Arobustmethodforcharacterizingthemineralogyofsuspendedsedimentincontinentalriversisintroduced.Itencompasses3steps:thefiltrationofafewmillilitersofwater,measurementsofX-rayenergydispersivespectrausingScanningElectronMicroscopy(SEM),androbustmachinelearningtoolsofclassification.ThemethodisappliedtosuspendedparticlescollectedfromvariousAmazonianrivers.Atotalofmorethan204,000particleswereanalyzedbySEM-EDXS(EnergyDispersiveX-raySpectroscopy),i.e.about15,700particlespersamplingstation,whichleadtotheidentificationof15distinctgroupsofmineralogicalphases.ThesizedistributionofparticlescollectedonthefilterswasderivedfromtheSEMmicrographstakeninthebackscatteredelectronimagingmodeandanalyzedwithImageJfreeware.Thedeterminationofthemainmineralogicalgroupscomposingthebulksedimentassociatedwithphysicalparameterssuchasparticlesizedistributionoraspectratioallowsaprecisecharacterizationoftheloadoftheterrigenousparticlesinriversorlakes.InthecaseoftheAmazonianriversinvestigated,theresultsshowthattheidentifiedmineralogiesareconsistentwithpreviousstudiesaswellasbetweenthedifferentsamplescollected.Themethodenabledtheevolutionofgrainsizedistributionfromfinetocoarsematerialtobedescribedinthewatercolumn.Implicationsabouthydrodynamicsortingofmineralparticlesinthewatercolumnarealsobrieflydiscussed.Theproposedmethodappearswellsuitedforintensiveroutinemonitoringofsuspendedsedimentinriversystems.
