Dongyun Xu

Soil spectroscopy offers a rapid, cost-effective alternative to traditional soil analyses for characterization and classification. Previous studies have mainly focused on predicting soil categories using single sensors, particularly visible–near-infrared (vis–NIR) or mid-infrared (MIR) spectroscopy. In this study, we evaluated the performance of vis–NIR, MIR, and their combined spectra for soil classification by partial least-squares discriminant analysis (PLSDA) and random forest (RF). Utilizing 60 typical soil profiles’ data of four soil classes from the global soil spectral library (GSSL), our results demonstrated that in PLSDA models, direct combination (optimal overall accuracy: 70.6%, kappa coefficient: 0.60) and outer product analysis (OPA) fused spectra (optimal overall accuracy: 68.1%, kappa coefficient: 0.57) outperformed vis–NIR (optimal overall accuracy: 62.2%, kappa coefficient: 0.49) but underperformed compared to MIR (optimal overall accuracy: 71.4%, kappa coefficient: 0.62). In RF models, classification accuracy using fused spectra was inferior to single spectral ranges, with MIR achieving the highest classification accuracy (optimal overall accuracy: 89.1%, kappa coefficient: 0.85). Therefore, MIR alone remains the most effective spectral range for accurate soil class discrimination. Our findings highlight the potential of MIR spectroscopy for enhancing global soil classification accuracy and efficiency, with important implications for soil resource management and agricultural planning across diverse environments.

Soil visible and near–infrared reflectance spectroscopy is an effective tool for the rapid estimation of soil organic carbon (SOC). The development of spectroscopic technology has increased the application of spectral libraries for SOC research. However, the direct application of spectral libraries for SOC prediction remains challenging due to the high variability in soil types and soil–forming factors. This study aims to address this challenge by improving SOC prediction accuracy through spectral classification. We utilized the European Land Use and Cover Area frame Survey (LUCAS) large–scale spectral library and employed a geographically weighted principal component analysis (GWPCA) combined with a fuzzy c–means (FCM) clustering algorithm to classify the spectra. Subsequently, we used partial least squares regression (PLSR) and the Cubist model for SOC prediction. Additionally, we classified the soil data by land cover types and compared the classification prediction results with those obtained from spectral classification. The results showed that (1) the GWPCA–FCM–Cubist model yielded the best predictions, with an average accuracy of R² = 0.83 and RPIQ = 2.95, representing improvements of 10.33% and 18.00% in R² and RPIQ, respectively, compared to unclassified full sample modeling. (2) The accuracy of spectral classification modeling based on GWPCA–FCM was significantly superior to that of land cover type classification modeling. Specifically, there was a 7.64% and 14.22% improvement in R² and RPIQ, respectively, under PLSR, and a 13.36% and 29.10% improvement in R² and RPIQ, respectively, under Cubist. (3) Overall, the prediction accuracy of Cubist models was better than that of PLSR models. These findings indicate that the application of GWPCA and FCM clustering in conjunction with the Cubist modeling technique can significantly enhance the prediction accuracy of SOC from large–scale spectral libraries.

The authors wish to make the following correction to this paper

Soil pollution by potentially toxic elements (PTEs) has become a core issue around the world. Knowledge of the spatial distribution of PTEs in soil is crucial for soil remediation. Portable X-ray fluorescence spectroscopy (p-XRF) provides a cost-saving alternative to the traditional laboratory analysis of soil PTEs. In this study, we collected 293 soil samples from Fuyang County in Southeast China. Subsequently, we used several geostatistical methods, such as inverse distance weighting (IDW), ordinary kriging (OK), and empirical Bayesian kriging (EBK), to estimate the spatial variability of soil PTEs measured by the laboratory and p-XRF methods. The final maps of soil PTEs were outputted by the model averaging method, which combines multiple maps previously created by IDW, OK, and EBK, using both lab and p-XRF data. The study results revealed that the mean PTE content measured by the laboratory methods was as follows: Zn (127.43 mg kg⁻¹) > Cu (31.34 mg kg⁻¹) > Ni (20.79 mg kg⁻¹) > As (10.65 mg kg⁻¹) > Cd (0.33 mg kg⁻¹). p-XRF measurements showed a spatial prediction accuracy of soil PTEs similar to that of laboratory analysis measurements. The spatial prediction accuracy of different PTEs outputted by the model averaging method was as follows: Zn (R² = 0.71) > Cd (R² = 0.68) > Ni (R² = 0.67) > Cu (R² = 0.62) > As (R² = 0.50). The prediction accuracy of the model averaging method for five PTEs studied herein was improved compared with that of the laboratory and p-XRF methods, which utilized individual geostatistical methods (e.g., IDW, OK, EBK). Our results proved that p-XRF was a reliable alternative to the traditional laboratory analysis methods for mapping soil PTEs. The model averaging approach improved the prediction accuracy of the soil PTE spatial distribution and reduced the time and cost of monitoring and mapping PTE soil contamination.

