Assessing the tundra-boreal transition zone with Sentinel-1/2–derived landcover data

Geographical and ecological research of Arctic tundra requires a clear definition of the biome boundaries. A commonly used dataset for the Arctic is the tundra biome boundary mapped by the Circumpolar Arctic Vegetation Mapping Project (CAVM, 2003). Given recent advances in satellite-image resolution and data availability, in the age of the rapidly changing Arctic climate this previously established delineation needs to be re-evaluated and, where necessary, updated. An updated boundary can support applications such as Arctic climate modelling and assessments of potential disturbances in permafrost rich regions.

A recently developed landcover dataset was investigated for this study. The Circumpolar Landcover Unit (CALU) Database provides highly detailed landcover information with a spatial resolution of 10 meters and consists of 23 thematic units, including 12 units representing tundra but also 3 forest classes. The used retrieval scheme of landcover units employed provides an unprecedented level of detail. The landcover units have been derived by fusion of satellite data using Sentinel-1 (synthetic aperture radar) and Sentinel-2 (multispectral). These units reflect gradients in moisture and vegetation structure. The available spatial detail of CALU has been already shown to provide the means to assess the complexity of lowland permafrost regions.

The original CALU database of version 1.0 covered the Arctic within the CAVM extent only. The latest version 2.0 partially extends further south, providing additional detail within the transition zone for many areas.

The aim of this study was to assess the southern boundary of the CAVM and to identify regions where further developments of the CALU dataset may aid to establish a new boundary. Spatial statistics were collected within selected buffer areas of the CAVM boundary. In addition, longitudinal zones were generated to test whether forest-related CALU classes systematically peak south of the currently mapped boundary.

Based on these statistics, in regions such as Alaska and the European part of Russia, the CAVM boundary generally corresponds well with CALU, with forest-related classes mostly dominating within the buffer area. In parts of Siberia and Canada, however, shrub-tundra classes are more prevalent, while forest-related classes occur farther south. This mismatch may reflect regional differences in vegetation structure and terrain-driven zonation, suggesting that a single latitudinal boundary product may not capture local transitions equally well everywhere.

Preliminary results indicate that in several regions the CALU database should be extended further south, because current coverage does not fully include forest-related classes. This limitation affects the use of CALU for the tundra-boreal biome boundary evaluation and for applications that require a consistent representation of tundra–taiga transitions.

CALU: Bartsch, A., Khairullin, R., Efimova, A., Widhalm, B., Muri, X., von Baeckmann, C., Bergstedt, H., Ermokhina, K., Hugelius, G., Heim, B., Leibman, M., & Gruber, C. (2024). Circumarctic Landcover Units (2.0) [Data set]. Zenodo. https://doi.org/10.5281/zenodo.14235736