Mapping fine-scale vegetation patterns as ecological indicators for peatland biodiversity and carbon sequestration

Northern peatlands provide key climate regulating services by sequestering and storing atmospheric carbon as peat, but they are also habitat for highly specialised plant and animal species. Habitat suitability and peat accumulation rate in peatlands are strongly related to vegetation structure (species composition, biomass) and its spatial organisation (microforms). Diversity in vegetation patterns therefore act as an ecological indicator for peatland functioning.

Although microforms and their associated plant species only occur at fine spatial scales (varying from 1–10m to 0.01–1m respectively), their patterning is often repeated on the scale of whole peatlands. Consequently, remote sensing applications have recently gained much attention in this ecosystem for their potential role in landscape-scale mapping and monitoring of fine-scale vegetation patterns and functions. However, standardized methods to optimize such approaches are currently lacking or non-existent. For this reason, we set out to develop remote sensing methodology that can accurately, efficiently, flexibly, and cheaply map the distribution of microforms and plant functional types for a variety of peatland types, spatial scales, and research goals. To this end, we collected very high-resolution drone imagery (spectral and topographical) from eight Irish peatlands in 2021 and 2022 (from 1–250ha) using a consumer-grade drone with RGB camera sensor. Hereafter, we thoroughly evaluated to what extent classification accuracy and total processing time from imagery capture to final map was affected by various flight parameters (flight altitude, image overlap), image processing parameters (spatial resolution, segmentation scale, training/testing sample size), and pattern complexity (spatial patch characteristics).

The results of our study show that peatland vegetation patterns could both accurately and efficiently be classified and mapped using drone imagery, independent of pattern complexity. We also found that flying at the maximum legal flight altitude of 120m is significantly more efficient than flying at any lower altitudes because the spatial resolution of drone imagery at 120m is most often much higher than the size of peatland vegetation patterns. Flying at lower altitudes thus introduces more internal heterogeneity within plants.  However, our results also indicate that minimum spatial resolution requirements for mapping microforms and plant functional types varied notably among the studied peatlands (ranging from 0.1–1m), and showed strong relationships with spatial patch characteristics of microforms. This suggest that spatial resolution requirements in heterogeneous landscapes are not only simply driven by the types of vegetation that are present, but also by their spatial organisation.

Taken together, our results show that peatlands lend themselves very well for drone-based, landscape-scale mapping and monitoring of vegetation patterns because of the affordability, flexibility, and ease by which drones can collect and process very high-resolution spectral and topographical data. Yet, given the tremendous scale at which peatlands can in the landscape, we urge development of nested drone-satellite approaches to further improve upscaling of fine-scale vegetation patterns and their functions to the regional and global scale.