Site-Independent Mapping of Clay Content in Vineyard Soils via Mobile Proximal Gamma-Ray Spectrometry and Machine Learning Calibrations.

Vineyards hold considerable soil variability between regions and plots, and there is frequently large soil heterogeneity within plots. Clay content in vineyard soils is of interest with respect to soil management, environmental monitoring, and wine quality. However, spatially resolved clay mapping is laborious and expensive. Gamma-ray spectrometry (GS) is a suitable tool for predicting clay content in precision agriculture when locally calibrated, but it has scarcely been tested site-independently and in vineyards. This study evaluated GS to predict clay content with a site-independent calibration and four machine learning algorithms (Support Vector Machines, Random Forest, k-Nearest Neighbors, and Bayesian regulated neuronal networks) in eight vineyards from four German vine-growing regions. Clay content in the studied soils ranged from 62 to 647 g kg-1. The Random Forest calibration was most suitable. Test set evaluation revealed good model performance for the entire dataset with RPIQ = 4.64, RMSEP = 56.7 g kg-1, and R2 = 0.87; however, prediction quality varied between the sites. Overall, GS with the Random Forest model calibration was appropriate to predict the clay content and its spatial distribution, even for heterogeneous geopedological settings and in individual plots. Therefore, GS is considered a valuable tool for soil mapping in vineyards, where clay content and product quality are closely linked.

Evaluation of Mid-Infrared and X-ray Fluorescence Data Fusion Approaches for Prediction of Soil Properties at the Field Scale.

Previous studies investigating multi-sensor fusion for the collection of soil information have shown variable improvements, and the underlying prediction mechanisms are not sufficiently understood for spectrally-active and -inactive properties. Our objective was to study prediction mechanisms and benefits of model fusion by measuring mid-infrared (MIR) and X-ray fluorescence (XRF) spectra, texture, total and labile organic carbon (OC) and nitrogen (N) content, pH, and cation exchange capacity (CEC) for n = 117 soils from an arable field in Germany. Partial least squares regression models underwent a three-fold training/testing procedure using MIR spectra or elemental concentrations derived from XRF spectra. Additionally, two sequential hybrid and two high-level fusion approaches were tested. For the studied field, MIR was superior for organic properties (ratio of prediction to interquartile distance of validation (RPIQV) for total OC = 7.7 and N = 5.0)), while XRF was superior for inorganic properties (RPIQV for clay = 3.4, silt = 3.0, and sand = 1.8). Even the optimal fusion approach brought little to no accuracy improvement for these properties. The high XRF accuracy for clay and silt is explained by the large number of elements with variable importance in the projection scores >1 (Fe ≈ Ni > Si ≈ Al ≈ Mg > Mn ≈ K ≈ Pb (clay only) ≈ Cr) with strong spearman correlations (±0.57 < rs < ±0.90) with clay and silt. For spectrally-inactive properties relying on indirect prediction mechanisms, the relative improvements from the optimal fusion approach compared to the best single spectrometer were marginal for pH (3.2% increase in RPIQV versus MIR alone) but more pronounced for labile OC (9.3% versus MIR) and CEC (12% versus XRF). Dominance of a suboptimal spectrometer in a fusion approach worsened performance compared to the best single spectrometer. Granger-Ramanathan averaging, which weights predictions according to accuracy in training, is therefore recommended as a robust approach to capturing the potential benefits of multiple sensors.

Spatial Interpolation of Gravimetric Soil Moisture Using EM38-mk Induction and Ensemble Machine Learning (Case Study from Dry Steppe Zone in Volgograd Region).

