Abstract We examine the feasibility of retrieving root zone soil moisture and partitioning of surface fluxes through a model inversion technique using surface measurements. Using a four-layer land surface model and observed datasets from Field Experiment (FIFE) 1987, we show that the sensitivities of surface soil moisture to deeper layer soil moisture are different to those of surface fluxes. Consequently, if one chooses the initial soil moisture profile that optimizes surface soil moisture, in a root mean square error sense, it may not lead to optimal estimation of surface fluxes. We also show that the accuracy of soil moisture profile retrieval from surface measurements depends strongly on the initial surface soil moisture conditions. For wetter surface conditions, an initialization based on remotely sensed surface soil moisture appears to be adequate for the retrieval of the soil moisture profile. For drier surface conditions, however, a decoupling of surface and deeper layer soil moisture might be triggered and an initialization based on surface soil moisture would lead to larger error.