A physically-based model for estimating frost penetration is used to develop a climatology of soil frost penetration extremes. Since the model is based on meteorological variables which are observed at U.S. Cooperative Network stations, frost depth statistics can be derived for approximately 500 sites within the northeastern United States.
The theoretical basis for the frost penetration model is that the process is driven primarily by thermal diffusion. At the lower boundary an approximately constant temperature and negligible heat flux are assumed. The upper boundary condition is given by the observed average daily air temperature. Layers of variable depth in the snow/soil system are then defined by frozen and unfrozen zones, the boundaries of which are at 0ūC. Imbalances between the resulting vertical heat fluxes are made up through either freezing or thawing a sufficient depth of soil.
Using model simulations, a suite of candidate extreme value distributions are screened for use in representing annual frost penetration extremes. This screening is based on a bootstrapping procedure in which data from stations with relatively long climatological records are used to construct a series of model estimates. Random samples of 30 are then repeatedly withdrawn from these records and used to fit each candidate distribution. Extropolations to the exceedence probabilies of the largest points in the parent data series are made and used to evaluate the bias and precision of each distribution. Based on this analysis the Gumbel distribution was chosen as providing the best representation of the soil freezing extremes.
Art DeGaetano Northeast Regional Climate Center 1115 Bradfield Hall Cornell University Ithaca, NY 14853 Telephone: (607)255-0385 Fax: (607)255-2106 E-mail:degaetano@metvax.cit.cornell.edu Daniel S. Wilks Department of Soil Crop and Atmospheric Sciences 1113 Bradfield Hall Cornell University Ithaca, NY 14853 Megan McKay Northeast Regional Climate Center 1123 Bradfield Hall Cornell University Ithaca, NY 14853