Quantitative analysis of fertilizer using laser-induced breakdown spectroscopy combined with random forest algorithm

Lai Wei; Yu Ding; Jing Chen; Linyu Yang; Jinyu Wei; Yinan Shi; Zigao Ma; Zhiying Wang; Wenjie Chen; Xingqiang Zhao

doi:10.3389/fchem.2023.1123003

Quantitative analysis of fertilizer using laser-induced breakdown spectroscopy combined with random forest algorithm

Front Chem. 2023 Jan 13:11:1123003. doi: 10.3389/fchem.2023.1123003. eCollection 2023.

Authors

Lai Wei^{1

2

3}, Yu Ding^{1

2

3}, Jing Chen^{1

2

3}, Linyu Yang^{1

2

3}, Jinyu Wei^{1

2

3}, Yinan Shi^{1

2

3}, Zigao Ma^{1

2

3}, Zhiying Wang^{1

2

3}, Wenjie Chen^{1

2

3}, Xingqiang Zhao^{1

2

3}

Affiliations

¹ Jiangsu Key Laboratory of Big Data Analysis Technology, Nanjing University of Information Science and Technology, Nanjing, China.
² Jiangsu Collaborative Innovation Center on Atmospheric Environment and Equipment Technology, Nanjing University of Information Science and Technology, Nanjing, China.
³ School of Automation, Nanjing University of Information Science and Technology, Nanjing, China.

Abstract

Chemical fertilizers are important for effectively improving soil fertility, promoting crop growth, and increasing grain yield. Therefore, methods that can quickly and accurately measure the amount of fertilizer in the soil should be developed. In this study, 20 groups of soil samples were analyzed using laser-induced breakdown spectroscopy, and partial least squares (PLS) and random forest (RF) models were established. The prediction performances of the models for the chemical fertilizer content and pH were analyzed as well. The experimental results showed that the R ² and root mean square error (RMSE) of the chemical fertilizer content in the soil obtained using the full-spectrum PLS model were .7852 and 2.2700 respectively. The predicted R ² for soil pH was .7290, and RMSE was .2364. At the same time, the full-spectrum RF model showed R ² of .9471 (an increase of 21%) and RMSE of .3021 (a decrease of 87%) for fertilizer content. R ² for the soil pH under the RF model was .9517 (an increase of 31%), whereas RMSE was .0298 (a decrease of 87%). Therefore, the RF model showed better prediction performance than the PLS model. The results of this study show that the combination of laser-induced breakdown spectroscopy with RF algorithm is a feasible method for rapid determination of soil fertilizer content.

Keywords: LIBS; PLS; RF; fertilizer; pH.

Grants and funding

The research is funded by National Natural Science Foundation of China (Nos. 62105160).