Robust models for predicting soil salinity that use visible and near-infrared (vis–NIR) reflectance spectroscopy are needed to better quantify soil salinity in agricultural fields. Currently available models are not sufficiently robust for variable soil moisture contents. Thus, we used external parameter orthogonalization (EPO), which effectively projects spectra onto the subspace orthogonal to unwanted variation, to remove the variations caused by an external factor, e.g., the influences of soil moisture on spectral reflectance. In this study, 570 spectra between 380 and 2400 nm were obtained from soils with various soil moisture contents and salt concentrations in the laboratory; 3 soil types × 10 salt concentrations × 19 soil moisture levels were used. To examine the effectiveness of EPO, we compared the partial least squares regression (PLSR) results established from spectra with and without EPO correction. The EPO method effectively removed the effects of moisture, and the accuracy and robustness of the soil salt contents (SSCs) prediction model, which was built using the EPO-corrected spectra under various soil moisture conditions, were significantly improved relative to the spectra without EPO correction. This study contributes to the removal of soil moisture effects from soil salinity estimations when using vis–NIR reflectance spectroscopy and can assist others in quantifying soil salinity in the future.
Core Ideas The spectra of unground samples can be used to predict soil organic C (SOC) and clay. Vis‐NIR spectroscopy models for ground samples can predict SOC and clay of unground samples. Using unground samples will save time and labor when estimating SOC and clay. Visible and near‐infrared spectroscopy (Vis‐NIR) can accurately predict soil organic C (SOC) and clay from the spectra of air‐dried and ground (DG) samples. However, grinding generally requires a lot of time and labor. In this study, air‐dried and unground (DU) samples were used to exploit the time and accuracy advantages of Vis‐NIR. The reflectance of 117 samples using DG and DU pretreatments was measured in the laboratory. Five spectral pretreatments were used: no‐pretreatment (NP), Savitzky–Golay (SG) smoothing, first derivative (FD), standard normal variate transformation, and multiplicative scatter correction (MSC). When calibrations and validations used the same sample pretreatment (DG or DU), good predictions (R2 > 0.75; RPD > 2.0) could be obtained for both SOC and clay. There were no sample × spectral pretreatments interaction effects on the prediction accuracy, whereas the main effects of the pretreatments were significant for clay but not for SOC. However, when calibrations derived from the DG samples were applied to the DU samples, the prediction accuracy decreased compared with that of the DG samples regardless of the SOC and clay. In this case, good predictions could be acquired for clay using the FD spectra and satisfactory predictions for SOC (0.60 < R2 < 0.75; 1.4 < RPD < 2.0) were obtained from the NP, SG, and MSC spectra. These results indicate that it is viable to predict SOC and clay content using DU samples with acceptable accuracy while saving a lot of time and labor compared with that needed for DG samples.
Previous studies for retrieving soil moisture content (SMC) from visible and nearinfrared hyperspectral data over vegetation-covered surfaces using spectral unmixing, non-negative matrix factorization, and albedo/vegetation coverage in trapezoid spaces have required mass spectral preprocessing and offered only limited improvements in prediction accuracy. Recently, deep learning has triggered some improvements in soil properties prediction from hyperspectral data because of its automatic feature extraction and high accuracy. In this study, hyperspectral data in a simulation experiment with different vegetation coverages, SMCs, and soil types were acquired. Deep learning models, one-dimensional convolutional neural network (1D-CNN), and long short-term memory network (LSTM) are proposed to predict SMC. The results showed that two deep learning models achieved excellent predictions (residual prediction deviation [RPD] > 2.5) using the unpreprocessed mixed spectra and partial least squares regression (PLSR) had a good prediction (RPD = 1.88). The 1D-CNN (R 2 p = .91) and LSTM (R 2 p = .90) significantly outperformed PLSR (R 2 p = .72), which demonstrated that deep learning could improve SMC prediction over partially vegetation-covered surfaces. However, when only using bare soil spectra, the prediction accuracy was commensurate, whether through the 1D-CNN, LSTM, or PLSR models; additionally, 1D-CNN and LSTM had better performance on all mixed spectra than bare soil spectra. These results indicated that deep learning had no advantage on smaller datasets. We also found that SMC prediction with deep learning was affected by vegetation coverage and soil type but was still very good. The 1D-CNN and LSTM are effective models for predicting SMC with large hyperspectral datasets acquired from complex soil surface conditions.
