As we observe increasing Arctic warming, we see that Arctic ecosystems are affected in many different ways. One of these ways is increasing frequency and intensity of extreme weather. The heterogeneity in vegetation responses to these changes is recognized but not fully understood. Arctic vegetation change occurs not only heterogeneous spatially, but also temporally, in for instance changes to the growing season length and timing of plant growth. Both species’ differing survival strategies and microclimatic differences have been raised as potentially important drivers to this complexity. At tthree Arctic sites, northern Sweden (Latnjajaure), Svalbard (Adventdalen) and Greenland (Blaesedalen), we have set up our experiment to run for multiple years to monitor both above and belowground microclimate and phenology. We use in-situ logger data, field observations and time-lapse cameras to observe our plots and study micro-climatic and phonologic variation. As we are collecting large amounts of images through the time-lapse cameras we are developing a machine learning algorithm to analyse for plant phenology. The algorithm will be based on a CNN structure and be trained for each specific species of flowering plants. The goal is to be able to analyse Arctic plant phenology data over larger timespans and higher landscape variability than previously possible due to restrictions caused by the remote nature and changeable weather conditions in the Arctic.