# Assessing Species Fractional Cover and α‐Diversity in Boreal Peatlands Across Trophic Levels Using Hyperspectral Data

## TL;DR

This study explores how hyperspectral data can help map and monitor biodiversity in boreal peatlands, which are crucial for carbon storage.

## Contribution

The paper introduces a novel approach using close-range spectral data to predict species cover and diversity in peatlands.

## Key findings

- Hyperspectral data can predict species fractional cover with moderate accuracy (R² = 0.58).
- Multispectral data sometimes outperformed hyperspectral data in predicting litter cover in ombrotrophic peatlands.
- Hyperspectral data better predicted α-diversity in ombrotrophic than minerotrophic habitats (R² = 0.44 vs. 0.22).

## Abstract

Boreal peatlands, which act as significant sinks and storage of global soil organic carbon, are increasingly threatened by the changing climate conditions as well as land use changes. Despite the importance of these ecosystems, their vegetation and ecological features remain poorly mapped compared to other terrestrial ecosystems. Hyperspectral satellite imaging shows promise for detailed vegetation mapping and biodiversity monitoring of boreal peatlands. However, its effective application requires a fundamental understanding of the spectral properties of the vegetation communities of boreal peatlands. To address this, we combined newly available, open‐source data consisting of close‐range sensed spectral libraries of boreal peatland vegetation communities and single species. Our aim was to examine the extent to which close‐range spectral data can be used to predict species‐specific fractional cover in minerotrophic and ombrotrophic peatland habitats using hyperspectral and multispectral data, and to assess the connection between spectral signatures and α‐diversity of the vegetation communities. Our findings show that hyperspectral data can be used to predict the fractional cover of certain plant species with moderate accuracy (R
2 = 0.58). When comparing data types, hyperspectral data typically produced slightly better model fits for species with larger sample sizes, appearing to be superior to multispectral data. However, in certain cases, such as in the prediction of litter cover in ombrotrophic peatland habitats, multispectral data yielded marginally better results (R
2 = 0.4–0.45). Furthermore, using hyperspectral data, we observed that the prediction of α‐diversity of the ombrotrophic habitats was moderately better (R
2 = 0.44) than that of the minerotrophic habitats (R
2 = 0.22). These results enhance our understanding of the spectral properties of the complex, multilayered vegetation communities and thus aid in the mapping of these vital ecosystems.

The reflectance spectra of minerotrophic and ombrotrophic peatland vegetation communities vary notably.