Multi-spatial-temporal remote sensing and machine learning for mapping management impacts on peatlands in Ireland

Abstract

Peatlands, constituting over half of terrestrial wetland ecosystems across the globe, hold critical ecological significance and are large stores of Carbon (C). In Ireland, the wetland landscape is dominated by rare oceanic raised bogs and blanket bogs, covering ~21% (1.46 Mha) of the land surface area. These ecosystems account for 50-75% of the soil organic carbon stock of Ireland and offer numerous ecosystem services including C storage, biodiversity support and water regulation. However, more than 90% of these ecosystems have been degraded over centuries due to artificial drainage ditches followed by land use activities like peat extraction, afforestation, and agriculture. This has an overall negative impact on the functioning of these ecosystems, shifting them from a moderate sink of C to a large source of C. Recognising the importance of these ecosystems, efforts are underway to conserve them through measures such as rewetting, restoration, and rehabilitation. Furthermore, Ireland needs to meet its 2030 climate energy framework targets related to GHG emissions from land use, land use change, and forestry, including wetlands. Despite Ireland's voluntary decision to include peatlands in this system in 2020, spatially explicit information on land use activities and associated GHG emissions from peatlands is lacking. The implementation of conservation efforts and accurate accounting of C stores and GHG emissions requires accurate identification and mapping of artificial drainage ditches and associated land use activities. The main objective of this study is to develop a robust set of methods capable of mapping and monitoring peatland land use changes using novel cloud computing, artificial intelligence (Machine Learning (ML) and Deep Learning (DL)) and multi-temporal-spatial remote sensing. The study begins with the development of a multi-level land use classification taxonomy, namely Land Use Classification for Irish Peatlands (LUCIP). The four-level classification system was used to identify land use activities. Each level of LUCIP corresponds to the spatial resolution of the sensor employed in the mapping of land use activities at that level. By utilising cloud computing, temporal mosaicking, Landsat and Sentinel data, this study developed a robust methodology that overcame cloud contamination and produced the first peatland land use and artificial drainage maps of Ireland with wall-to-wall coverage. At LUCIP level-2 four peatland land use classes were identified: industrial peat extraction, forest, grassland, and residual peatland. These were then mapped and assessed using medium resolution (30 m) Landsat (5 and 8) data combined with Google Earth Engine (GEE) and random forest ML. These four land uses were assessed across three time periods: 1990, 2005 and 2019. The overall accuracy of the classification was 86% and 85% for the 2005 and 2019 maps, respectively. The accuracy of the 1990 dataset could not be assessed due to the unavailability of high-resolution reference data. The results indicate that extensive management activities have taken place in peatlands over the past three decades. There is no notable change between 1990 and 2005. However drastic changes were observed between 2005 - 2019, with exposed peat area on Bord na Mona (BnM) landholding decreasing while afforestation and grassland increasing on these sites. Overall, there is an increase in forests and a decrease in grasslands on peatlands. The study then scopes down further to raised bogs. At LUCIP level-3 seven land use activities on Irish raised bogs peatlands were identified: cutaway, cutover, forest, grassland, remnant peatlands, waterbodies, and built-up. These were mapped using high-resolution (10m) Sentinel-2 imagery, random forest ML and GEE. The results revealed that agricultural grassland comprised 43% of the land use on raised bogs, followed by, forestry (21%), cutover (11%), cutaway (10%) remnant peatlands (13%), waterbodies and built-up ~1% each. The overall accuracy of the map was 89%. The map was also used to estimate CO2 emissions for four classes constituting 85% of raised bogs: cutover, cutaway, grassland, and forestry using the IPCC wetlands supplement and literature-based emission factors, estimated emissions were at ~1.92 and ~0.68 Mt CO2-C-yr-1 respectively. This is the first study to spatially quantify land use and related emissions from raised bogs. The results revealed widespread degradation of these globally rare oceanic raised bogs habitats, making them net emitters of CO2. Lastly, the study utilises very high-resolution (0.25m) aerial imagery and a DL approach based on U-net to map the artificial drainage (LUCIP level 4) on 523,000 ha of raised bogs at a national scale. The results show that -drainage is widespread with approximately 20,000 km of drains mapped. This is the first study to map diverse sizes and patterns (regular and irregular) of artificial drains using DL methods on three peatland land use categories i.e., cutaway, cutover, and remnant peatlands. The results show that over 70% (1300 km) of the artificial drainage ditches exist on Bord na Móna-based raised bogs alone. The overall accuracy of the model was 80% on an independent testing dataset. The data was also used to derive the ¿Frac¿_ditch which was 0.03 (fraction of artificial drainage on industrial peat extraction sites). This is lower than IPCC Tier 1 ¿Frac¿_ditch and can be used for IPCC Tier 2 reporting for Ireland. The degradation of peatlands in Ireland through land use activities is widely acknowledged. However, due to the unavailability of spatially explicit information on artificial drainage ditches and land use, there is an overall ambiguity in the accurate accounting and reporting of emissions from these ecosystems and the status of land use activities. The study develops a robust set of methods capable of mapping and monitoring the artificial drainage and land use activities on Irish peatlands. The maps developed here are vital for the conservation of these ecosystems, and the methodologies can also be applied to other regions with similar peatland land use issues. This also has the potential for regional and global applications, providing maps that could help understand unsustainable management practices on peatlands and their impact on GHG emissions.