Statistical Modelling of Permafrost Extent in the Circum-Arctic Region

Permafrost regions cover approximately a quarter of the Northern Hemisphere and thawing has been recorded in many different locations. The thawing process is likely to continue given that the Northern Latitudes will experience increased warming, which is known as the Arctic Amplification Factor. Changes in permafrost regions can cause changes in hydrology, biogeochemical cycles and ecosystems. Furthermore, infrastructure built on permafrost grounds such as cites and oil and gas pipelines are at high risk of collapse in the event of thawing. Additionally, there are large carbon pools stored in regions of frozen ground and release of carbon dioxide and methane could enhance the effect of global warming. While decrease in permafrost and its hazards have widely been recognized there is uncertainty about the extent of loss in permafrost area under different warming scenarios suggested by the IPCC. Models differ by their application so that global models are generally process-based and models in mountain areas often use statistical models. A previous study found a strong correlation between mean annual air temperature and current permafrost extent. This thesis investigates the relationship between growing and freezing degree days, seasonality and soil organic carbon content with regard to current permafrost extent and predicts permafrost loss for the RCP4.5 scenario. All variables are expected to show better or equal correlation coefficients as the mean annual air temperature as they give a more precise indication of the freezing conditions, which is supported by findings of previous local studies. The results show that none of the variables are a better indicator for permafrost. However, the predicted loss of permafrost by growing degree days with base temperature 5°C falls within the confidence interval of the permafrost-mean annual air temperature prediction.Permafrost describes ground that is frozen continuously for two years or more and occurs in in polar and mountainous regions in both hemispheres. Decrease in permafrost has been recorded in many regions and will continue in a warming climate. This is problematic as the thawing of these grounds can lead to changes in biogeochemical cycles, hydrology and ecosystems. Furthermore, soils in the Northern Latitudes store large amounts or carbon which are released in form of carbon dioxide and methane. This can further enhance global warming. Additionally, many cities as well as oil and gas bases are built on permafrost ground, and thawing has in some regions already lead to damage and collapse of these structures. While decrease of permafrost is very certain, studies differ widely in the extent of decrease their predictions. Therefore, many models have been developed for different geographic scales and regions. This thesis used a statistical approach to examine the correlations between mean annual air temperature, growing and freezing degree days, seasonality and soil organic carbon content to model the permafrost fractions per area. The growing degree days for base temperature 5°C and 0°C were used to make predictions of permafrost extent under the RCP4.5 scenario, which predicts a 1.5°C warming by 2100, while also accounting for the Arctic Amplification Factor. The predictions were compared to a study that calculated loss of permafrost area based the mean annual air temperature. The results show that mean annual air temperature has the strongest correlation with permafrost presence. However, both predictions for area loss with the growing degree days variables fall within the confidence interval of the predicted area loss of the mean annual air temperature. For a comprehensive assessment of the variables with regard to their application for predictions in the Circum-Polar region a dataset of a higher resolution would be needed