Prediction of Soil Water Content Based on Hyperspectral Reflectance Combined with Competitive Adaptive Reweighted Sampling and Random Frog Feature Extraction and the Back-Propagation Artificial Neural Network Method

Abstract

The soil water content (SWC) is a critical factor in agricultural production. To achieve real-time and nondestructive monitoring of the SWC, an experiment was conducted to measure the hyperspectral reflectance of soil samples with varying levels of water content. The soil samples were divided into two parts, SWC higher than field capacity (super-θf) and SWC lower than field capacity (sub-θf), and the outliers were detected by Monte Carlo cross-validation (MCCV). The raw spectra were processed using Savitzky–Golay (SG) smoothing and then the spectral feature variable of SWC was extracted by using a combination of competitive adaptive reweighted sampling (CARS) and random frog (Rfrog). Based on the extracted feature variables, an extreme learning machine (ELM), a back-propagation artificial neural network (BPANN), and a support vector machine (SVM) were used to establish the prediction model. The results showed that the accuracy of retrieving the SWC using the same model was poor, under two conditions, i.e., SWC above and below θf, mainly due to the influence of the lower accuracy of the super-θf part. The number of feature variables extracted by the sub-θf and super-θf datasets were 25 and 18, respectively, accounting for 1.85% and 1.33% of the raw spectra, and the variables were widely distributed in the NIR range. Among the models, the best results were achieved by the BPANN model for both the sub-θf and the super-θf datasets; the R2p, RMSEp, and RRMSE of the sub-θf samples were 0.941, 1.570%, and 6.685%, respectively. The R2p, RMSEp, and RRMSE of the super-θf samples were 0.764, 1.479%, and 4.205%, respectively. This study demonstrates that the CARS–Rfrog–BPANN method was reliable for the prediction of SWC.

Integration of Vis–NIR Spectroscopy and Machine Learning Techniques to Predict Eight Soil Parameters in Alpine Regions

“…At the same time, the correct model and processing method need to be used, making prediction difficult. For the modeling methods, the overall ranking of model performance is as follows: RF > BPNN > SVM > PLSR, which is consistent with the findings of previous studies, but the order of XGBoost is usually not fixed [37,45,52,53]. The prediction of different soil parameters is different for different data preprocessing methods and modeling approaches.…”

“…Kawamura et al [35,44] found that the machine learning model showed higher accuracy than the traditional linear model (e.g., multiple linear stepwise regression and partial least squares) when predicting soil parameters (e.g., N and P) using Vis-NIR. Previous research in their study on predicting soil TN using Vis-NIR technology combined with support vector machines, random forests, extreme gradient boosting (XGBoost), and backpropagation neural networks discovered that all machine learning methods could achieve accurate TN predictions, with the model accuracy ranking as follows: RF > BPNN > SVM [37,45,52,53]. XGBoost's performance may be higher or somewhere in between.…”

Advancing Digital Image-Based Recognition of Soil Water Content: A Case Study in Bailu Highland, Shaanxi Province, China

“…Water, an essential element in various forms and compositions, gives rise to intricate mechanisms and pathways that significantly influence the soil environment at multiple levels [2,3]. In the field of agriculture, soil water content (SWC) is vital for plant growth and development [4]. Adequate water supply at different growth stages is essential for any crop, as both too much and too little water can negatively impact crop yield and quality [5].…”