Abstract
Phosphorus (P) is an essential plant nutrient and plays a vital role in achieving potential yields in rice-wheat sequence. conventional analytical techniques for assessing plant available P in soil are time consuming, relatively expensive and are not environment friendly. Diffuse reflectance spectroscopy particularly visible-near infrared (VIS-NIR) reflectance spectroscopy is an effective alternative means for rapid, nondestructive and eco-friendly assessment of available phosphorus and other soil properties. Geo-referenced surface soil (0–15 cm) samples (170) were collected from a rice-wheat field in Ludhiana district of Punjab state for performing both chemical (using Olsen method) and hyper-spectral analysis (350–2500 nm). A multivariate modeling technique, namely partial least square regression (PLSR) was employed to develop spectral model. Statistical diagnostics of coefficient of determination (R2), root mean square of error of prediction (RMSEP) and ratio of performance deviation (RPD) were used to evaluate the predictive performance of the developed spectral model. the R2, RMseP and RPd values were 0.42, 1.15 and 1.40, respectively for separate validation dataset of the PLsR model. however, the R2 and RMseP were better when the same set was used as validation set by using one-leave-out method. the RPd value suggested acceptable prediction accuracy of the spectral model. however, low R2 value and somewhat higher RMseP suggested exploring new modelling or data preprocessing techniques to enhance prediction accuracy.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have