Related Experiment Video
Updated: Jun 10, 2026
Dependence of Laser-induced Breakdown Spectroscopy Results on Pulse Energies and Timing Parameters Using Soil Simulants
Published on: September 23, 2013
Laser-induced breakdown spectroscopy (LIBS) coupled with machine learning for classification of paddy soils from
Zubaidah Yusop1,2, Khairul Fikri Tamrin3,4
1Department of Mechanical and Manufacturing Engineering, Faculty of Engineering, Universiti Malaysia Sarawak (UNIMAS), Kota Samarahan, Sarawak, 94300, Malaysia.
Abstract:
Soil nutrients variability plays a crucial role in paddy growth and yield. Insufficient information on soil nutrient content can lead to inefficient fertilizers application, resulting in either nutrient deficiency or excess in the soil. This study investigates the potential of laser-induced breakdown spectroscopy (LIBS) combined with machine learning for rapid classification of paddy soils from non-granary cultivation systems in Sarawak, Malaysia. Soil samples from irrigated lowland, rainfed lowland, and upland areas were analysed using conventional physicochemical methods and LIBS to evaluate variability in soil properties and spectral characteristics. The LIBS spectra revealed distinct multi-elemental signatures dominated by Ca, Fe, K, and N, while P exhibited weak emission intensity due to both low concentration and intrinsic plasma emission limitations. Principal Component Analysis (PCA) reduced spectral dimensionality, with the first four components explaining 74.38% of the total variance. However, PCA score plots showed substantial overlap among soil groups, indicating limited separability using unsupervised analysis. To address this, a supervised classification approach using Support Vector Machine (SVM) was implemented. The PCA-SVM model achieved an average classification accuracy of 76.42 ± 15.17% under repeated sample-level hold-out validation, which improved to 87.33 ± 12.80% using Leave-One-Sample-Out Cross-Validation (LOSOCV). These results demonstrate that, although intrinsic spectral differences among paddy cultivation categories are subtle, the integration of LIBS with machine learning facilitates effective extraction of discriminative spectral patterns. This study highlights the potential applicability of LIBS as a rapid, multi-element analytical approach for preliminary classification of non-granary paddy cultivation systems, particularly in heterogeneous and resource-limited agricultural environments.
