Predict Soil Organic Carbon Content Research Articles

Predicting soil organic carbon content in Spain by combining Landsat TM and ALOS PALSAR images

Soil organic carbon (SOC) is the largest carbon pool and a key property of ecosystems. Compared with traditional field surveys, remote sensing (RS) represents a more efficient approach to mapping SOC, especially in larger-scale areas. Hyperspectral imagery provides a great potential for SOC prediction, but predicting the SOC content at a regional scale for a given year remains a great challenge. Multispectral RS images (e.g., Landsat images) with middle spatial resolution can be used as an alternative to predict SOC amongst other images. However, multispectral images are commonly affected by cloud cover and lack information on the shallow soil surface and soil surface with vegetation. By contrast, synthetic aperture radar (SAR) images can capture shallow surface information, penetrate through vegetation and are not affected by cloud cover. In this study, a new approach was evaluated for the prediction of SOC content by integrating Landsat Thematic Mapper (TM) and Advanced Land Observing Satellite (ALOS) Phased Arrayed L-band Synthetic Aperture Radar (PALSAR) images. Mainland Spain (hereafter Spain) was used as the study site. Soil samples from the Land Use and Coverage Area frame Survey (LUCAS) European topsoil dataset in 2009 were utilized to fit a random forest (RF) model of the relationships between SOC content and several covariates in Spain. SOC prediction using only the Landsat TM image and only the ALOS PALSAR image was established as benchmarks to verify the efficiency of the proposed approach. The results for Spain showed that the integrated approach holds the highest prediction accuracy, with R2, RPD and RMSE of 0.59, 1.98 and 9.27 g kg−1, respectively. The correlations between the SOC content and various covariates were investigated and discussed, and the derived indices presented high relationships with SOC. The proposed method can accurately map the SOC content covering a large area by combining spectral information on land cover from the Landsat TM and topsoil information from the ALOS PALSAR. Land covers and elevation are closely related to SOC prediction, where lands with high vegetation canopy density have high SOC content, and the SOC content in areas with slopes within 15° is mainly concentrated between 3 and 26 g kg−1.

When does stratification of a subtropical soil spectral library improve predictions of soil organic carbon content?

More accurate models for the prediction of soil organic carbon (SOC) by visible-near-infrared (Vis-NIR) spectroscopy remains a challenging task, especially when the soil spectral libraries (SSL) is composed of soils with a high pedological variation. One proposition to increase the models accuracy is to reduce the SSL variance, which can be achieved by stratifying the library into sub-libraries. Thus, the main objective of this study was to evaluate whether the stratification of a SSL by environmental, pedological and Vis-NIR spectral criteria results in greater accuracy of spectroscopic models than to general models for prediction of SOC content. The performance of the models was evaluated considering the variance of soil components and sample number. In addition, we tested the effect of two spectral preprocessing techniques and two multivariate calibration methods on spectroscopic modeling. For these purposes, a SSL composed of 2471 samples from Southern Brazil was stratified based on i) physiographic region; ii) land-use/land-cover; iii) soil texture, and iv) spectral class. Two spectral processing techniques: Savitzky-Golay - 1st derivative (SGD) and continuum removed reflectance (CRR) and two multivariate methods (partial least squares regression - PLSR and Cubist) were used to fit the models. The best performances for the global and local models were achieved with the CRR spectral processing associated with the Cubist method. The stratification of the SSL in more homogeneous sample groups by environmental criteria (physiographic regions and land-use/land-cover) improved the accuracy of SOC predictions compared to pedological (soil texture) and Vis-NIR spectral (spectral classes) criteria. The reduction in the number of samples negatively affected the performance of models for sub-libraries with high pedological and spectral variation. Stratification criteria were proposed in a flowchart to assist in decision making in future studies. Our findings suggest the importance of sample balance across environmental, pedological and spectral strata, in order to optimize SOC predictions.

