Prediction of soil Sand and clay contents via visible and near-infrared (Vis-NIR) spectroscopy
Küçük resim yok
Tarih
2017
Kurum Yazarları
Mouazen, Abdul M.
Kim, P.
Analide, C.
Süreli Yayın başlığı
Süreli Yayın ISSN
Cilt Başlığı
Yayınevi
Ios Press
Özet
Visible and near infrared (Vis-NIR) spectroscopy is a non-destructive analytical method that can be used to complement, enhance or potentially replace conventional methods of soil analysis. The aim of this research was to predict the particle size distribution (PSD) of soils using a Vis-NIR) spectrophotometry in one irrigate field having a vertisol clay texture in the Karacabey district of Bursa Province, Turkey. A total of 86 soil samples collected from the study area were subjected to optical scanning in the laboratory with a portable, fiber-type Vis-NIR spectrophotometer (AgroSpec, tec5 Technology for Spectroscopy, Germany). Before the partial least square regression (PLSR) analysis, the entire reflectance spectra were randomly split into calibration (80%) and validation (20%) sets. A leave-one-out cross-validation PLSR analysis was carried out using the calibration set with Unscrambler (R) software, whereas the model prediction ability was tested using the validation (prediction) set. Models developed were used to predict sand and clay content using on-line collected spectra from the field. Results showed an "excellent" laboratory prediction performance for both sand (R-2 = 0.81, RMSEP = 3.84% and RPD = 2.32 in cross-validation; R-2 = 0.90, RMSEP = 2.91% and RPD = 2.99 in the prediction set) and clay (R-2 = 0.86, RMSEP = 3.4% and RPD = 2.66 in cross validation; R-2 = 0.92, RMSEP = 2.67% and RPD = 3.14 in the prediction set). Modelling of silt did not result in any meaningful correlations. Less accurate on-line predictions were recorded compared to the laboratory results, although the on-line predictions were very good (RPD = 2.24-2.31). On-line predicted maps showed reasonable spatial similarity to corresponding laboratory measured maps. This study proved that soil sand and clay content can be successfully measured and mapped using Vis-NIR spectroscopy under both laboratory and on-line scanning conditions.
Açıklama
Bu çalışma, 21-22, Ağustus 2017 tarihlerinde Seul[Güney Kore]’de düzenlenen 13. International Conference on Intelligent Environments (IE) Kongresi‘nde bildiri olarak sunulmuştur.
Anahtar kelimeler
Computer science, PLS regression analysis, Sand, Clay, Vis-NIR spectroscopy, Diffuse-reflectance spectroscopy, Artificial neural-network, Moisture-content, Organic-carbon, Least-squares, Quality, Ph, Spectra
Alıntı
Tümsavaş, Z. vd. (2017). ''Prediction of soil Sand and clay contents via visible and near-infrared (Vis-NIR) spectroscopy''. ed, C. Anelide ve P. Kim. Ambient Intelligence and Smart Environments, Intelligent Environments 2017, 22, 29-38.