• JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
 
  Bookmark and Share
 
 
Doctoral Thesis
DOI
https://doi.org/10.11606/T.18.2023.tde-19022024-102606
Document
Author
Full name
Ana Carolina Martins Cidin
Institute/School/College
Knowledge Area
Date of Defense
Published
São Carlos, 2023
Supervisor
Committee
Crestana, Silvio (President)
Boas, Paulino Ribeiro Villas
Ferreira, Ednaldo Jose
Mauad, Frederico Fabio
Moraes, Jose Reinaldo da Silva Cabral de
Title in Portuguese
Desenvolvimento de funções de pedotransferência para estimativa de teor de carbono em solos brasileiros
Keywords in Portuguese
carbono orgânico
Geoestatística
modelo linear
solos tropicais
Abstract in Portuguese
O setor agrícola é a maior fonte de emissão de dióxido de carbono (CO2) no país e a estratégia global é ampliar a adoção de tecnologias agropecuárias sustentáveis, como a redução das emissões de carbono (C) pelas atividades de uso e manejo do solo, promovendo o aumento dos estoques de C no solo. Em razão disso, a necessidade de medir a quantidade de C nos solos tem sido discutida, com o intuito de se obter estimativas mais precisas de valores de C armazenados para compor inventários e subsidiar pesquisas e ampliar a agricultura de baixo C. O objetivo deste trabalho foi desenvolver funções de pedotransferências (PTFs) para estimar os teores de C no solo, a partir de um banco de dados de solos brasileiros. Foram utilizados teores de areia, silte e pH (H2O) como variáveis preditoras e as PTFs foram desenvolvidas a partir de modelo linear (LM), modelo linear generalizado (GLM) e modelo misto linear generalizado com efeitos aleatórios espacialmente correlacionados (GLMMs), considerando distribuição gama (GAM) e gaussiana (GAU) e função de ligação log e de identidade, respectivamente. Foram desenvolvidas PTFs, considerando a base de dados completa, os dados agrupados por cada bioma brasileiro, os dados agrupados por bioma e classe de solo e os dados agrupados por bioma, classe de solo e uso do solo. As metodologias GLMMs apresentaram melhores resultados, com R2 ajustado que variou de 0,40 a 0,96, comparados com LM e GLM. As PTFs para Amazônia, Cerrado, Mata Atlântica subgrupos MA2 e MA3 apresentaram melhores resultados desenvolvidas a partir do GLMMs GAM, refletindo em menores valores de RMSE de calibração e validação. As PTFs para Caatinga, Mata Atlântica subgrupo MA1, Pampa e Pantanal apresentaram melhores resultados quando desenvolvidas por GLMMs GAU. Nas amostras agrupadas por biomas e classes de solo, os resultados das PTFs nas metodologias GLMMs foram superiores aos observados nas PTFs desenvolvidas somente com amostras agrupadas por biomas, com exceção da PTF para Latossolo na Amazônia e Luvissolo na Caatinga. As PTFs apresentaram resultados com tendência a superestimar os teores de C no solo.
Title in English
Development of pedotransfer functions for estimating carbon content in Brazilian soils
Keywords in English
Geostatistics
linear model
organic carbon
tropical soils
Abstract in English
The agricultural sector is the largest source of carbon dioxide (CO2) emissions in the country, and the global strategy is to expand the adoption of sustainable agricultural technologies, such as reducing carbon (C) emissions from land use and management activities, promoting increased soil C stocks. As a result, the need to measure the amount of C in soils has been discussed to obtain more accurate estimates of stored C values for inventory purposes and to support research and expand low-carbon agriculture. The aim of this study was to develop pedotransfer functions (PTFs) for estimate soil C content based on a database of Brazilian soils. Sand, silt, and pH (H2O) contents were used as predictor variables, and the PTFs were developed using linear models (LM), generalized linear models (GLM), and generalized linear mixed models with spatially correlated random effects (GLMMs), considering gamma (GAM) and Gaussian (GAU) distributions, and using log and identity link functions, respectively. PTFs were developed for the complete database, grouped by each Brazilian biome, data grouped by biome and soil class, and data grouped by biome, soil class, and land use. The GLMMs methodologies showed better fit, with adjusted R2 ranging from 0.40 to 0.96, compared to LM and GLM. The PTFs for the Amazon, Cerrado, and Mata Atlântica subgroups MA2 and MA3 performed better when developed using GLMMs GAM, resulting in lower calibration and validation root mean square errors (RMSE). The PTFs for Caatinga, Mata Atlântica subgroup MA1, Pampa, and Pantanal performed better when developed using GLMMs GAU. In the samples grouped by biomes and soil classes, the results of the PTFs using GLMMs methodologies were better than those observed in the PTFs developed only with samples grouped by biomes, except for the PTF for Ferralsols in the Amazon and Luvisols in the Caatinga. The PTFs tended to overestimate soil C content.
 
WARNING - Viewing this document is conditioned on your acceptance of the following terms of use:
This document is only for private use for research and teaching activities. Reproduction for commercial use is forbidden. This rights cover the whole data about this document as well as its contents. Any uses or copies of this document in whole or in part must include the author's name.
Publishing Date
2024-03-04
 
WARNING: Learn what derived works are clicking here.
All rights of the thesis/dissertation are from the authors
CeTI-SC/STI
Digital Library of Theses and Dissertations of USP. Copyright © 2001-2024. All rights reserved.