Applying object-based segmentation in the temporal domain to characterise snow seasonality

Abstract In the context of a changing climate it is important to be able to monitor and map descriptors of snow seasonality. Because of its relatively low elevation range, Australia’s alpine bioregion is a marginal area for seasonal snow-cover with high inter-annual variability. It has been predicted that snow-cover will become increasingly ephemeral within the alpine bioregion as warming continues. To assist the monitoring of snow seasonality and ephemeral snow-cover, a remote sensing method is proposed. The method adapted principles of object-based image analysis that have traditionally be used in the spatial domain and applied them in the temporal domain. The method allows for a more comprehensive characterisation of snow seasonality relative to other methods. Using high-temporal resolution (daily) MODIS image time-series, remotely sensed descriptors were derived and validated using in situ observations. Overall, moderate to strong relationships were observed between the remotely sensed descriptors of the persistent snow-covered period (start r  = 0.70, p  r  = 0.88, p  r  = 0.88, p  in situ counterparts. Although only weak correspondence ( r  = 0.39, p  in situ observations relative to the remotely sense observations. For 2009, the mapped results for the number of snow-cover events suggested that snow-cover between 1400 and 1799 m was characterised by a high numbers of ephemeral events.