The implementation of the sustainable management of the interaction between agriculture and the environment requires an increasingly deep understanding and numerical description of the soil genesis and properties of soils. One of the areas of application of relevant knowledge is digital irrigated agriculture. During the development of such technologies, the traditional methods of soil research can be quite expensive and time consuming. Proximal soil sensing in combination with predictive soil mapping can significantly reduce the complexity of the work. In this study, we used topographic variables and data from the Electromagnetic Induction Meter (EM38-mk) in combination with soil surface hydrological variables to produce cartographic models of the gravimetric soil moisture for a number of depth intervals. For this purpose, in dry steppe zone conditions, a test site was organized. It was located at the border of the parcel containing the irrigated soybean crop, where 50 soil samples were taken at different points alongside electrical conductivity data (ECa) measured in situ in the field. The modeling of the gravimetric soil moisture was carried out with the stepwise inclusion of independent variables, using methods of ensemble machine learning and spatial cross-validation. The obtained cartographic models showed satisfactory results with the best performance R2cv 0.59-0.64. The best combination of predictors that provided the best results of the model characteristics for predicting gravimetric soil moisture were geographical variables (buffer zone distances) in combination with the initial variables converted into the principal components. The cartographic models of the gravimetric soil moisture variability obtained this way can be used to solve the problems of managed irrigated agriculture, applying fertilizers at variable rates, thereby optimizing the use of resources by crop producers, which can ultimately contribute to the sustainable management of natural resources.

Geochemical fingerprinting and magnetic susceptibility to unravel the heterogeneous composition of urban soils.

The typically high heterogeneity of urban soil properties challenges their characterization and interpretation. The objective of this study was to investigate if proximally sensed volume-specific magnetic susceptibility and/or geochemical soil properties can uncover differences in anthropogenic, lithogenic and pedological contributions in, and between, urban soils. We also tested if volume-specific magnetic susceptibility can predict heavy metal enrichment. Data on 29 soil properties of 103 soil horizons from 16 soils from Ghent, Belgium, were analyzed by factor analysis. A correlation analysis, and in-depth analysis of five contrasting urban soils supplemented insights gained from the factor analysis. The factor analysis extracted four factors: 29.2 % of the soil property variability was attributed to fossil fuel combustion and industrial processes, with high (>0.80) loadings for S, organic carbon, magnetic susceptibility, and Zn. Furthermore, 26.0 % of the variability was linked to parent material differences, with high loadings (>0.80) for K, Rb and Ti. In absence of geogenic carbonates, increased soil alkalinity due to anthropogenic input of CaCO3 explained 17.0 % of the variability. Lastly, 4.7 % of the variability resulted from variable Zr contents by local geology. Elemental analysis by XRF, possibly combined with magnetic susceptibility measurements, helped to explain lateral or vertical differences related to (1) the nature of anthropogenic influence, for instance burning (e.g., by the S and Zn content) or the incorporation of building rubble (e.g., by the Ca content); (2) the particle size distribution (e.g., by the K, Rb or Ti content); (3) lithology (e.g., by the Zr content); or (4) pedology, such as organic matter build-up (e.g., by the S content) or leaching of alkalis (e.g., by the Ca content). Even though artifacts and soil translocation were common in the studied soils, volume-specific soil magnetic susceptibility measured on fine earth predicted the total heavy metal pollution loading index well (Pearson correlation = 0.85).

A loss function to evaluate agricultural decision-making under uncertainty: a case study of soil spectroscopy.

Modern sensor technologies can provide detailed information about soil variation which allows for more precise application of fertiliser to minimise environmental harm imposed by agriculture. However, growers should lose neither income nor yield from associated uncertainties of predicted nutrient concentrations and thus one must acknowledge and account for uncertainties. A framework is presented that accounts for the uncertainty and determines the cost-benefit of data on available phosphorus (P) and potassium (K) in the soil determined from sensors. For four fields, the uncertainty associated with variation in soil P and K predicted from sensors was determined. Using published fertiliser dose-yield response curves for a horticultural crop the effect of estimation errors from sensor data on expected financial losses was quantified. The expected losses from optimal precise application were compared with the losses expected from uniform fertiliser application (equivalent to little or no knowledge on soil variation). The asymmetry of the loss function meant that underestimation of P and K generally led to greater losses than the losses from overestimation. This study shows that substantial financial gains can be obtained from sensor-based precise application of P and K fertiliser, with savings of up to £121 ha-1 for P and up to £81 ha-1 for K, with concurrent environmental benefits due to a reduction of 4-17 kg ha-1 applied P fertiliser when compared with uniform application. Supplementary Information: The online version contains supplementary material available at 10.1007/s11119-022-09887-2.

Assessment of Soil Fertility Using Induced Fluorescence and Machine Learning.

