The thermal state of permafrost is a crucial indicator of environmental changes occurring in the Arctic. The monitoring of ground temperatures in Svalbard has been carried out in instrumented boreholes, although only few are deeper than 10 m and none are located in southern part of Spitsbergen. Only one of them, Janssonhaugen, located in central part of the island, provides the ground temperature data down to 100 m. Recent studies have proved that significant warming of the ground surface temperatures, observed especially in the last three decades, can be detected not only just few meters below the surface, but reaches much deeper layers. The aim of this paper is evaluation of the permafrost state in the vicinity of the Polish Polar Station in Hornsund using the numerical heat transfer model CryoGrid 2. The model is calibrated with ground temperature data collected from a 2 m deep borehole established in 2013 and then validated with data from the period 1990-2014 from five depths up to 1 m, measured routinely at the Hornsund meteorological station. The study estimates modelled ground thermal profile down to 100 m in depth and presents the evolution of the ground thermal regime in the last 25 years. The simulated subsurface temperature trumpet shows that multiannual variability in that period can reach 25 m in depth. The changes of the ground thermal regime correspond to an increasing trend of air temperatures observed in Hornsund and general warming across Svalbard.

In the framework of non-destructive evaluation (NDE), an accurate and precise characterization of defects is fundamental. This paper proposes a novel method for characterization of partial detachment of thermal barrier coatings from metallic surfaces, using the long pulsed thermography (LPT). There exist many applications, in which the LPT technique provides clear and intelligible thermograms. The introduced method comprises a series of post-processing operations of the thermal images. The purpose is to improve the linear fit of the cooling stage of the surface under investigation in the logarithmic scale. To this end, additional fit parameters are introduced. Such parameters, defined as damage classifiers, are represented as image maps, allowing for a straightforward localization of the defects. The defect size information provided by each classifier is, then, obtained by means of an automatic segmentation of the images. The main advantages of the proposed technique are the automaticity (due to the image segmentation procedures) and relatively limited uncertainties in the estimation of the defect size.