Epsilon Archive for Student Projects

Abstract

Small scale variations in mire surface elevation referred to as microtopography are fundamental characteristics of mire ecosystems especially in the boreal region. Microtopography commonly classified into hummocks and hollows play a major role in several ecological, hydrological and biogeochemical processes including vegetation composition and carbon and methane dynamics. This makes microforms a crucial factor to account on when aiming for modeling ecosystem fluxes and monitoring peatland ecosystem change and resilience under shifting climatic conditions. However, quantitively modelling approaches using the technological advantages of remote sensing applications are limited so far and current methods lack of simplicity and straight forward mapping ability.

In this study a new novel and simple modelling approach for classifying mire microtopography, only based on a digital elevation model (DEM), got applied and tested on four study sites of Kulbäcksliden peatland research infrastructure (northern Sweden). Furthermore, a vegetation classification was performed on the same sites using random forest (RF) classifiers with and without microtopography as an input to evaluate the effect of microforms on the classification accuracy results.

The results indicate promising tendencies for the applicability of the new microtopographic approach even though the accuracy results point out an over estimation of hummock and hollow features, which could be resolved by adapting new height thresholds. The highest overall accuracy of for the vegetation classification was reached using all possible input parameters including microtopography. Still only minor improvements can be observed using microtopography with regards to fine resolution spectral data and the need of optimized height thresholds for microtopography.