Techniques such as proximal soil sampling are investigated to increase the sampling density and hence the resolution at which nutrient prescription maps are developed. With the advent of a commercial mobile fluorescence sensor, this study assessed the potential of fluorescence to estimate soil chemical properties and fertilizer recommendations. This experiment was conducted over two years at nine sites on 168 soil samples and used random forest regression to estimate soil properties, fertility classes, and recommended N rates for maize production based on induced fluorescence of air-dried soil samples. Results showed that important soil properties such as soil organic matter, pH, and CEC can be estimated with a correlation of 0.74, 0.75, and 0.75, respectively. When attempting to predict fertility classes, this approach yielded an overall accuracy of 0.54, 0.78, and 0.69 for NO3-N, SOM, and Zn, respectively. The N rate recommendation for maize can be directly estimated by fluorescence readings of the soil with an overall accuracy of 0.78. These results suggest that induced fluorescence is a viable approach for assessing soil fertility. More research is required to transpose these laboratory-acquired soil analysis results to in situ readings successfully.

Evaluation of Two Portable Hyperspectral-Sensor-Based Instruments to Predict Key Soil Properties in Canadian Soils.

In contrast with classic bench-top hyperspectral (multispectral)-sensor-based instruments (spectrophotometers), the portable ones are rugged, relatively inexpensive, and simple to use; therefore, they are suitable for field implementation to more closely examine various soil properties on the spot. The purpose of this study was to evaluate two portable spectrophotometers to predict key soil properties such as texture and soil organic carbon (SOC) in 282 soil samples collected from proportional fields in four Canadian provinces. Of the two instruments, one was the first of its kind (prototype) and was a mid-infrared (mid-IR) spectrophotometer operating between ~5500 and ~11,000 nm. The other instrument was a readily available dual-type spectrophotometer having a spectral range in both visible (vis) and near-infrared (NIR) regions with wavelengths ranging between ~400 and ~2220 nm. A large number of soil samples (n = 282) were used to represent a wide variety of soil textures, from clay loam to sandy soils, with a considerable range of SOC. These samples were subjected to routine laboratory soil analysis before both spectrophotometers were used to collect diffuse reflectance spectroscopy (DRS) measurements. After data collection, the mid-IR and vis-NIR spectra were randomly divided into calibration (70%) and validation (30%) sets. Partial least squares regression (PLSR) was used with leave one out cross-validation techniques to derive the spectral calibrations to predict SOC, sand, and clay content. The performances of the calibration models were reevaluated on the validation set. It was found that sand content can be predicted more accurately using the portable mid-IR spectrophotometer and clay content is better predicted using the readily available dual-type vis-NIR spectrophotometer. The coefficients of determination (R2) and root mean squared error (RMSE) were determined to be most favorable for clay (0.82 and 78 g kg-1) and sand (0.82 and 103 g kg-1), respectively. The ability to predict SOC content precisely was not particularly good for the dataset of soils used in this study with an R2 and RMSE of 0.54 and 4.1 g kg-1. The tested method demonstrated that both portable mid-IR and vis-NIR spectrophotometers were comparable in predicting soil texture on a large soil dataset collected from agricultural fields in four Canadian provinces.

Spectral data of tropical soils using dry-chemistry techniques (VNIR, XRF, and LIBS): A dataset for soil fertility prediction.

Proximal soil sensing technologies, such as visible and near infrared diffuse reflectance spectroscopy (VNIR), X-ray fluorescence spectroscopy (XRF), and laser-induced breakdown spectroscopy (LIBS), are dry-chemistry techniques that enable rapid and environmentally friendly soil fertility analyses. The application of XRF and LIBS sensors in an individual or combined manner for soil fertility prediction is quite recent, especially in tropical soils. The shared dataset presents spectral data of VNIR, XRF, and LIBS sensors, even as the characterization of key soil fertility attributes (clay, organic matter, cation exchange capacity, pH, base saturation, and exchangeable P, K, Ca, and Mg) of 102 soil samples. The samples were obtained from two Brazilian agricultural areas and have a wide variation of chemical and textural attributes. This is a pioneer dataset of tropical soils, with potential to be reused for comparative studies with other datasets, e.g., comparing the performance of sensors, instrumental conditions, and/or predictive models on different soil types, soil origin, concentration range, and agricultural practices. Moreover, it can also be applied to compose soil spectral libraries that use spectral data collected under similar instrumental conditions.

