Permafrost Distribution Modeling in the Semi-arid Chilean Andes
Abstract
lMountain permafrost and rock glaciers in the Andes are of growing interest due to the increase in human activities in this remote area and new environmental regulations. Empirical models of mountain permafrost distribution based on the spatial analysis of intact and relict rock glaciers and Mean Annual Air Temperature (MAAT) have been established as a tool for regional-scale assessments of permafrost favorability across entire mountain ranges. In this approach, bias adjustments are made to account for differences in permafrost presence within rock glaciers and in general debris-covered high-mountain terrain (Boeckli et al., 2012). This method allows for a rapid initial assessment of favorability for the presence of mountain permafrost in data-scarce vast mountain regions. In the present study, this methodology is applied to map a Permafrost Favorability Index (PFI) throughout the semi-arid Andes of central Chile (29° S-32°S), excluding areas of exposed bedrock. For this purpose, an inventory of relict and intact rock glaciers was compiled for the study region, MAAT was regionalized based on scarce meteorological data available in the study region, especially for high altitude area, and permafrost favorability was spatially modeled empirically based on MAAT and potential incoming solar radiation (PISR). The compiled inventory includes 3575 rock glaciers in the, Huasco, Elqui, Limarí and Choapa watersheds in Central Chile (total area: 38651 km 2 , including low elevation areas), 1664 of which were classified as relict and therefore as indicative of permafrost absence; 51 rock glaciers were removed before modeling due to their exceptional locations. Glaciers in the study region are limited to the highest parts of the Elqui and Huasco catchments. The estimated total ice volume of rock glaciers in the study region is about 2.5 times as large as the total glacier ice volume (Azócar and Brenning, 2010). MAAT for the 1981-2010 period was regionalized based on 116 station-years of annual air temperatures at 11 weather stations located between 2150 and 4000 m a.s.l. Due to the heterogeneity of time periods covered, a linear mixed-effects model with elevation and latitude as fixed-effects predictors and year as a random effect was used to account for spatial trends and for year-to-year variations, and MAAT was predicted from this model by averaging over the random effects. Figure 1: Permafrost favorability index (PFI) in the semi-arid Chilean Andes based on the permafrost distribution model for debris areas A Generalized Additive Model (GAM) was built to model the activity status of 3524 rock glaciers based on the (nonlinear) interaction term of regionalized MAAT and PISR derived from the DEM. A Perma-frost Favorability Index (PFI: ranges from 0 to 1) was obtained by adjusting model predictions. A temperature offset was applied to correct overestimation of 1005