Scale determines the detection of icing effects on tundra productivity in Svalbard
Global warming increases the frequency and severity of thaw-freeze cycles and rain-on-snow events in the Arctic, leading to the encasement of vegetation in basal ice (icing) rather than insulating snow. Plot-scale experimental icing has shown to initially delay and eventually enhance tundra productivity, but we do not yet know whether this response can also be detected in iced tundra at the landscape scale. Here, I analyse NDVI metrics of iced and ice-free tundra at the plot scale from an in-situ icing experiment over six years in Svalbard. I further assess whether plot-scale NDVI metrics agree with metrics obtained from Sentinel 2 imagery for similar habitat. I then analyse NDVI metrics from Sentinel 2 imagery over three years at locations with recorded ice presence and absence at the landscape scale. I found that plot-scale icing slightly increased and consistently delayed peak NDVI by four days. Plot NDVI metrics of ice-free tundra agreed well with Sentinel 2 metrics of the exact same locations, but differed for Sentinel 2 metrics of similar habitat in surrounding locations. At the landscape scale, ice presence marginally increased and advanced peak NDVI from Sentinel 2 imagery, but only at high elevations and in certain vegetation categories. My findings suggest that the 10 m spatial resolution of Sentinel 2 imagery, or the temporal span of three years used here, might not be enough to detect the effect of localised icing in the Arctic and thus prove unsuitable to upscale directly from plot to landscape. As the paces of both climate change and technological advances in remote sensing continue to accelerate, we need to further link in-situ and remotely sensed observations to capture the full picture of heterogenous change in Arctic tundra productivity.