Evaluation of Optimized Preprocessing and Modeling Algorithms for Prediction of Soil Properties Using VIS-NIR Spectroscopy.

The absorbance spectra for air-dried and ground soil samples from Ontario, Canada were collected in the visible and near-infrared (VIS-NIR) region from 343 to 2200 nm. The study examined thirteen combination of six preprocessing (1st derivative, 2nd derivative, Savitzky-Golay, Gap, SNV and Detrend) method included in 'prospectr' R package along with four modeling approaches: partial least square regression (PLSR), cubist, random forest (RF), and extreme learning machine (ELM) for prediction of the soil organic matter (SOM). The 1st derivative + gap, 2nd derivative + gap and standard normal variance (SNV) were the best preprocessing algorithms. Thus, only these three preprocessing algorithms along with four modeling approaches were used for prediction of soil pH, electrical conductively (EC), %sand, %silt, %clay, %very coarse sand (VCS), %coarse sand (CS), %medium sand (ms) and %fine sand (fs). The results showed that OM, pH, %sand, %silt and %CS were all predicted with confidence (R2 > 0.60) and the combination of 1st derivative + gap and RF gained the best performance. A detailed comparison of the preprocessing and modeling algorithms for various soil properties in this study demonstrate that for better prediction of soil properties using VIS-NIR spectroscopy requires different preprocessing and modeling algorithms. However, in general RF and 1st derivative + gap can be labeled at the best combination of preprocessing and modelling algorithms.

Application of Low-Cost MEMS Spectrometers for Forest Topsoil Properties Prediction.

Increasing temperatures and drought occurrences recently led to soil moisture depletion and increasing tree mortality. In the interest of sustainable forest management, the monitoring of forest soil properties will be of increasing importance in the future. Vis-NIR spectroscopy can be used as fast, non-destructive and cost-efficient method for soil parameter estimations. Microelectromechanical system devices (MEMS) have become available that are suitable for many application fields due to their low cost as well as their small size and weight. We investigated the performance of MEMS spectrometers in the visual and NIR range to estimate forest soil samples total C and N content of Ah and Oh horizons at the lab. The results were compared to a full-range device using PLSR and Cubist regression models at local (2.3 ha, n: Ah = 60, Oh = 50) and regional scale (State of Saxony, Germany, 184,000 km2, n: Ah = 186 and Oh = 176). For each sample, spectral reflectance was collected using MEMS spectrometer in the visual (Hamamatsu C12880MA) and NIR (NeoSpetrac SWS62231) range and using a conventional full range device (Veris Spectrophotometer). Both data sets were split into a calibration (70%) and a validation set (30%) to evaluate prediction power. Models were calibrated for Oh and Ah horizon separately for both data sets. Using the regional data, we also used a combination of both horizons. Our results show that MEMS devices are suitable for C and N prediction of forest topsoil on regional scale. On local scale, only models for the Ah horizon yielded sufficient results. We found moderate and good model results using MEMS devices for Ah horizons at local scale (R2≥ 0.71, RPIQ ≥ 2.41) using Cubist regression. At regional scale, we achieved moderate results for C and N content using data from MEMS devices in Oh (R2≥ 0.57, RPIQ ≥ 2.42) and Ah horizon (R2≥ 0.54, RPIQ ≥2.15). When combining Oh and Ah horizons, we achieved good prediction results using the MEMS sensors and Cubist (R2≥ 0.85, RPIQ ≥ 4.69). For the regional data, models using data derived by the Hamamatsu device in the visual range only were least precise. Combining visual and NIR data derived from MEMS spectrometers did in most cases improve the prediction accuracy. We directly compared our results to models based on data from a conventional full range device. Our results showed that the combination of both MEMS devices can compete with models based on full range spectrometers. MEMS approaches reached between 68% and 105% of the corresponding full ranges devices R2 values. Local models tended to be more accurate than regional approaches for the Ah horizon. Our results suggest that MEMS spectrometers are suitable for forest soil C and N content estimation. They can contribute to improved monitoring in the future as their small size and weight could make in situ measurements feasible.

