Analysis of the global ESA GlobPermafrost map for Scandinavia

Abstract: Due to its high vulnerability, permafrost is one of the key features studied in the field of climate change impacts. Permafrost is widespread in the Arctic region. The majority of the area underlain by permafrost is however difficult to access for in-situ monitoring and it is difficult to get an overview of the current state of permafrost in many areas. Permafrost modeling provides a solution which overcomes this difficulty and allows studies on permafrost distribution as well as some characteristics, i.e. ground temperatures over large remote areas. Temperature at the top of the permafrost (TTOP) is one of several permafrost modeling approaches which conceptually represents a steady-state equilibrium model. In this study, two TTOP-based models were used; the GlobPermafrost model which was used to produce the most recent global permafrost map (Alfred-Wegener-Institut) and a local Scandinavian model. The aim of this study was twofold; firstly, the performance of the GlobPermafrost model in Scandinavia was analyzed by comparing the model output with the output from the local Scandinavian model. Secondly, the role of land cover data as an input variable in the TTOP model was investigated. The TTOP-based GlobPermafrost model was run with different land cover input data to evaluate this. In general, the GlobPermafrost model underestimated permafrost occurrence in Scandinavia (overall r2 being 0.39). The lowest underestimation is located in the regions with little or no permafrost. The biggest underestimations are found in peatlands and mountainous areas with more likely permafrost occurrence. Unexpected underestimation of permafrost was observed in the forests. This exposed the weaknesses of regional permafrost model overestimating permafrost occurrence in forests. The rerun of the GlobPermafrost model with three times more detailed land cover input data did surprisingly not have a great effect on the model performance (r2 only changed by 8%). The small changes detected in the GlobPermafrost output could be explained by the changes in wetland fraction between the two land cover datasets used as input to the GlobPermafrost model. The overall conclusions from this study are 1) that the GlobPermafrost model underestimates the amount of permafrost in the study area, especially in the mountains and 2) that improved input land cover data was only of minor importance to the TTOP model performance and future research should hence focus on other forcing input data to improve model performance.