Global soil moisture observations, which stand to improve flood and drought applications, are currently being produced by multiple satellite missions. One such mission is Soil Moisture Ocean Salinity (SMOS), an L-band satellite with a spatial resolution of roughly 40 km and a revisit time of less than 3 days. SMOS is both too dry and too noisy (bias = -0.072 m3 m−3, ubRMSE = 0.061 m3 m−3) during the growing season (Apr – Oct, 2013 – 2015) over an in situ soil moisture network in the South Fork of the Iowa River (SFIR) watershed. The mission accuracy goals are to have a zero-bias and an ubRMSE less than 0.04 m3 m−3 . We hypothesized that the SMOS dry bias could be caused by: the inclusion of invalid retrievals; bias in the auxiliary surface temperature input; errors in auxiliary soil textural maps; and the use of a non-representative parameterization of scattering in the canopy. Following the examination of SMOS theta v retrieval validity, we implemented two end-user filters: a strict instantaneous radio frequency interference (IRRFI) filter and a X² probability filter. The use of these filters restricts the number of the theta v retrievals to 25 per pixel per month (unfiltered: 32 per pixel per month). Bias in the effective ground temperature (Tg), derived from the “AUX_ECMWF” product, would need to be greater than -1.5 K to create a dry theta v bias. Few individual months had Tg biases large enough to impact theta v retrieval; the average bias was 0.25 K (RMSE = 1.4 K). The SMOS soil textural maps, updated in May 2015 for inter-mission comparability, corrected errors in the clay fraction over the SFIR that had previously been artificially wetting theta v retrievals (by 0.01 – 0.03 m3 m−3). Finally, scattering within the canopy, while relevant for crops such as corn, is not accounted for by default in the SMOS retrieval algorithm. Introducing a non-zero value of the single scattering albedo (omega = 0.05) dried the theta v bias by an additional 0.03 m3 m−3 during the two-month test case (Jul – Aug, 2015). While we were unable to identify the source of the SMOS dry bias in the SFIR, we made remarkable progress in understanding how the retrieval algorithm handles agricultural land surfaces. We intend to investigate soil surface roughness as another potential source of the dry bias in the near future.