Each soil moisture data set is characterized by its specific mean value, variability, and dynamical range. For the assimilation of soil moisture observations into numerical models observation operators have to be developed, which reduce systematic differences. In this study, cumulative distribution function (CDF) matching is used to derive observation operators for TRMM Microwave Imager (TMI) derived soil moisture for the southern US, modeled soil moisture fields from the European Centre for Medium-Range Weather Forecasts (ECMWF), and model output from the Variable Infiltration Capacity model (VIC). It is found that the transferability of these observation operators in space and time strongly depends on the geographical region. In the Central US, where the assimilation of soil moisture is most promising, the observation operator exhibits little variability in time. The temporal variability in the observation operator can result in substantial differences between the modeled field and the observation. In Numerical Weather Prediction (NWP) applications, where the model tends to be updated on a regular basis, dynamic observation operators will be necessary to assimilate soil moisture. For climate studies or re-analysis projects long time series are required to define an observation operator, which correctly reproduces interannual variability.