Comparison of Cubist models for soil organic carbon prediction via portable XRF measured data.

Soil organic carbon (SOC) tends to form complexes with most metallic ions within the soil system. Relatively few studies compare SOC predictions via portable X-ray fluorescence (pXRF) measured data coupled with the Cubist algorithm. The current study applied three different Cubist models to estimate SOC while using several pXRF measured data. Soil samples (n = 158) were collected from the Litavka floodplain area during two separate sampling campaigns in 2018. Thirteen pXRF data or predictors (K, Ca, Rb, Mn, Fe, As, Ba, Th, Pb, Sr, Ti, Zr, and Zn) were selected to develop the proposed models. Validation and comparison of the models applied the mean absolute error (MAE), root mean square error (RMSE), and coefficient of determination (R2). The results revealed that Cubist 1, utilizing all the predictors yielded the best model outcome (MAE = 0.51%, RMSE = 0.68%, R2 = 0.78) followed by Cubist 2, using predictors with relatively high importance (VarImp. predictors) (MAE = 0.64%, RMSE = 0.82%, R2 = 0.68), and lastly Cubist 3 with predictors showing a significantly positive correlation (MAE = 0.69%, RMSE = 0.90%, R2 = 0.62). The Cubist 1 model was considered more promising for explaining the complex relationships between SOC and the pXRF data used. Moreover, for the estimation of SOC in temperate floodplain soils all the Cubist models gave an acceptable model. However, future research should focus on using other auxiliary data [e.g., soil properties, data from other sensors (e.g., FieldSpec)] as well as extend the study area to cover more soil types hence improve model robustness as well as parsimoniousness.

Prediction of Soil Clay Content and Cation Exchange Capacity Using Visible Near-Infrared Spectroscopy, Portable X-ray Fluorescence, and X-ray Diffraction Techniques.

This article investigates a novel data fusion method to predict clay content and cation exchange capacity using visible near-infrared (visNIR) spectroscopy, portable X-ray fluorescence (pXRF), and X-ray diffraction (XRD) techniques. A total of 367 soil samples from two study areas in regional Australia were analyzed and intra- and interarea calibration options were explored. Cubist models were constructed using information from each device independently and in combination. pXRF produced the most accurate predictions of any individual device. Models based on fused data significantly improved the accuracy of predictions compared with those based on individual devices. The combination of pXRF and visNIR had the greatest performance. Overall, the relative increase in Lin's concordance correlation coefficient ranged from 1% to 12% and the corresponding decrease in root-mean-square error (RMSE) ranged from 10% to 46%. Provision of XRD data resulted in a decrease in observed RMSE values, although differences were not significant. Validation metrics were less promising when models were calibrated in one study area and then transferred to the other. Observed RMSE values were ∼2 to 3 times larger under this model transfer scenario and independent use of XRD was found to have the best overall performance.

Comparison of Multi-Frequency and Multi-Coil Electromagnetic Induction (EMI) for Mapping Properties in Shallow Podsolic Soils.

Electromagnetic induction (EMI) technique is an established method to measure the apparent electrical conductivity (ECa) of soil as a proxy for its physicochemical properties. Multi-frequency (MF) and multi-coil (MC) are the two types of commercially available EMI sensors. Although the working principles are similar, their theoretical and effective depth of investigation and their resolution capacity can vary. Given the recent emphasis on non-invasive mapping of soil properties, the selection of the most appropriate instrument is critical to support robust relationships between ECa and the targeted properties. In this study, we compared the performance of MC and MF sensors by their ability to define relationships between ECa (i.e., MF-ECa and MC-ECa) and shallow soil properties. Field experiments were conducted under wet and dry conditions on a silage-corn field in western Newfoundland, Canada. Relationships between temporally stable properties, such as texture and bulk density, and temporally variable properties, such as soil water content (SWC), cation exchange capacity (CEC) and pore water electrical conductivity (ECw) were investigated. Results revealed significant (p < 0.05) positive correlations of ECa to silt content, SWC and CEC for both sensors under dry conditions, higher correlated for MC-ECa. Under wet conditions, correlation of MF-ECa to temporally variable properties decreased, particularly to SWC, while the correlations to sand and silt increased. We concluded that the MF sensor is more sensitive to changes in SWC which influenced its ability to map temporally variable properties. The performance of the MC sensor was less affected by variable weather conditions, providing overall stronger correlations to both, temporally stable or variable soil properties for the tested Podzol and hence the more suitable sensor toward various precision agricultural practices.

