The capacitance sensor Diviner 2000 was used in an open field to monitor the volumetric soil water content (θ_ν) in the Minqin Oasis of Northwest China. The objectives of the work were to evaluate the accuracy of the factory calibration equation; to obtain a soil-specific calibration as a linear and a power equation with both a saline and a tapwater reference; and to obtain a general equation for multiple soil layers. Calibration was done during an irrigation experiment by using water with different salinities. The soil was characterised by different salinities and textures across the plots and soil layers. The calibration was carried out using water of the same salinity and composition as the local irrigation water. This procedure would yield a better normalisation of the scaled frequency (SF) for each irrigation treatment. For the treatments under study, the accuracy of the factory calibration equation was not good (range of the root mean square error, RMSE: 0.048–0.171 cm³ cm^–3). The accuracy of the soil-specific field calibration was higher. We therefore concluded that a soil-specific calibration is necessary. The RMSE shows no difference for the calibration equation with either tap or saline water reference. The linear and power equations, determined with a tap and saline water reference for each treatment and layer, gave larger differences in θ_ν, Δθ_ν, at high SF than at low SF values. The Δθ_v values increased with increasing salinity, i.e. from the control to the most saline water irrigation treatment. The linear equation had a significantly different slope for the 0.3–0.6 m layer, which has higher clay content. Pooling in a single equation of the data from multiple layers and/or treatments was satisfactory for the 0.3–0.6 m layer of the treatments W₂ and W₅, with RMSE of 0.026, with the higher clay content.