Analysis of Precision Agriculture Datasets for on -Farm Research

"Data collected with ""Precision Agriculture"" (PA) technologies can provide new insights into soil plant relationships and how these relationships are affected by management practices. However, to analyze these datasets requires the consideration of limitations due to differential spatial scaling, multicollinearity, and autocorrelation. The main objective of the presented studies is to evaluate methods of analysis for large datasets generated on farm using PA technologies. In order to achieve my objective, I presented two major research studies. The study presented in Chapter II, focused on the delineation of zones of consistent productivity and their relationships with site properties. Furthermore, the study evaluated the effectiveness of two discriminant analysis methods, one of the parametric and the other non-parametric, to predict these areas using site properties. Site attributes explained clusters of consistent low (cluster 4) and high (cluster 1) yields differentially for the fields under study. In the Bellflower, IL, field, cluster 4 areas had greater soil reflectance in all the spectral bands and higher elevation associated with low soil organic matter (SOM). In both Centralia fields, MO, clusters 4 showed higher electrical conductivity (ECdeep and ECshallow ) values than clusters 1. This difference is explained by the presence of shallow claypans that impede root development and restrict nutrient and water availability. The study explained in Chapter III, predicted soybean composition using site attributes and hyperspectral remote sensing images. This study showed techniques to overcome issues of multicollinearity of the independent variables and the autocorrelation in the error term. Areas of greater SOM electrical conductivity, lower elevation, lower slope, and lower soil reflectance were related to greater soybean protein and lower oil seed concentrations, inconstantly. Information integrating site and season conditions as the vegetation indices from remote sensing images collected at the end of the season were more important predictors of soybean oil and protein seed concentration."

Name thesis:degree_name

Ph.D.

Level thesis:degree_level

Dissertation

Discipline thesis:degree_discipline

Crop Sciences

Grantor

University of Illinois at Urbana-Champaign

Year dc:date

2015

Author dc:creator

• Martin, Nicolas Federico

Contributors dc:contributor

• Bullock, Donald G.

• Agriculture, Agronomy

Language dc:language

eng

Identifier

(MiAaPQ)AAI3242931

OAI identifier oai:identifier

oai:www.ideals.illinois.edu:2142/85029