The potential of combining frequency domain reflectometry (FDR) and visible and near infrared spectroscopy (vis-NIRS) was shown to successfully assess soil bulk density (BD). However, the accuracy of both sensing tools was reported to be influenced by the soil moisture content (MC). The aim of the paper is to evaluate the influence of soil MC level on accuracy of BD assessment under semi field conditions by the combination of FDR and vis-NIRS data. Measurements of volumetric moisture content (θv) in the field and gravimetric moisture content (ω) in the laboratory using a FDR and a vis-NIRS (350–2500nm) techniques were conducted in five arable fields of different soil texture classes in Silsoe, Bedfordshire, UK. In order to account for different MC levels, three measurement campaigns were carried out during the period from July 2011 to October 2012, where a total of 300 soil samples were used, representing three natural MC levels of low (L1), medium (L2) and high (L3). L1, L2 and L3 datasets were subjected to artificial neural network (ANN) analysis to predict ω and θv based on fusion of vis-NIRS spectra and FDR output voltage, and subsequently the predicted values were substituted into a model to assess BD.Results showed that MC has large influence on both the vis-NIRS and FDR sensors for measuring ω and θv, respectively. The accuracy of BD assessment improved with soil MC increase, with root mean square error of prediction (RMSEp) values of 0.079, 0.072 and 0.061 g cm−3, for average ω of 0.106 (L1), 0.197 (L2) and 0.28 (L3) gg−1, respectively. The accuracy of θv measurement with the FDR depended on ensuring good contact with the soil, which is not the case for dry soil conditions, at which accuracy of θv measurement and BD assessment was deteriorated. It is recommended to set an optimal MC range (depending of soil texture), over which precise soil BD estimation can be certain.