Banca de DEFESA: GUILHERME FREITAS DA SILVA

Uma banca de DEFESA de MESTRADO foi cadastrada pelo programa.
STUDENT : GUILHERME FREITAS DA SILVA
DATE: 17/12/2024
TIME: 09:00
LOCAL: PPGMEG
TITLE:

DIGITAL MAPPING OF SOIL CLASSES USING TOPOGRAPHIC AND AERO GEOPHYSICAL DATA IN THE HYDROLOGICAL PLANNING UNIT OF THE GUANDU-MIRIM RIVER AND COASTAL BASINS, RJ

KEY WORDS:

Pedometry; machine learning; gamma spectrometric data; source material.

PAGES: 56
BIG AREA: Ciências Exatas e da Terra
AREA: Geociências
SUBÁREA: Geologia
SPECIALTY: Cartografia Geológica
SUMMARY:

Soil mapping is fundamental for the proper planning of activities related to land use, such as agriculture, the construction of roads and buildings, or the zoning of industrial areas, among others. However, soil mapping involves extensive and costly work, which often results in inefficient results. In this sense, digital soil mapping tools have been used to improve quality, measure the intrinsic error in soil maps, and with a better cost-benefit ratio. Some of the main characteristics and classifications of soils are influenced by factors in their formation, known as S.C.O.R.P.A.N. factors, which stand for Soil, Climate, Organisms, Relief, Parent Material, Age and Position. Digital elevation models (DEMs) and their derived data are the most widely used covariates to represent topographic conditions, playing an important role in digital soil mapping. Topographic features such as altimetry, slope, curvature and others that represent the relief of an area are effectively used to predict soil characteristics. As most existing soils are mineral soils, and their mineral composition reflects their properties and characteristics, the use of covariates that directly reflect characteristics of the parent material, such as Aero Geophysical Data (AGD), is justified in order to improve the quality of digital soil mapping (DSM). The aim of this study was to use topographic and aero-geophysical data to map the spatial distribution of soil classes in the Hydrological Planning Unit of the Guandu-Mirim River Basin and Coastal Basins, in Rio de Janeiro, Brazil. The Decision Tree and Random Forest predictive models that obtained the best statistical results used DAG as covariates and obtained overall accuracies of 0.87 and 0.97 respectively. Among the covariates derived from DAGs, uranium appeared among the 10 most important covariates for all the models that used DAGs. The results obtained indicate that the use of DAG as a representative of the source material formation factor can contribute to obtaining more accurate maps.

COMMITTEE MEMBERS:
Presidente - 2223668 - HELENA SARAIVA KOENOW PINHEIRO
Interno - 386982 - ALEXIS ROSA NUMMER
Externo à Instituição - MATEUS MARQUES BUENO