Automatic road extraction from very-high-resolution remote sensing images has become a popular topic in a wide range of fields. Convolutional neural networks are often used for this purpose. However, many network models do not achieve satisfactory extraction results because of the elongated nature and varying sizes of roads in images. To improve the accuracy of road extraction, this paper proposes a deep learning model based on the structure of Deeplab v3. It incorporates squeeze-and-excitation (SE) module to apply weights to different feature channels, and performs multi-scale upsampling to preserve and fuse shallow and deep information. To solve the problems associated with unbalanced road samples in images, different loss functions and backbone network modules are tested in the model’s training process. Compared with cross entropy, dice loss can improve the performance of the model during training and prediction. The SE module is superior to ResNext and ResNet in improving the integrity of the extracted roads. Experimental results obtained using the Massachusetts Roads Dataset show that the proposed model (Nested SE-Deeplab) improves F1-Score by 2.4% and Intersection over Union by 2.0% compared with FC-DenseNet. The proposed model also achieves better segmentation accuracy in road extraction compared with other mainstream deep-learning models including Deeplab v3, SegNet, and UNet.

Wise soil management requires detailed soil information, but conventional soil class mapping in a rather coarse spatial resolution cannot meet the demand for precision agriculture. With the advantages of non-destructiveness, rapid cost-efficiency, and labor savings, the spectroscopic technique has proved its high potential for success in soil classification. Previous studies mainly focused on predicting soil classes using a single sensor. In this study, we attempted to compare the predictive ability of visible near infrared (vis-NIR) spectra, mid-infrared (MIR) spectra, and their fused spectra for soil classification. A total of 146 soil profiles were collected from Zhejiang, China, and the soil properties and spectra were measured by their genetic horizons. Along with easy-to-measure auxiliary soil information (soil organic matter, soil texture, color and pH), four spectral data, including vis-NIR, MIR, their simple combination (vis-NIR-MIR), and outer product analysis (OPA) fused spectra, were used for soil classification using a multiple objectives mixed support vector machine model. The independent validation results showed that the classification model using MIR (accuracy of 64.5%) was slightly better than that using vis-NIR (accuracy of 64.2%). The predictive model built on vis-NIR-MIR did not improve the classification accuracy, having the lowest accuracy of 61.1%, which likely resulted from an over-fitting problem. The model based on OPA fused spectra performed best with an accuracy of 68.4%. Our results prove the potential of fusing vis-NIR and MIR using OPA for improving prediction ability for soil classification.

To verify the feasibility of portable X-ray fluorescence (PXRF) for rapidly analyzing, assessing and improving soil heavy metals mapping, 351 samples were collected from Fuyang District, Hangzhou City, in eastern China. Ordinary kriging (OK) and co-ordinary kriging (COK) combined with PXRF measurements were used to explore spatial patterns of heavy metals content in the soil. The Getis-Ord index was calculated to discern hot spots of heavy metals. Finally, multi-variable indicator kriging was conducted to obtain a map of multi-heavy metals pollution. The results indicated Cd is the primary pollution element in Fuyang, followed by As and Pb. Application of PXRF measurements as covariates in COK improved model accuracy, especially for Pb and Cd. Heavy metals pollution hot spots were mainly detected in northern Fuyang and plains along the Fuchun River in southern Fuyang because of mining, industrial and traffic activities, and irrigation with polluted water. Area with high risk of multi-heavy metals pollution mainly distributed in plain along the Fuchun River and the eastern Fuyang. These findings certified the feasibility of using PXRF as an efficient and reliable method for soil heavy metals pollution assessment and mapping, which could contribute to reduce the cost of surveys and pollution remediation.

