02775naa a2200361 a 450000100080000000500110000800800410001902400540006010000160011424501900013026000090032052015780032965000190190765300280192665300180195465300190197265300340199165300280202565300180205365300210207165300310209265300170212365300310214070000220217170000260219370000280221970000210224770000220226870000270229070000160231770000260233377300540235921426782022-05-05       2022    bl uuuu     u00u1 u    #d7 ahttps://doi.org/10.1016/j.microc.2022.1075442DOI1 aMAIA, A. J.  aNear-infrared spectroscopy for prediction of potentially toxic elements in soil and sediments from a semiarid and coastal humid tropical transitional river basin.h[electronic resource]  c2022  aThe input of potentially toxic elements (PTE) in river basins is a major environmental problem. PTE concentrations are determined using traditional analytical methods, which are mostly time consuming, expensive, and reliant on hazardous reagents. An alternative method to traditional chemical analysis is near-infrared (NIR) spectroscopy, which allows for the quantification of several PTEs through chemometric models. The aim of this study is to apply NIR spectroscopy for the prediction of PTE concentrations in the Ipojuca river basin, an area exposed to potential pollutant activities. We collected 145 soil samples and 33 bed sediments samples. All samples were ground and sieved at &lt- 100-um, and then analyzed for Al, Ce, Co, Cr, Fe, La, Mn, Mo, Ni, Pr, Sc, Sm, Sn, Sr, Th, Ti, V, and Y concentrations by inductively coupled plasma optical emission spectroscopy. Spectral data were retrieved from all samples using an FT-IR/NIR spectrometer in the range of 1000 - 2500 nm. The samples were subdivided into two sets: (i) soil and (ii) soil and bed sediments. Prediction models were built using the random forest algorithm (RF) and partial least squares regression (PLS) combined with different spectral preprocessing methods. Satisfactory results were obtained for Al, Ti, Sc, and V, and reasonable results for Fe, La, Mn, Pr, Sm, Sr, and Th. Our findings indicate that RF models obtains generally better results than PLS, and also that NIR spectroscopy can be a viable alternative assessment tool even in large areas with geochemical and pedological heterogeneity.  aContaminação  aAprendizado de máquina  aContamination  aEspectroscopia  aGestão de recursos hídricos  aHydrological monitoring  aInfravermelho  aMachine learning  aMonitoramento hidrológico  aSoil sensing  aWater resources management1 aNASCIMENTO, R. C.1 aSILVA, Y. J. A. B. da1 aNASCIMENTO, C. W. A. do1 aMENDES, W. de S.1 aVERAS NETO, J. G.1 aARAUJO FILHO, J. C. de1 aTIECHER, T.1 aSILVA, Y. J. A. B. da  tMicrochemical Journalgv. 179, 107544, Aug. 2022.