Abstract/Summary

Peatlands are carbon-rich wetland ecosystems and represent the largest terrestrial carbon store. Although they are natural carbon sinks, damage, drainage and extraction over past decades have turned peatlands into a global carbon source. To tackle this nearly irreversible loss, peatland conservation and restoration projects on global and national levels have been increasing in numbers. High water table depth (WTD) is a highly important factor that influences peatland condition, resilience and ability to accumulate carbon. Given the extent of peatlands, a regular physical collection of data in situ, looking forward, would be impractical and difficult to accomplish, and the development of a remote sensing methods for peatland WTD monitoring would be highly beneficial. The accessibility to satellite data along with advancements in sensors, both in variety - optical, microwave, thermal, and their resolutions - spatial, spectral, and temporal, has greatly increased in the last decade. Combined with advances in image processing using cloud computing and machine learning, it has made it easier to access and process remotely sensed data. Synthetic aperture radar (SAR), with its ability to provide data regardless of the weather, has emerged as an important source of data for environmental applications. This project aimed to advance the usage of remotely sensed SAR data to predict peatland water table depth. First, a unique high resolution laboratory study was completed confirming SAR backscatter sensitivity to changes in peatland soil moisture and water table depth. This was followed by a case study for the Forsinard Flows area, where Sentinel-1 SAR data were used to build and test three models of different complexity for WTD prediction. The random forest model was found to be the most suited with an overall good temporal fit, highest correlation scores and lowest RMSE values. The model was later tested on a wider Peatland ACTION dataset, reaching an even higher score, affirming its applicability to peatlands in various conditions (near natural, degraded and undergoing restoration). In the final section of the thesis, up to twenty year-long time series of remote sensing data were analysed to investigate trends and change points in peatland restoration areas. The trends found using lower resolution satellite data from MODIS gave mixed results and would only be indicative of very abrupt changes, such as tree felling. The trends from the modelled WTD series based on Sentinel-1 data were indicative of positive trajectories towards higher WTD, following restoration. The results from this thesis suggest that remotely sensed data can be informative about changes in the WTD and overall peatland condition, can be used to look at seasonal change, and can be indicative of restoration progress and response to droughts. Recent studies have shown a close link between greenhouse gasses and peatland WTD, therefore, if methods of predicting WTD based on remotely sensed data are developed further, they ultimately could be used as a proxy for greenhouse gas emission reporting.