Predicting Soil Organic Carbon Content Using Hyperspectral Remote Sensing in a Degraded Mountain Landscape in Lesotho

Soil organic carbon constitutes an important indicator of soil fertility. The purpose of this study was to predict soil organic carbon content in the mountainous terrain of eastern Lesotho, southern Africa, which is an area of high endemic biodiversity as well as an area extensively used for small-scale agriculture. An integrated field and laboratory approach was undertaken, through measurements of reflectance spectra of soil using an Analytical Spectral Device (ASD) FieldSpec® 4 optical sensor. Soil spectra were collected on the land surface under field conditions and then on soil in the laboratory, in order to assess the accuracy of field spectroscopy-based models. The predictive performance of two different statistical models (random forest and partial least square regression) was compared. Results show that random forest regression can most accurately predict the soil organic carbon contents on an independent dataset using the field spectroscopy data. In contrast, the partial least square regression model overfits the calibration dataset. Important wavelengths to predict soil organic contents were localised around the visible range (400–700 nm). This study shows that soil organic carbon can be most accurately estimated using derivative field spectroscopy measurements and random forest regression.

Improving the Spatial Prediction of Soil Organic Carbon Content in Two Contrasting Climatic Regions by Stacking Machine Learning Models and Rescanning Covariate Space

Understanding the spatial distribution of soil organic carbon (SOC) content over different climatic regions will enhance our knowledge of carbon gains and losses due to climatic change. However, little is known about the SOC content in the contrasting arid and sub-humid regions of Iran, whose complex SOC–landscape relationships pose a challenge to spatial analysis. Machine learning (ML) models with a digital soil mapping framework can solve such complex relationships. Current research focusses on ensemble ML models to increase the accuracy of prediction. The usual ensemble method is boosting or weighted averaging. This study proposes a novel ensemble technique: the stacking of multiple ML models through a meta-learning model. In addition, we tested the ensemble through rescanning the covariate space to maximize the prediction accuracy. We first applied six state-of-the-art ML models (i.e., Cubist, random forests (RF), extreme gradient boosting (XGBoost), classical artificial neural network models (ANN), neural network ensemble based on model averaging (AvNNet), and deep learning neural networks (DNN)) to predict and map the spatial distribution of SOC content at six soil depth intervals for both regions. In addition, the stacking of multiple ML models through a meta-learning model with/without rescanning the covariate space were tested and applied to maximize the prediction accuracy. Out of six ML models, the DNN resulted in the best modeling accuracies, followed by RF, XGBoost, AvNNet, ANN, and Cubist. Importantly, the stacking of models indicated a significant improvement in the prediction of SOC content, especially when combined with rescanning the covariate space. For instance, the RMSE values for SOC content prediction of the upper 0–5 cm of the soil profiles of the arid site and the sub-humid site by the proposed stacking approaches were 17% and 9% respectively, less than that obtained by the DNN models—the best individual model. This indicates that rescanning the original covariate space by a meta-learning model can extract more information and improve the SOC content prediction accuracy. Overall, our results suggest that the stacking of diverse sets of models could be used to more accurately estimate the spatial distribution of SOC content in different climatic regions.

Predictive mapping of soil organic carbon in Northeast Algeria

Soils of semi-arid climates undergo organic carbon loss, which in turn affects their agricultural potential. Geostatistics is often used as an interpolation tool to thoroughly describe SOC spatial distribution. To focus on soil organic carbon (SOC) depletion, the Tiffech watershed (Northeast of Algeria), an economically important agricultural area, was chosen due to intensive agricultural practices, decline of forests and occurrence of erosion.The present study aimed to predict the spatial variation of SOC in Tiffech watershed using geostatistics and a Geographical Information System (GIS) software, comparing the performance of two geostatistical methods—Ordinary Kriging (OK) and Regression Kriging (RK)—also assessing the role of auxiliary variables in improving the prediction accuracy and highlighting the role of land cover in SOC storage.The SOC content in Tiffech soils was determined on 42 soil samples from the surface layer (0–10 cm) collected all over the study area and the results were used to estimate SOC density in non-sampled locations.The prediction efficiency of the two methods was evaluated by calculating the Mean Error (ME), the Root Mean Square Error (RMSE) and the Root Mean Square Standardized Error (RMSSE).The interpolation results showed that SOC distribution in the study area was correlated to the topography, the clay index, and general landscape features. SOC content increased northwards in the area, ranging from 0.53 to 6.9 kg·m−2 in relation to land use. As expected, maps figured a good conservation status of SOC stocks in areas with dense vegetation; conversely poor SOC contents were estimated where land degradation factors take place.Cross-validation results showed an outperformance in the interpolation accuracy of RK on OK after the introduction of environmental variables, with an RMSE value of 0.02 versus 0.81. This suggests a higher efficiency of RK in predicting SOC content across the Tiffech area in comparison with OK, confirming that introducing some auxiliary data correlated to the target variable in SOC estimation, considerably improved the interpolation accuracy.

