Open Access

As part of the BonaRes research initiative, funded by the BMBF, strategies are being developed to use soil as a sustainable resource in the bio economy. The interdisciplinary subproject I4S - “Intelligence for soil” - is responsible for the development of an integrated system for site-specific management of soil fertility. For this purpose, a platform is constructed and various sensors are installed. Real-time data will be summarised in models and decision-making algorithms will be used to control fertilisation and accordingly improve soil functions. Aim of the BAM in the frame of I4S is the characterisation of an energy-dispersive X-ray fluorescence (EDXRF) based sensor for robust determination of plant essential nutrients in soil. First a principal component analysis (PCA) was used to identify outliers and to observe the largest variance within the German soil samples. It could be monitored that splitting of the samples was due to their iron content. Given that clay samples contain high amounts of iron and sandy samples low amounts, a classification of the samples by their soil texture according to VD LUFA was possible. Considering the complex composition of soil, a matrix-specific calibration was carried out by univariate and multivariate data analysis. The figures of merit demonstrated that a more robust calibration model with negligible matrix effects can be obtained by a multivariate approach using partial least squares regression (PLSR). A better correlation between predicted values compared to reference values for German soil samples was observed for the chemometric calibration model than for the univariate one. Different factors can affect the received calibration models such as moisture and particle size distribution which is especially important due to later online analysis. In first studies the influence of moisture on the detection of plant essential nutrients was investigated. With increasing water content, the characteristic fluorescence peaks decrease and start to increase again at a water content of 15 %. With lower moisture content the soil agglomerates which leads to lower packing of the sample, resulting in a rougher surface which negatively influence the signals. Whereas, agglomerates are not formed at higher water content. This allows the sample to be packed more tightly thus a smoother surface and a better homogeneity is obtained. Furthermore, particle size distribution leads to significantly higher uncertainties and lower signals when comparing grounded (< 500 μm) and not grounded (< 2 mm) samples. This can be explained by amplifying of the already known inhomogeneity of soils. Both factors must be included in the chemometric PLSR to obtain robust calibration models for each macro and micro nutrient.

Within the framework of precision agriculture, the determination of various soil properties is moving into focus, especially the demand for sensors suitable for in-situ measurements. Energy-dispersive X-ray fluorescence (EDXRF) can be a powerful tool for this purpose. In this study a huge diverse soil set (n = 598) from 12 different study sites in Germany was analysed with EDXRF. First, a principal component analysis (PCA) was performed to identify possible similarities among the sample set. Clustering was observed within the four texture classes clay, loam, silt and sand, as clay samples contain high and sandy soils low iron mass fractions. Furthermore, the potential of uni- and multivariate data evaluation with partial least squares regression (PLSR) was assessed for accurate Determination of nutrients in German agricultural samples using two calibration sample sets. Potassium and iron were chosen for testing the performance of both models. Prediction of these nutrients in 598 German soil samples with EDXRF was more accurate using PLSR which is confirmed by a better overall averaged deviation and PLSR should therefore be preferred.

As part of the BonaRes research initiative, funded by the BMBF, strategies are being developed to use soil as a sustainable resource in the bio economy. The interdisciplinary subproject I4S - “Intelligence for soil” - is responsible for the development of an integrated system for site-specific management of soil fertility. For this purpose, a platform is constructed and various sensors are installed. Real-time data will be summarised in models and decision-making algorithms will be used to control fertilisation and accordingly improve soil functions. Aim of the BAM in the frame of I4S is the characterisation of an energy-dispersive X-ray fluorescence (EDXRF) based sensor for robust determination of plant essential nutrients in soil. First a principal component analysis (PCA) was used to identify outliers and to observe the largest variance within the German soil samples. It could be monitored that splitting of the samples was due to their iron content. Given that clay samples contain high amounts of iron and sandy samples low amounts, a classification of the samples by their soil texture according to VD LUFA was possible. Considering the complex composition of soil, a matrix-specific calibration was carried out by univariate and multivariate data analysis. The figures of merit demonstrated that a more robust calibration model with negligible matrix effects can be obtained by a multivariate approach using partial least squares regression (PLSR). A better correlation between predicted values compared to reference values for German soil samples was observed for the chemometric calibration model than for the univariate one. Different factors can affect the received calibration models such as moisture and particle size distribution which is especially important due to later online analysis. In first studies the influence of moisture on the detection of plant essential nutrients was investigated. With increasing water content, the characteristic fluorescence peaks decrease and start to increase again at a water content of 15 %. With lower moisture content the soil agglomerates which leads to lower packing of the sample, resulting in a rougher surface which negatively influence the signals. Whereas, agglomerates are not formed at higher water content. This allows the sample to be packed more tightly thus a smoother surface and a better homogeneity is obtained. Furthermore, particle size distribution leads to significantly higher uncertainties and lower signals when comparing grounded (< 500 μm) and not grounded (< 2 mm) samples. This can be explained by amplifying of the already known inhomogeneity of soils. Both factors must be included in the chemometric PLSR to obtain robust calibration models for each macro and micro nutrient.

