Non-Parametric Responses of Aboveground Biomass and NDVI to Land Surface Parameters in Arctic-Alpine Environments

Abstract:

Aboveground biomass (AGB) is an important carbon pool and it affects various phenomena in Arctic and alpine areas, e.g. biodiversity, surface albedo and soil conditions. The growing availability of high-resolution digital elevation models (DEM) makes it possible to utilize topographical information for modeling local ground surface conditions globally.

We investigated the effect of topography on field measured AGB (n = 359) and its commonly used proxy, the Normalized Difference Vegetation Index (NDVI) calculated from SPOT 5 imagery. The study area located in an Arctic-alpine treeline environment (69 °N, 21 °E). We performed the analyses with boosted regression trees method by using elevation and four land surface parameters (LSPs), derived from 10 m DEM, as predictors. The LSPs were namely Potential Incoming Solar Radiation (PISR, MJ m^-2 a^-1), Topographic Position Index (TPI, r = 300 m), Slope (angle in degrees) and Topographic Wetness Index (TWI).

AGB varied from 0 to 5647 g m^-2, while median AGB of the data was 449 g m^-2. The explained deviance of the AGB and NDVI models were 53 % and 65 %, respectively. Elevation and PISR were the most important predictors. Their interaction was also significant in both cases as the highest AGB were at low-elevation, high-radiation sites, which implicates that PISR significantly improves the modelling of temperature related growing conditions. TWI had no clear effect to AGB nor to NDVI. TPI and Slope had a minor effect on AGB, but no effect to NDVI. Areas lower than their surroundings (negative TPI) had relatively high AGB. Furthermore, steeper slopes had higher AGB compared to flat sites. This is probably caused by the presence of mountain birch (Betula pubescens ssp. czerepanovii), which favors protected and steeper topography.

Local topography is an important driver of the fine scale AGB patterns. Thus, DEM derived LSPs should be taken into account when modelling current and future biomass distributions in Arctic and alpine areas. Moreover, LSPs can be useful when downscaling coarse resolution NDVI data.