We examined three methods of analyzing dual-probe heat-pulse (DPHP) data for estimating volumetric water content of near-surface materials where diurnal temperature variations can approach ± 20°C. The three methods of analyzing thermal responses include: (i) the single-point method (SPM); (ii) a method that uses a Levenburg–Marquardt optimization of thermal conductivity, volumetric heat capacity, and drift in ambient temperature with subsequent calculation of water content (LM-Uncorrected); and (iii) an expansion of the LM-Uncorrected method that incorporates temperature dependencies of soil thermal conductivity and optimizes on volumetric water content, apparent needle spacing, and drift in ambient temperature during the measurement (LM-Corrected). Ambient temperature measurements were made with either a soil thermistor installed away from the DPHP sensor (SPM) or from the temperature response of the DPHP sensor itself (LM-Uncorrected and LM-Corrected). Methods were tested based on measurements made in April and August 2006 from sensors installed at ground surface in the Mojave Desert. The SPM results showed that large (summertime) diurnal temperature variations led to significant oscillations of water content, recorded as high as ± 0.10 m3 m−3. The LM-Uncorrected method showed a significant reduction in water content oscillations, but variations were still large. The LM-Corrected method showed that the time series of water content was significantly smoother than uncorrected water contents (i.e., ± 0.005 m3 m−3 in August). The DPHP-derived water contents compared favorably to data from water content reflectometers installed nearby. The results show that soil water content estimates from the LM-Corrected method were substantially less affected by large ambient temperature variations.