Soil Organic Carbon Prediction by Vis-NIR Spectroscopy: Case Study the Kur-Aras Plain, Azerbaijan

ABSTRACTVisible and near infrared (Vis-NIR) spectroscopy is a rapid, accurate, cost-effective, and nondestructive alternative method for soil analysis. The aim of this study was to examine the potential of Vis-NIR spectroscopy to predict soil organic carbon (SOC) content. In total, 194 soil samples were collected from 46 soil profiles in the Kur-Aras plain dominated by Calcisols, Gleysols, and Anthrosols, Azerbaijan. SOC content was predicted by partial least squares regression (PLSR) using spectral reflectance data. In the modeling phase, spectral data were involved in preprocessing and transformation techniques. The dataset was randomly separated into two subsets: a calibration subset and a validation subset that was independent of the calibration subset. Modeling results were evaluated based on coefficient of determination (R2), root mean square error of prediction (RMSE), and residual prediction deviation (RPD). Most optimal calibration model was considered when the prediction of the leave-one-out cross-validation showed best performance. This study found that different preprocessing procedures effect model performance. SOC content significantly effected spectral reflectance from soil, where a decrease in reflectance with an increase in SOC was determined through the entire wavelength. We obtained an accurate prediction model for SOC with the R2, RMSE, and RPD values of cross validation 0.85, 3.77 g/kg, and 2.54, respectively. This model is valid over the range 3.80–6.71 g/kg of SOC content. Further attempts need to be given to different calibration strategies to more accurately predict SOC content.

The Impact of Acquisition Date on the Prediction Performance of Topsoil Organic Carbon from Sentinel-2 for Croplands

The spatial assessment of soil organic carbon (SOC) is a major environmental challenge, notably for evaluating soil carbon stocks. Recent works have shown the capability of Sentinel-2 optical data to predict SOC content over temperate agroecosystems characterized by annual crops, using a single acquisition date. Considering a Sentinel-2 time series, this work intends to analyze the impact of acquisition date, and related weather and soil surface conditions on the prediction performance of topsoil SOC content (plough layer). A Sentinel-2 time-series was gathered, comprised of the dates corresponding to both the maximum of bare soil coverage and minimum of cloud coverage. Cross-validated partial least squares regression (PLSR) models were constructed between soil reflectance image spectra, and SOC content analyzed from 329 top soil samples collected over the study area. Cross-validation R2 ranged from 0.005 to 0.58, root mean square error from 5.86 to 3.02 g·kg−1 and residual prediction deviation values from 1.0 to 1.5 (without unit), according to date. The main factors influencing these differences were soil roughness, in conjunction with soil moisture, and the cloud and cloud shadow cover of the entire tile. The best performing dates were spring dates characterized by both lowest soil surface roughness and moisture content. Normalized difference vegetation index (NDVI) values below 0.35 did not influence prediction performance. This consolidates the previous results obtained during single date acquisitions and offers wider perspectives for the further use of Sentinel-2 into multidate mosaics for digital soil mapping.

