Variability in vegetation cover and albedo across latitudinal tree line

Latitudinal tree line represents the interface between boreal forests and tundra ecosystems, and is the northern limit of the world's forest. The transition between tundra and forest is often gradual, occurring over tens to hundreds of kilometers. This gradient represents a substantial change in the biophysical properties of the earth surface, one that is particularly important in ecosystems that are snow covered for much of the year. Tree line is, however, commonly delineated by the point of the northern most tree, with no gradient between forested and non-forested ecosystems. As a consequence crisp delineations of tree line incorporated into models introduce error in surface radiation budgets due to inaccurate albedo representations. Errors in modeled carbon and water fluxes are likely as well.

Here we use satellite observations to quantify several key biophysical properties across latitudinal tree line for a series of sites throughout the pan-boreal region We find decreases in NDVI and increases in albedo across the transition from boreal forest to tundra, as expected. However, in the absence of topographical barriers we find that these transitions can occur over upwards of 100 km, and that biophysical properties characteristic of tundra ecosystems can occur as far as 100 km south of tree line. Albedo is compared to NDVI and several satellite derived proxies for tree density (e.g. Vegetation Continuous Fields) in order to test for the existence of a sharp gradient that may constitute of biophysical tree line. Our results indicate that the biophysical characteristics of latitudinal vegetation gradients are poorly characterized by traditional tree line representations. Consequently, it is likely that land surface models overestimate surface radiation budgets and carbon fluxes in the boreal biome. When satellite observations of albedo are compared to GCM's from the AR5 model archive we find large differences between modeled and observed albedo under both snow covered and snow-free conditions. Model errors are generally not systematic, varying between models, and regionally within models. These results suggest that land surface characterizations are likely to cause errors in modeled energy budgets for the Arctic, with potential implications for regional climate feedbacks. This will be particularly important for modeling the climate feed back effects associated with vegetation change in the Arctic.