Proximal soil sensing: quantification of physical and chemical soil attributes

The objective of this work was to investigate soil sensing techniques and to analyze the potential for their use directly in the field. Four distinct steps were developed to meet the following objectives: a) to compare and evaluate the potential of predicting soil attributes with three portable spectrometers (vis-NIR) in a controlled environment; B) to evaluate the on-the-go prediction of pH, P and K using optical sensors (vis-NIR reflectance) and electrochemical techniques (ion-selective pH and K electrodes) in an experimental area with induced variability; C) assemble and test a field platform with electrical, electrochemical and optical sensors; D) to evaluate the potential of predicting soil texture using a portable X-ray fluorescence spectroscopy equipment. In the first step the vis-NIR spectral reflectance readings of the three equipments evaluated were very similar, with correlation coefficients above 0.86 in the 400 to 1800 nm range. When compared in this spectral region, the equipment produced very similar prediction models, with slight superiority for the FieldSpec system. The models showed to be more promising for the prediction of soil texture in relation to chemical attributes. In the second step the field readings using ion-selective pH and K electrodes presented a high correlation with the laboratory analyzes. The evaluations showed that soil conditions with low moisture significantly affect the readings. Despite the high correlation, the field values need to be corrected for the desired laboratory methodology. Prediction models of P, K and pH using field-vis-NIR spectrometry showed low precision. The tests of the Multisensors Soil Platform (MSP) in the third stage demonstrated that it is possible to use electric, electrochemical and optical sensors in the same platform. The electrical conductivity (EC) readings showed that this parameter was related with soil texture, acting as an indicator of variability and allowing the identification of texture transitions. The pH measured by MSP exhibited correlations below those verified in the second step. However, some atypical results were verified, such as the higher correlation between pH in CaCl2 and MSP than in H2O. The vis-NIR reflectance readings using the MSP resulted in good sand and clay prediction models, allowing the creation of high resolution maps of these parameters. Portable X-ray fluorescence spectroscopy was efficient for estimating soil texture. The sand and clay contents were estimated both by simple linear regressions and multiple regressions with R2 values above 0.60. Total Fe was the main element used in these regression models. (AU)