With increased concern about the nature of global climate change, a program has been established to develop techniques for monitoring changes in cryosphere parameters using remotely sensed data as a primary data source. This research program, known as CRYSYS (CRYospheric SYStem), is focussing on monitoring surface snow and ice conditions in order to monitor climate-driven processes that influence the cryosphere. One component of the cryosphere that is of particular interest is permafrost. The primary objective of this research was to evaluate three supervised classification schemes (maximum likelihood, evidential reasoning, and a neural network) in the prediction and mapping of depth to late-summer frozen ground (DTFG) in the widespread discontinuous permafrost zone of the boreal forest of central Yukon. Source imagery used in the classifications was composed of TM- and DEM-derived data known to be correlated with DTFG. Results of a two-class DTFG experiment indicate that all tested classifiers are suitable for generating two-class correlative DTFG data products in the Mayo region. The neural network classifier was found to be most successful, producing a two-class DTFG image with a 93% agreement rate between predicted and field-measured DTFG classes. Land cover and equivalent latitude were consistently found to be especially useful sources for use in the classifications. When three DTFG classes were used, agreement rates greatly decreased for all classifiers, supporting field observations that suggest that only two DTFG classes exist in the study area.