Achtung: hohPublica wurde am 18.11.2024 aktualisiert. Falls Sie auf Darstellungsfehler stoßen, löschen Sie bitte Ihren Browser-Cache (Strg + Umschalt + Entf). *** Attention: hohPublica was last updated on November 18, 2024. If you encounter display errors, please delete your browser cache (Ctrl + Shift + Del).

Institut für Pflanzenproduktion und Agrarökologie in den Tropen und Subtropen

Permanent URI for this collectionhttps://hohpublica.uni-hohenheim.de/handle/123456789/15

Browse

Browsing Institut für Pflanzenproduktion und Agrarökologie in den Tropen und Subtropen by Subject "Afrika"

Type the first few letters and click on the Browse button
Now showing 1 - 1 of 1
  • Thumbnail Image
    Publication
    Prediction of soil properties for agricultural and environmental applications from infrared and X-ray soil spectral properties
    (2013) Towett, Erick Kibet; Cadisch, Georg
    Many of today?s most pressing problems facing developing countries, such as food security, climate change, and environmental protection, require large area data on soil functional capacity. Conventional assessments (methods and measurements) of soil capacity to perform specific agricultural and environmental functions are time consuming and expensive. In addition, repeatability, reproducibility and accuracy of conventional soil analytical data are major challenges. New, rapid methods to quantify soil properties are needed, especially in developing countries where reliable data on soil properties is sparse, and to take advantage of new opportunities for digital soil mapping. Mid infrared diffuse reflectance spectroscopy (MIR) has already shown promise as a rapid analytical tool and there are new opportunities to include other high-throughput techniques, such as total X-ray fluorescence (TXRF), and X-ray diffraction (XRD) spectroscopy. In this study TXRF and XRD were tested in conjunction with IR to provide powerful diagnostic capabilities for the direct prediction of key soil properties for agricultural and environmental applications especially for Sub-Saharan Africa (SSA) soils. Optimal combinations of spectral methods for use in pedotransfer functions for low cost, rapid prediction of chemical and physical properties of African soils as well as prediction models for soil organic carbon and soil fertility properties (soil extractable nutrients, pH and exchangeable acidity) were tested in this study. These state-of-the-art methods for large-area soil health measurement and monitoring will aid in accelerating economic development in developing sub-Saharan Africa countries with regards to climate change, increasing water scarcity and impacts on local and global food security as well as sustainable agricultural production and ecosystem resilience in the tropics. This study has developed and tested a method for the use of TXRF for direct quantification of total element concentrations in soils using a TXRF (S2 PICOFOXTM) spectrometer and demonstrated that TXRF could be used as a rapid screening tool for total element concentrations in soils assuming sufficient calibration measures are followed. The results of the current study have shown that TXRF can provide efficient chemical fingerprinting which could be further tested for inferring soil chemical and physical functional properties which is of interest in the African soil context for agricultural and environmental management at large scale. Further, this thesis has helped to improve understanding of the variation and patterns of element concentration data for 1034 soil samples from 34 stratified randomly-located 100-km2 ?sentinel? sites across SSA and explored the link between variability of soil properties and climate, parent material, vegetation types and land use patterns with the help of Random Forests statistics. Our results of total element concentration were within the range reported globally for soil Cr, Mn, Zn, Ni, V, Sr, and Y and in the high range for Al, Cu, Ta, Pb, and Ga. There were significant variations (P < 0.05) in total element composition within and between the sites for all the elements analysed. In addition, the greatest proportion of total variance and number of significant variance components occurred at the site (55-88%) followed by the cluster nested within site levels (10-40%). Our results also indicated that the strong observed within site as well as between site variations in many elements can serve to diagnose their soil fertility potential. Explorations of the relationships between element composition data and other site factors using ?randomForest? statistics have demonstrated that all site and soil-forming factors have important influence on total elemental concentrations in the soil with the most important variables explaining the main patterns of variation in total element concentrations being cluster, topography, landuse, precipitation and temperature. However, the importance of cluster can be explained by spatial correlation at distances of <1 km. This study has also analysed the potential of combining analyses undertaken using MIR spectroscopy and TXRF on 700 soil samples from 44 ?sentinel? sites distributed across SSA. MIR prediction models for soil organic carbon, and other soil fertility properties (such as soil extractable nutrients, pH, exchangeable acidity and soil texture) were developed using Random Forests (RF) regression and the current study has added total element concentration data to the residuals of the MIRS predictions to test how they can improve the MIR prediction accuracies. The RF approach out-perfomed the conventional partial least squares regression (PLSR) on simultaneous determination of soil properties; and in addition, RF results were also easily interpretable, computationally much faster and did not rely on data transformations or any other assumptions about data distributions compared to PLSR. With respect to the potential of combining TXRF and MIR spectra, including total element concentration data from TXRF analysis in the RF models significantly reduced root mean square error of prediction by 63% for Ecd, 54% for Mehlich-3 S, and 53% for Mehlich-3 Na. Thus, TXRF spectra were a useful supplement to improve prediction of soil properties not well predicted by MIRS. The prediction improvement from including TXRF was due to detection of a few outliers that did not appear as MIR spectral outliers. MIR showed remarkable ability to capture total elemental composition effects on physico-chemical soil properties but TXRF may have potential for outlier detection in large studies. This study has also helped to develop high-throughput spectral analytical methods and provided recommendations on optimal spectral analytical methods for the Globally Integrated Africa Soil Information Service (AfSIS) Project. Successfully developed methods in this study will become part of the standard AfSIS procedures.