Estimating forest soil organic carbon content using vis-NIR spectroscopy: Implications for large-scale soil carbon spectroscopic assessment

Large-scale soil organic carbon (SOC) stock assessment is expensive as a large number of samples must be collected and then their time-consuming measurements must be made in the laboratory. Previous studies have shown that visible-near-infrared reflectance (vis-NIR) spectroscopy can quickly predict SOC content at a low cost. However, the application of this method at the large scale remains challenging due to the high spatial heterogeneity of SOC and the spatially dependent relationships of soil spectra and SOC content. Here, we conducted large-scale soil sampling across China's forests and established the Chinese forest soil spectral library (CFSSL) by measuring SOC content and scanning the vis-NIR reflectance of 11, 213 soil samples. Compared with the traditional global partial least squares regression (PLSR) modeling method (R2 = 0.75, RPIQ = 1.95), the clustering by fast research and find of density peak in combination with the Cubist model significantly improved the prediction ability of SOC content (R2 = 0.96, RPIQ = 5.83). This study provided a cost-efficient spectroscopic methodology, including measurement and prediction modeling, for large-scale SOC estimation.

Assessing the Performance of UAS-Compatible Multispectral and Hyperspectral Sensors for Soil Organic Carbon Prediction

Laboratory spectroscopy has proved its reliability for estimating soil organic carbon (SOC) by exploiting the relationship between electromagnetic radiation and key spectral features of organic carbon located in the VIS-NIR-SWIR (350–2500 nm) region. While this approach provides SOC estimates at specific sampling points, geo-statistical or interpolation techniques are required to infer continuous spatial information. UAS-based proximal or remote sensing has the potential to provide detailed and spatially explicit spectral sampling of the topsoil at the field or even watershed scale. However, the factors affecting the quality of spectral acquisition under outdoor conditions need to be considered. In this study, we investigate the capabilities of two portable hyperspectral sensors (STS-VIS and STS-NIR), and two small-form multispectral cameras with narrow bands in the VIS-NIR region (Parrot Sequoia and Mini-MCA6), to predict SOC content. We collected spectral data under both controlled laboratory and outdoor conditions, with the latter being affected by variable illumination and atmospheric conditions and sensor-sample distance. We also analysed the transferability of the prediction models between different measurement setups by aligning spectra acquired under different conditions (laboratory and outdoor) or by different instruments. Our results indicate that UAS-compatible small-form sensors can be used to reliably estimate SOC. The results show that: (i) the best performance for SOC estimation under outdoor conditions was obtained using the VIS-NIR range, while the addition of the SWIR region decreased the prediction accuracy; (ii) prediction models using only the narrow bands of multispectral cameras gave similar or better performances than those using continuous spectra from the STS hyperspectral sensors; and (iii) when used in outdoor conditions, the micro hyperspectral sensors substantially benefitted from a laboratory model calibration followed by a spectral transfer using an internal soil standard. Based on this analysis, we recommend VIS-NIR portable instruments for estimating spatially distributed SOC data. The integration of these sensors in UAS-mapping devices could represent a cost-effective solution for soil research and precision farming applications when high resolution data are required.

Predicting soil organic carbon content in croplands using crop rotation and Fourier transform decomposed variables

Previous studies on soil organic carbon content or stock mapping mostly use natural environmental covariates and do not consider the soil management practice factor. However, human activities have become an important influencing factor for soil organic carbon, especially for agricultural soils. Crop species/crop rotations and management practices significantly affect the amount and spatial variation of soil organic carbon in croplands, but have not been considered for mapping soil organic carbon. In this study, we used direct crop rotation information and variables generated using Fourier transform on HJ-1A/1B NDVI time series data to capture the periodic effect of crop rotation, and explored the effectiveness of incorporating such information in predicting topsoil organic carbon content in cropland. A case study applied such method in a largely agricultural area in Anhui province, China. Crop rotation information was obtained through field investigation. Various combinations of predictive environmental variables were experimented for mapping soil organic carbon. The results were validated using field samples. Results showed that the combination of natural environment variables with both crop rotation type and variables derived through Fourier transform yielded the highest accuracy. In addition, only using the Fourier decomposed variables and crop rotation information were able to achieve a similar accuracy with using only soil formative natural environmental variables. This indicates that crop rotation information has comparable predictive power of soil organic carbon as natural environment variables. This study demonstrates the effectiveness of including agricultural practice information in digital soil mapping in agricultural landscapes with differences in crop rotation.

