Banca de DEFESA: JOÃO PEDRO BESSA LARANGEIRA

Uma banca de DEFESA de MESTRADO foi cadastrada pelo programa.
STUDENT : JOÃO PEDRO BESSA LARANGEIRA
DATE: 30/04/2025
TIME: 14:00
LOCAL: UFRRJ-Depto de Solos
TITLE:

An automated system for the classification of diagnostic horizons in the brazilian soil classification system (sibcs)

KEY WORDS:

SiBCS, soil classification, Python, expert systems, pedology.

PAGES: 58
BIG AREA: Ciências Agrárias
AREA: Agronomia
SUBÁREA: Ciência do Solo
SPECIALTY: Gênese, Morfologia e Classificação dos Solos
SUMMARY:

This study presents an innovative computational model for the automated classification of soil horizons according to the Brazilian Soil Classification System (SiBCS). The research is justified by the complexity and subjectivity inherent in manual classification processes, which often lead to errors and inconsistencies, particularly given the periodic updates to SiBCS. The main objective was to develop a rule-based system using the Python programming language to identify and classify surface and subsurface diagnostic horizons, with the capability to integrate into the MultiSoils platform to broaden its applicability. The methodology involved encoding SiBCS criteria in Python, with emphasis on the py_rules library, followed by model validation through 18 tests based on real soil profiles, including Oxisols, Inceptisols, Ultisols, Gleysols, Alfisols, Vertisols, Planosols, and Plinthosols, adapted from previous studies and Brazilian Soil Classification and Correlation Meetings (RCCs). The results demonstrated the system's effectiveness in consistently classifying horizons, although challenges were identified, particularly in parameterizing spodic horizons due to ambiguities in SiBCS criteria. The study underscores the importance of objective diagnostic attributes to facilitate automation, as well as the potential for integration with existing tools such as Embrapa's SmartSolos. It is concluded that the developed model is viable and functional, representing a significant advancement for digital pedology, with applications in precision agriculture, environmental management, and education. However, the need for continuous system updates to align with SiBCS revisions and improve accuracy is emphasized. This work contributes to the literature by proposing a technological solution that combines pedological and computational knowledge, reducing reliance on human experts and enhancing the reproducibility of soil classifications

COMMITTEE MEMBERS:
Presidente - 1220296 - MARCOS BACIS CEDDIA
Interno - 1060711 - MARCOS GERVASIO PEREIRA
Externo à Instituição - DIEGO NUNES BRANDÃO - CEFET/RJ