ResearchSpace

Mapping big tree presence in open savanna, using tree shadow and high resolution multispectral imagery

Show simple item record

dc.contributor.author Main, Russel S
dc.contributor.author Cho, Moses A
dc.contributor.author Mathieu, Renaud SA
dc.contributor.author Asner, G
dc.date.accessioned 2013-01-28T08:50:52Z
dc.date.available 2013-01-28T08:50:52Z
dc.date.issued 2012-10
dc.identifier.citation Main, R., Cho, M.A., Mathieu, R.S.A. and Asner, G. 2012. Mapping big tree presence in open savanna, using tree shadow and high resolution multispectral imagery. African Association of Remote Sensing of Environment (AARSE) 2012, El Jadida, Morroco, 28 October-2 November 2012 en_US
dc.identifier.uri http://hdl.handle.net/10204/6480
dc.description African Association of Remote Sensing of Environment (AARSE) 2012, El Jadida, Morroco, 28 October-2 November 2012 en_US
dc.description.abstract Large scattered trees play an important role in the functioning of savanna landscapes. They act as focal points for various organism activities, which influence the distribution of nutrients and water within the landscape, which in turn influences savanna patch dynamics. They generally are considered as prominent keystone structures associated with a stable ecological stage. A variety of anthropogenic land use and management activities are however putting increasing pressure on the big tree abundance, habitat structure, and ultimately the ecological function of the African savanna biome. Mapping these trees at individual level over large areas could yield valuable information for landscape ecology studies. Recent concerns related to the decrease of large trees in Kruger National Park prompted the development of field protocols for monitoring changes in large trees (Druce et al 2008). Airborne multispectral and LiDAR surveys are technically best suited for this application, but are expensive for large scale studies. Multispectral imagery affords the possibility of regional scale studies, but often lacks the spatial resolution for discriminating tree canopies. High spatial resolution multispectral sensors (i.e. 2-5 m pixel size) do however provide the opportunity to extract large scattered trees (>5 m height and 25m2 canopy surface) using their projected shadow. Available SPOT5 imagery was pan-sharpened (to 2.5m) and then subjected to various image transformation techniques, which all aimed to enhance the shadow. The products of the various image transformations were then used in an object based classifier to produce shadow maps. The shadow maps were validated against tree maps derived from a 3D discrete Carnegie Airborne Observatory LiDAR dataset. Tempered by a challenging accuracy assessment scenario, the methods achieved promising user’s accuracy results, which ranged between 64 and 79% depending on tree densities. More research is needed into the factors affecting shadow detection, but we are encouraged about this method becoming an affordable method for mapping trees in savanna landscapes. en_US
dc.language.iso en en_US
dc.relation.ispartofseries Workflow;10090
dc.subject Savanna ecosystem en_US
dc.subject Large tree monitoring en_US
dc.subject African savanna biome en_US
dc.subject Savanna landscapes en_US
dc.subject SPOT5 imagery en_US
dc.subject LiDAR surveys en_US
dc.subject High resolution multispectral imagery en_US
dc.title Mapping big tree presence in open savanna, using tree shadow and high resolution multispectral imagery en_US
dc.type Conference Presentation en_US


Files in this item

This item appears in the following Collection(s)

Show simple item record