WAVELET-BASED SPECTRAL ANALYSIS FOR SOIL NITRATE CONTENT MEASUREMENT

This article investigates the use of a relatively new signal analysis technique for the prediction of soil nitrate based on mid-infrared spectroscopy. Wavelet analysis was applied to soil Fourier transform infrared (FTIR) attenuated total reflectance (ATR) spectral data in order to predict nitrate contents. Two sets of experiments were conducted. One consisted of artificially adding known amounts of nitrate, carbonate, and humic acid to Yolo loam soil in a laboratory. Nitrate concentrations for this experiment varied from approximately 0 to 2400 ppm NO3-N. Results indicated that wavelet analysis was able to decompose the spectra to allow for identification of the nitrate peak separate from other interfering peaks such as carbonate and humic acid. The volume of the nitrate peak for each sample was correlated to nitrate concentration. The correlation coefficients between nitrate peak volumes and nitrate concentrations were found to be in the range of 0.94 to 0.98, with standard errors varying from 32 to 79 ppm NO3-N. To determine if wavelet analysis could distinguish nitrate peaks in actual field conditions with low nitrate concentrations, experiments were conducted involving Capay clay and Yolo loam soils treated with nitrate fertilizers in-situ in the presence of interfering compounds such as carbonate and humic acid. Nitrate concentrations varied from approximately 0 to 140 ppm NO3-N. As with the laboratory experiments, wavelet analysis was successful at separating the nitrate peak from adjacent and overlapping peaks in the field experiments. R2 values ranged from 0.93 for pooled data to 0.99 for individual experiments, with standard errors of 5.8 to 9.5 ppm NO3-N for pooled data.