Habitat classification using airborne and spaceborne remote sensing for biodiversity assessment in Wales

Abstract

Biodiversity and its conservation are an important subject as human pressure on natural resources increases continuously. Without accurate means of measuring biodiversity, however, monitoring is very di cult and conservation e orts might not be targeted e ectively. There is a great demand for biodiversity assessment on a regional scale in order to support national conservation aims as stated, for example, in the UK Biodiversity Action Plan. Remote sensing lends itself to interpretation at the landscape scale and this study aims to assess a variety of optical and laser remote sensing data with regard to their usefulness for biodiversity assessment in Wales. The study was divided into four distinct areas to evaluate di erent remote sensing data with regard to their utility for facilitating the measurement and assessment of distinct elements of biodiversity. These components are vegetation composition and condition, land cover on a regional scale, three-dimensional woodland structure and the interaction of ora and fauna within the landscape structure. Methodological advances include a novel land cover mapping approach from multispectral remote sensing data comparable to traditional manual habitat surveys as well as an analysis of forest vertical pro le under consideration of bird habitat preferences. Remote sensing data investigated included airborne hyperspectral data, multispectral satellite imagery and airborne LiDAR. The potential of hyperspectral data for the di erentiation of grasslands of varying levels of improvement was tested at two experimental grassland study sites and the results suggest a strong correlation between biomass and the red-edge region of the electromagnetic spectrum. A relationship between the presence of nonphotosynthetic vegetation and the level of agricultural improvement was further established and utilized in the formulation of rules for the classi cation of grassland habitats. The outcomes of this study were used to support the landscape-scale land cover mapping of the extent of 38 classes from a multi-temporal combination of two spaceborne multispectral sensors (SPOT 5 HRG and IRS LISS IV). The derived maps achieved a moderate accuracy of 64%, though individual classes, especially woodlands and bogs, exceeded this value. The ability of Light Detection and Ranging (LiDAR) and terrestrial laser scanner data to capture the three-dimensional structure of forests was investigated. It was found that both sensor types were limited in their ability to accurately represent forest vertical pro le due to respective downward and upward signal attenuation through the canopy. However, both provided an accurate digital terrain model and correlated well in their estimation of canopy height. Despite the limitations of vertical forest structure assessment from airborne Li- DAR, observation of bird species could be linked to distinct forest vertical pro les. Specialist woodland species were found to have the strongest habitat preferences with regard to the vertical forest structure. This project has achieved advancements in the mapping of agricultural land and habitats in Wales, using remote sensing data, speci cally in the di erentiation of grassland improvement levels and tree species discrimination from multispectral satellite imagery. Furthermore, a strong correspondence between airborne and terrestrial laser scanner outputs has been established and LiDAR forest pro les have been shown to relate well to known woodland bird habitat preferences. The added value derived from examining these four research areas as part of a single study, consists of the knowledge gained in how best to harness the respective remote sensing methods for the evaluation of very di erent aspects of biodiversity. It has further been shown that it is possible to use optical remote sensing data at a high spatial and spectral resolution, but low availability to inform and improve the utilization of more widely accessible, but less detailed images. Furthermore, a method has been developed which allows the interpolation of avian diversity from the assessment vertical forest structure. As biodiversity consists of many di erent elements at a wide variety of scales it is crucial to be able to perform such integrated analyses of its various components. However, only a combined approach towards the utilization of remote sensing, as demonstrated in this study, is likely to gain the necessary data. The outcomes of this research support Wales-wide assessment of biodiversity and facilitate the production of regional or national vegetation maps as well as structural attributes for input into models. Components of the study can be used to support, for example, climate change research, assessments of biodiversity and policy decisions. Optical and laser remote sensing data can be successfully utilized for Wales-wide biodiversity components analysis.