Combined Use of Vis-NIR and XRF Sensors for Tropical Soil Fertility Analysis: Assessing Different Data Fusion Approaches.

Visible and near infrared (vis-NIR) diffuse reflectance and X-ray fluorescence (XRF) sensors are promising proximal soil sensing (PSS) tools for predicting soil key fertility attributes. This work aimed at assessing the performance of the individual and combined use of vis-NIR and XRF sensors to predict clay, organic matter (OM), cation exchange capacity (CEC), pH, base saturation (V), and extractable (ex-) nutrients (ex-P, ex-K, ex-Ca, and ex-Mg) in Brazilian tropical soils. Individual models using the data of each sensor alone were calibrated using multiple linear regressions (MLR) for the XRF data, and partial least squares (PLS) regressions for the vis-NIR data. Six data fusion approaches were evaluated and compared against individual models using relative improvement (RI). The data fusion approaches included (i) two spectra fusion approaches, which simply combined the data of both sensors in a merged dataset, followed by support vector machine (SF-SVM) and PLS (SF-PLS) regression analysis; (ii) two model averaging approaches using the Granger and Ramanathan (GR) method; and (iii) two data fusion methods based on least squares (LS) modeling. For the GR and LS approaches, two different combinations of inputs were used for MLR. The GR2 and LS2 used the prediction of individual sensors, whereas the GR3 and LS3 used the individual sensors prediction plus the SF-PLS prediction. The individual vis-NIR models showed the best results for clay and OM prediction (RPD ≥ 2.61), while the individual XRF models exhibited the best predictive models for CEC, V, ex-K, ex-Ca, and ex-Mg (RPD ≥ 2.57). For eight out of nine soil attributes studied (clay, CEC, pH, V, ex-P, ex-K, ex-Ca, and ex-Mg), the combined use of vis-NIR and XRF sensors using at least one of the six data fusion approaches improved the accuracy of the predictions (with RI ranging from 1 to 21%). In general, the LS3 model averaging approach stood out as the data fusion method with the greatest number of attributes with positive RI (six attributes; namely, clay, CEC, pH, ex-P, ex-K, and ex-Mg). Meanwhile, no single approach was capable of exploiting the synergism between sensors for all attributes of interest, suggesting that the selection of the best data fusion approach should be attribute-specific. The results presented in this work evidenced the complementarity of XRF and vis-NIR sensors to predict fertility attributes in tropical soils, and encourage further research to find a generalized method of data fusion of both sensors data.

Data fusion for the measurement of potentially toxic elements in soil using portable spectrometers.

Soil contamination posed by potentially toxic elements is becoming more serious under continuously development of industrialization and the abuse of fertilizers and pesticides. The investigation of soil potentially toxic elements is therefore urgently needed to ensure human and other organisms' health. In this study, we investigated the feasibility of the separate and combined use of portable X-ray fluorescence (pXRF) and visible near-infrared reflectance (vis-NIR) sensors for measuring eight potentially toxic elements in soil. Low-level fusion was achieved by the direct combination of the pXRF and vis-NIR spectra; middle-level fusion was achieved by the combination of selected bands of the pXRF and vis-NIR spectra using the Boruta feature selection algorithm; and high-level fusion was conducted by outer-product analysis (OPA) and Granger-Ramanathan averaging (GRA). The estimation accuracy for the eight considered elements were in the following order: Zn > Cu > Ni > Cr > As > Cd > Pb > Hg. The measurement for Cu and Zn could be achieved by pXRF spectra alone with Lin's concordance correlation coefficient (LCCC) values of 0.96 and 0.98, and ratio of performance to interquartile distance (RPIQ) values of 2.36 and 2.69, respectively. The measurement of Ni had the highest model performance for high-level fusion GRA with LCCC of 0.89 and RPIQ of 3.42. The measurements of Cr using middle- and high-level fusion were similar, with LCCC of 0.86 and RPIQ of 2.97. The best estimation accuracy for As, Cd, and Pb were obtained by high-level fusion using OPA, with LCCC >0.72 and RPIQ >1.2. However, Hg measurement by these techniques failed, having an unacceptable performance of LCCC <0.20 and RPIQ <0.75. These results confirm the effectiveness of using portable spectrometers to determine the contents of several potentially toxic elements in soils.

