Svalbard is a hotspot of climate change in the rapidly warming Arctic. The strong air temperature warming coincides with a multitude of changes in other climate variables such as liquid precipitation, snow cover, and the surface energy budget components. These changes have highly complex effects on the soil temperature and freezing conditions. We investigate seasonal patterns of change in climate and soil conditions at the Bayelva study site close to Ny-Ålesund, Svalbard for the period 1998-2020. We use Bayesian inference to estimate trends in monthly mean values of air and soil temperature, radiation fluxes, sensible and latent heat flux, liquid precipitation, snow depth, and soil moisture. We then apply PCMCI+, a recently developed causal inference framework, in order to quantify the contributions of all meteorological variables to soil warming. Air temperature at the Bayelva site rose in all months of the year in the last 23 years (1998-2020). This trend has been particularly strong in April (1.3°C/10years), September (1.5°C/10years) and October (1.9°C/10years). The strong changes in spring and autumn led to earlier snowmelt (-14 days/10 years, 2007-2020) and more snow free days (+26 days/10years, 2007-2020). We observe later soil freezing in October and lower snow depth. Furthermore, strong rain events have become more frequent in winter, which contributed to soil warming. As a result of changes in air temperature, water fluxes, and the energy budget, top soil temperature increased in particular during spring (May/June 1.4°C/10years, 1998-2020). Our results illustrate how rapid climate change drives soil warming and permafrost thaw. They can help to validate results from climate and land surface models as well as aid in future predictions of landscape changes in Svalbard.