Revealing Spatio-Temporal Dynamics of Arctic Shrub Expansion: Utilizing Vegetation Cover Fractions from Landsat Time Series

julia.boike [ at ]


Warming induced rapid and extensive alterations of tundra ecosystem form and functioning, including an increased abundance of shrub species, entails profound implications on pan-Arctic and global scale. In order to disentangle carbon and energy fluxes and assess climate feedbacks, we require a better understanding of tundra vegetation composition and structure in general, and of shrub expansion in particular. Remote sensing provides the unique opportunity of assessing ecological change at high spatial and temporal detail within the vast and remote landscapes of the Arctic tundra biome. Yet current satellite-based approaches are often restrained by scarce field observations, constraining the spatio-temporal dimensions that are needed to assess shrubification processes. Here, I estimated the fractional cover of four major Plant Functional Types (PFTs), including shrubs, and other land cover classes using multi-seasonal image features derived from Landsat acquisitions within the greater Mackenzie Delta region for 1984–1988, 1999–2002 and 2017–2020. I deployed regression-based unmixing using Kernel Ridge Regression (KRR) and Random Forest Regression (RFR) in combination with multi-temporal synthetic training data generated from pure endmember spectra obtained directly from the imagery. The method facilitates capturing the fine-scale heterogeneous nature of Arctic vegetation types across space and time, acknowledging differences between algorithm choice and target classes. Independent validation based on very-high-resolution airborne and drone acquisitions suggests that KRR outperformed RFR with a good prediction accuracy for shrubs (MAE = 11.7 %) and other land cover classes (MAEs = 1.0–11.9 %). The multitemporal predictions revealed intense shrub expansion of on average 2.3 –4.7 % per decade across much of the study area. The spatio-temporal patterns suggest that the mechanisms and hotspots of shrubification have shifted from an infilling of existing patches in the shrub dominated tundra, towards a latitudinal expansion into low-statured tundra communities in recent decades. Simultaneously, I found a widespread decline in herbaceous vegetation cover, partly in conjunction with shrub expansion, corroborating evidence and projections of the replacement and homogenisation of vegetation communities facilitated by the competitive advantage of certain shrub species under a warming climate. At large, the method applied, and the maps generated, initiate new opportunities for mapping past and present land cover fractions and advance our spatio-temporal understanding of Arctic shrub expansion within the vast and heterogeneous tundra biome.

Item Type
Thesis (Master)
Primary Division
Primary Topic
Helmholtz Cross Cutting Activity (2021-2027)
Publication Status
Eprint ID
Cite as
Nill, L. (2021): Revealing Spatio-Temporal Dynamics of Arctic Shrub Expansion: Utilizing Vegetation Cover Fractions from Landsat Time Series , Master thesis, Geographisches Institut der Humboldt-Universität zu Berlin.


Geographical region

Research Platforms


Edit Item Edit Item