Comparison of Calibration Approaches in Laser-Induced Breakdown Spectroscopy for Proximal Soil Sensing in Precision Agriculture.

The lack of soil data, which are relevant, reliable, affordable, immediately available, and sufficiently detailed, is still a significant challenge in precision agriculture. A promising technology for the spatial assessment of the distribution of chemical elements within fields, without sample preparation is laser-induced breakdown spectroscopy (LIBS). Its advantages are contrasted by a strong matrix dependence of the LIBS signal which necessitates careful data evaluation. In this work, different calibration approaches for soil LIBS data are presented. The data were obtained from 139 soil samples collected on two neighboring agricultural fields in a quaternary landscape of northeast Germany with very variable soils. Reference analysis was carried out by inductively coupled plasma optical emission spectroscopy after wet digestion. The major nutrients Ca and Mg and the minor nutrient Fe were investigated. Three calibration strategies were compared. The first method was based on univariate calibration by standard addition using just one soil sample and applying the derived calibration model to the LIBS data of both fields. The second univariate model derived the calibration from the reference analytics of all samples from one field. The prediction is validated by LIBS data of the second field. The third method is a multivariate calibration approach based on partial least squares regression (PLSR). The LIBS spectra of the first field are used for training. Validation was carried out by 20-fold cross-validation using the LIBS data of the first field and independently on the second field data. The second univariate method yielded better calibration and prediction results compared to the first method, since matrix effects were better accounted for. PLSR did not strongly improve the prediction in comparison to the second univariate method.

Simplifying Sample Preparation for Soil Fertility Analysis by X-ray Fluorescence Spectrometry.

Portable X-ray fluorescence (pXRF) sensors allow one to collect digital data in a practical and environmentally friendly way, as a complementary method to traditional laboratory analyses. This work aimed to assess the performance of a pXRF sensor to predict exchangeable nutrients in soil samples by using two contrasting strategies of sample preparation: pressed pellets and loose powder (<2 mm). Pellets were prepared using soil and a cellulose binder at 10% w w-1 followed by grinding for 20 min. Sample homogeneity was probed by X-ray fluorescence microanalysis. Exchangeable nutrients were assessed by pXRF furnished with a Rh X-ray tube and silicon drift detector. The calibration models were obtained using 58 soil samples and leave-one-out cross-validation. The predictive capabilities of the models were appropriate for both exchangeable K (ex-K) and Ca (ex-Ca) determinations with R2 ≥ 0.76 and RPIQ > 2.5. Although XRF analysis of pressed pellets allowed a slight gain in performance over loose powder samples for the prediction of ex-K and ex-Ca, satisfactory performances were also obtained with loose powders, which require minimal sample preparation. The prediction models with local samples showed promising results and encourage more detailed investigations for the application of pXRF in tropical soils.

A multi-source data fusion approach to assess spatial-temporal variability and delineate homogeneous zones: A use case in a table grape vineyard in Greece.

Precision Viticulture requires very fine-scale spatial and temporal resolution to assess quite accurately variation in a vineyard. Many studies have used proximal sensing technology and spatial-temporal data analysis to characterize the local variation of plant vigour over time. The objective of this study was to present the potential of multivariate geostatistical techniques to fuse multi-temporal data from a multi-band radiometer and a geophysical sensor with different support for delineation of a vineyard into homogeneous zones, to be submitted to differential agricultural management. The study was conducted in a commercial table grape vineyard located in southern Greece during the years 2016 and 2017. Soil electrical conductivity was measured using an EM38 sensor, while Crop Circle canopy sensor, with the sensor located at 1.5 m height from the soil surface and 1.2 m horizontally from the vines, was used for scanning the side canopy area at different crop stages. The temporal multi-sensor data were analysed with the geostatistical data fusion techniques of block cokriging, to produce thematic maps, and factorial block cokriging to estimate synthetic scale-dependent regionalized factors. The factor maps at different scales are characterised by random variability with several micro-structures of different plant and soil properties, which leads to difficulties in delineating macro-areas with homogeneous features. In such conditions, high resolution VRA technology should be preferred to management by homogeneous zones for precision viticulture. The results have shown the potential of the proposed approach to deal with multi-source data in precision viticulture. However, further statistical research on data fusion of the outcomes from different sensors is still needed.