A multi-factor weighted regression approach for estimating the spatial distribution of soil organic carbon in grasslands

Soil organic carbon (SOC) plays an important role in global carbon cycling and is increasingly important to the ecosystem. An accurate SOC content map would significantly contribute to the proper application of ecological modeling. Therefore, there is a need to accurately estimate and map SOC content in grasslands. We evaluated various methods for estimating the SOC content of grasslands using field soil sampling data and auxiliary data in the pastoral area. The results showed that (1) SOC is affected by various factors, including geographic location, soil, topography, and climate. Single-variable SOC models account for 2%–72% of the variations in the grassland SOC. (2) Based on the correlation of environmental variables of SOC, normalized difference vegetation index, annual precipitation, annual average temperature, elevation, and moisture index were explored as critical auxiliary data to predict SOC content. We established multi-factor weighted regression model (MWRM). (3) We compared three spatial estimation methods, including inverse distance weighting, regression kriging, and MWRM, to determine a suitable method for SOC mapping. Our results indicate that among the three spatial estimation methods, MWRM provides the lowest prediction error (RMSE = 4.85 g/kg; MAE = 3.47 g/kg; MRE = 24.04%) and highest R2 (0.89) and Lin's concordance (0.94) values in the spatial estimation at a 0–10 cm soil layer. (4) Therefore, we applied MWRM to predict SOC content at various layers, and its SOC content prediction in the topsoil (0–20 cm) is better than that in the subsurface (20–30 cm) and subsoil (30–40 cm). The SOC content spatial distribution demonstrated a similar pattern for each soil layer and the SOC content gradually decreased with increasing soil depth.

Digital mapping of soil properties using multiple machine learning in a semi-arid region, central Iran

Knowledge about distribution of soil properties over the landscape is required for a variety of land management applications and resources, modeling, and monitoring practices. The main aim of this research was to conduct a spatially prediction of the top soil properties such as soil organic carbon (SOC), calcium carbonate equivalent (CCE), and clay content using digital soil mapping (DSM) approaches in Borujen region, Chaharmahal-Va-Bakhtiari province, central Iran. To achieve this goal, a total of 334 soil samples were collected from 0 to 30 cm depth. Three non-linear models including Cubist (Cu), Random Forest (RF), Regression Tree (RT) and a Multiple Linear Regression (MLR) were used to link environmental covariates and the studied soil properties. The environmental covariates were obtained from a digital elevation model (DEM) and satellite imagery (Landsat Enhanced Thematic Mapper; ETM). The model was calibrated and validated by the 10-fold cross-validation approach. Root mean square error (RMSE) and coefficient of determination (R2) were used to determine the performance of the models, and relative RMSE (RMSE%) was used to define prediction accuracy. According to the RMSE and R2, Cu and RF resulted in the most accurate predictions for CCE (R2 = 0.30 and RMSE = 9.52) and clay contents (R2 = 0.15 and RMSE = 7.86), respectively, while both of RF and Cu models showed the highest performance to predict SOC content (R2 = 0.55). Results showed that remote sensing covariates (Ratio Vegetation Index and band 4) were the most important variables to explain the variability of SOC and CCE content, but only topographic attributes were responsible for clay content variation. According to RMSE% results, it could be concluded that the best model is not necessarily able to make the most accurate estimation. This study recommended that more observations and denser sampling should be carried out in the entire study area. Alternatively, stratified sampling by elevation in homogeneous sub-areas was recommended. The stratified sampling probably will increase the performance of models.

