Refined data analysis models for soil carbon and nitrogen measurements using laser-induced breakdown spectroscopy

Laser-induced breakdown spectroscopy (LIBS) is an atomic emission spectroscopy technique that uses a pulsed laser to generate plasma on a target material. The elemental composition of the target material can be determined by spectrally resolving plasma emissions. The application of LIBS for soil carbon and nitrogen analysis is important globally for fertilizer optimization and sustainable food production in precision agriculture. However, determining soil carbon and nitrogen content under normal atmospheric conditions is difficult because of LIBS spectrum fluctuation, matrix impact, soil heterogeneity, and nitrogen interference from the air. Consequently, the precise measurement of soil carbon and nitrogen necessitates optimization of LIBS experimental parameters and the development of refined machine-learning models. We demonstrate that a 532 nm laser wavelength significantly enhances soil carbon measurement accuracy by generating a higher plasma density than achieved at 1064 nm. Using this LIBS system, we have collected 12900 spectra from 43 soil samples. A machine learning-based calibration is externally validated by randomly selecting eight independent soil samples from two different locations and plots of agricultural lands. Despite significantly different particle sizes and locations of soil samples, the model can determine soil nitrogen under atmospheric conditions with a root mean squared error (RMSE) of 0.03% with a limit of detection (LOD) of 0.05%. For soil carbon, the RMSE is 0.3-0.6% with an LOD of 0.6% carbon content.