THE EFFECTS OF VEGETATION ON THE ABILITY TO MAP SOILS USING IMAGING SPECTROMETER DATA

Remote sensing applied to tasks of mapping soil and rock surfaces must address the problem of vegetation cover in all but the most arid terr ain. Masking out pixels with a high proportion of vegetation using a t hreshold on the near-infrared/red ratio is a popular strategy for live vegetation. The important effects of dead vegetation on the SWIR refl ectance is usually ignored. Data gathered by the GER-II imaging spectr ometer over a semi-arid area near Almaden, south central Spain were us ed to test the sensitivity of thematic soil mapping to variable cover of live and dead vegetation. After calibration to reflectance a least- squares unmixing analysis was performed using image end-members and pr oportions maps of vegetation and soil/rock components generated. Despi te a low signal-to-noise ratio, three soil/rock and four vegetation en dmembers were successfully mapped and validated from field estimates. A quantitative assessment was made of the effects of live and dead veg etation on the ability of the unmixing analysis to distinguish between granite and shale soils using synthetically mixed spectra gathered us ing field spectroradiometry and statistical analysis of the imaging sp ectrometer data. Dead vegetation was shown to have a greater impact on soil spectra than live vegetation. The ability to distinguish between the soils was lost at 50-60 per cent vegetation cover.