Expanding cryospheric landform inventories

Abstract

There is a clear spatial discrepancy between the area potentially underlain by permafrost and the landforms recorded in the national inventory of cryospheric landforms in the Dry Andes of Argentina (∼22°–35°S). In the periglacial belt around 30°S, these areas are often covered by extensive block- and talus slopes, whose distribution and potential permafrost content have received little attention so far. We present the first geomorphological mapping and predictive modeling of these underestimated landforms in a semi-arid high Andean catchment with representative cryospheric landform cover (30°S, 69°W). Random forest models produce robust and transferable predictions of both target landforms, demonstrating a high predictive power (mean AUROC values ≥0.95 using non-spatial validation and ≥0.83 using spatial validation). By combining geomorphological mapping, predictive modeling, and geostatistical analysis of block- and talus slopes, we enhance our knowledge of their distribution characteristics, formative controls and potential ground ice content. While both landforms provide suitable site conditions for permafrost occurrence, talus slopes are expected to contain significantly higher ground ice content than blockslopes due to their more favorable characteristics for ice formation and preservation. Given their widespread distribution across almost 79% of the modeled area, block- and talus slopes constitute potentially important ground ice storages and runoff contributors that are not included in current hydrological assessments of mountain permafrost. Our results underscore the need to expand existing cryospheric landform inventories to achieve a more comprehensive quantification of underrepresented periglacial landforms and thus a realistic acquisition of cryospheric water resources in high mountain environments. The newly compiled inventories can serve as a basis for further investigations (e.g., geophysical surveys, hydrochemical analysis, permafrost distribution models) at different spatial scales.