are similar to those observed in ground data and land surface reanalysis, with an average error less than 0.003 m3 m−3 y−1. These results strengthen the basis of using the product for long-term studies and demonstrate the necessity of homogeneity testing of multisatellite CDRs in general.