The long–range goal of this study was to develop a portable, soil nitrate sensor to determine soil nitrate content in–situ. The immediate objective of this study was to develop a rapid technique to determine soil nitrate content using diffuse reflectance spectroscopy in the mid–infrared (MIR) range. A Fourier Transform Infrared (FTIR) spectrophotometer was used to determine the MIR response of various concentrations of calcium nitrate solution in water. The results clearly showed the existence of a strong nitrate absorption peak at 7194 nm (1390 cm –1 ). For KBr–diluted soil samples, the ratio of the area under the nitrate absorbance peak (1360–1390 cm –1 ) to the water absorbance peak (1640–1660 cm –1 ) was proportional to nitrate concentration. Mid–infrared diffuse reflectance spectroscopy detected soil nitrate content at low concentrations in KBr–diluted soil samples. However, when undiluted soil samples containing varying amounts of nitrate content were tested in the MIR range, it was necessary to further condition the data with noise filtering techniques. Three noise removal techniques were compared to improve the spectral quality: (1) the Savitzky and Golay, (2) FFT filtering, and (3) wavelet denoising. The wavelet denoising technique resulted in the highest coefficient of determination (R 2 = 0.856) for nitrate prediction. These noise removal techniques could not remove specular reflection effects and band overlap. To overcome these problems, a continuous wavelet transform, which could remove specular reflection as well as minimize noise, was used to deconvolute the soil spectral data, resulting in a high correlation (R 2 = 0.878) between soil nitrate content and MIR diffuse spectral reflectance.
Read full abstract