Quantifying Overestimated Permafrost Extent Driven by Rock Glacier Inventory

Rock glaciers (RGs) are normally used as ground-truth observations to indicate the presence of permafrost, and hence extensively used in training permafrost distribution models. However, the unique structure and composition of RGs enhance ground cooling effects, leading to more favorable conditions for permafrost presence than in adjacent ground. We therefore hypothesized and confirmed that permafrost extent is overestimated using RG-driven models. The results indicate that the permafrost zonation index was overestimated by about 8.4%-13.1% in a model driven by RG observations compared to a model using in situ measurements of permafrost presence/absence. The bias is particularly pronounced in discontinuous permafrost region, where it is thought to be highly sensitive to climate change, resulting in about a 41.8%-90.8% overestimation in permafrost region and 7.0%-18.6% misclassification. In order to better use the large RG datasets available to understand permafrost conditions, we provide a method to correct this bias in a fundamental model.

Automated summary

Rock glaciers used as ground-truth observations can lead to overestimation of permafrost extent in models, particularly in discontinuous permafrost regions. Correcting this bias is crucial for better understanding permafrost conditions.