Reducing Uncertainties Associated with Deforestation Estimation in Ireland using Optical and Radar Sensors
Devaney, John1; Barrett, Brian2; Redmond, John3; Cawkwell, Fiona2; O'Halloran, John1
1School of Biological, Earth and Environmental Sciences, University College Cork (UCC), IRELAND; 2School of Geography and Archaeology, University College Cork (UCC), IRELAND; 3Forest Service, Department of Agriculture, Food and the Marine, IRELAND

Since its establishment in 1994, the United Nations Framework Convention on Climate Change has required annex 1 parties to provide an annual inventory of greenhouse gas (GHG) emissions and removals. As forests act as an important and manageable carbon sink, a common challenge is to quantify spatial and temporal patterns of forest carbon. Over the last half of the 20th century, forest cover in Ireland has increased from less than 1% to 11% and it is a government aim to increase forest cover to 17% by 2030. However, preliminary results from Ireland’s National Forest Inventory (NFI) suggest an increase in the rate of deforestation over the last decade. Such deforestation may have a significant impact on GHG emissions, as well as negatively affecting biodiversity and ecosystem services provided by forest habitats.

In this study, the potential use of radar remote sensing for identifying forest change in Ireland is investigated. Optical remote sensing has been used to detect and quantify deforestation but its application is limited in countries such as Ireland with near-constant cloud coverage. Synthetic Aperture Radar (SAR) satellites provide an ideal method of routinely tracking areas of forests for changes, irrespective of weather conditions or time of day. The study area covers two counties located in the midlands and northwest of Ireland. Multitemporal ALOS PALSAR imagery is used to estimate changes in forest area between 2008 and 2010. This is compared with existing vector based forest cover statistics and spatially explicit maps of deforestation for the region, derived from aerial photography and supporting field measurements. Pixel-based supervised machine learning is carried out to classify pixels of the multi-mode SAR dataset and both a post-classification comparison method and image differencing are adopted for a change detection analysis. The results of this study will be useful for understanding the applicability of SAR to detect and map forest cover removal in Ireland and the uncertainties associated with current estimates of deforestation.