Although many Soil Spectral Libraries (SSLs) have been created globally, these libraries still have not been operationalized for end-users. To address this limitation, this study created an online Brazilian Soil Spectral Service (BraSpecS). The system was based on the Brazilian Soil Spectral Library (BSSL) with samples collected in the Visible–Near–Short-wave infrared (vis–NIR–SWIR) and Mid-infrared (MIR) ranges. The interactive platform allows users to find spectra, act as custodians of the data, and estimate several soil properties and classification. The system was tested by 500 Brazilian and 65 international users. Users accessed the platform (besbbr.com.br), uploaded their spectra, and received soil organic carbon (SOC) and clay content prediction results via email. The BraSpecS prediction provided good results for Brazilian data, but performed variably for other countries. Prediction for countries outside of Brazil using local spectra (External Country Soil Spectral Libraries, ExCSSL) mostly showed greater performance than BraSpecS. Clay R2 ranged from 0.5 (BraSpecS) to 0.8 (ExCSSL) in vis–NIR–SWIR, but BraSpecS MIR models were more accurate in most situations. The development of external models based on the fusion of local samples with BSSL formed the Global Soil Spectral Library (GSSL). The GSSL models improved soil properties prediction for different countries. Nevertheless, the proposed system needs to be continually updated with new spectra so they can be applied broadly. Accordingly, the online system is dynamic, users can contribute their data and the models will adapt to local information. Our community-driven web platform allows users to predict soil attributes without learning soil spectral modeling, which will invite end-users to utilize this powerful technique.
It is critical to identify the assembly processes and determinants of soil microbial communities to better predict soil microbial responses to environmental change in arid and semiarid areas. Here, soils from 16 grassland-only, 9 paired grassland and farmland, and 16 farmland-only sites were collected across the central Inner Mongolia Plateau covering a steep environmental gradient. Through analyzing the paired samples, we discovered that land uses had strong effects on soil microbial communities, but weak effects on their assembly processes. For all samples, although no environmental variables were significantly correlated with the net relatedness index (NRI), both the nearest taxon index (NTI) and the β-nearest taxon index (βNTI) were most related to mean annual precipitation (MAP). With the increase of MAP, soil microbial taxa at the tips of the phylogenetic tree were more clustered, and the contribution of determinism increased. Determinism (48.6%), especially variable selection (46.3%), and stochasticity (51.4%) were almost equal in farmland, while stochasticity (75.0%) was dominant in grassland. Additionally, Mantel tests and redundancy analyses (RDA) revealed that the main determinants of soil microbial community structure were MAP in grassland, but mean annual temperature (MAT) in farmland. MAP and MAT were also good predictors of the community composition (the top 200 dominant OTUs) in grassland and farmland, respectively. Collectively, in arid and semiarid areas, soil microbial communities were more sensitive to environmental change in farmland than in grassland, and unlike the major impact of MAP on grassland microbial communities, MAT was the primary driver of farmland microbial communities. Importance As one of the most diverse organisms, soil microbes play indispensable roles in many ecological processes in arid and semiarid areas with limited macrofaunal and plant diversity, yet the mechanisms underpinning soil microbial community are not fully understood. In this study, soil microbial communities were investigated along a 500 km transect covering a steep environmental gradient across farmland and grassland in the areas. The results showed that precipitation was the main factor mediating the assembly processes. Determinism was more influential in farmland, and variable selection of farmland was twice that of grassland. Temperature mainly drove farmland microbial communities, while precipitation mainly affected grassland microbial communities. These findings provide new information about the assembly processes and determinants of soil microbial communities in arid and semiarid areas, consequently improving the predictability of the community dynamics, which have implications for sustaining soil microbial diversity and ecosystem functioning, particularly under global climate change conditions.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.