Abstract
Hyperspectral remote sensing technology can be used to monitor the soil nutrient changes in a rapid, real-time, and non-destructive manner, which is of great significance to promote the development of precision agriculture. In this paper, 225 soil samples were studied. The effects of different water treatments on soil organic carbon (SOC) content, and the relationship between SOC content and spectral reflectance (350–2500 nm) were studied. 17 kinds of preprocessing algorithm were performed on the original spectral (R), and the five allocation ratios of calibration to verification sets were set. Finally, the model was constructed by partial least squares regression (PLSR). The results showed that the effects of water treatment on SOC content were different in different growth stages of winter wheat. Results of correlation analysis showed that the differential transformation can refine the spectral characteristics, and improve the correlation between SOC content and spectral reflectance. Results of model construction showed that the models constructed by second-order differential transformation were not good. But the ratio of standard deviation to the standard prediction error (RPD) values of the models were constructed by simple mathematical transformation (T0–T5) and first-order differential transformation (T6–T11) can reach more than 1.4. The simple mathematical transformation (T0–T2, T4–T5) and the first-order differential transformation (T6–T10) resulted in the highest RPD in mode 5 and mode 2, respectively. Among all the models, the model of T7 in mode 2 reach the highest accuracy with a RPD value of 1.9861. Therefore, it is necessary to consider the data preprocessing algorithm and allocation ratio in the process of constructing the hyperspectral monitoring model of SOC.
Highlights
Hyperspectral remote sensing technology can be used to monitor the soil nutrient changes in a rapid, real-time, and non-destructive manner, which is of great significance to promote the development of precision agriculture
The results showed that the differential transformation may enlarge the spectral characteristics in some wave bands which have a great correlation with soil organic carbon (SOC) content
The prediction accuracy of the hyperspectral monitoring model of SOC content is greatly affected by soil spectral preprocessing mathematical algorithms and the ratio of the calibration set to validation set
Summary
Hyperspectral remote sensing technology can be used to monitor the soil nutrient changes in a rapid, real-time, and non-destructive manner, which is of great significance to promote the development of precision agriculture. It is necessary to consider the data preprocessing algorithm and allocation ratio in the process of constructing the hyperspectral monitoring model of SOC. Amin et al.[36] constructed a hyperspectral monitoring model of SOC content based on PLSR in Azerbaijan, and found that the model constructed after Savitzky-Golay smoothing can reach the highest accuracy with R 2 and RPD are 0.85 and 2.54, respectively. Yu et al.[37] constructed a hyperspectral monitoring model of soil organic matter content by using PLSR after preprocessing the spectral data, and the results showed that the model besed on continuous removal (CR) preprocessing had the best accuracy. Ji et al.[18] used a variety of modeling methods to predict soil organic matter content based on different data allocation ratios, it was found that when the ratios of calibration to validation sets were different, the accuracy of model constructed by different modeling methods was different.
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
