Near-surface measurements of soil water content (&thgr;) using time domain reflectometry (TDR) may exhibit anomalous behavior in the presence of diurnal temperature (T) fluctuations. Experimental results obtained in a companion paper led to the hypothesis that the observed bulk dielectric permittivity (ϵb) is determined by an interplay between two competing phenomena: (1) the reduction in the dielectric permittivity of bulk water with increased T; and (2) the increase in TDR-measured ϵb with increased T due to release of bound water. In this study we develop a physically based model for the temperature dependency of TDR-measured soil bulk dielectric permittivity and propose practical correction factors. The model considers the modified properties of water near solid surfaces to define a layer of rotationally hindered water (within the TDR frequency bandwidth) having a temperature dependent thickness. Changes in measured ϵb(T) are thus attributed to variations in the thickness of the rotationally hindered layer which has a lower dielectric permittivity than free water and hence is less visible to travel-time-based TDR waveform analyses. The model is sensitive to the soil specific surface area and the water content, both of which determine the ratio of bound to bulk soil water. Comparisons with experimental data covering a wide range of soils, water contents, and temperatures showed good agreement. Further studies are needed to evaluate some of the model's critical parameters such as the cutoff frequency below which water is considered bound. A temperature correction approximation is based on analytical expressions for TDR-measured bulk dielectric permittivity and requires estimates of soil specific surface area and bulk density, which may be estimated from soil texture. The thermodielectric sensitivity of TDR-measured bulk dielectric permittivity and water content may serve as a basis for estimating soil specific surface area. ¿ 1999 American Geophysical Union