Abstract

Peatlands are key reservoirs of belowground carbon (C) stock and their monitoring is important to assess the rapid changes in the C cycle caused by climate change and anthropogenic impacts. Frequently, information of peatland area and vegetation type estimated by remote sensing are used along with soil measurements and allometric functions to estimate belowground C stock. Despite the accuracy of such approaches, there is still the need of finding mappable proxies that allows an easier prediction with remote sensing data alone to reduce field and laboratory efforts. Therefore, we assessed the use of aboveground vegetation attributes as proxies to predict peatland belowground C stock. First, the ecological relations between remotely detectable vegetation attributes (i.e. vegetation heights, aboveground biomass, species richness and floristic composition) and belowground C stock were obtained using structural equation modeling (SEM). SEM was formulated using expert knowledge and trained and validated using plot-based information. Second, the SEM latent vectors were spatially mapped using random forest regressions with UAV hyperspectral and structural information. Finally, this enabled us to map belowground C stock using the SEM functions parameterized with the random forest derived maps. This approach resulted in higher accuracies in relation to a direct application of random forest with UAV data, with improvements from r² of 0.39 to 0.54, normalized RMSE of 31.33% to 20.24% and bias of -0.73 to 0.05. Our case study showed that (1) vegetation attributes, especially vegetation height, species richness and aboveground biomass, can be good proxies to estimate peatland belowground C stock as they are relatively easy to obtain using remote sensing data, but may hold strong relationships with the belowground C gradient; and (2) SEM is good to include mechanistic knowledge in empirical modeling approaches.