Digital mapping of a soil profile

SummaryDetailed information on soil profiles is required for site characterization, soil mapping, studies on pedogenesis, soil–landscape modelling and soil classification. In this study, we used digital soil mapping techniques to map the profile wall of an Alfisol (90‐cm depth × 100‐cm width). Geochemical data (sampled at 10 cm × 10 cm) and digital images (resolution 1 cm × 1 cm) were used as input data in models that predicted a range of properties across the soil profile. Fuzzy c‐means clustering was applied to the profile and colour maps, and a confusion index was calculated. The colour coordinates were strongly correlated with SOC (soil organic carbon), sand and silt contents, and weathering indices. Random forest using the RGB model (R2 = 0.57, RMSE (root mean square error) = 2.41) showed less accurate prediction of SOC content than random forest using the CIE L*a*b* (R2 = 0.84, RMSE = 1.57) and HSV, hue, saturation, chroma (R2 = 0.85, RMSE = 1.46) models. Detailed profile maps showed considerable within‐horizon variation and identified animal holes, roots and limestone fragments. Colour and profile maps of soil properties and weathering indices matched field‐delineated soil horizons. Clusters obtained by soil properties and weathering indices represented the field‐delineated horizons and had the largest overall accuracy (P = 0.85 and 0.86, respectively). The confusion index showed that the horizons had wavy and somewhat irregular boundaries. We conclude that colour coordinates are useful for predicting and mapping soil properties in soil profiles. Random forest can be used to produce high‐resolution soil profile maps, and fuzzy c‐means clustering is useful for delineating soil horizons.HighlightsWe investigated the use of digital images in soil horizon delineation and within soil horizon variation We used various colour models to map soil profile properties and define horizons Colour and profile maps of soil properties and weathering indices matched field delineated soil horizons. Random forest can produce high‐resolution soil profile maps, and fuzzy c‐means clustering can delineate soil horizons

Building an Open Science Framework to Model Soil Organic Carbon.

As funding agencies embrace open science principles that encourage sharing data and computer code developed to produce research outputs, we must respond with new modes of publication. Furthermore, as we address the expanding reproducibility crisis in the sciences, we must work to release research materials in ways that enable reproducibility-publishing data, computer code, and research products in addition to the traditional journal article. Toward addressing these needs, we present an example framework to model and map soil organic carbon (SOC) in the cereal grains production region of the northwestern United States. Primarily associated with soil organic matter, SOC relates to many soil properties that influence resiliency and soil health for agriculture. It is also critical for understanding soil-atmospheric C flux, a significant part of the overall C budget of the Earth. The technique for modeling soil properties uses seven categories of environmental input data to make predictions: known soil attributes, climatic values, organisms present, relief, parent material, age, and spatial location. We gather data representing these categories from various public sources. The map is produced using a random forest statistical model with inputs to predict SOC content on a 30-m spatial grid. All modeling components including input data, metadata, computer code, and output products are made freely available under an explicit open-source license. In this way, reproducibility is supported, the methods and code released are available to be reused by other researchers, and the research products are open to critical review and improvement.

Prediction of soil organic carbon in a coal mining area by Vis-NIR spectroscopy.

Coal mining has led to increasingly serious land subsidence, and the reclamation of the subsided land has become a hot topic of concern for governments and scholars. Soil quality of reclaimed land is the key indicator to the evaluation of the reclamation effect; hence, rapid monitoring and evaluation of reclaimed land is of great significance. Visible-near infrared (Vis-NIR) spectroscopy has been shown to be a rapid, timely and efficient tool for the prediction of soil organic carbon (SOC). In this study, 104 soil samples were collected from the Baodian mining area of Shandong province. Vis-NIR reflectance spectra and soil organic carbon content were then measured under laboratory conditions. The spectral data were first denoised using the Savitzky-Golay (SG) convolution smoothing method or the multiple scattering correction (MSC) method, after which the spectral reflectance (R) was subjected to reciprocal, reciprocal logarithm and differential transformations to improve spectral sensitivity. Finally, regression models for estimating the SOC content by the spectral data were constructed using partial least squares regression (PLSR). The results showed that: (1) The SOC content in the mining area was generally low (at the below-average level) and exhibited great variability. (2) The spectral reflectance increased with the decrease of soil organic carbon content. In addition, the sensitivity of the spectrum to the change in SOC content, especially that in the near-infrared band of the original reflectance, decreased when the SOC content was low. (3) The modeling results performed best when the spectral reflectance was preprocessed by Savitzky-Golay (SG) smoothing coupled with multiple scattering correction (MSC) and first-order differential transformation (modeling R2 = 0.86, RMSE = 2.00 g/kg, verification R2 = 0.78, RMSE = 1.81 g/kg, and RPD = 2.69). In addition, the first-order differential of R combined with SG, MSC with R, SG together with MSC and R also produced better modeling results than other pretreatment combinations. Vis-NIR modeling with specific spectral preprocessing methods could predict SOC content effectively.

