@InProceedings{GrisDalMouRosFlo:2017:PrDaEs,
               author = "Gris, Diego Jose and Dalmolin, Ricardo Sim{\~a}o Diniz and 
                         Moura-Bueno, Jean Michel and Rosin, Nicolas Augusto and Flores, 
                         Jo{\~a}o Pedro Moro",
                title = "Pr{\'e}-processamento dos dados de espectroscopia na regi{\~a}o 
                         do Vis-NIR melhoram a predi{\c{c}}{\~a}o do carbono 
                         org{\^a}nico do solo?",
            booktitle = "Anais...",
                 year = "2017",
               editor = "Gherardi, Douglas Francisco Marcolino and Arag{\~a}o, Luiz 
                         Eduardo Oliveira e Cruz de",
                pages = "6304--6311",
         organization = "Simp{\'o}sio Brasileiro de Sensoriamento Remoto, 18. (SBSR)",
            publisher = "Instituto Nacional de Pesquisas Espaciais (INPE)",
              address = "S{\~a}o Jos{\'e} dos Campos",
             abstract = "Soil is one of the most important carbon (C) reservoirs in the 
                         environment, acting as a sink or source of C to the atmosphere, 
                         depending on its management. Therefore, data on soil organic 
                         carbon (SOC) is important for managing C in the ecosystem, but 
                         traditional methods for mapping SOC are highly demanding in time 
                         and resources. In this scenario, diffuse reflectance spectroscopy 
                         (DRS) emerges as a more efficient technique for SOC assessment. 
                         This study evaluates preprocessing techniques for soil spectra and 
                         models for SOC prediction. The study was conducted at Giru{\'a}, 
                         State of Rio Grande do Sul (RS), Brazil, where 841 soil samples 
                         were collected at 261 sites, in an area of 940 ha. SOC was 
                         determined analytically through wet combustion. Soil spectra were 
                         obtained using a FieldSpec Pro spectroradiometer. Preprocessing 
                         techniques included: smoothing (SMO), continuum removal (CRR), 
                         standard normal variate (SNV), detrend (DET), Savitzky-Golay first 
                         derivative (SGD), and Savitzky-Golay first derivative after 
                         transformation to absorbance (ASG). The samples were divided in 
                         training (70%) and validation (30%) sets. Two models were tested 
                         for SOC prediction: partial least squares regression (PLSR) and 
                         random forest (RF). SMO was the best preprocessing technique for 
                         PLSR, while SGD performed best for RF. PLSR had the best 
                         performance for SOC prediction (Rv2 = 0.72; RMSEv = 0.52% and 
                         RPIQv = 2.23) when compared to RF (Rv2 = 0.71; RMSEv = 0.53% and 
                         RPIQv = 2.20).",
  conference-location = "Santos",
      conference-year = "28-31 maio 2017",
                 isbn = "978-85-17-00088-1",
                label = "59674",
             language = "pt",
         organisation = "Instituto Nacional de Pesquisas Espaciais (INPE)",
                  ibi = "8JMKD3MGP6W34M/3PSMCLB",
                  url = "http://urlib.net/ibi/8JMKD3MGP6W34M/3PSMCLB",
           targetfile = "59674.pdf",
                 type = "Solos e umidade do solo",
        urlaccessdate = "16 jun. 2024"
}