Statistical modelling of mountain permafrost distribution: Local calibration and incorporation of remotely sensed data

Field mapping of mountain permafrost is laborious and is generally based on interpolation between point information. A spatial model that is based on elevation and a parameterization of solar radiation during summer is presented here. It allows estimation of permafrost distribution and can be calibrated locally, based on bottom temperature of snow (BTS) measurements or other indicators such as mapped features of permafrost creep. Local calibration makes this approach flexible and allows application in various mountain ranges. Model output consists of a continuous field of simulated BTS values that are subsequently divided into the classes 'permafrost likely','permafrost possible' and 'no permafrost' following the rules of thumb established for BTS field measurements in the Alps. Additionally, the simulated BTS values can be interpreted as a crude proxy for ground temperature regime and sensitivity to permafrost degradation. A map of vegetation abundance derived from atmospherically and topographically corrected satellite imagery was incorporated into this model to enhance the accuracy of the prediction. Based on the same corrected satellite image, a map of albedo was derived and used to calculate net short-wave radiation, in an attempt to increase model accuracy. However, the statistical relationship with BTS did not improve. This is probably due to the correlation of short-wave solar radiation with snow-melt patterns or other factors of permafrost distribution which are being influenced differently by the introduction of albedo. Copyright (C) 2001 John Wiley & Sons, Ltd.