Soil salinization is a global issue resulting in soil degradation, arable land loss and ecological environmental deterioration. Over the decades, multispectral and hyperspectral remote sensing have enabled efficient and cost-effective monitoring of salt-affected soils. However, the potential of hyperspectral sensors installed on an unmanned aerial vehicle (UAV) to estimate and map soil salinity has not been thoroughly explored. This study quantitatively characterized and estimated field-scale soil salinity using an electromagnetic induction (EMI) equipment and a hyperspectral camera installed on a UAV platform. In addition, 30 soil samples (0~20 cm) were collected in each field for the lab measurements of electrical conductivity. First, the apparent electrical conductivity (EC_a) values measured by EMI were calibrated using the lab measured electrical conductivity derived from soil samples based on empirical line method. Second, the soil salinity was quantitatively estimated using the random forest (RF) regression method based on the reflectance factors of UAV hyperspectral images and satellite multispectral data. The performance of models was assessed by Lin’s concordance coefficient (CC), ratio of performance to deviation (RPD), and root mean square error (RMSE). Finally, the soil salinity of three study fields with different land cover were mapped. The results showed that bare land (field A) exhibited the most severe salinity, followed by dense vegetation area (field C) and sparse vegetation area (field B). The predictive models using UAV data outperformed those derived from GF-2 data with lower RMSE, higher CC and RPD values, and the most accurate UAV-derived model was developed using 62 hyperspectral bands of the image of the field A with the RMSE, CC, and RPD values of 1.40 dS m⁻¹, 0.94, and 2.98, respectively. Our results indicated that UAV-borne hyperspectral imager is a useful tool for field-scale soil salinity monitoring and mapping. With the help of the EMI technique, quantitative estimation of surface soil salinity is critical to decision-making in arid land management and saline soil reclamation.

Soil organic matter (SOM) and pH are essential soil fertility indictors of paddy soil in the middle-lower Yangtze Plain. Rapid, non-destructive and accurate determination of SOM and pH is vital to preventing soil degradation caused by inappropriate land management practices. Visible-near infrared (vis-NIR) spectroscopy with multivariate calibration can be used to effectively estimate soil properties. In this study, 523 soil samples were collected from paddy fields in the Yangtze Plain, China. Four machine learning approaches—partial least squares regression (PLSR), least squares-support vector machines (LS-SVM), extreme learning machines (ELM) and the Cubist regression model (Cubist)—were used to compare the prediction accuracy based on vis-NIR full bands and bands reduced using the genetic algorithm (GA). The coefficient of determination (R²), root mean square error (RMSE), and ratio of performance to inter-quartile distance (RPIQ) were used to assess the prediction accuracy. The ELM with GA reduced bands was the best model for SOM (SOM: R² = 0.81, RMSE = 5.17, RPIQ = 2.87) and pH (R² = 0.76, RMSE = 0.43, RPIQ = 2.15). The performance of the LS-SVM for pH prediction did not differ significantly between the model with GA (R² = 0.75, RMSE = 0.44, RPIQ = 2.08) and without GA (R² = 0.74, RMSE = 0.45, RPIQ = 2.07). Although a slight increase was observed when ELM were used for prediction of SOM and pH using reduced bands (SOM: R² = 0.81, RMSE = 5.17, RPIQ = 2.87; pH: R² = 0.76, RMSE = 0.43, RPIQ = 2.15) compared with full bands (R² = 0.81, RMSE = 5.18, RPIQ = 2.83; pH: R² = 0.76, RMSE = 0.45, RPIQ = 2.07), the number of wavelengths was greatly reduced (SOM: 201 to 44; pH: 201 to 32). Thus, the ELM coupled with reduced bands by GA is recommended for prediction of properties of paddy soil (SOM and pH) in the middle-lower Yangtze Plain.