Search

Soil analysis using near-infrared spectroscopy has shown great potential to be an alternative to traditional laboratory analysis, and there is continuously increasing interest in building large-scale soil spectral libraries (SSLs). However, due to issues such as high non-linearity in soil spectral data and complexity in soil spatial variation, the establishment of robust prediction models for soil spectral libraries remains a challenge. This study aimed to investigate the performance of deep learning algorithms, including long short-term memory (LSTM) and LSTM–convolutional neural networks (LSTM–CNN) integrated models, to predict the soil organic matter (SOM) of a provincial-scale SSL, and compare it to the normally used local weighted regression (LWR) model. The Hebei soil spectral library (HSSL) contains 425 topsoil samples (0–20 cm), of which every 3 soil samples were collected from dry land, irrigated land, and paddy fields, respectively, in different counties of Hebei Province, China. The results show that the accuracy of the validation dataset rank as follows: LSTM–CNN (R2p = 0.96, RMSEp = 1.66 g/kg) > LSTM (R2p = 0.83, RMSEp = 3.42 g/kg) > LWR (R2p = 0.82, RMSEp = 3.79 g/kg). The LSTM–CNN model performed the best, mainly due to its comprehensive ability to effectively extract spatial and temporal features. Meanwhile, the LSTM model achieved higher accuracy than the LWR model, owing to its built-in memory unit and its advantage of faster feature band extraction. Thus, it was suggested to use deep learning algorithms for SOM predictions in SSLs. However, their performance on larger-scale SSLs such as continental/global SSLs still needs to be further investigated.

As a precious soil resource, black soils in Northeast China are currently facing severe land degradation. Visible and near-infrared spectroscopy (vis-NIR, 350–2500 nm) and mid-infrared spectroscopy (MIR, 2500–25,000 nm) have shown great potential to predict soil properties. However, there is still limited research on using MIR in situ. The aim of this study was to explore the feasibility of in situ MIR for the prediction of soil total nitrogen (TN) and total phosphorus (TP) and to compare its performance with the use of laboratory MIR, as well as the use of in situ and laboratory vis-NIR. A total of 450 samples from 90 soil profiles, along with their in situ and laboratory spectra of MIR and vis-NIR, were collected in a field with ten different tillage and management practices in a typical black soil area of Northeast China. Partial least square regression (PLSR), random forest (RF) and multivariate adaptive regression splines (MARS) were used to generate the calibrations between the spectra and the two properties. The results showed that both MIR and vis-NIR were able to predict the TN whether in laboratory or in situ conditions, but neither of them could predict the TP quantitatively since there was no sensitive band on both spectra regarding the TP. The prediction accuracy of the TN with laboratory spectra was higher than that with in situ spectra, for both vis-NIR and MIR. The optimal prediction accuracy of the TN with laboratory MIR (RMSE = 0.11 g/kg, RPD = 3.12) was higher than that of laboratory vis-NIR (RMSE = 0.14 g/kg, RPD = 2.45). The optimal prediction accuracy of in situ MIR (RMSE = 0.20 g/kg, RPD = 1.80) was lower than that of in situ vis-NIR (RMSE = 0.16 g/kg, RPD = 2.14). The prediction performance of the spectra followed laboratory MIR > laboratory vis-NIR > in situ vis-NIR > in situ MIR. The performance of in situ MIR was relatively poor, mainly due to the fact that MIR was more influenced by soil moisture. This study verified the feasibility of in situ MIR for soil property prediction and provided an approach for obtaining rapid soil information and a reference for soil research and management in black soil areas.