We propose a methodology to blend soil moisture retrievals from passive (SSM/I, TRMM and AMSR-E) and active (ERS and ASCAT) microwave satellite observations and obtain an improved long-term (1992-2008) global product. The first step is to merge the three passive microwave data sets into a single "passive" product and two active microwave data sets into a single "active" product. This avoids the issue of differences in sensor specifications resulting in different absolute soil moisture values. Next, both data sets are adjusted against a common reference (GLDAS-Noah), which scales both data sets to the same dynamic range and allows the generation of a blended product. The proposed approach has the potential to be applied to existing as well as new missions.