Effect of soil sample geometry on indoor water content measurement using visible-NIR.

摘要

Visible and near infrared reflectance spectroscopy offers the potential for rapid and cost-effective soil analysis. Their prediction model is merely valid to their sampling pretreatment and the experiment proceedings, and this is believed to be due to the influence of the physical properties of the soil sample and the geometric conditions of the sampling accessory. This study was conducted to evaluate the influence of the container to soil spectra in a given distance and angle of the spectra detector, and their ability to soil water content prediction, six petri dishes with different diameter and height (d12.5 cm, h2.2 cm; d14.7 cm, h2.9 cm; d15.7 cm, h2.8 cm; d10.4 cm, h1.7 cm; d9.3 cm, h1.5 cm; d9.7 cm, h1.9 cm) were applied for analysis. A total of 106 soil samples (Aeric Endoaquepts) obtained in Hangzhou, Zhejiang (China) were analysed with rest to their spectra features and the water content. Principal component analysis-artificial neural network was used to build prediction models with a calibration data set of 81 randomly chosen samples. The remaining 25 soil samples were used to validate the prediction model. Accurate water content prediction was obtained when the diameter of the soil container was 12.5 cm with a correlation coefficient of 0.91. The results were consistent with the geometric analysis..