Evaluating Testability of Permafrost Models Through Physical and Thermal Testing in Complex Mountain Terrain, Yukon, Canada

Due to its subsurface nature, permafrost cannot be directly observed with the naked eye or optical remote sensing. Consequently, accurately describing its distribution and thermal state is challenging. This is especially true in vast, remote environments, where obtaining comprehensive field data is demanding or improbable. This results in a reliance on models, which are constrained by limited, spatiotemporally fragmented baseline data, and which are rarely validated against actual field data. While such models may be sufficient in homogeneous permafrost environments, or to capture general trends over large areas, their accuracy is limited at finer scales in thermally heterogeneous environments. To explore and conceptualize this issue, we conducted a field sampling campaign in two thermally complex valleys in the Ogilvie Mountains. The study area exhibits strong ground temperature variability over short distances due to surface‐based temperature inversions, extreme aspects, and complex land cover types. Our objective was to qualitatively assess our ability to collect in situ permafrost thermal data (“testability”) using simple, repeatable, low‐cost methods. Although the study area is located within the zone of continuous permafrost, only nine (18.3%) of the 49 Cryotic Assessment Sites (CAS) produced cryotic temperatures in situ, due to substrate clast density and active layer thickness. Testability outcomes were used to develop a generalized linear model (PTEST), which predicts low testability at higher elevations and higher testability in valley bottoms, indicating a substantial potential sampling bias for model calibration and validation. Comparisons with two local permafrost models highlight persistent uncertainty in permafrost characterization in mountainous environments.