Doctoral Degrees (Soil, Crop and Climate Sciences)
Permanent URI for this collection
Browse
Browsing Doctoral Degrees (Soil, Crop and Climate Sciences) by Subject "Apparent electrical conductivity"
Now showing 1 - 1 of 1
Results Per Page
Sort Options
Item Open Access Quantifying soil fertility parameters with electromagnetic induction, infrared reflectance spectroscopy and conventional chemistry procedures for wheat and maize under irrigation in arid climate(University of the Free State, 2021-04) Gura, Isaac; Du Preez, C. C.; Van Rensburg, L. D.; Barnard, J. H.Current global challenges, such as food security and soil quality, cannot be solved without up-to-date, high-quality, high-resolution, spatio-temporal, and continuous soil and environmental data that characterize soil and cropping ecosystems. Therefore, accurate and precise assessments of soil and crop characteristics are critical for site-specific management, vibrant soil condition and environmental sustainability. The inability to evaluate soil and crop characteristics quickly and inexpensively remains one of the main challenges of precision agriculture. Therefore, the ultimate aim of this study was to evaluate the use of soil sensors, viz the mid-infrared (MIR) sensor and the apparent electrical conductivity (ECa) sensor, in quantifying multiple soil fertility properties and their variability under irrigation. The study also attempted to apply the sensor data fusion approach to improve the assessment of multiple soil quality indicators and the overall soil quality under irrigation using the Soil Management Assessment Framework (SMAF). The established international ECa-directed soil sampling design approach was employed at each of the seven fields of interest by measuring apparent soil electrical conductivity (ECa) with a Geonics EM38-MK2 sensor (non-invasive geophysical electromagnetic induction, EMI). A “Response Surface Sampling Design” (RSSD) sampling methodology in the “Electrical Conductivity Sampling Assessment and Prediction” (ESAP) software was used to direct soil and crop sampling based on the degree of ECa variability. This methodology reduced sampling points from each field to 12 sampling points after an initial ECa survey. Soil samples from each field were analysed in the laboratory for various soil properties that are related to soil fertility. Wheat and maize were also sampled from the ECa directed sampling points at each field at the end of the winter 2016 season and 2016/17 summer season, respectively. The MIR spectra was obtained in the laboratory from the soil and crop samples from each field using a sensor iS50 Nicolet Fourier Transform Infrared (FTIR) (Thermo Fisher Scientific Inc., Waltham, MA) equipped with an accessory for attenuated reflectance acquisition (iS50 FTIR-ATR). The MIR sensor and EM38 sensor showed different levels of accuracy with respect to predicting soil fertility properties under irrigation. The study results demonstrated the effectiveness and usefulness of the MIR attenuated total reflectance (ATR) technique coupled with partial squares least regression (PLSR) in quantitative analysis of soil fertility properties. In contrast, the EM38 sensor modelled accurately only a few soil properties per site at a given sampling time. Comparatively, the model results from both sensors show that the MIR sensor produced better prediction models for most of the measured soil fertility properties than the EM38 sensor. For quantifying nutrient accumulation in wheat and maize, the MIR sensor technique produced more excellent predictive models for the nutrient concentrations in wheat samples than in maize. The results from the in-field spatial characterization of plant nutrient levels and crop yields at the study sites showed that although ECa readings may be useful for the spatial characterization of some soil fertility properties in non-saline and non-sodic soils in South Africa, the results showed many inconsistencies between sites and between the centre pivots. The limitations of quantifying soil properties and overall soil quality using a single soil sensor can be overcome by integrating data from conceptually different sensing techniques to improve model accuracy and robustness. The findings in this study demonstrated that models for most of the soil properties obtained based on step-wise multiple linear regression (SMLR) fusion of data from MIR sensor and EM38 sensor measurements were more robust as compared to models from individual sensors. The SMLR sensor fusion technique failed to improve the models of some soil properties at the selected fields as well as the overall SMAF soil quality index at the Douglas 40 ha field. A more robust fusion technique such as PLSR can be used to implement the data fusion for these properties. The sensor data fusion results demonstrate the superiority and efficiency of the sensor data fusion approach in the measurement of soil fertility properties and overall soil quality in irrigation systems of South Africa. Based on the findings of this study, for soil fertility evaluation and quantification, it is recommended to use the MIR technique coupled with PLSR and alternatively, the ECa measurements as complementary information to provide extended attribute coverage and increased capacity of the sensor data fusion.