Mit wachsenden Bevölkerungszahlen steigt auch der Rohstoffkonsum, und der nachhaltigere und effizientere Umgang mit knappen Ressourcen wie Wasser und Boden rückt in den Fokus. Im Rahmen der vom Bundesministerium für Bildung und Forschung (BMBF) geförderten Forschungsinitiative BonaRes werden Strategien entwickelt, um Boden als nachhaltige Ressource in der Bioökonomie zu nutzen. Das interdisziplinäre Teilprojekt I4S – Intelligence for soil – ist dabei für die Entwicklung eines integrierten Systems zum ortsspezifischen Management der Bodenfruchtbarkeit zuständig. Hierfür wird eine Plattform gebaut, auf der verschiedene Sensoren installiert werden sollen, deren in Echtzeit erhaltene Messdaten in Modelle und Entscheidungsalgorithmen zur Steuerung der Düngung und dementsprechend Verbesserung der Bodenfunktionen einfließen sollen. So wären Untersuchungen in engmaschigen, dynamischen Rastern und schnelle Analysen großer Flächen möglich, um höhere Erträge zu erwirtschaften. Einer dieser Sensoren, auf dessen Grundlage ein robustes Online-Verfahren zur Bestimmung der Makro- und Mikronährstoffe, wie Ca, K und Mn in Böden entwickelt werden soll, ist die Röntgenfluoreszenzanalyse (RFA). Die RFA eignet sich vor allem durch ihre schnelle, kontaktlose, simultane Multielementanalyse und wird bereits zur Bestimmung von Schwermetallen in Böden eingesetzt. Ein weiterer Vorteil der RFA ist die geringe Probenvorbereitung. Um Messungen auf einem Feld zu simulieren, wurde in einen Laboraufbau ein Probenteller installiert, der sich bei verschiedenen Winkelgeschwindigkeiten bewegen lässt und so dynamisches Messen der als Spur aufgetragenen Probe ermöglicht. Gerade bei einer komplexen Matrix wie Boden, die eine breite Elementverteilung aufweist, ist die Datenauswertung ein wichtiger Faktor. Die bisher verwendete univariate Datenanalyse liefert gute Ergebnisse und zeigt, dass durch Kalibrierung der RFA mit 14 Referenzmaterialien große Ackerflächen ausgewertet werden können. Dies funktioniert aber nur, solange die Elementgehalte der Proben mit denen der Referenzmaterialien vergleichbar sind. Die zusätzlich verwendete multivariate Datenanalyse bietet hingegen die Möglichkeit, Modellrechnungen der Böden durchzuführen und Böden mithilfe der Hauptkomponentenanalyse (PCA) besser zu klassifizieren. Des Weiteren können basierend auf der Partial Least Squares Regression (PLSR) Kalibriermodelle erstellt werden, welche Prognosen zu den Elementgehalten unbekannter Ackerböden ermöglichen. Zusätzlich bietet die multivariate Datenanalyse die Möglichkeit, Störgrößen wie unterschiedliche Korngrößenverteilung und Feuchtigkeitsgrad der Probe in das Modell miteinzubeziehen. Beide Auswertemethoden sollen anhand statisch gemessener Proben verglichen werden und durch Messung einer Vielzahl realer Bodenproben von unterschiedlichen Standorten erweitert werden. Im weiteren Verlauf müssen beide Modelle auf dynamisch bewegte Proben übertragen werden.

Aim of the Federal Institute for Materials Research and Testing (BAM) in the frame of I4S - “Intelligence for Soil” is the characterization of an X-ray fluorescence (XRF) based sensor for robust online-application of arable land.Fast soil mapping for agricultural purpose allows the site-specific optimized introduction of plant essential nutrients like S, K, Ca, and Fe. This is important given that the distribution of minor and trace elements varies widely. The non-destructive and contactless XRF is suitable for rapid in-situ analysis on the field due to minimal sample preparation and simultaneous multi-element analysis. Soils are already considered as a complex matrix due to their wide range of elements in different contents, especially light elements with low atomic numbers (Z<19). Problems by measuring soil samples also arise from heterogeneity of the sample and matrix effects. Large grain size distribution causes strong inhomogeneity and matrix effects occur through physical properties related to high concentration of main components. Matrix-specific calibration strategies for determination of total major and minor plant essential nutrients are particularly important regarding these difficulties. For accurate calibration, data treatment and evaluation must also be considered. Empirical univariate and multivariate data analysis were compared regarding their analytical figures of merit. Using principal component analysis (PCA) it was possible to classify German soils in different groups as sand, clay and silt. A calibration curve was obtained by partial least squares regression (PLSR) and the elemental content of German soils was predicted. Elemental distribution maps for different German arable lands were created and the results compared to reference measurements. The correlation between predicted values and reference values were in good agreement for most major and minor nutrients.