Please use this identifier to cite or link to this item:
|Title:||Chemometric Modelling and Remote Sensing of Arable Land Soil Organic Carbon as Mediterranean Land Degradation Indicator - A Case Study in Southern Italy|
|Authors:||BOETTCHER Kristin; KEMPER Thomas; MACHWITZ Miriam; MEHL Wolfgang; SOMMER Stefan|
|Other Identifiers:||EUR 23513 EN|
|Type:||EUR - Scientific and Technical Research Reports|
|Abstract:||The application of chemometric models for the quantitative estimation of soil organic matter (SOM) from laboratory reflectance data from samples taken on the regional/national level from Italian sites is explored in Part 1 of this report. In addition, the possibility to transfer the developed models from the spectral resolution of lab/field instrumentation to the one of operational satellite systems has been evaluated, by using the laboratory spectra to simulate the respective soil reflectance signatures of Landsat-TM, MODIS and MERIS. Soil physical and chemical laboratory analyses results were provided by the JRC-IES SOIL action (formerly JRC FP6 MOSES action). The 376 soil samples, used in this study, were collected for previous projects of the IES SOIL action and its partners within a wide range of environmental settings in Italy. Reflectance measurements were obtained on disturbed soil samples using an ASD Field Spec Pro spectro-radiometer. Data transformation methods (standardisation, vector-normalisation and first and second order derivatives) have been applied on the spectral data. The transformed spectral data have been used for the prediction of SOM and carbonate content using the partial least squares regression (PLSR). The results (R2 between 0.57 and 0.8) demonstrate the successful application of reflectance spectroscopy combined with chemometric modelling for the estimation of SOM and carbonate content. The calibration models demonstrated a tolerable stability over a variety of different soil types, which is a positive factor for opening the opportunity to use this methodology for monitoring larger areas. Furthermore it could be shown, that the spectral resolution of the MERIS sensor is sufficient for approximation of the SOC/SOM content from pure soil spectra. Consequently the second part of the study focused on the use of MERIS satellite data for the estimation of soil organic carbon content of bare soils at regional scale. The study concentrated on the Apulia region, where we had high density of available field sampling sites, and on parts of the coastal areas of the Abruzzi region South of Pescara, which are known to be amongst the more critical areas in Italy suffering from land degradation problems and desertification risk. For specific morphological-lithological units simple spectral models, based on soil colour and spectral shape attributes, were built to derive soil organic carbon content. In order to apply these models to MERIS satellite data, a time series of images covering the years 2003 and 2004 were acquired for Southern Italy. Pre-processing of image data aimed at extracting those pixels with negligible vegetation abundance at least at one date of observation per year, i.e. practically showing pure bare soil signatures only, and consisted of: ¿ geometrical co-registration and superposition of images from different acquisition dates ¿ the derivation of minimum vegetation composites for each year applying simple minimum value criteria for MERIS vegetation indices ¿ the determination of soil and vegetation abundance at sub-pixel level based on spectral mixture modelling. ¿ the removal of residual vegetation influence from image spectra Soil colour attributes (soil lightness, R coordinate of R-G-B model) and coefficients of a second order polynomial fitted through the pixel reflectance signatures were derived from the minimum vegetation composites of both years. The spatial distribution of soil organic carbon was estimated for each year within specific morphological-lithological units in the Apulia region. In addition models could be applied to other regions in Southern Italy. Estimation results showed good agreement with independent field data and the pedo-transfer rules based estimations of Jones et. al. (2004, 2005).|
|JRC Institute:||Institute for Environment and Sustainability|
Items in repository are protected by copyright, with all rights reserved, unless otherwise indicated.