Quantitative knowledge of the spatial distribution of tissue temperature is an essential indicator of thermal therapy progress, and is needed to assure treatment safety and efficacy. Measured changes in ultrasonic travel time from an imaging transducer to backscattering sites in and around the heated region can provide clinically useful temperature estimates from which thermal dose throughout the imaged region can be obtained. Previous studies have noted the difficulty in inverting travel time to obtain temperature, due to lack of sensitivity over a temperature interval often encompassing the therapeutic range. It is shown, through in vitro experiments, that temperature rise can be accurately obtained from ultrasonic measurements during therapy delivery and post-treatment cool down phases, using RF backscatter data, collected by a commercial scanner, and a heat transfer model. The temperature estimation problem is divided into two parts: noninvasive HIFU calibration experiments are conducted prior to therapy to estimate local model-relevant tissue properties; temperature rise is estimated during therapy. Variability in the observed dose response is modeled as a directly related change in the magnitude of the HIFU heat source, while assuming that the acoustic beam pattern is constant.