An efficient methodology to simulate mixed spectral signatures of land covers through Field Radiometry data

Abstract

An efficient methodology to simulate mixed spectral signatures of land covers, from endmember data, using linear statistical modelling based on the least squares estimation approach, is proposed. The optimal set of endmember has been obtained by measurements in situ with a field spectroradiometer GER 1500. Also, it is proposed the use of new sub-pixel methods based on statistics and certain “units of sampling” to apply to the landscapes. The resultant point estimations for these new units will be the “observations” and all of them will carry out an special role to simulate the final spectral signature. This methodology is used to simulate spectral signatures of a Mediterranean forest landscape near to Madrid (Spain). Furthermore the spectral signature model obtained through Field Radiometry data will be correlated with the image data of the same zone provided by the Landsat 7 Enhaced Thematic Mapper Plus (ETM+) sensor once corrected. The results obtained in correlation studies seem to conclude its efficiency. At the same time, the results open new research guidelines.