Improving In-Situ Estimation of Soil Profile Properties Using a Multi-Sensor Probe.

Optical diffuse reflectance spectroscopy (DRS) has been used for estimating soil physical and chemical properties in the laboratory. In-situ DRS measurements offer the potential for rapid, reliable, non-destructive, and low cost measurement of soil properties in the field. In this study, conducted on two central Missouri fields in 2016, a commercial soil profile instrument, the Veris P4000, acquired visible and near-infrared (VNIR) spectra (343⁻2222 nm), apparent electrical conductivity (ECa), cone index (CI) penetrometer readings, and depth data, simultaneously to a 1 m depth using a vertical probe. Simultaneously, soil core samples were obtained and soil properties were measured in the laboratory. Soil properties were estimated using VNIR spectra alone and in combination with depth, ECa, and CI (DECS). Estimated soil properties included soil organic carbon (SOC), total nitrogen (TN), moisture, soil texture (clay, silt, and sand), cation exchange capacity (CEC), calcium (Ca), magnesium (Mg), potassium (K), and pH. Multiple preprocessing techniques and calibration methods were applied to the spectral data and evaluated. Calibration methods included partial least squares regression (PLSR), neural networks, regression trees, and random forests. For most soil properties, the best model performance was obtained with the combination of preprocessing with a Gaussian smoothing filter and analysis by PLSR. In addition, DECS improved estimation of silt, sand, CEC, Ca, and Mg over VNIR spectra alone; however, the improvement was more than 5% only for Ca. Finally, differences in estimation accuracy were observed between the two fields despite them having similar soils, with one field demonstrating better results for all soil properties except silt. Overall, this study demonstrates the potential for in-situ estimation of profile soil properties using a multi-sensor approach, and provides suggestions regarding the best combination of sensors, preprocessing, and modeling techniques for in-situ estimation of profile soil properties.

Estimating lead and zinc concentrations in peri-urban agricultural soils through reflectance spectroscopy: Effects of fractional-order derivative and random forest.

Heavy metal contamination of peri-urban agricultural soil is detrimental to soil environmental quality and human health. A rapid assessment of soil pollution status is fundamental for soil remediation. Heavy metals can be monitored by visible and near-infrared spectroscopy coupled with chemometric models. First and second derivatives are two commonly used spectral preprocessing methods for resolving overlapping peaks. However, these methods may lose the detailed spectral information of heavy metals. Here, we proposed a fractional-order derivative (FOD) algorithm for preprocessing reflectance spectra. A total of 170 soil samples were collected from a typical peri-urban agricultural area in Wuhan City, Hubei Province. The reflectance spectra and lead (Pb) and zinc (Zn) concentrations of the samples were obtained in the laboratory. Two calibration methods, namely, partial least square regression and random forest (RF), were used to establish the relation between the spectral data and the two heavy metals. In addition, we aimed to explore the use of spectral estimation mechanism to predict the Pb and Zn concentrations. Three model evaluation parameters, namely, coefficient of determination (R2), root mean squared error, and ratio of performance to inter-quartile range (RPIQ), were used. Overall, the spectral reflectance decreased with the increase in Pb and Zn contents. The FOD algorithm gradually removed spectral baseline drifts and overlapping peaks. However, the spectral strength slowly decreased with the increase in fractional order. High fractional-order spectra underwent more spectral noises than low fractional-order spectra. The optimal prediction accuracies were achieved by the 0.25- and 0.5-order reflectance RF models for Pb (validation R2 = 0.82, RPIQ = 2.49) and Zn (validation R2 = 0.83, RPIQ = 2.93), respectively. A spectral detection of Pb and Zn mainly relied on their covariation with soil organic matter, followed by Fe. In summary, our results provided theoretical bases for the rapid investigation of Pb and Zn pollution areas in peri-urban agricultural soils.