OConnell Presentation

Radiometric normalisation of Landsat ETM+ data for a change detection study

Jerome O’Connell, Nicholas Holden, John Connolly

Biosystems Engineering, University College Dublin; jerome.oconnell@ucd.ie

Living biomass is fundamental to the carbon cycle of peatlands. In Ireland, the peatland carbon stock is estimated at 1071 Mt. Disturbance of vegetation (whether by anthropogenic influences or climate change) will potentially change peatland carbon balance. The aim of this study is to develop a remote sensing tool that can be used to monitor and assess change in peatland vegetation over time. Correction of remotely sensed imagery, be that radiometric, geometric or topographical, is a vital stage in any change detection study. In this study a database of geometrically corrected Landsat ETM+ data located over the Wicklow mountains was converted from DN (Digital Number) to at-sensor reflectance using Chavez (1996) COST model. Topographical normalisation was completed via a DEM with a two stage slope matching correction procedure (Nichol et al, 2006). Radiometric normalisation was applied by linear regression using Temporally Invariant Clusters (TIC). This method, which has been adapted from Chen et al (2005), extracted two spectrally invariant objects from each image and plots the pixels within both objects on a 2D scatterplot. Mean center location then established the center point for both clusters and the resulting center to center line formed the regression equation between both images. In Ireland deep water bodies and urban areas provide the most stable invariant targets (McGovern et al, 2002). However the scale and spectral heterogeneity of Irish urban environments meant that this study had to derive a more specific approach than that proposed by Chen et al (2005). The results showed a significant improvement on invariant target selection, especially for urban environments. The effect of increasing temporal scale on the spectral characteristic between the master and slave images was assessed through image subtraction of absolute normalised and relatively normalised images. This version or TIC normalisation provides a simple, accurate and easily repeatable method radiometric standardisation for Irish environments.

Access presentation here (5 meg)

Return to Presentation Programme Page