VIII Congreso Internacional de Investigación REDU
← vista completa
Publicado el 25 de abril de 2022 | http://doi.org/10.5867/Medwave.2022.S1.CI22
Predicción del COS mediante espectroscopia Vis-NIR. Una comparación de dispositivos para el registro de espectros
Prediction of SOC using Vis-NIR spectroscopy. A comparison of devices for spectra recording
Lizardo Reyna-Bowen , María Teresa Hidalgo, Ramón Leal Murillo, José Alfonso Gómez Calero, Pilar Fernández Rebollo
Palabras clave SOC, organic-matter, Walkley & Black, reflectance, PLSR model
|
Introducción
Soil analysis is at the heart of precision agriculture applications and there is a need to find alternatives to, or complementary methods for, traditional laboratory analysis.
Objetivos
The present work, our objectives were: (i) to compare the accuracy of two different spectra recording devices used with prediction models of soil organic carbon (SOC). (ii) To predict the SOC concentration of a similar soil type but a different land use with the model accurately.
Método
440 soil samples were taken at different depth from farms with different land use. The soil samples were dried at 40 ◦C and passed through a 2 mm sieve. Soil reflectance spectra were measured with LabSpec 5000 spectrophotometer. Spectra were recorded with: (i) a low-intensity contact probe; and (ii) a muglight accessory with a circular tray adapter. SOC concentration was determined by Walkley & Black method. The accuracy of the prediction models was verified by cross-validation and, using the coefficient of determination (R2), the root mean squared error (RMSE), the residual predictive deviation (RPD) and range error ratio (RER). Concerning cross-validation, the accuracy of SOC models was excellent.
Principales Resultados
R2 and RPD were always higher than 0.95 and 4, respectively. In addition, RER reached values higher than 20. External validation provided more conservative accuracy metrics. Although RPD indexes were above 2.5, indicating excellent predictions, muglight model performed slightly better than contact probe model due to lower RMSE (0.222 vs 0.244) and higher R2 (0.90 vs 0.89).
Conclusiones
The devices for record the spectra, the muglight and the contact-probe, provided models of high and similar accuracy. The use of the contact-probe reduces the time required to record the spectra of soil samples compared to the use of muglight device and may therefore reduce the analytical cost and improved the financial feasibility of soil spectroscopy. Based on our statistical comparison, SOC could be predicted accurately using the contact probe, which reduces the time needed to record the spectrum of a sample.
|
Esta obra de Medwave está bajo una licencia Creative Commons Atribución-NoComercial 3.0 Unported. Esta licencia permite el uso, distribución y reproducción del artículo en cualquier medio, siempre y cuando se otorgue el crédito correspondiente al autor del artículo y al medio en que se publica, en este caso, Medwave.