Abstract
Soil temperature is very important in regulating ecosystem processes, yet it is often difficult and costly to measure. Most models that have endeavored to predict soil temperature have either a long time step or several complicated independent variables. Daily mean air and soil temperatures were measured from 1989-1997 in four northern hardwood sites along a 500 km latitudinal gradient in Michigan. These data were used to derive a simple method to predict daily mean soil temperature (depth of 15 cm) using the daily mean air temperature from the previous day and a cosine function of Julian date (R super(2) identical with 0.93-0.96; SEM identical with 0.98-1.40 degree C). Predicted values were compared with actual recorded soil temperatures from 1997 at each of the sites, and the average difference between the observed and predicted values ranged from 0.11 to 0.39 degree C. Different coefficients were estimated for each of the sites; however, this general method of predicting soil temperature appears applicable to any site. Once calibrated for a given site, soil temperature may be simply estimated, thus reducing the need for extended monitoring efforts. This method also allows the reconstruction of soil temperature records beyond the monitoring period. Projecting long-term trends in soil temperature may help to further elucidate several ecosystem processes and also may provide more information on how a changing global climate will impact forest ecosystems.