Local modeling approaches for estimating soil properties in selected Indian soils using diffuse reflectance data over visible to near-infrared region

Robust calibration algorithms are needed for the accurate assessment of soil properties in the diffuse reflectance spectroscopy (DRS) approach. Despite several studies on different calibration algorithms, the prediction accuracy of soil properties using DRS need to be improved. Specifically, the utility of local modeling approaches for small spectral libraries is less examined compared with global modeling approaches. In this study, we compared global modeling approaches such as partial-least-squares regression (PLSR), lasso, ridge regression with several locally-weighted PLSR (PLSRLW) approaches. We also examined seven different distance measures: Euclidean distance, covariance-based distance, correlation-based distance, surface difference spectrum, information-based distance, optimized principal component Mahalanobis, and locally-linear embeddings used in the PLSRLW approach for their effectiveness in modeling soil properties using DRS. A total of 954 soil samples were collected from three different states of India: West Bengal, Odisha, and Karnataka. Five soil properties such as sand content, clay content, soil organic carbon (SOC) content, extractable iron (Fe) content and extractable zinc (Zn) content were predicted using reflectance spectra over 350–2500 nm. Root-mean-squared error (RMSE) and the coefficient of determination (R2) were used as performance statistics. Among the global modeling approaches, lasso performed better than the PLSR although it is computationally more intensive than the PLSR. In general, the correlation-based PLSRLW performed significantly better than the global approaches. Specifically, the test R2 values increased from 0.66 to 0.72 for prediction of sand content, from 0.59 to 0.70 for prediction of SOC content, and from 0.70 to 0.74 for prediction of Fe content by using the PLSRLW over PLSR. We also used depth of absorption peak of spectra at approximately 1400, 1900 and 2200 nm for mineralogical characterization of soil samples. The results suggested that the improvement in prediction accuracy of soil properties using the PLSRLW was achieved because calibration samples which had same mineralogy as the test sample were given higher weights. These results suggest that the prediction accuracy of soil properties may also be improved in small spectral libraries if an appropriate local modeling scheme is selected.

Stocks, composition and vulnerability to loss of soil organic carbon predicted using mid-infrared spectroscopy

Developing a routine and cost effective capability for measuring soil organic carbon (SOC) content and composition will allow identification of land management practices with a potential to maintain or enhance SOC stocks. Coupling SOC content data and mid-infrared (MIR) spectra through the application of partial least-squares regression (PLSR) analyses has been used to develop such a prediction capability. The objective of this study was to determine whether MIR/PLSR analyses provide accurate estimates of the content and composition of SOC that can be used to quantify SOC stocks and its potential vulnerability to loss. Soil was collected from a field trial incorporating a range of land use (pasture, arable cropping and bare fallow) and tillage (intensive, minimum and no tillage) treatments over a nine-year period. The SOC content was measured by dry combustion analysis. Particulate organic carbon was separated from other forms of carbon on the basis of particle size (SOC in the >50 µm fraction). Resistant organic carbon was quantified using solid-state 13C nuclear magnetic resonance. The MIR/PLSR algorithms were successfully developed to predict the natural logarithms of the contents of SOC and POC in the collected soils. With initial calibration, a single MIR analysis could be used in conjunction with PLSR algorithms to predict the content of SOC and its allocation to component fractions. The MIR/PLSR predicted SOC contents provided reliable estimates of the impact of agricultural management on the 0–25-cm SOC stocks, as well as an indication of the vulnerability of SOC to loss. Development of this capability will facilitate the rapid and cost effective collection of SOC content data for detecting the impact of agricultural management treatments on SOC stocks, composition and potential vulnerability to change.

