Ambientipagem no melhoramento de plantas e Plataforma 3,5K de covariáveis ambientípicas

Abstract

Envirotyping and enviromics have gained prominence for enabling the detailed characterization of agricultural systems based on multiscale information about ecological conditions. This work aimed to integrate environmental data from public platforms (NASA POWER, SoilGrids, and WorldClim) through an automated and scalable pipeline, targeting the construction of continuous spatial grids of agricultural-space covariates and territorial analysis. Initially, a literature review is presented in chapter format, as a Portuguese version of the article published in the journal Molecular Plant, which discusses the conceptual foundations of enviromics, the potential of remote sensing, the challenges of data integration, and future perspectives in the field. Subsequently, a pipeline developed in Python was implemented, capable of performing automated acquisition, standardization, spatial interpolation, and integration of multiple sources of environmental data. From the generated matrix, Principal Component Analyses (PCA) were applied, highlighting ecological gradients relevant for genotype recommendation. The proposed methodology enabled the organization of 3,507 environmental layers and the generation of a georeferenced dataset with pixel-wise extraction of covariates, suitable for analyses focused on genotypes × environment (G×E) interaction. The resulting matrix structure provides an accessible interface for breeders, even without advanced knowledge in Geographic Information Systems (GIS). Thus, this dissertation offers both conceptual and practical contributions to the advancement of envirotyping in the context of genetic improvement, promoting the integration of statistics, geoinformation, and envirotyping platforms for practical application within plant breeding programs.