Banca de QUALIFICAÇÃO: MICKAELLE MARIA DE ALMEIDA PEREIRA

Uma banca de QUALIFICAÇÃO de DOUTORADO foi cadastrada pelo programa.
STUDENT : MICKAELLE MARIA DE ALMEIDA PEREIRA
DATE: 25/08/2023
TIME: 09:00
LOCAL: https://meet.google.com/crq-foty-txx
TITLE:

Application of functional data analysis to CH4, CO2 and N2O emissions from different land uses

KEY WORDS:

functional curves, derivatives, flux, land use, continuous variation.

PAGES: 54
BIG AREA: Ciências Agrárias
AREA: Agronomia
SUMMARY:

Improper management of agricultural systems, as well as changes in land cover, are significant factors in the increase of greenhouse gases (GHGs). Statistical tools exist for measuring GHG fluxes, based on the temporal variation of concentration as a function of time. However, these analyses may not cover features arising from the variation and randomness present in the phenomenon. Thus, functional data analysis (FDA) consolidates a new perspective for deriving models and optimizing techniques in exploratory data analysis, expressing remarkable potential in the study of the variations of a given variable over time, considering both the continuous variation (linear or non-linear) and its randomness. The objective of the study was to evaluate the dynamics and complexity of carbon dioxide (CO2), methane (CH4) and nitrous oxide (N2O) emissions in four types of land use (extensive pasture, intensive pasture, row and between-row sugarcane) using FDA techniques from univariate and multivariate perspectives. Initially, gas flux estimates were measured using linear, exponential and functional models. The functional model showed values closer to the linear model than the exponential model, showing a better correlation and agreement between these two models. Next, the data were analyzed by functional statistical methods. The discrete observations were restructured as functions, using B-splines smoothing. Consequently, derivatives of the functions were applied to calculate the variation of concentrations as a function of time (fluxes), finding great variability for the land uses, alternating between effluxes and influxes. With the functional analysis of variance (FANOVA) it was observed that daily functional averages of gas emissions in the distinct soils do not differ significantly for CH4, adopting significance level at 5% probability. For CO2 and N2O there was a statistical difference. The first three multivariate principal functional components (MFPCA) cumulatively captured more than 90% of the total variation present in the data. Cluster analysis, referring to the principal component scores, separated the observations into four groups. Therefore, by encompassing the continuous nature of the system, the FDA accurately represented the process of gas exchange between soil and atmospheric air.

COMMITTEE MEMBERS:
Externa à Instituição - JANINA BRAGA DO CARMO - UFSCAR
Externo à Instituição - JOSE ROMUALDO DE SOUSA LIMA
Interno - ***.221.388-** - LUIZ ANTONIO MARTINELLI - USP
Presidente - PAULO JOSE DUARTE NETO
Interno - TIAGO ALESSANDRO ESPINOLA FERREIRA