Evaluation of two methods to eliminate the effect of water from soil vis–NIR spectra for predictions of organic carbon

Visible near infrared reflectance spectroscopy (vis–NIR) is an increasingly popular measurement method that can provide cheaper and faster predictions of soil properties, including soil organic carbon content (SOC). The spectroscopic prediction method relies significantly on the development of regressions of data in spectral databases or libraries. While the vis–NIR estimation of SOC was developed in controlled laboratory conditions, its natural development in recent years has been to perform the vis–NIR measurements in situ, where soil spectra are recorded under field conditions. However, environmental factors, such as soil moisture content, have been shown to affect soil spectra, making the use of regressions derived using soil spectral libraries difficult. Direct standardization (DS) and external parameter orthogonalisation (EPO) are two methods that were proposed for the correction of variable moisture conditions and other environmental factors. In this study, we compared DS and EPO on a set of 150 soil samples (3 depths from each of 50 soil cores) from a farm in New Zealand. The samples were re-wetted under controlled conditions, and spectra were recorded at nine different moisture levels. Our results show that DS and EPO are two effective strategies to mitigate the effects of soil water content on vis–NIR spectra. While DS and EPO results were similar when a large number of soil cores were reserved for calibrating the moisture correction methods, SOC predictions using the EPO correction significantly outperformed those using the DS correction for a lower number of cores (5 cores, 15 samples).

Bayesian geostatistical prediction of soil organic carbon contents of solonchak soils in nor-thern Tarim Basin, Xinjiang, China.

Accurate prediction of soil organic carbon (SOC) distribution is crucial for soil resources utilization and conservation, climate change adaptation, and ecosystem health. In this study, we selected a 1300 m×1700 m solonchak sampling area in northern Tarim Basin, Xinjiang, China, and collected a total of 144 soil samples (5-10 cm). The objectives of this study were to build a Baye-sian geostatistical model to predict SOC content, and to assess the performance of the Bayesian model for the prediction of SOC content by comparing with other three geostatistical approaches [ordinary kriging (OK), sequential Gaussian simulation (SGS), and inverse distance weighting (IDW)]. In the study area, soil organic carbon contents ranged from 1.59 to 9.30 g·kg-1 with a mean of 4.36 g·kg-1 and a standard deviation of 1.62 g·kg-1. Sample semivariogram was best fitted by an exponential model with the ratio of nugget to sill being 0.57. By using the Bayesian geostatistical approach, we generated the SOC content map, and obtained the prediction variance, upper 95% and lower 95% of SOC contents, which were then used to evaluate the prediction uncertainty. Bayesian geostatistical approach performed better than that of the OK, SGS and IDW, demonstrating the advantages of Bayesian approach in SOC prediction.

Using an inexpensive color sensor for rapid assessment of soil organic carbon

Quantifying soil organic carbon (SOC) is important for soil management, precision agriculture, soil mapping and carbon dynamics research. Inexpensive sensor technologies offer the potential for rapid quantification of SOC in laboratory samples as well as in the field. The objective of this study was to use a commercially-available color sensor to develop SOC prediction models for both dry and moist soils from the Piedmont region of South Carolina. Thirty-one soil samples were analyzed for lightness to darkness, redness to greenness, and yellowness to blueness (CIEL*a*b*) color using a Nix Pro™ color sensor. Soil color was measured under both dry and moist soil conditions and the depth of each soil sample was also recorded. Using L*, a*, b* and soil depth for each sample as initial predictors, regression analyses were conducted to develop SOC prediction models for dry and moist soils. The resulting residual plots, root mean squared errors (RMSE), and coefficients of determination (R2) were used to assess model fits for predicting the SOC content of soil. Cross validation was conducted to determine the efficiency of the predictive models and the mean squared prediction error (MSPE) was calculated. The final models included soil depth, L*, and a* as independent variables (dry soils R2=0.7978 and MSPE=0.0819, moist soils R2=0.7254 and MSPE=0.1536). The results suggest that soil color sensors have potential for rapid SOC determination, and soil depth and color are useful in predicting SOC content in soils.