Do vegetation boundaries display smooth or abrupt spatial transitions along environmental gradients? Evidence from the prairie–forest biome boundary of historic Minnesota, USA




Two alternative mechanisms of abrupt vegetation change across ecological boundaries have been proposed: (1) concomitantly abrupt gradients in physical environmental variables and vegetation across the boundary, and (2) gradual environmental gradients that vegetation responds to in a non-linear or threshold manner. Here, we evaluate spatial patterns of climate and vegetation across a grassland–forest biome boundary to examine evidence in favour of either of these alternatives.


Minnesota, USA.


Vegetation data represented the presence of prairie vs. forest vegetation in Minnesota from 1847 to 1908, generally prior to European settlement of the region, while the climatic variables represented an index of long-term average moisture availability (precipitation minus potential evapotranspiration (P – PET). Using linear and sigmoidal regression models, we evaluated spatial patterns of change in vegetation, climate and vegetation–climate relationships across 22 transects (170–400 km) oriented perpendicular to the biome boundary. We also evaluated boundary characteristics in light of dominant topographical controls and position along the boundary.


Vegetation followed a sigmoidal pattern of change across the boundary, with mean boundary width of ca. 100 km. The P – PET increased by ca. 100 mm across the boundary following a comparatively smooth pattern of change. Climate–vegetation relationships were clearly non-linear across the boundary, indicating these variables did not change in a common spatial pattern. Regional topographical controls modified relationships between vegetation and climate along the length of the boundary.


Our results document strong non-linear relationships between the presence of forest vegetation and its dominant climate control across a grassland–forest biome boundary. An average change of ca. 100 mm in P – PET moving across the boundary is about 40% of the long-term mean annual range of this variable, suggesting that modest changes to P – PET may potentially cause substantial shifts in the location of the prairie–forest boundary.