SPSoil DataBank : Estimation of soil properties using spectral measurements from the AGROSAVIA soil bank Proyecto de investigación uri icon

Descripción

  • Information of soil properties can be determinant in crop management. Correct and adequate availability of nutrients is an essential factor to ensure good economic performance in any agricultural production. Conventional methods of soil analysis in the laboratory constitute one of the limitations to carry out this type of study due to the high cost and time required, make them impractical for precision agriculture, refraining farmers from implementing optimal practices in their crop. We propose a spectral method for land property assessment based on near-infrared measurements of soil samples. We will apply this approach to the available soil sample bank in Agrosavia. More than 3K available standard lab results and their corresponding spectral signature will be use to train, test and validate machine learning predictions to assess estimation accuracy, precision and limitations. Our proposed tool will be available for any farmer, professional or scientist aiming to improve yield in crops

Estado Proyecto

  • Cerrado

Objetivo General

  • Perform an exploratory analysis of the conventional lab test results and NIRS spectra available at the Agrosavia soil lab facilities

Objetivos Específicos

  • -Build and test machine learning models for accuracy and precision in estimating soil properties from NIRS measurements. -Perform an exploratory analysis of the conventional soil lab test results, searching for combinations of soil properties that could potentially be useful in estimating a selected target property. -Labeling the data using known soil property standards given by our soil experts

Tipo Proyecto

  • I+D+i

Intervalo de fecha / hora

  • 2016-01-01T00:00:00