نوع مقاله : علمی - پژوهشی
نویسندگان
1 گروه مهندسی مکانیک بیوسیستم، دانشکده کشاورزی، دانشگاه تربیت مدرس، تهران، ایران
2 بخش تحقیقات هوشمندسازی کشاورزی، موسسه تحقیقات فنی و مهندسی کشاورزی، سازمان تحقیقات، آموزش و ترویج کشاورزی، کرج، ایران.
3 گروه علوم خاک، دانشکده کشاورزی، دانشگاه تربیت مدرس، تهران، ایران
4 شورای ملی تحقیقات ایتالیا، موسسه روش شناسی تجزیه و تحلیل محیطی، پوتنزا، ایتالیا
چکیده
کلیدواژه‌ها
عنوان مقاله [English]
نویسندگان [English]
Abstract
Introduction: Determination of variation of organic carbon content in the field is important due to the importance and role of soil organic carbon content, including its effective role in increasing soil resistance against wind and water erosion. In this research, the ability of reflectance visible-near infrared (Vis/NIR) spectrometry for measurement and prediction of soil organic carbon content and the effect of the type of spectral preprocessing on the accuracy of multivariable predictive models was studied.
Material and methods: In this research, spectroscopy of soil samples was performed at 7 moisture levels in the interactance measurement mode in the 350-2500 nm spectral range using a contact probe. Spectrophotometry of 5 different sections of each soil sample was carried out and the data were processed and analyzed using a contact probe. Spectral data obtained from the spectrophotometer included unwanted information, background and noise in addition to the information of the samples. In order to arrive at accurate and reliable analytical models, pre-processing of the spectral data was required prior to regression model simulation. Multivariate calibration models of partial least squares (PLS) were developed based on the reference measurements and the information of the preprocessed spectra using a combination of different methods for assessment and prediction of soil organic carbon content. These included: smoothing (moving average (MA), and Savitzky-Golay (SG)); normalizing (multiplicative scatter correction (MSC), standard normal Variate (SNV)); as well as increasing the spectral resolution (first and second derivatives (D1, D2)).
Results and discussion: Results showed that NIR spectroscopy is a suitable method for measurement of organic carbon content in soil samples. Prediction utilizing the data analyzed using the PLS model based on SG + MSC, produced the best detection results. Thus, SG+MSC preprocessing (Rc2 =0.81, RMSEC = 0.239, Rp2 = 0.79, RMSEP = 0.252) is suitable for predicting the amount of soil OC with high accuracy (SDR= 3.191). Results showed that reflectance rate diminishes with increasing moisture content reducing the ability of the PLS model to predict organic carbon content. This is true across all the preprocessing methods. In addition, the determined index values and validation criteria showed that prediction of organic carbon content with the PLS model using SG+D1+MSC, SG+MSC, SG+MSC, SG+D1+MSC, SG+SNV and SG+SNV combinations gives the best detection results for the following moisture levels, respectively: 6, 12, 18, 24, 30 and 36%.
Conclusion: Vis/NIR spectroscopy can be used as an alternative method for conventional laboratory methods for determination of soil organic carbon content. The research results showed that the use of Vis/NIR spectroscopy for determination of soil organic carbon content can be considered in the site-specific management of fields, which it leads to saving inputs and reducing the pressure on the environment, ultimately.
Results and discussion: Results showed that NIR spectroscopy is a suitable method for measurement of organic carbon content in soil samples. Prediction utilizing the data analyzed using the PLS model based on SG + MSC, produced the best detection results. Thus, SG+MSC preprocessing (Rc2 =0.81, RMSEC = 0.239, Rp2 = 0.79, RMSEP = 0.252) is suitable for predicting the amount of soil OC with high accuracy (SDR= 3.191). Results showed that reflectance rate diminishes with increasing moisture content reducing the ability of the PLS model to predict organic carbon content. This is true across all the preprocessing methods. In addition, the determined index values and validation criteria showed that prediction of organic carbon content with the PLS model using SG+D1+MSC, SG+MSC, SG+MSC, SG+D1+MSC, SG+SNV and SG+SNV combinations gives the best detection results for the following moisture levels, respectively: 6, 12, 18, 24, 30 and 36%.
Conclusion: Vis/NIR spectroscopy can be used as an alternative method for conventional laboratory methods for determination of soil organic carbon content. The research results showed that the use of Vis/NIR spectroscopy for determination of soil organic carbon content can be considered in the site-specific management of fields, which it leads to saving inputs and reducing the pressure on the environment, ultimately.
کلیدواژه‌ها [English]
)