Abstract
Three practical schemes for computing the snow surface temperature T~s~, i.e. the force–restore method (FRM), the surface conductance method (SCM), and the Kondo and Yamazaki method (KYM), were assessed with respect to T~s~ retrieved from cloud‐free, NOAA‐AVHRR satellite data for three land‐cover types of the Paddle River basin of central Alberta. In terms of R^2^, the mean T~s~, the t‐test and F‐test, the FRM generally simulated more accurate T~s~ than the SCM and KYM. The bias in simulated T~s~ is usually within several degrees Celsius of the NOAA‐AVHRR T~s~ for both the calibration and validation periods, but larger errors are encountered occasionally, especially when T~s~ is substantially above 0 °C. Results show that the simulated T~s~ of the FRM is more consistent than that of the SCM, which in turn was more consistent than that of the KYM. This is partly because the FRM considers two aspects of heat conduction into snow, a stationary‐mean diurnal (sinusoidal) temperature variation at the surface coupled to a near steady‐state ground heat flux, whereas the SCM assumes a near steady‐state, simple heat conduction, and other simplifying assumptions, and the KYM does not balance the snowpack heat fluxes by assuming the snowpack having a vertical temperature profile that is linear. Copyright © 2005 John Wiley & Sons, Ltd.