Local scale assessment of peatland restoration priorities using multi-sensor airborne remote sensing and machine learning

• Multi-sensor airborne data support monitoring of peatlands at a local scale. • Peatland plant communities mapped into 22 classes with 79% overall accuracy. • Novel desiccation index quantifies short-term peatland desiccation (∼16 % RMSE, ∼12.5 % MAE) • Bog-type peatlands show lowest resilience, fen-type peatlands show highest. • Results identify priority sites for targeted peatland restoration actions. Advances in remote sensing now enable comprehensive monitoring of peatlands, providing detailed data on vegetation composition and hydrological conditions. The main objective of this study was to assess the potential of multi-sensor airborne data for monitoring the short-term drought resilience of peat-forming vegetation at the local scale. The study covered 248 peatlands within the Tuchola Forest (northern Poland), for which three types of airborne data were acquired in 2022 using sensor fusion: LiDAR, hyperspectral, and thermal imagery, along with ground-based reference data. Machine learning techniques were applied to develop: (a) a map of peatland plant communities comprising 22 legend classes, which achieved an overall accuracy of 79 %, and (b) a spatial desiccation model of peatland vegetation − constructed for a selected subset of eight peatlands − based on the novel Multifactorial Peatland Desiccation Index. The model achieved a Root Mean Squared Error (RMSE) of 1.898 (∼16 %) and a Mean Absolute Error (MAE) of 1.536 (∼12.5 %). Classification results revealed that non-peat-forming vegetation types cover nearly 67 % of the delineated peatland area, reflecting its degradation. Analysis of desiccation across peat-forming plant communities shows that the lowest short-term drought resilience is observed in Ledo-Sphagnetum magellanici and Sphagno recurvi-Eriophoretum vaginati communities, typical of bog-type peatlands. In contrast, the highest resilience was found in Scorpidio-Caricetum diandrae and Menyantho-Sphagnetum teretis communities, characteristic of fen-type peatlands. Assessment of the extent and desiccation status of valuable peat-forming communities enabled the identification of peatlands for highest-priority restoration.