Fingerprinting selection for agroenvironmental catchment studies: EDXRF analysis for solving complex artificial mixtures

TORRES ASTORGA, R; H. VELASCO; DERCON, G.; L. MABIT

Vienna

Conferencia; EGU General Assemblies; 2017

European Geophysical Unioin

Soil erosion and associated sediment transportation and deposition processes are key environmental problems inCentral Argentinian watersheds. Several land use practices - such as intensive grazing and crop cultivation - areconsidered likely to increase significantly land degradation and soil/sediment erosion processes.Characterized by highly erodible soils, the sub catchment Estancia Grande (12.3 km2) located 23 km north east ofSan Luis has been investigated by using sediment source fingerprinting techniques to identify critical hot spots ofland degradation.The authors created 4 artificial mixtures using known quantities of the most representative sediment sources ofthe studied catchment. The first mixture was made using four rotation crop soil sources. The second and the thirdmixture were created using different proportions of 4 different soil sources including soils from a feedlot, a rotationcrop, a walnut forest and a grazing soil. The last tested mixture contained the same sources as the third mixture butwith the addition of a fifth soil source (i.e. a native bank soil).The Energy Dispersive X Ray Fluorescence (EDXRF) analytical technique has been used to reconstruct the sourcesediment proportion of the original mixtures. Besides using a traditional method of fingerprint selection such asKruskal?Wallis H-test and Discriminant Function Analysis (DFA), the authors used the actual source proportionsin the mixtures and selected from the subset of tracers that passed the statistical tests specific elemental tracers thatwere in agreement with the expected mixture contents. The selection process ended with testing in a mixing modelall possible combinations of the reduced number of tracers obtained. Alkaline earth metals especially Strontium(Sr) and Barium (Ba) were identified as the most effective fingerprints and provided a reduced Mean AbsoluteError (MAE) of approximately 2% when reconstructing the 4 artificial mixtures.This study demonstrates that the EDXRF fingerprinting approach performed very well in reconstructing our originalmixtures especially in identifying and quantifying the contribution of the 4 rotation crop soil sources in the firstmixture.